Add files using upload-large-folder tool
Browse files- .gitattributes +1 -0
- README.md +40 -0
- chat_template.jinja +86 -0
- config.json +65 -0
- configuration_llada2_moe.py +85 -0
- generation_config.json +7 -0
- model-00001-of-00012.safetensors +3 -0
- model-00002-of-00012.safetensors +3 -0
- model-00003-of-00012.safetensors +3 -0
- model-00004-of-00012.safetensors +3 -0
- model-00005-of-00012.safetensors +3 -0
- model-00006-of-00012.safetensors +3 -0
- model-00007-of-00012.safetensors +3 -0
- model-00008-of-00012.safetensors +3 -0
- model-00009-of-00012.safetensors +3 -0
- model-00010-of-00012.safetensors +3 -0
- model-00011-of-00012.safetensors +3 -0
- model-00012-of-00012.safetensors +3 -0
- model.safetensors.index.json +964 -0
- modeling_llada2_moe.py +1621 -0
- special_tokens_map.json +37 -0
- tokenizer.json +3 -0
- tokenizer_config.json +2115 -0
.gitattributes
CHANGED
|
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
+
tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
README.md
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
library_name: mlx
|
| 4 |
+
tags:
|
| 5 |
+
- dllm
|
| 6 |
+
- diffusion
|
| 7 |
+
- llm
|
| 8 |
+
- text_generation
|
| 9 |
+
- mlx
|
| 10 |
+
base_model: inclusionAI/LLaDA2.0-flash-preview
|
| 11 |
+
pipeline_tag: text-generation
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
+
# mlx-community/LLaDA2.0-flash-preview-4bit
|
| 15 |
+
|
| 16 |
+
This model [mlx-community/LLaDA2.0-flash-preview-4bit](https://huggingface.co/mlx-community/LLaDA2.0-flash-preview-4bit) was
|
| 17 |
+
converted to MLX format from [inclusionAI/LLaDA2.0-flash-preview](https://huggingface.co/inclusionAI/LLaDA2.0-flash-preview)
|
| 18 |
+
using mlx-lm version **0.28.4**.
|
| 19 |
+
|
| 20 |
+
## Use with mlx
|
| 21 |
+
|
| 22 |
+
```bash
|
| 23 |
+
pip install mlx-lm
|
| 24 |
+
```
|
| 25 |
+
|
| 26 |
+
```python
|
| 27 |
+
from mlx_lm import load, generate
|
| 28 |
+
|
| 29 |
+
model, tokenizer = load("mlx-community/LLaDA2.0-flash-preview-4bit")
|
| 30 |
+
|
| 31 |
+
prompt = "hello"
|
| 32 |
+
|
| 33 |
+
if tokenizer.chat_template is not None:
|
| 34 |
+
messages = [{"role": "user", "content": prompt}]
|
| 35 |
+
prompt = tokenizer.apply_chat_template(
|
| 36 |
+
messages, add_generation_prompt=True
|
| 37 |
+
)
|
| 38 |
+
|
| 39 |
+
response = generate(model, tokenizer, prompt=prompt, verbose=True)
|
| 40 |
+
```
|
chat_template.jinja
ADDED
|
@@ -0,0 +1,86 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{% set thinking_option = 'off' %}
|
| 2 |
+
{{- '<role>SYSTEM</role>' }}
|
| 3 |
+
{%- if messages[0].role == 'system' %}
|
| 4 |
+
{{- messages[0].content + '\n' }}
|
| 5 |
+
{%- endif %}
|
| 6 |
+
{%- if tools %}
|
| 7 |
+
{{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
|
| 8 |
+
{%- for tool in tools %}
|
| 9 |
+
{{- "\n" }}
|
| 10 |
+
{{- tool | tojson }}
|
| 11 |
+
{%- endfor %}
|
| 12 |
+
{{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n" }}
|
| 13 |
+
{%- endif %}
|
| 14 |
+
{{- 'detailed thinking ' + thinking_option + '<|role_end|>' }}
|
| 15 |
+
{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
|
| 16 |
+
{%- for message in messages[::-1] %}
|
| 17 |
+
{%- set index = (messages|length - 1) - loop.index0 %}
|
| 18 |
+
{%- if ns.multi_step_tool and message.role == "user" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
|
| 19 |
+
{%- set ns.multi_step_tool = false %}
|
| 20 |
+
{%- set ns.last_query_index = index %}
|
| 21 |
+
{%- endif %}
|
| 22 |
+
{%- endfor %}
|
| 23 |
+
{%- for message in messages %}
|
| 24 |
+
{%- if message.content is string %}
|
| 25 |
+
{%- set content = message.content %}
|
| 26 |
+
{%- else %}
|
| 27 |
+
{%- set content = '' %}
|
| 28 |
+
{%- endif %}
|
| 29 |
+
{%- if message.role == "user" %}
|
| 30 |
+
{{- '<role>HUMAN</role>' + message.content + '<|role_end|>' }}
|
| 31 |
+
{%- elif message.role == "system" and not loop.first %}
|
| 32 |
+
{{- '<role>SYSTEM</role>' + message.content + '<|role_end|>' }}
|
| 33 |
+
{%- elif message.role == "assistant" %}
|
| 34 |
+
{%- set reasoning_content = '' %}
|
| 35 |
+
{%- if message.reasoning_content is string %}
|
| 36 |
+
{%- set reasoning_content = message.reasoning_content %}
|
| 37 |
+
{%- else %}
|
| 38 |
+
{%- if '</think>' in content %}
|
| 39 |
+
{%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
|
| 40 |
+
{%- set content = content.split('</think>')[-1].lstrip('\n') %}
|
| 41 |
+
{%- endif %}
|
| 42 |
+
{%- endif %}
|
| 43 |
+
{%- if loop.index0 > ns.last_query_index %}
|
| 44 |
+
{%- if reasoning_content %}
|
| 45 |
+
{{- '<role>ASSISTANT</role>' + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
|
| 46 |
+
{%- else %}
|
| 47 |
+
{{- '<role>ASSISTANT</role>' + content }}
|
| 48 |
+
{%- endif %}
|
| 49 |
+
{%- else %}
|
| 50 |
+
{{- '<role>ASSISTANT</role>' + content }}
|
| 51 |
+
{%- endif %}
|
| 52 |
+
{%- if message.tool_calls %}
|
| 53 |
+
{%- for tool_call in message.tool_calls %}
|
| 54 |
+
{%- if (loop.first and content) or (not loop.first) %}
|
| 55 |
+
{{- '\n' }}
|
| 56 |
+
{%- endif %}
|
| 57 |
+
{%- if tool_call.function %}
|
| 58 |
+
{%- set tool_call = tool_call.function %}
|
| 59 |
+
{%- endif %}
|
| 60 |
+
{{- '<tool_call>\n{"name": "' }}
|
| 61 |
+
{{- tool_call.name }}
|
| 62 |
+
{{- '", "arguments": ' }}
|
| 63 |
+
{%- if tool_call.arguments is string %}
|
| 64 |
+
{{- tool_call.arguments }}
|
| 65 |
+
{%- else %}
|
| 66 |
+
{{- tool_call.arguments | tojson }}
|
| 67 |
+
{%- endif %}
|
| 68 |
+
{{- '}\n</tool_call>' }}
|
| 69 |
+
{%- endfor %}
|
| 70 |
+
{%- endif %}
|
| 71 |
+
{{- '<|role_end|>' }}
|
| 72 |
+
{%- elif message.role == "tool" %}
|
| 73 |
+
{%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
|
| 74 |
+
{{- '<role>OBSERVATION</role>' }}
|
| 75 |
+
{%- endif %}
|
| 76 |
+
{{- '\n<tool_response>\n' }}
|
| 77 |
+
{{- content }}
|
| 78 |
+
{{- '\n</tool_response>' }}
|
| 79 |
+
{%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
|
| 80 |
+
{{- '<|role_end|>' }}
|
| 81 |
+
{%- endif %}
|
| 82 |
+
{%- endif %}
|
| 83 |
+
{%- endfor %}
|
| 84 |
+
{%- if add_generation_prompt %}
|
| 85 |
+
{{- '<role>ASSISTANT</role>' }}
|
| 86 |
+
{%- endif %}
|
config.json
ADDED
|
@@ -0,0 +1,65 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"LLaDA2MoeModelLM"
|
| 4 |
+
],
|
| 5 |
+
"attention_dropout": 0.0,
|
| 6 |
+
"auto_map": {
|
| 7 |
+
"AutoConfig": "configuration_llada2_moe.LLaDA2MoeConfig",
|
| 8 |
+
"AutoModel": "modeling_llada2_moe.LLaDA2MoeModel",
|
| 9 |
+
"AutoModelForCausalLM": "modeling_llada2_moe.LLaDA2MoeModelLM"
|
| 10 |
+
},
|
| 11 |
+
"embedding_dropout": 0.0,
|
| 12 |
+
"first_k_dense_replace": 1,
|
| 13 |
+
"head_dim": 128,
|
| 14 |
+
"hidden_act": "silu",
|
| 15 |
+
"hidden_size": 4096,
|
| 16 |
+
"initializer_range": 0.02,
|
| 17 |
+
"intermediate_size": 9216,
|
| 18 |
+
"max_position_embeddings": 16384,
|
| 19 |
+
"max_window_layers": 28,
|
| 20 |
+
"model_type": "llada2_moe",
|
| 21 |
+
"moe_intermediate_size": 1024,
|
| 22 |
+
"moe_router_enable_expert_bias": true,
|
| 23 |
+
"n_group": 8,
|
| 24 |
+
"norm_head": false,
|
| 25 |
+
"norm_softmax": false,
|
| 26 |
+
"norm_topk_prob": true,
|
| 27 |
+
"num_attention_heads": 32,
|
| 28 |
+
"num_experts": 256,
|
| 29 |
+
"num_experts_per_tok": 8,
|
| 30 |
+
"num_hidden_layers": 32,
|
| 31 |
+
"num_key_value_heads": 4,
|
| 32 |
+
"num_shared_experts": 1,
|
| 33 |
+
"output_dropout": 0.0,
|
| 34 |
+
"output_router_logits": false,
|
| 35 |
+
"pad_token_id": 156892,
|
| 36 |
+
"partial_rotary_factor": 0.5,
|
| 37 |
+
"quantization": {
|
| 38 |
+
"group_size": 64,
|
| 39 |
+
"bits": 4,
|
| 40 |
+
"mode": "affine"
|
| 41 |
+
},
|
| 42 |
+
"quantization_config": {
|
| 43 |
+
"group_size": 64,
|
| 44 |
+
"bits": 4,
|
| 45 |
+
"mode": "affine"
|
| 46 |
+
},
|
| 47 |
+
"rms_norm_eps": 1e-06,
|
| 48 |
+
"rope_scaling": null,
|
| 49 |
+
"rope_theta": 600000,
|
| 50 |
+
"rotary_dim": 64,
|
| 51 |
+
"routed_scaling_factor": 2.5,
|
| 52 |
+
"router_dtype": "fp32",
|
| 53 |
+
"score_function": "sigmoid",
|
| 54 |
+
"tie_word_embeddings": false,
|
| 55 |
+
"topk_group": 4,
|
| 56 |
+
"torch_dtype": "bfloat16",
|
| 57 |
+
"transformers_version": "4.52.3",
|
| 58 |
+
"use_bias": false,
|
| 59 |
+
"use_cache": true,
|
| 60 |
+
"use_qkv_bias": false,
|
| 61 |
+
"use_rmsnorm": true,
|
| 62 |
+
"use_sliding_window": false,
|
| 63 |
+
"using_split_qkv_in_self_attention": false,
|
| 64 |
+
"vocab_size": 157184
|
| 65 |
+
}
|
configuration_llada2_moe.py
ADDED
|
@@ -0,0 +1,85 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""LLaDA2 MoE model configuration"""
|
| 2 |
+
|
| 3 |
+
from transformers.configuration_utils import PretrainedConfig
|
| 4 |
+
|
| 5 |
+
|
| 6 |
+
class LLaDA2MoeConfig(PretrainedConfig):
|
| 7 |
+
model_type = "llada2_moe"
|
| 8 |
+
|
| 9 |
+
def __init__(
|
| 10 |
+
self,
|
| 11 |
+
vocab_size=30592,
|
| 12 |
+
hidden_size=1024,
|
| 13 |
+
intermediate_size=None,
|
| 14 |
+
num_hidden_layers=24,
|
| 15 |
+
num_attention_heads=16,
|
| 16 |
+
num_key_value_heads=0,
|
| 17 |
+
hidden_act="silu",
|
| 18 |
+
use_qkv_bias=False, # llada2 only
|
| 19 |
+
use_qk_norm=False,
|
| 20 |
+
use_bias=True, # llada2 only
|
| 21 |
+
rms_norm_eps=1e-05,
|
| 22 |
+
norm_head=False, # llada2 only
|
| 23 |
+
tie_word_embeddings=False, # PretrainedConfig key, here change default value.
|
| 24 |
+
embedding_dropout=0.1,
|
| 25 |
+
attention_dropout=0.1,
|
| 26 |
+
output_dropout=0.1,
|
| 27 |
+
initializer_range=0.02,
|
| 28 |
+
max_position_embeddings=16384,
|
| 29 |
+
rope_theta=10000.0,
|
| 30 |
+
use_cache=True,
|
| 31 |
+
use_sliding_window=False,
|
| 32 |
+
sliding_window=4096,
|
| 33 |
+
max_window_layers=28,
|
| 34 |
+
rope_scaling=None,
|
| 35 |
+
pad_token_id=126081,
|
| 36 |
+
num_experts=16,
|
| 37 |
+
num_shared_experts=0,
|
| 38 |
+
num_experts_per_tok=2,
|
| 39 |
+
n_group=8,
|
| 40 |
+
topk_group=4,
|
| 41 |
+
routed_scaling_factor=2.5,
|
| 42 |
+
moe_intermediate_size=None,
|
| 43 |
+
first_k_dense_replace=0,
|
| 44 |
+
head_dim=None,
|
| 45 |
+
output_router_logits=False,
|
| 46 |
+
partial_rotary_factor=0.5,
|
| 47 |
+
**kwargs,
|
| 48 |
+
):
|
| 49 |
+
self.num_hidden_layers = num_hidden_layers
|
| 50 |
+
self.vocab_size = vocab_size
|
| 51 |
+
self.hidden_size = hidden_size
|
| 52 |
+
self.intermediate_size = intermediate_size
|
| 53 |
+
self.num_attention_heads = num_attention_heads
|
| 54 |
+
self.num_key_value_heads = num_key_value_heads
|
| 55 |
+
self.hidden_act = hidden_act
|
| 56 |
+
self.use_qkv_bias = use_qkv_bias
|
| 57 |
+
self.use_bias = use_bias
|
| 58 |
+
self.norm_head = norm_head
|
| 59 |
+
self.rms_norm_eps = rms_norm_eps
|
| 60 |
+
self.embedding_dropout = embedding_dropout
|
| 61 |
+
self.attention_dropout = attention_dropout
|
| 62 |
+
self.output_dropout = output_dropout
|
| 63 |
+
self.initializer_range = initializer_range
|
| 64 |
+
self.max_position_embeddings = max_position_embeddings
|
| 65 |
+
self.rope_theta = rope_theta
|
| 66 |
+
self.use_cache = use_cache
|
| 67 |
+
self.use_sliding_window = use_sliding_window
|
| 68 |
+
self.sliding_window = sliding_window
|
| 69 |
+
self.max_window_layers = max_window_layers
|
| 70 |
+
self.head_dim = head_dim or self.hidden_size // self.num_attention_heads
|
| 71 |
+
self.rope_scaling = rope_scaling
|
| 72 |
+
|
| 73 |
+
# MoE configs
|
| 74 |
+
self.num_experts = num_experts
|
| 75 |
+
self.num_shared_experts = num_shared_experts
|
| 76 |
+
self.num_experts_per_tok = num_experts_per_tok
|
| 77 |
+
self.n_group = n_group
|
| 78 |
+
self.topk_group = topk_group
|
| 79 |
+
self.moe_intermediate_size = moe_intermediate_size
|
| 80 |
+
self.first_k_dense_replace = first_k_dense_replace
|
| 81 |
+
self.output_router_logits = output_router_logits
|
| 82 |
+
self.routed_scaling_factor = routed_scaling_factor
|
| 83 |
+
self.partial_rotary_factor = partial_rotary_factor
|
| 84 |
+
|
| 85 |
+
super().__init__(pad_token_id=pad_token_id, tie_word_embeddings=tie_word_embeddings, **kwargs)
|
generation_config.json
ADDED
|
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"eos_token_id": 156892,
|
| 4 |
+
"pad_token_id": 156892,
|
| 5 |
+
"transformers_version": "4.46.3",
|
| 6 |
+
"use_cache": false
|
| 7 |
+
}
|
model-00001-of-00012.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:df4e5996435cea36414d5692e9250803b6bd6f233839ae8012fca067b1ca4e7b
|
| 3 |
+
size 5361039659
|
model-00002-of-00012.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a32787dbd411295a89a2f244a6f178572f273bb63d87d7d26fe5f94fd147b927
|
| 3 |
+
size 4923127174
|
model-00003-of-00012.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8805f5a89b172b56268c31d80bb887fbd81834916f18529d8a80083c874b59a4
|
| 3 |
+
size 4923127124
|
model-00004-of-00012.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b6ad8e6a5b81fb1104920feeda61c801402614298549c4e653ae1351c2e12eea
|
| 3 |
+
size 4892698602
|
model-00005-of-00012.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8e1fb404fb2f8b15a459c68a2123d58021a3346f9140dee09416156338ab9b1f
|
| 3 |
+
size 4923127211
|
model-00006-of-00012.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:add8219901584580cdce6174c6003979b501985b51277e890b40f85e6ac8f3cf
|
| 3 |
+
size 4923127211
|
model-00007-of-00012.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b4573ef38c48a34057cb4f8213ad7d2e0df43ca6b9b66858d7ada0c36ed27b12
|
| 3 |
+
size 4892698606
|
model-00008-of-00012.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:985f865279b57dfebd4136c6d0cc947d4ef5f42d4843d1b20fbc5cf9f40e026c
|
| 3 |
+
size 4923127249
|
model-00009-of-00012.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3caadd31e1fb62f4c80994df380e34a7c6e44453dcf8d66ed2fb41685f17f819
|
| 3 |
+
size 4923127209
|
model-00010-of-00012.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9ee757a2019b0338d08c92214757dabf32115284ea1a6eb3157c586db43c3d53
|
| 3 |
+
size 4892698622
|
model-00011-of-00012.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ed7a0418518c0a5aab6543b07cbbe4749c8f6e42aa329b8eb8b9d2045126f410
|
| 3 |
+
size 4923127271
|
model-00012-of-00012.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b9391137d5e618e7ba870b6b1227620c5ee4fbf0ed431ebcef7c20814ae9d2fa
|
| 3 |
+
size 3421683448
|
model.safetensors.index.json
ADDED
|
@@ -0,0 +1,964 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"metadata": {
|
| 3 |
+
"total_size": 57922592256,
|
| 4 |
+
"total_parameters": 102889705216
|
| 5 |
+
},
|
| 6 |
+
"weight_map": {
|
| 7 |
+
"lm_head.biases": "model-00012-of-00012.safetensors",
|
| 8 |
+
"lm_head.scales": "model-00012-of-00012.safetensors",
|
| 9 |
+
"lm_head.weight": "model-00012-of-00012.safetensors",
|
| 10 |
+
"model.layers.0.attention.dense.biases": "model-00001-of-00012.safetensors",
|
| 11 |
+
"model.layers.0.attention.dense.scales": "model-00001-of-00012.safetensors",
|
| 12 |
+
"model.layers.0.attention.dense.weight": "model-00001-of-00012.safetensors",
|
| 13 |
+
"model.layers.0.attention.key_layernorm.weight": "model-00001-of-00012.safetensors",
|
| 14 |
+
"model.layers.0.attention.query_key_value.biases": "model-00001-of-00012.safetensors",
|
| 15 |
+
"model.layers.0.attention.query_key_value.scales": "model-00001-of-00012.safetensors",
|
| 16 |
+
"model.layers.0.attention.query_key_value.weight": "model-00001-of-00012.safetensors",
|
| 17 |
+
"model.layers.0.attention.query_layernorm.weight": "model-00001-of-00012.safetensors",
|
| 18 |
+
"model.layers.0.input_layernorm.weight": "model-00001-of-00012.safetensors",
|
| 19 |
+
"model.layers.0.mlp.down_proj.biases": "model-00001-of-00012.safetensors",
|
| 20 |
+
"model.layers.0.mlp.down_proj.scales": "model-00001-of-00012.safetensors",
|
| 21 |
+
"model.layers.0.mlp.down_proj.weight": "model-00001-of-00012.safetensors",
|
| 22 |
+
"model.layers.0.mlp.gate_proj.biases": "model-00001-of-00012.safetensors",
|
| 23 |
+
"model.layers.0.mlp.gate_proj.scales": "model-00001-of-00012.safetensors",
|
| 24 |
+
"model.layers.0.mlp.gate_proj.weight": "model-00001-of-00012.safetensors",
|
| 25 |
+
"model.layers.0.mlp.up_proj.biases": "model-00001-of-00012.safetensors",
|
| 26 |
+
"model.layers.0.mlp.up_proj.scales": "model-00001-of-00012.safetensors",
|
| 27 |
+
"model.layers.0.mlp.up_proj.weight": "model-00001-of-00012.safetensors",
|
| 28 |
+
"model.layers.0.post_attention_layernorm.weight": "model-00001-of-00012.safetensors",
|
| 29 |
+
"model.layers.1.attention.dense.biases": "model-00001-of-00012.safetensors",
|
| 30 |
+
"model.layers.1.attention.dense.scales": "model-00001-of-00012.safetensors",
|
| 31 |
+
"model.layers.1.attention.dense.weight": "model-00001-of-00012.safetensors",
|
| 32 |
+
"model.layers.1.attention.key_layernorm.weight": "model-00001-of-00012.safetensors",
|
| 33 |
+
"model.layers.1.attention.query_key_value.biases": "model-00001-of-00012.safetensors",
|
| 34 |
+
"model.layers.1.attention.query_key_value.scales": "model-00001-of-00012.safetensors",
|
| 35 |
+
"model.layers.1.attention.query_key_value.weight": "model-00001-of-00012.safetensors",
|
| 36 |
+
"model.layers.1.attention.query_layernorm.weight": "model-00001-of-00012.safetensors",
|
| 37 |
+
"model.layers.1.input_layernorm.weight": "model-00001-of-00012.safetensors",
|
| 38 |
+
"model.layers.1.mlp.gate.expert_bias": "model-00001-of-00012.safetensors",
|
| 39 |
+
"model.layers.1.mlp.gate.weight": "model-00001-of-00012.safetensors",
|
| 40 |
+
"model.layers.1.mlp.shared_experts.down_proj.biases": "model-00001-of-00012.safetensors",
|
| 41 |
+
"model.layers.1.mlp.shared_experts.down_proj.scales": "model-00001-of-00012.safetensors",
|
| 42 |
+
"model.layers.1.mlp.shared_experts.down_proj.weight": "model-00001-of-00012.safetensors",
|
| 43 |
+
"model.layers.1.mlp.shared_experts.gate_proj.biases": "model-00001-of-00012.safetensors",
|
| 44 |
+
"model.layers.1.mlp.shared_experts.gate_proj.scales": "model-00001-of-00012.safetensors",
|
| 45 |
+
"model.layers.1.mlp.shared_experts.gate_proj.weight": "model-00001-of-00012.safetensors",
|
| 46 |
+
"model.layers.1.mlp.shared_experts.up_proj.biases": "model-00001-of-00012.safetensors",
|
| 47 |
+
"model.layers.1.mlp.shared_experts.up_proj.scales": "model-00001-of-00012.safetensors",
|
| 48 |
+
"model.layers.1.mlp.shared_experts.up_proj.weight": "model-00001-of-00012.safetensors",
|
| 49 |
+
"model.layers.1.mlp.switch_mlp.down_proj.biases": "model-00001-of-00012.safetensors",
|
| 50 |
+
"model.layers.1.mlp.switch_mlp.down_proj.scales": "model-00001-of-00012.safetensors",
|
| 51 |
+
"model.layers.1.mlp.switch_mlp.down_proj.weight": "model-00001-of-00012.safetensors",
|
| 52 |
+
"model.layers.1.mlp.switch_mlp.gate_proj.biases": "model-00001-of-00012.safetensors",
|
| 53 |
+
"model.layers.1.mlp.switch_mlp.gate_proj.scales": "model-00001-of-00012.safetensors",
|
| 54 |
+
"model.layers.1.mlp.switch_mlp.gate_proj.weight": "model-00001-of-00012.safetensors",
|
| 55 |
+
"model.layers.1.mlp.switch_mlp.up_proj.biases": "model-00001-of-00012.safetensors",
|
| 56 |
+
"model.layers.1.mlp.switch_mlp.up_proj.scales": "model-00001-of-00012.safetensors",
|
| 57 |
+
"model.layers.1.mlp.switch_mlp.up_proj.weight": "model-00001-of-00012.safetensors",
|
| 58 |
+
"model.layers.1.post_attention_layernorm.weight": "model-00001-of-00012.safetensors",
|
| 59 |
+
"model.layers.10.attention.dense.biases": "model-00004-of-00012.safetensors",
|
| 60 |
+
"model.layers.10.attention.dense.scales": "model-00004-of-00012.safetensors",
|
| 61 |
+
"model.layers.10.attention.dense.weight": "model-00004-of-00012.safetensors",
|
| 62 |
+
"model.layers.10.attention.key_layernorm.weight": "model-00004-of-00012.safetensors",
|
| 63 |
+
"model.layers.10.attention.query_key_value.biases": "model-00004-of-00012.safetensors",
|
| 64 |
+
"model.layers.10.attention.query_key_value.scales": "model-00004-of-00012.safetensors",
|
| 65 |
+
"model.layers.10.attention.query_key_value.weight": "model-00004-of-00012.safetensors",
|
| 66 |
+
"model.layers.10.attention.query_layernorm.weight": "model-00004-of-00012.safetensors",
|
| 67 |
+
"model.layers.10.input_layernorm.weight": "model-00004-of-00012.safetensors",
|
| 68 |
+
"model.layers.10.mlp.gate.expert_bias": "model-00004-of-00012.safetensors",
|
| 69 |
+
"model.layers.10.mlp.gate.weight": "model-00004-of-00012.safetensors",
|
| 70 |
+
"model.layers.10.mlp.shared_experts.down_proj.biases": "model-00004-of-00012.safetensors",
|
| 71 |
+
"model.layers.10.mlp.shared_experts.down_proj.scales": "model-00004-of-00012.safetensors",
|
| 72 |
+
"model.layers.10.mlp.shared_experts.down_proj.weight": "model-00004-of-00012.safetensors",
|
| 73 |
+
"model.layers.10.mlp.shared_experts.gate_proj.biases": "model-00004-of-00012.safetensors",
|
| 74 |
+
"model.layers.10.mlp.shared_experts.gate_proj.scales": "model-00004-of-00012.safetensors",
|
| 75 |
+
"model.layers.10.mlp.shared_experts.gate_proj.weight": "model-00004-of-00012.safetensors",
|
| 76 |
+
"model.layers.10.mlp.shared_experts.up_proj.biases": "model-00004-of-00012.safetensors",
|
| 77 |
+
"model.layers.10.mlp.shared_experts.up_proj.scales": "model-00004-of-00012.safetensors",
|
| 78 |
+
"model.layers.10.mlp.shared_experts.up_proj.weight": "model-00004-of-00012.safetensors",
|
| 79 |
+
"model.layers.10.mlp.switch_mlp.down_proj.biases": "model-00004-of-00012.safetensors",
|
| 80 |
+
"model.layers.10.mlp.switch_mlp.down_proj.scales": "model-00004-of-00012.safetensors",
|
| 81 |
+
"model.layers.10.mlp.switch_mlp.down_proj.weight": "model-00004-of-00012.safetensors",
|
| 82 |
+
"model.layers.10.mlp.switch_mlp.gate_proj.biases": "model-00004-of-00012.safetensors",
|
| 83 |
+
"model.layers.10.mlp.switch_mlp.gate_proj.scales": "model-00004-of-00012.safetensors",
|
| 84 |
+
"model.layers.10.mlp.switch_mlp.gate_proj.weight": "model-00004-of-00012.safetensors",
|
| 85 |
+
"model.layers.10.mlp.switch_mlp.up_proj.biases": "model-00004-of-00012.safetensors",
|
| 86 |
+
"model.layers.10.mlp.switch_mlp.up_proj.scales": "model-00004-of-00012.safetensors",
|
| 87 |
+
"model.layers.10.mlp.switch_mlp.up_proj.weight": "model-00004-of-00012.safetensors",
|
| 88 |
+
"model.layers.10.post_attention_layernorm.weight": "model-00004-of-00012.safetensors",
|
| 89 |
+
"model.layers.11.attention.dense.biases": "model-00004-of-00012.safetensors",
|
| 90 |
+
"model.layers.11.attention.dense.scales": "model-00004-of-00012.safetensors",
|
| 91 |
+
"model.layers.11.attention.dense.weight": "model-00004-of-00012.safetensors",
|
| 92 |
+
"model.layers.11.attention.key_layernorm.weight": "model-00004-of-00012.safetensors",
|
| 93 |
+
"model.layers.11.attention.query_key_value.biases": "model-00004-of-00012.safetensors",
|
| 94 |
+
"model.layers.11.attention.query_key_value.scales": "model-00004-of-00012.safetensors",
|
| 95 |
+
"model.layers.11.attention.query_key_value.weight": "model-00004-of-00012.safetensors",
|
| 96 |
+
"model.layers.11.attention.query_layernorm.weight": "model-00004-of-00012.safetensors",
|
| 97 |
+
"model.layers.11.input_layernorm.weight": "model-00005-of-00012.safetensors",
|
| 98 |
+
"model.layers.11.mlp.gate.expert_bias": "model-00005-of-00012.safetensors",
|
| 99 |
+
"model.layers.11.mlp.gate.weight": "model-00005-of-00012.safetensors",
|
| 100 |
+
"model.layers.11.mlp.shared_experts.down_proj.biases": "model-00005-of-00012.safetensors",
|
| 101 |
+
"model.layers.11.mlp.shared_experts.down_proj.scales": "model-00005-of-00012.safetensors",
|
| 102 |
+
"model.layers.11.mlp.shared_experts.down_proj.weight": "model-00005-of-00012.safetensors",
|
| 103 |
+
"model.layers.11.mlp.shared_experts.gate_proj.biases": "model-00005-of-00012.safetensors",
|
| 104 |
+
"model.layers.11.mlp.shared_experts.gate_proj.scales": "model-00005-of-00012.safetensors",
|
| 105 |
+
"model.layers.11.mlp.shared_experts.gate_proj.weight": "model-00005-of-00012.safetensors",
|
| 106 |
+
"model.layers.11.mlp.shared_experts.up_proj.biases": "model-00005-of-00012.safetensors",
|
| 107 |
+
"model.layers.11.mlp.shared_experts.up_proj.scales": "model-00005-of-00012.safetensors",
|
| 108 |
+
"model.layers.11.mlp.shared_experts.up_proj.weight": "model-00005-of-00012.safetensors",
|
| 109 |
+
"model.layers.11.mlp.switch_mlp.down_proj.biases": "model-00005-of-00012.safetensors",
|
| 110 |
+
"model.layers.11.mlp.switch_mlp.down_proj.scales": "model-00005-of-00012.safetensors",
|
| 111 |
+
"model.layers.11.mlp.switch_mlp.down_proj.weight": "model-00005-of-00012.safetensors",
|
| 112 |
+
"model.layers.11.mlp.switch_mlp.gate_proj.biases": "model-00004-of-00012.safetensors",
|
| 113 |
+
"model.layers.11.mlp.switch_mlp.gate_proj.scales": "model-00004-of-00012.safetensors",
|
| 114 |
+
"model.layers.11.mlp.switch_mlp.gate_proj.weight": "model-00004-of-00012.safetensors",
|
| 115 |
+
"model.layers.11.mlp.switch_mlp.up_proj.biases": "model-00004-of-00012.safetensors",
|
| 116 |
+
"model.layers.11.mlp.switch_mlp.up_proj.scales": "model-00004-of-00012.safetensors",
|
| 117 |
+
"model.layers.11.mlp.switch_mlp.up_proj.weight": "model-00004-of-00012.safetensors",
|
| 118 |
+
"model.layers.11.post_attention_layernorm.weight": "model-00005-of-00012.safetensors",
|
| 119 |
+
"model.layers.12.attention.dense.biases": "model-00005-of-00012.safetensors",
|
| 120 |
+
"model.layers.12.attention.dense.scales": "model-00005-of-00012.safetensors",
|
| 121 |
+
"model.layers.12.attention.dense.weight": "model-00005-of-00012.safetensors",
|
| 122 |
+
"model.layers.12.attention.key_layernorm.weight": "model-00005-of-00012.safetensors",
|
| 123 |
+
"model.layers.12.attention.query_key_value.biases": "model-00005-of-00012.safetensors",
|
| 124 |
+
"model.layers.12.attention.query_key_value.scales": "model-00005-of-00012.safetensors",
|
| 125 |
+
"model.layers.12.attention.query_key_value.weight": "model-00005-of-00012.safetensors",
|
| 126 |
+
"model.layers.12.attention.query_layernorm.weight": "model-00005-of-00012.safetensors",
|
| 127 |
+
"model.layers.12.input_layernorm.weight": "model-00005-of-00012.safetensors",
|
| 128 |
+
"model.layers.12.mlp.gate.expert_bias": "model-00005-of-00012.safetensors",
|
| 129 |
+
"model.layers.12.mlp.gate.weight": "model-00005-of-00012.safetensors",
|
| 130 |
+
"model.layers.12.mlp.shared_experts.down_proj.biases": "model-00005-of-00012.safetensors",
|
| 131 |
+
"model.layers.12.mlp.shared_experts.down_proj.scales": "model-00005-of-00012.safetensors",
|
| 132 |
+
"model.layers.12.mlp.shared_experts.down_proj.weight": "model-00005-of-00012.safetensors",
|
| 133 |
+
"model.layers.12.mlp.shared_experts.gate_proj.biases": "model-00005-of-00012.safetensors",
|
| 134 |
+
"model.layers.12.mlp.shared_experts.gate_proj.scales": "model-00005-of-00012.safetensors",
|
| 135 |
+
"model.layers.12.mlp.shared_experts.gate_proj.weight": "model-00005-of-00012.safetensors",
|
| 136 |
+
"model.layers.12.mlp.shared_experts.up_proj.biases": "model-00005-of-00012.safetensors",
|
| 137 |
+
"model.layers.12.mlp.shared_experts.up_proj.scales": "model-00005-of-00012.safetensors",
|
| 138 |
+
"model.layers.12.mlp.shared_experts.up_proj.weight": "model-00005-of-00012.safetensors",
|
| 139 |
+
"model.layers.12.mlp.switch_mlp.down_proj.biases": "model-00005-of-00012.safetensors",
|
| 140 |
+
"model.layers.12.mlp.switch_mlp.down_proj.scales": "model-00005-of-00012.safetensors",
|
| 141 |
+
"model.layers.12.mlp.switch_mlp.down_proj.weight": "model-00005-of-00012.safetensors",
|
| 142 |
+
"model.layers.12.mlp.switch_mlp.gate_proj.biases": "model-00005-of-00012.safetensors",
|
| 143 |
+
"model.layers.12.mlp.switch_mlp.gate_proj.scales": "model-00005-of-00012.safetensors",
|
| 144 |
+
"model.layers.12.mlp.switch_mlp.gate_proj.weight": "model-00005-of-00012.safetensors",
|
| 145 |
+
"model.layers.12.mlp.switch_mlp.up_proj.biases": "model-00005-of-00012.safetensors",
|
| 146 |
+
"model.layers.12.mlp.switch_mlp.up_proj.scales": "model-00005-of-00012.safetensors",
|
| 147 |
+
"model.layers.12.mlp.switch_mlp.up_proj.weight": "model-00005-of-00012.safetensors",
|
| 148 |
+
"model.layers.12.post_attention_layernorm.weight": "model-00005-of-00012.safetensors",
|
| 149 |
+
"model.layers.13.attention.dense.biases": "model-00005-of-00012.safetensors",
|
| 150 |
+
"model.layers.13.attention.dense.scales": "model-00005-of-00012.safetensors",
|
| 151 |
+
"model.layers.13.attention.dense.weight": "model-00005-of-00012.safetensors",
|
| 152 |
+
"model.layers.13.attention.key_layernorm.weight": "model-00005-of-00012.safetensors",
|
| 153 |
+
"model.layers.13.attention.query_key_value.biases": "model-00005-of-00012.safetensors",
|
| 154 |
+
"model.layers.13.attention.query_key_value.scales": "model-00005-of-00012.safetensors",
|
| 155 |
+
"model.layers.13.attention.query_key_value.weight": "model-00005-of-00012.safetensors",
|
| 156 |
+
"model.layers.13.attention.query_layernorm.weight": "model-00005-of-00012.safetensors",
|
| 157 |
+
"model.layers.13.input_layernorm.weight": "model-00005-of-00012.safetensors",
|
| 158 |
+
"model.layers.13.mlp.gate.expert_bias": "model-00005-of-00012.safetensors",
|
| 159 |
+
"model.layers.13.mlp.gate.weight": "model-00005-of-00012.safetensors",
|
| 160 |
+
"model.layers.13.mlp.shared_experts.down_proj.biases": "model-00005-of-00012.safetensors",
|
| 161 |
+
"model.layers.13.mlp.shared_experts.down_proj.scales": "model-00005-of-00012.safetensors",
|
| 162 |
+
"model.layers.13.mlp.shared_experts.down_proj.weight": "model-00005-of-00012.safetensors",
|
| 163 |
+
"model.layers.13.mlp.shared_experts.gate_proj.biases": "model-00005-of-00012.safetensors",
|
| 164 |
+
"model.layers.13.mlp.shared_experts.gate_proj.scales": "model-00005-of-00012.safetensors",
|
| 165 |
+
"model.layers.13.mlp.shared_experts.gate_proj.weight": "model-00005-of-00012.safetensors",
|
| 166 |
+
"model.layers.13.mlp.shared_experts.up_proj.biases": "model-00005-of-00012.safetensors",
|
| 167 |
+
"model.layers.13.mlp.shared_experts.up_proj.scales": "model-00005-of-00012.safetensors",
|
| 168 |
+
"model.layers.13.mlp.shared_experts.up_proj.weight": "model-00005-of-00012.safetensors",
|
| 169 |
+
"model.layers.13.mlp.switch_mlp.down_proj.biases": "model-00005-of-00012.safetensors",
|
| 170 |
+
"model.layers.13.mlp.switch_mlp.down_proj.scales": "model-00005-of-00012.safetensors",
|
| 171 |
+
"model.layers.13.mlp.switch_mlp.down_proj.weight": "model-00005-of-00012.safetensors",
|
| 172 |
+
"model.layers.13.mlp.switch_mlp.gate_proj.biases": "model-00005-of-00012.safetensors",
|
| 173 |
+
"model.layers.13.mlp.switch_mlp.gate_proj.scales": "model-00005-of-00012.safetensors",
|
| 174 |
+
"model.layers.13.mlp.switch_mlp.gate_proj.weight": "model-00005-of-00012.safetensors",
|
| 175 |
+
"model.layers.13.mlp.switch_mlp.up_proj.biases": "model-00005-of-00012.safetensors",
|
| 176 |
+
"model.layers.13.mlp.switch_mlp.up_proj.scales": "model-00005-of-00012.safetensors",
|
| 177 |
+
"model.layers.13.mlp.switch_mlp.up_proj.weight": "model-00005-of-00012.safetensors",
|
| 178 |
+
"model.layers.13.post_attention_layernorm.weight": "model-00005-of-00012.safetensors",
|
| 179 |
+
"model.layers.14.attention.dense.biases": "model-00005-of-00012.safetensors",
|
| 180 |
+
"model.layers.14.attention.dense.scales": "model-00005-of-00012.safetensors",
|
| 181 |
+
"model.layers.14.attention.dense.weight": "model-00005-of-00012.safetensors",
|
| 182 |
+
"model.layers.14.attention.key_layernorm.weight": "model-00005-of-00012.safetensors",
|
| 183 |
+
"model.layers.14.attention.query_key_value.biases": "model-00005-of-00012.safetensors",
|
| 184 |
+
"model.layers.14.attention.query_key_value.scales": "model-00005-of-00012.safetensors",
|
| 185 |
+
"model.layers.14.attention.query_key_value.weight": "model-00005-of-00012.safetensors",
|
| 186 |
+
"model.layers.14.attention.query_layernorm.weight": "model-00005-of-00012.safetensors",
|
| 187 |
+
"model.layers.14.input_layernorm.weight": "model-00006-of-00012.safetensors",
|
| 188 |
+
"model.layers.14.mlp.gate.expert_bias": "model-00006-of-00012.safetensors",
|
| 189 |
+
"model.layers.14.mlp.gate.weight": "model-00006-of-00012.safetensors",
|
| 190 |
+
"model.layers.14.mlp.shared_experts.down_proj.biases": "model-00006-of-00012.safetensors",
|
| 191 |
+
"model.layers.14.mlp.shared_experts.down_proj.scales": "model-00006-of-00012.safetensors",
|
| 192 |
+
"model.layers.14.mlp.shared_experts.down_proj.weight": "model-00006-of-00012.safetensors",
|
| 193 |
+
"model.layers.14.mlp.shared_experts.gate_proj.biases": "model-00006-of-00012.safetensors",
|
| 194 |
+
"model.layers.14.mlp.shared_experts.gate_proj.scales": "model-00006-of-00012.safetensors",
|
| 195 |
+
"model.layers.14.mlp.shared_experts.gate_proj.weight": "model-00006-of-00012.safetensors",
|
| 196 |
+
"model.layers.14.mlp.shared_experts.up_proj.biases": "model-00006-of-00012.safetensors",
|
| 197 |
+
"model.layers.14.mlp.shared_experts.up_proj.scales": "model-00006-of-00012.safetensors",
|
| 198 |
+
"model.layers.14.mlp.shared_experts.up_proj.weight": "model-00006-of-00012.safetensors",
|
| 199 |
+
"model.layers.14.mlp.switch_mlp.down_proj.biases": "model-00006-of-00012.safetensors",
|
| 200 |
+
"model.layers.14.mlp.switch_mlp.down_proj.scales": "model-00006-of-00012.safetensors",
|
| 201 |
+
"model.layers.14.mlp.switch_mlp.down_proj.weight": "model-00006-of-00012.safetensors",
|
| 202 |
+
"model.layers.14.mlp.switch_mlp.gate_proj.biases": "model-00005-of-00012.safetensors",
|
| 203 |
+
"model.layers.14.mlp.switch_mlp.gate_proj.scales": "model-00005-of-00012.safetensors",
|
| 204 |
+
"model.layers.14.mlp.switch_mlp.gate_proj.weight": "model-00005-of-00012.safetensors",
|
| 205 |
+
"model.layers.14.mlp.switch_mlp.up_proj.biases": "model-00006-of-00012.safetensors",
|
| 206 |
+
"model.layers.14.mlp.switch_mlp.up_proj.scales": "model-00006-of-00012.safetensors",
|
| 207 |
+
"model.layers.14.mlp.switch_mlp.up_proj.weight": "model-00006-of-00012.safetensors",
|
| 208 |
+
"model.layers.14.post_attention_layernorm.weight": "model-00006-of-00012.safetensors",
|
| 209 |
+
"model.layers.15.attention.dense.biases": "model-00006-of-00012.safetensors",
|
| 210 |
+
"model.layers.15.attention.dense.scales": "model-00006-of-00012.safetensors",
|
| 211 |
+
"model.layers.15.attention.dense.weight": "model-00006-of-00012.safetensors",
|
| 212 |
+
"model.layers.15.attention.key_layernorm.weight": "model-00006-of-00012.safetensors",
|
| 213 |
+
"model.layers.15.attention.query_key_value.biases": "model-00006-of-00012.safetensors",
|
| 214 |
+
"model.layers.15.attention.query_key_value.scales": "model-00006-of-00012.safetensors",
|
| 215 |
+
"model.layers.15.attention.query_key_value.weight": "model-00006-of-00012.safetensors",
|
| 216 |
+
"model.layers.15.attention.query_layernorm.weight": "model-00006-of-00012.safetensors",
|
| 217 |
+
"model.layers.15.input_layernorm.weight": "model-00006-of-00012.safetensors",
|
| 218 |
+
"model.layers.15.mlp.gate.expert_bias": "model-00006-of-00012.safetensors",
|
| 219 |
+
"model.layers.15.mlp.gate.weight": "model-00006-of-00012.safetensors",
|
| 220 |
+
"model.layers.15.mlp.shared_experts.down_proj.biases": "model-00006-of-00012.safetensors",
|
| 221 |
+
"model.layers.15.mlp.shared_experts.down_proj.scales": "model-00006-of-00012.safetensors",
|
| 222 |
+
"model.layers.15.mlp.shared_experts.down_proj.weight": "model-00006-of-00012.safetensors",
|
| 223 |
+
"model.layers.15.mlp.shared_experts.gate_proj.biases": "model-00006-of-00012.safetensors",
|
| 224 |
+
"model.layers.15.mlp.shared_experts.gate_proj.scales": "model-00006-of-00012.safetensors",
|
| 225 |
+
"model.layers.15.mlp.shared_experts.gate_proj.weight": "model-00006-of-00012.safetensors",
|
| 226 |
+
"model.layers.15.mlp.shared_experts.up_proj.biases": "model-00006-of-00012.safetensors",
|
| 227 |
+
"model.layers.15.mlp.shared_experts.up_proj.scales": "model-00006-of-00012.safetensors",
|
| 228 |
+
"model.layers.15.mlp.shared_experts.up_proj.weight": "model-00006-of-00012.safetensors",
|
| 229 |
+
"model.layers.15.mlp.switch_mlp.down_proj.biases": "model-00006-of-00012.safetensors",
|
| 230 |
+
"model.layers.15.mlp.switch_mlp.down_proj.scales": "model-00006-of-00012.safetensors",
|
| 231 |
+
"model.layers.15.mlp.switch_mlp.down_proj.weight": "model-00006-of-00012.safetensors",
|
| 232 |
+
"model.layers.15.mlp.switch_mlp.gate_proj.biases": "model-00006-of-00012.safetensors",
|
| 233 |
+
"model.layers.15.mlp.switch_mlp.gate_proj.scales": "model-00006-of-00012.safetensors",
|
| 234 |
+
"model.layers.15.mlp.switch_mlp.gate_proj.weight": "model-00006-of-00012.safetensors",
|
| 235 |
+
"model.layers.15.mlp.switch_mlp.up_proj.biases": "model-00006-of-00012.safetensors",
|
| 236 |
+
"model.layers.15.mlp.switch_mlp.up_proj.scales": "model-00006-of-00012.safetensors",
|
| 237 |
+
"model.layers.15.mlp.switch_mlp.up_proj.weight": "model-00006-of-00012.safetensors",
|
| 238 |
+
"model.layers.15.post_attention_layernorm.weight": "model-00006-of-00012.safetensors",
|
| 239 |
+
"model.layers.16.attention.dense.biases": "model-00006-of-00012.safetensors",
|
| 240 |
+
"model.layers.16.attention.dense.scales": "model-00006-of-00012.safetensors",
|
| 241 |
+
"model.layers.16.attention.dense.weight": "model-00006-of-00012.safetensors",
|
| 242 |
+
"model.layers.16.attention.key_layernorm.weight": "model-00006-of-00012.safetensors",
|
| 243 |
+
"model.layers.16.attention.query_key_value.biases": "model-00006-of-00012.safetensors",
|
| 244 |
+
"model.layers.16.attention.query_key_value.scales": "model-00006-of-00012.safetensors",
|
| 245 |
+
"model.layers.16.attention.query_key_value.weight": "model-00006-of-00012.safetensors",
|
| 246 |
+
"model.layers.16.attention.query_layernorm.weight": "model-00006-of-00012.safetensors",
|
| 247 |
+
"model.layers.16.input_layernorm.weight": "model-00006-of-00012.safetensors",
|
| 248 |
+
"model.layers.16.mlp.gate.expert_bias": "model-00006-of-00012.safetensors",
|
| 249 |
+
"model.layers.16.mlp.gate.weight": "model-00006-of-00012.safetensors",
|
| 250 |
+
"model.layers.16.mlp.shared_experts.down_proj.biases": "model-00006-of-00012.safetensors",
|
| 251 |
+
"model.layers.16.mlp.shared_experts.down_proj.scales": "model-00006-of-00012.safetensors",
|
| 252 |
+
"model.layers.16.mlp.shared_experts.down_proj.weight": "model-00006-of-00012.safetensors",
|
| 253 |
+
"model.layers.16.mlp.shared_experts.gate_proj.biases": "model-00006-of-00012.safetensors",
|
| 254 |
+
"model.layers.16.mlp.shared_experts.gate_proj.scales": "model-00006-of-00012.safetensors",
|
| 255 |
+
"model.layers.16.mlp.shared_experts.gate_proj.weight": "model-00006-of-00012.safetensors",
|
| 256 |
+
"model.layers.16.mlp.shared_experts.up_proj.biases": "model-00006-of-00012.safetensors",
|
| 257 |
+
"model.layers.16.mlp.shared_experts.up_proj.scales": "model-00006-of-00012.safetensors",
|
| 258 |
+
"model.layers.16.mlp.shared_experts.up_proj.weight": "model-00006-of-00012.safetensors",
|
| 259 |
+
"model.layers.16.mlp.switch_mlp.down_proj.biases": "model-00006-of-00012.safetensors",
|
| 260 |
+
"model.layers.16.mlp.switch_mlp.down_proj.scales": "model-00006-of-00012.safetensors",
|
| 261 |
+
"model.layers.16.mlp.switch_mlp.down_proj.weight": "model-00006-of-00012.safetensors",
|
| 262 |
+
"model.layers.16.mlp.switch_mlp.gate_proj.biases": "model-00006-of-00012.safetensors",
|
| 263 |
+
"model.layers.16.mlp.switch_mlp.gate_proj.scales": "model-00006-of-00012.safetensors",
|
| 264 |
+
"model.layers.16.mlp.switch_mlp.gate_proj.weight": "model-00006-of-00012.safetensors",
|
| 265 |
+
"model.layers.16.mlp.switch_mlp.up_proj.biases": "model-00006-of-00012.safetensors",
|
| 266 |
+
"model.layers.16.mlp.switch_mlp.up_proj.scales": "model-00006-of-00012.safetensors",
|
| 267 |
+
"model.layers.16.mlp.switch_mlp.up_proj.weight": "model-00006-of-00012.safetensors",
|
| 268 |
+
"model.layers.16.post_attention_layernorm.weight": "model-00006-of-00012.safetensors",
|
| 269 |
+
"model.layers.17.attention.dense.biases": "model-00006-of-00012.safetensors",
|
| 270 |
+
"model.layers.17.attention.dense.scales": "model-00006-of-00012.safetensors",
|
| 271 |
+
"model.layers.17.attention.dense.weight": "model-00006-of-00012.safetensors",
|
| 272 |
+
"model.layers.17.attention.key_layernorm.weight": "model-00006-of-00012.safetensors",
|
| 273 |
+
"model.layers.17.attention.query_key_value.biases": "model-00006-of-00012.safetensors",
|
| 274 |
+
"model.layers.17.attention.query_key_value.scales": "model-00006-of-00012.safetensors",
|
| 275 |
+
"model.layers.17.attention.query_key_value.weight": "model-00006-of-00012.safetensors",
|
| 276 |
+
"model.layers.17.attention.query_layernorm.weight": "model-00006-of-00012.safetensors",
|
| 277 |
+
"model.layers.17.input_layernorm.weight": "model-00007-of-00012.safetensors",
|
| 278 |
+
"model.layers.17.mlp.gate.expert_bias": "model-00007-of-00012.safetensors",
|
| 279 |
+
"model.layers.17.mlp.gate.weight": "model-00007-of-00012.safetensors",
|
| 280 |
+
"model.layers.17.mlp.shared_experts.down_proj.biases": "model-00007-of-00012.safetensors",
|
| 281 |
+
"model.layers.17.mlp.shared_experts.down_proj.scales": "model-00007-of-00012.safetensors",
|
| 282 |
+
"model.layers.17.mlp.shared_experts.down_proj.weight": "model-00007-of-00012.safetensors",
|
| 283 |
+
"model.layers.17.mlp.shared_experts.gate_proj.biases": "model-00007-of-00012.safetensors",
|
| 284 |
+
"model.layers.17.mlp.shared_experts.gate_proj.scales": "model-00007-of-00012.safetensors",
|
| 285 |
+
"model.layers.17.mlp.shared_experts.gate_proj.weight": "model-00007-of-00012.safetensors",
|
| 286 |
+
"model.layers.17.mlp.shared_experts.up_proj.biases": "model-00007-of-00012.safetensors",
|
| 287 |
+
"model.layers.17.mlp.shared_experts.up_proj.scales": "model-00007-of-00012.safetensors",
|
| 288 |
+
"model.layers.17.mlp.shared_experts.up_proj.weight": "model-00007-of-00012.safetensors",
|
| 289 |
+
"model.layers.17.mlp.switch_mlp.down_proj.biases": "model-00007-of-00012.safetensors",
|
| 290 |
+
"model.layers.17.mlp.switch_mlp.down_proj.scales": "model-00007-of-00012.safetensors",
|
| 291 |
+
"model.layers.17.mlp.switch_mlp.down_proj.weight": "model-00007-of-00012.safetensors",
|
| 292 |
+
"model.layers.17.mlp.switch_mlp.gate_proj.biases": "model-00007-of-00012.safetensors",
|
| 293 |
+
"model.layers.17.mlp.switch_mlp.gate_proj.scales": "model-00007-of-00012.safetensors",
|
| 294 |
+
"model.layers.17.mlp.switch_mlp.gate_proj.weight": "model-00007-of-00012.safetensors",
|
| 295 |
+
"model.layers.17.mlp.switch_mlp.up_proj.biases": "model-00007-of-00012.safetensors",
|
| 296 |
+
"model.layers.17.mlp.switch_mlp.up_proj.scales": "model-00007-of-00012.safetensors",
|
| 297 |
+
"model.layers.17.mlp.switch_mlp.up_proj.weight": "model-00007-of-00012.safetensors",
|
| 298 |
+
"model.layers.17.post_attention_layernorm.weight": "model-00007-of-00012.safetensors",
|
| 299 |
+
"model.layers.18.attention.dense.biases": "model-00007-of-00012.safetensors",
|
| 300 |
+
"model.layers.18.attention.dense.scales": "model-00007-of-00012.safetensors",
|
| 301 |
+
"model.layers.18.attention.dense.weight": "model-00007-of-00012.safetensors",
|
| 302 |
+
"model.layers.18.attention.key_layernorm.weight": "model-00007-of-00012.safetensors",
|
| 303 |
+
"model.layers.18.attention.query_key_value.biases": "model-00007-of-00012.safetensors",
|
| 304 |
+
"model.layers.18.attention.query_key_value.scales": "model-00007-of-00012.safetensors",
|
| 305 |
+
"model.layers.18.attention.query_key_value.weight": "model-00007-of-00012.safetensors",
|
| 306 |
+
"model.layers.18.attention.query_layernorm.weight": "model-00007-of-00012.safetensors",
|
| 307 |
+
"model.layers.18.input_layernorm.weight": "model-00007-of-00012.safetensors",
|
| 308 |
+
"model.layers.18.mlp.gate.expert_bias": "model-00007-of-00012.safetensors",
|
| 309 |
+
"model.layers.18.mlp.gate.weight": "model-00007-of-00012.safetensors",
|
| 310 |
+
"model.layers.18.mlp.shared_experts.down_proj.biases": "model-00007-of-00012.safetensors",
|
| 311 |
+
"model.layers.18.mlp.shared_experts.down_proj.scales": "model-00007-of-00012.safetensors",
|
| 312 |
+
"model.layers.18.mlp.shared_experts.down_proj.weight": "model-00007-of-00012.safetensors",
|
| 313 |
+
"model.layers.18.mlp.shared_experts.gate_proj.biases": "model-00007-of-00012.safetensors",
|
| 314 |
+
"model.layers.18.mlp.shared_experts.gate_proj.scales": "model-00007-of-00012.safetensors",
|
| 315 |
+
"model.layers.18.mlp.shared_experts.gate_proj.weight": "model-00007-of-00012.safetensors",
|
| 316 |
+
"model.layers.18.mlp.shared_experts.up_proj.biases": "model-00007-of-00012.safetensors",
|
| 317 |
+
"model.layers.18.mlp.shared_experts.up_proj.scales": "model-00007-of-00012.safetensors",
|
| 318 |
+
"model.layers.18.mlp.shared_experts.up_proj.weight": "model-00007-of-00012.safetensors",
|
| 319 |
+
"model.layers.18.mlp.switch_mlp.down_proj.biases": "model-00007-of-00012.safetensors",
|
| 320 |
+
"model.layers.18.mlp.switch_mlp.down_proj.scales": "model-00007-of-00012.safetensors",
|
| 321 |
+
"model.layers.18.mlp.switch_mlp.down_proj.weight": "model-00007-of-00012.safetensors",
|
| 322 |
+
"model.layers.18.mlp.switch_mlp.gate_proj.biases": "model-00007-of-00012.safetensors",
|
| 323 |
+
"model.layers.18.mlp.switch_mlp.gate_proj.scales": "model-00007-of-00012.safetensors",
|
| 324 |
+
"model.layers.18.mlp.switch_mlp.gate_proj.weight": "model-00007-of-00012.safetensors",
|
| 325 |
+
"model.layers.18.mlp.switch_mlp.up_proj.biases": "model-00007-of-00012.safetensors",
|
| 326 |
+
"model.layers.18.mlp.switch_mlp.up_proj.scales": "model-00007-of-00012.safetensors",
|
| 327 |
+
"model.layers.18.mlp.switch_mlp.up_proj.weight": "model-00007-of-00012.safetensors",
|
| 328 |
+
"model.layers.18.post_attention_layernorm.weight": "model-00007-of-00012.safetensors",
|
| 329 |
+
"model.layers.19.attention.dense.biases": "model-00007-of-00012.safetensors",
|
| 330 |
+
"model.layers.19.attention.dense.scales": "model-00007-of-00012.safetensors",
|
| 331 |
+
"model.layers.19.attention.dense.weight": "model-00007-of-00012.safetensors",
|
| 332 |
+
"model.layers.19.attention.key_layernorm.weight": "model-00007-of-00012.safetensors",
|
| 333 |
+
"model.layers.19.attention.query_key_value.biases": "model-00007-of-00012.safetensors",
|
| 334 |
+
"model.layers.19.attention.query_key_value.scales": "model-00007-of-00012.safetensors",
|
| 335 |
+
"model.layers.19.attention.query_key_value.weight": "model-00007-of-00012.safetensors",
|
| 336 |
+
"model.layers.19.attention.query_layernorm.weight": "model-00007-of-00012.safetensors",
|
| 337 |
+
"model.layers.19.input_layernorm.weight": "model-00008-of-00012.safetensors",
|
| 338 |
+
"model.layers.19.mlp.gate.expert_bias": "model-00008-of-00012.safetensors",
|
| 339 |
+
"model.layers.19.mlp.gate.weight": "model-00008-of-00012.safetensors",
|
| 340 |
+
"model.layers.19.mlp.shared_experts.down_proj.biases": "model-00008-of-00012.safetensors",
|
| 341 |
+
"model.layers.19.mlp.shared_experts.down_proj.scales": "model-00008-of-00012.safetensors",
|
| 342 |
+
"model.layers.19.mlp.shared_experts.down_proj.weight": "model-00008-of-00012.safetensors",
|
| 343 |
+
"model.layers.19.mlp.shared_experts.gate_proj.biases": "model-00008-of-00012.safetensors",
|
| 344 |
+
"model.layers.19.mlp.shared_experts.gate_proj.scales": "model-00008-of-00012.safetensors",
|
| 345 |
+
"model.layers.19.mlp.shared_experts.gate_proj.weight": "model-00008-of-00012.safetensors",
|
| 346 |
+
"model.layers.19.mlp.shared_experts.up_proj.biases": "model-00008-of-00012.safetensors",
|
| 347 |
+
"model.layers.19.mlp.shared_experts.up_proj.scales": "model-00008-of-00012.safetensors",
|
| 348 |
+
"model.layers.19.mlp.shared_experts.up_proj.weight": "model-00008-of-00012.safetensors",
|
| 349 |
+
"model.layers.19.mlp.switch_mlp.down_proj.biases": "model-00008-of-00012.safetensors",
|
| 350 |
+
"model.layers.19.mlp.switch_mlp.down_proj.scales": "model-00008-of-00012.safetensors",
|
| 351 |
+
"model.layers.19.mlp.switch_mlp.down_proj.weight": "model-00008-of-00012.safetensors",
|
| 352 |
+
"model.layers.19.mlp.switch_mlp.gate_proj.biases": "model-00007-of-00012.safetensors",
|
| 353 |
+
"model.layers.19.mlp.switch_mlp.gate_proj.scales": "model-00007-of-00012.safetensors",
|
| 354 |
+
"model.layers.19.mlp.switch_mlp.gate_proj.weight": "model-00007-of-00012.safetensors",
|
| 355 |
+
"model.layers.19.mlp.switch_mlp.up_proj.biases": "model-00007-of-00012.safetensors",
|
| 356 |
+
"model.layers.19.mlp.switch_mlp.up_proj.scales": "model-00007-of-00012.safetensors",
|
| 357 |
+
"model.layers.19.mlp.switch_mlp.up_proj.weight": "model-00007-of-00012.safetensors",
|
| 358 |
+
"model.layers.19.post_attention_layernorm.weight": "model-00008-of-00012.safetensors",
|
| 359 |
+
"model.layers.2.attention.dense.biases": "model-00001-of-00012.safetensors",
|
| 360 |
+
"model.layers.2.attention.dense.scales": "model-00001-of-00012.safetensors",
|
| 361 |
+
"model.layers.2.attention.dense.weight": "model-00001-of-00012.safetensors",
|
| 362 |
+
"model.layers.2.attention.key_layernorm.weight": "model-00001-of-00012.safetensors",
|
| 363 |
+
"model.layers.2.attention.query_key_value.biases": "model-00001-of-00012.safetensors",
|
| 364 |
+
"model.layers.2.attention.query_key_value.scales": "model-00001-of-00012.safetensors",
|
| 365 |
+
"model.layers.2.attention.query_key_value.weight": "model-00001-of-00012.safetensors",
|
| 366 |
+
"model.layers.2.attention.query_layernorm.weight": "model-00001-of-00012.safetensors",
|
| 367 |
+
"model.layers.2.input_layernorm.weight": "model-00001-of-00012.safetensors",
|
| 368 |
+
"model.layers.2.mlp.gate.expert_bias": "model-00001-of-00012.safetensors",
|
| 369 |
+
"model.layers.2.mlp.gate.weight": "model-00001-of-00012.safetensors",
|
| 370 |
+
"model.layers.2.mlp.shared_experts.down_proj.biases": "model-00001-of-00012.safetensors",
|
| 371 |
+
"model.layers.2.mlp.shared_experts.down_proj.scales": "model-00001-of-00012.safetensors",
|
| 372 |
+
"model.layers.2.mlp.shared_experts.down_proj.weight": "model-00001-of-00012.safetensors",
|
| 373 |
+
"model.layers.2.mlp.shared_experts.gate_proj.biases": "model-00001-of-00012.safetensors",
|
| 374 |
+
"model.layers.2.mlp.shared_experts.gate_proj.scales": "model-00001-of-00012.safetensors",
|
| 375 |
+
"model.layers.2.mlp.shared_experts.gate_proj.weight": "model-00001-of-00012.safetensors",
|
| 376 |
+
"model.layers.2.mlp.shared_experts.up_proj.biases": "model-00001-of-00012.safetensors",
|
| 377 |
+
"model.layers.2.mlp.shared_experts.up_proj.scales": "model-00001-of-00012.safetensors",
|
| 378 |
+
"model.layers.2.mlp.shared_experts.up_proj.weight": "model-00001-of-00012.safetensors",
|
| 379 |
+
"model.layers.2.mlp.switch_mlp.down_proj.biases": "model-00001-of-00012.safetensors",
|
| 380 |
+
"model.layers.2.mlp.switch_mlp.down_proj.scales": "model-00001-of-00012.safetensors",
|
| 381 |
+
"model.layers.2.mlp.switch_mlp.down_proj.weight": "model-00001-of-00012.safetensors",
|
| 382 |
+
"model.layers.2.mlp.switch_mlp.gate_proj.biases": "model-00001-of-00012.safetensors",
|
| 383 |
+
"model.layers.2.mlp.switch_mlp.gate_proj.scales": "model-00001-of-00012.safetensors",
|
| 384 |
+
"model.layers.2.mlp.switch_mlp.gate_proj.weight": "model-00001-of-00012.safetensors",
|
| 385 |
+
"model.layers.2.mlp.switch_mlp.up_proj.biases": "model-00001-of-00012.safetensors",
|
| 386 |
+
"model.layers.2.mlp.switch_mlp.up_proj.scales": "model-00001-of-00012.safetensors",
|
| 387 |
+
"model.layers.2.mlp.switch_mlp.up_proj.weight": "model-00001-of-00012.safetensors",
|
| 388 |
+
"model.layers.2.post_attention_layernorm.weight": "model-00001-of-00012.safetensors",
|
| 389 |
+
"model.layers.20.attention.dense.biases": "model-00008-of-00012.safetensors",
|
| 390 |
+
"model.layers.20.attention.dense.scales": "model-00008-of-00012.safetensors",
|
| 391 |
+
"model.layers.20.attention.dense.weight": "model-00008-of-00012.safetensors",
|
| 392 |
+
"model.layers.20.attention.key_layernorm.weight": "model-00008-of-00012.safetensors",
|
| 393 |
+
"model.layers.20.attention.query_key_value.biases": "model-00008-of-00012.safetensors",
|
| 394 |
+
"model.layers.20.attention.query_key_value.scales": "model-00008-of-00012.safetensors",
|
| 395 |
+
"model.layers.20.attention.query_key_value.weight": "model-00008-of-00012.safetensors",
|
| 396 |
+
"model.layers.20.attention.query_layernorm.weight": "model-00008-of-00012.safetensors",
|
| 397 |
+
"model.layers.20.input_layernorm.weight": "model-00008-of-00012.safetensors",
|
| 398 |
+
"model.layers.20.mlp.gate.expert_bias": "model-00008-of-00012.safetensors",
|
| 399 |
+
"model.layers.20.mlp.gate.weight": "model-00008-of-00012.safetensors",
|
| 400 |
+
"model.layers.20.mlp.shared_experts.down_proj.biases": "model-00008-of-00012.safetensors",
|
| 401 |
+
"model.layers.20.mlp.shared_experts.down_proj.scales": "model-00008-of-00012.safetensors",
|
| 402 |
+
"model.layers.20.mlp.shared_experts.down_proj.weight": "model-00008-of-00012.safetensors",
|
| 403 |
+
"model.layers.20.mlp.shared_experts.gate_proj.biases": "model-00008-of-00012.safetensors",
|
| 404 |
+
"model.layers.20.mlp.shared_experts.gate_proj.scales": "model-00008-of-00012.safetensors",
|
| 405 |
+
"model.layers.20.mlp.shared_experts.gate_proj.weight": "model-00008-of-00012.safetensors",
|
| 406 |
+
"model.layers.20.mlp.shared_experts.up_proj.biases": "model-00008-of-00012.safetensors",
|
| 407 |
+
"model.layers.20.mlp.shared_experts.up_proj.scales": "model-00008-of-00012.safetensors",
|
| 408 |
+
"model.layers.20.mlp.shared_experts.up_proj.weight": "model-00008-of-00012.safetensors",
|
| 409 |
+
"model.layers.20.mlp.switch_mlp.down_proj.biases": "model-00008-of-00012.safetensors",
|
| 410 |
+
"model.layers.20.mlp.switch_mlp.down_proj.scales": "model-00008-of-00012.safetensors",
|
| 411 |
+
"model.layers.20.mlp.switch_mlp.down_proj.weight": "model-00008-of-00012.safetensors",
|
| 412 |
+
"model.layers.20.mlp.switch_mlp.gate_proj.biases": "model-00008-of-00012.safetensors",
|
| 413 |
+
"model.layers.20.mlp.switch_mlp.gate_proj.scales": "model-00008-of-00012.safetensors",
|
| 414 |
+
"model.layers.20.mlp.switch_mlp.gate_proj.weight": "model-00008-of-00012.safetensors",
|
| 415 |
+
"model.layers.20.mlp.switch_mlp.up_proj.biases": "model-00008-of-00012.safetensors",
|
| 416 |
+
"model.layers.20.mlp.switch_mlp.up_proj.scales": "model-00008-of-00012.safetensors",
|
| 417 |
+
"model.layers.20.mlp.switch_mlp.up_proj.weight": "model-00008-of-00012.safetensors",
|
| 418 |
+
"model.layers.20.post_attention_layernorm.weight": "model-00008-of-00012.safetensors",
|
| 419 |
+
"model.layers.21.attention.dense.biases": "model-00008-of-00012.safetensors",
|
| 420 |
+
"model.layers.21.attention.dense.scales": "model-00008-of-00012.safetensors",
|
| 421 |
+
"model.layers.21.attention.dense.weight": "model-00008-of-00012.safetensors",
|
| 422 |
+
"model.layers.21.attention.key_layernorm.weight": "model-00008-of-00012.safetensors",
|
| 423 |
+
"model.layers.21.attention.query_key_value.biases": "model-00008-of-00012.safetensors",
|
| 424 |
+
"model.layers.21.attention.query_key_value.scales": "model-00008-of-00012.safetensors",
|
| 425 |
+
"model.layers.21.attention.query_key_value.weight": "model-00008-of-00012.safetensors",
|
| 426 |
+
"model.layers.21.attention.query_layernorm.weight": "model-00008-of-00012.safetensors",
|
| 427 |
+
"model.layers.21.input_layernorm.weight": "model-00008-of-00012.safetensors",
|
| 428 |
+
"model.layers.21.mlp.gate.expert_bias": "model-00008-of-00012.safetensors",
|
| 429 |
+
"model.layers.21.mlp.gate.weight": "model-00008-of-00012.safetensors",
|
| 430 |
+
"model.layers.21.mlp.shared_experts.down_proj.biases": "model-00008-of-00012.safetensors",
|
| 431 |
+
"model.layers.21.mlp.shared_experts.down_proj.scales": "model-00008-of-00012.safetensors",
|
| 432 |
+
"model.layers.21.mlp.shared_experts.down_proj.weight": "model-00008-of-00012.safetensors",
|
| 433 |
+
"model.layers.21.mlp.shared_experts.gate_proj.biases": "model-00008-of-00012.safetensors",
|
| 434 |
+
"model.layers.21.mlp.shared_experts.gate_proj.scales": "model-00008-of-00012.safetensors",
|
| 435 |
+
"model.layers.21.mlp.shared_experts.gate_proj.weight": "model-00008-of-00012.safetensors",
|
| 436 |
+
"model.layers.21.mlp.shared_experts.up_proj.biases": "model-00008-of-00012.safetensors",
|
| 437 |
+
"model.layers.21.mlp.shared_experts.up_proj.scales": "model-00008-of-00012.safetensors",
|
| 438 |
+
"model.layers.21.mlp.shared_experts.up_proj.weight": "model-00008-of-00012.safetensors",
|
| 439 |
+
"model.layers.21.mlp.switch_mlp.down_proj.biases": "model-00008-of-00012.safetensors",
|
| 440 |
+
"model.layers.21.mlp.switch_mlp.down_proj.scales": "model-00008-of-00012.safetensors",
|
| 441 |
+
"model.layers.21.mlp.switch_mlp.down_proj.weight": "model-00008-of-00012.safetensors",
|
| 442 |
+
"model.layers.21.mlp.switch_mlp.gate_proj.biases": "model-00008-of-00012.safetensors",
|
| 443 |
+
"model.layers.21.mlp.switch_mlp.gate_proj.scales": "model-00008-of-00012.safetensors",
|
| 444 |
+
"model.layers.21.mlp.switch_mlp.gate_proj.weight": "model-00008-of-00012.safetensors",
|
| 445 |
+
"model.layers.21.mlp.switch_mlp.up_proj.biases": "model-00008-of-00012.safetensors",
|
| 446 |
+
"model.layers.21.mlp.switch_mlp.up_proj.scales": "model-00008-of-00012.safetensors",
|
| 447 |
+
"model.layers.21.mlp.switch_mlp.up_proj.weight": "model-00008-of-00012.safetensors",
|
| 448 |
+
"model.layers.21.post_attention_layernorm.weight": "model-00008-of-00012.safetensors",
|
| 449 |
+
"model.layers.22.attention.dense.biases": "model-00008-of-00012.safetensors",
|
| 450 |
+
"model.layers.22.attention.dense.scales": "model-00008-of-00012.safetensors",
|
| 451 |
+
"model.layers.22.attention.dense.weight": "model-00008-of-00012.safetensors",
|
| 452 |
+
"model.layers.22.attention.key_layernorm.weight": "model-00008-of-00012.safetensors",
|
| 453 |
+
"model.layers.22.attention.query_key_value.biases": "model-00008-of-00012.safetensors",
|
| 454 |
+
"model.layers.22.attention.query_key_value.scales": "model-00008-of-00012.safetensors",
|
| 455 |
+
"model.layers.22.attention.query_key_value.weight": "model-00008-of-00012.safetensors",
|
| 456 |
+
"model.layers.22.attention.query_layernorm.weight": "model-00008-of-00012.safetensors",
|
| 457 |
+
"model.layers.22.input_layernorm.weight": "model-00009-of-00012.safetensors",
|
| 458 |
+
"model.layers.22.mlp.gate.expert_bias": "model-00009-of-00012.safetensors",
|
| 459 |
+
"model.layers.22.mlp.gate.weight": "model-00009-of-00012.safetensors",
|
| 460 |
+
"model.layers.22.mlp.shared_experts.down_proj.biases": "model-00009-of-00012.safetensors",
|
| 461 |
+
"model.layers.22.mlp.shared_experts.down_proj.scales": "model-00009-of-00012.safetensors",
|
| 462 |
+
"model.layers.22.mlp.shared_experts.down_proj.weight": "model-00009-of-00012.safetensors",
|
| 463 |
+
"model.layers.22.mlp.shared_experts.gate_proj.biases": "model-00009-of-00012.safetensors",
|
| 464 |
+
"model.layers.22.mlp.shared_experts.gate_proj.scales": "model-00009-of-00012.safetensors",
|
| 465 |
+
"model.layers.22.mlp.shared_experts.gate_proj.weight": "model-00009-of-00012.safetensors",
|
| 466 |
+
"model.layers.22.mlp.shared_experts.up_proj.biases": "model-00009-of-00012.safetensors",
|
| 467 |
+
"model.layers.22.mlp.shared_experts.up_proj.scales": "model-00009-of-00012.safetensors",
|
| 468 |
+
"model.layers.22.mlp.shared_experts.up_proj.weight": "model-00009-of-00012.safetensors",
|
| 469 |
+
"model.layers.22.mlp.switch_mlp.down_proj.biases": "model-00009-of-00012.safetensors",
|
| 470 |
+
"model.layers.22.mlp.switch_mlp.down_proj.scales": "model-00009-of-00012.safetensors",
|
| 471 |
+
"model.layers.22.mlp.switch_mlp.down_proj.weight": "model-00009-of-00012.safetensors",
|
| 472 |
+
"model.layers.22.mlp.switch_mlp.gate_proj.biases": "model-00008-of-00012.safetensors",
|
| 473 |
+
"model.layers.22.mlp.switch_mlp.gate_proj.scales": "model-00008-of-00012.safetensors",
|
| 474 |
+
"model.layers.22.mlp.switch_mlp.gate_proj.weight": "model-00008-of-00012.safetensors",
|
| 475 |
+
"model.layers.22.mlp.switch_mlp.up_proj.biases": "model-00009-of-00012.safetensors",
|
| 476 |
+
"model.layers.22.mlp.switch_mlp.up_proj.scales": "model-00009-of-00012.safetensors",
|
| 477 |
+
"model.layers.22.mlp.switch_mlp.up_proj.weight": "model-00009-of-00012.safetensors",
|
| 478 |
+
"model.layers.22.post_attention_layernorm.weight": "model-00009-of-00012.safetensors",
|
| 479 |
+
"model.layers.23.attention.dense.biases": "model-00009-of-00012.safetensors",
|
| 480 |
+
"model.layers.23.attention.dense.scales": "model-00009-of-00012.safetensors",
|
| 481 |
+
"model.layers.23.attention.dense.weight": "model-00009-of-00012.safetensors",
|
| 482 |
+
"model.layers.23.attention.key_layernorm.weight": "model-00009-of-00012.safetensors",
|
| 483 |
+
"model.layers.23.attention.query_key_value.biases": "model-00009-of-00012.safetensors",
|
| 484 |
+
"model.layers.23.attention.query_key_value.scales": "model-00009-of-00012.safetensors",
|
| 485 |
+
"model.layers.23.attention.query_key_value.weight": "model-00009-of-00012.safetensors",
|
| 486 |
+
"model.layers.23.attention.query_layernorm.weight": "model-00009-of-00012.safetensors",
|
| 487 |
+
"model.layers.23.input_layernorm.weight": "model-00009-of-00012.safetensors",
|
| 488 |
+
"model.layers.23.mlp.gate.expert_bias": "model-00009-of-00012.safetensors",
|
| 489 |
+
"model.layers.23.mlp.gate.weight": "model-00009-of-00012.safetensors",
|
| 490 |
+
"model.layers.23.mlp.shared_experts.down_proj.biases": "model-00009-of-00012.safetensors",
|
| 491 |
+
"model.layers.23.mlp.shared_experts.down_proj.scales": "model-00009-of-00012.safetensors",
|
| 492 |
+
"model.layers.23.mlp.shared_experts.down_proj.weight": "model-00009-of-00012.safetensors",
|
| 493 |
+
"model.layers.23.mlp.shared_experts.gate_proj.biases": "model-00009-of-00012.safetensors",
|
| 494 |
+
"model.layers.23.mlp.shared_experts.gate_proj.scales": "model-00009-of-00012.safetensors",
|
| 495 |
+
"model.layers.23.mlp.shared_experts.gate_proj.weight": "model-00009-of-00012.safetensors",
|
| 496 |
+
"model.layers.23.mlp.shared_experts.up_proj.biases": "model-00009-of-00012.safetensors",
|
| 497 |
+
"model.layers.23.mlp.shared_experts.up_proj.scales": "model-00009-of-00012.safetensors",
|
| 498 |
+
"model.layers.23.mlp.shared_experts.up_proj.weight": "model-00009-of-00012.safetensors",
|
| 499 |
+
"model.layers.23.mlp.switch_mlp.down_proj.biases": "model-00009-of-00012.safetensors",
|
| 500 |
+
"model.layers.23.mlp.switch_mlp.down_proj.scales": "model-00009-of-00012.safetensors",
|
| 501 |
+
"model.layers.23.mlp.switch_mlp.down_proj.weight": "model-00009-of-00012.safetensors",
|
| 502 |
+
"model.layers.23.mlp.switch_mlp.gate_proj.biases": "model-00009-of-00012.safetensors",
|
| 503 |
+
"model.layers.23.mlp.switch_mlp.gate_proj.scales": "model-00009-of-00012.safetensors",
|
| 504 |
+
"model.layers.23.mlp.switch_mlp.gate_proj.weight": "model-00009-of-00012.safetensors",
|
| 505 |
+
"model.layers.23.mlp.switch_mlp.up_proj.biases": "model-00009-of-00012.safetensors",
|
| 506 |
+
"model.layers.23.mlp.switch_mlp.up_proj.scales": "model-00009-of-00012.safetensors",
|
| 507 |
+
"model.layers.23.mlp.switch_mlp.up_proj.weight": "model-00009-of-00012.safetensors",
|
| 508 |
+
"model.layers.23.post_attention_layernorm.weight": "model-00009-of-00012.safetensors",
|
| 509 |
+
"model.layers.24.attention.dense.biases": "model-00009-of-00012.safetensors",
|
| 510 |
+
"model.layers.24.attention.dense.scales": "model-00009-of-00012.safetensors",
|
| 511 |
+
"model.layers.24.attention.dense.weight": "model-00009-of-00012.safetensors",
|
| 512 |
+
"model.layers.24.attention.key_layernorm.weight": "model-00009-of-00012.safetensors",
|
| 513 |
+
"model.layers.24.attention.query_key_value.biases": "model-00009-of-00012.safetensors",
|
| 514 |
+
"model.layers.24.attention.query_key_value.scales": "model-00009-of-00012.safetensors",
|
| 515 |
+
"model.layers.24.attention.query_key_value.weight": "model-00009-of-00012.safetensors",
|
| 516 |
+
"model.layers.24.attention.query_layernorm.weight": "model-00009-of-00012.safetensors",
|
| 517 |
+
"model.layers.24.input_layernorm.weight": "model-00009-of-00012.safetensors",
|
| 518 |
+
"model.layers.24.mlp.gate.expert_bias": "model-00009-of-00012.safetensors",
|
| 519 |
+
"model.layers.24.mlp.gate.weight": "model-00009-of-00012.safetensors",
|
| 520 |
+
"model.layers.24.mlp.shared_experts.down_proj.biases": "model-00009-of-00012.safetensors",
|
| 521 |
+
"model.layers.24.mlp.shared_experts.down_proj.scales": "model-00009-of-00012.safetensors",
|
| 522 |
+
"model.layers.24.mlp.shared_experts.down_proj.weight": "model-00009-of-00012.safetensors",
|
| 523 |
+
"model.layers.24.mlp.shared_experts.gate_proj.biases": "model-00009-of-00012.safetensors",
|
| 524 |
+
"model.layers.24.mlp.shared_experts.gate_proj.scales": "model-00009-of-00012.safetensors",
|
| 525 |
+
"model.layers.24.mlp.shared_experts.gate_proj.weight": "model-00009-of-00012.safetensors",
|
| 526 |
+
"model.layers.24.mlp.shared_experts.up_proj.biases": "model-00009-of-00012.safetensors",
|
| 527 |
+
"model.layers.24.mlp.shared_experts.up_proj.scales": "model-00009-of-00012.safetensors",
|
| 528 |
+
"model.layers.24.mlp.shared_experts.up_proj.weight": "model-00009-of-00012.safetensors",
|
| 529 |
+
"model.layers.24.mlp.switch_mlp.down_proj.biases": "model-00009-of-00012.safetensors",
|
| 530 |
+
"model.layers.24.mlp.switch_mlp.down_proj.scales": "model-00009-of-00012.safetensors",
|
| 531 |
+
"model.layers.24.mlp.switch_mlp.down_proj.weight": "model-00009-of-00012.safetensors",
|
| 532 |
+
"model.layers.24.mlp.switch_mlp.gate_proj.biases": "model-00009-of-00012.safetensors",
|
| 533 |
+
"model.layers.24.mlp.switch_mlp.gate_proj.scales": "model-00009-of-00012.safetensors",
|
| 534 |
+
"model.layers.24.mlp.switch_mlp.gate_proj.weight": "model-00009-of-00012.safetensors",
|
| 535 |
+
"model.layers.24.mlp.switch_mlp.up_proj.biases": "model-00009-of-00012.safetensors",
|
| 536 |
+
"model.layers.24.mlp.switch_mlp.up_proj.scales": "model-00009-of-00012.safetensors",
|
| 537 |
+
"model.layers.24.mlp.switch_mlp.up_proj.weight": "model-00009-of-00012.safetensors",
|
| 538 |
+
"model.layers.24.post_attention_layernorm.weight": "model-00009-of-00012.safetensors",
|
| 539 |
+
"model.layers.25.attention.dense.biases": "model-00009-of-00012.safetensors",
|
| 540 |
+
"model.layers.25.attention.dense.scales": "model-00009-of-00012.safetensors",
|
| 541 |
+
"model.layers.25.attention.dense.weight": "model-00009-of-00012.safetensors",
|
| 542 |
+
"model.layers.25.attention.key_layernorm.weight": "model-00009-of-00012.safetensors",
|
| 543 |
+
"model.layers.25.attention.query_key_value.biases": "model-00009-of-00012.safetensors",
|
| 544 |
+
"model.layers.25.attention.query_key_value.scales": "model-00009-of-00012.safetensors",
|
| 545 |
+
"model.layers.25.attention.query_key_value.weight": "model-00009-of-00012.safetensors",
|
| 546 |
+
"model.layers.25.attention.query_layernorm.weight": "model-00009-of-00012.safetensors",
|
| 547 |
+
"model.layers.25.input_layernorm.weight": "model-00010-of-00012.safetensors",
|
| 548 |
+
"model.layers.25.mlp.gate.expert_bias": "model-00010-of-00012.safetensors",
|
| 549 |
+
"model.layers.25.mlp.gate.weight": "model-00010-of-00012.safetensors",
|
| 550 |
+
"model.layers.25.mlp.shared_experts.down_proj.biases": "model-00010-of-00012.safetensors",
|
| 551 |
+
"model.layers.25.mlp.shared_experts.down_proj.scales": "model-00010-of-00012.safetensors",
|
| 552 |
+
"model.layers.25.mlp.shared_experts.down_proj.weight": "model-00010-of-00012.safetensors",
|
| 553 |
+
"model.layers.25.mlp.shared_experts.gate_proj.biases": "model-00010-of-00012.safetensors",
|
| 554 |
+
"model.layers.25.mlp.shared_experts.gate_proj.scales": "model-00010-of-00012.safetensors",
|
| 555 |
+
"model.layers.25.mlp.shared_experts.gate_proj.weight": "model-00010-of-00012.safetensors",
|
| 556 |
+
"model.layers.25.mlp.shared_experts.up_proj.biases": "model-00010-of-00012.safetensors",
|
| 557 |
+
"model.layers.25.mlp.shared_experts.up_proj.scales": "model-00010-of-00012.safetensors",
|
| 558 |
+
"model.layers.25.mlp.shared_experts.up_proj.weight": "model-00010-of-00012.safetensors",
|
| 559 |
+
"model.layers.25.mlp.switch_mlp.down_proj.biases": "model-00010-of-00012.safetensors",
|
| 560 |
+
"model.layers.25.mlp.switch_mlp.down_proj.scales": "model-00010-of-00012.safetensors",
|
| 561 |
+
"model.layers.25.mlp.switch_mlp.down_proj.weight": "model-00010-of-00012.safetensors",
|
| 562 |
+
"model.layers.25.mlp.switch_mlp.gate_proj.biases": "model-00010-of-00012.safetensors",
|
| 563 |
+
"model.layers.25.mlp.switch_mlp.gate_proj.scales": "model-00010-of-00012.safetensors",
|
| 564 |
+
"model.layers.25.mlp.switch_mlp.gate_proj.weight": "model-00010-of-00012.safetensors",
|
| 565 |
+
"model.layers.25.mlp.switch_mlp.up_proj.biases": "model-00010-of-00012.safetensors",
|
| 566 |
+
"model.layers.25.mlp.switch_mlp.up_proj.scales": "model-00010-of-00012.safetensors",
|
| 567 |
+
"model.layers.25.mlp.switch_mlp.up_proj.weight": "model-00010-of-00012.safetensors",
|
| 568 |
+
"model.layers.25.post_attention_layernorm.weight": "model-00010-of-00012.safetensors",
|
| 569 |
+
"model.layers.26.attention.dense.biases": "model-00010-of-00012.safetensors",
|
| 570 |
+
"model.layers.26.attention.dense.scales": "model-00010-of-00012.safetensors",
|
| 571 |
+
"model.layers.26.attention.dense.weight": "model-00010-of-00012.safetensors",
|
| 572 |
+
"model.layers.26.attention.key_layernorm.weight": "model-00010-of-00012.safetensors",
|
| 573 |
+
"model.layers.26.attention.query_key_value.biases": "model-00010-of-00012.safetensors",
|
| 574 |
+
"model.layers.26.attention.query_key_value.scales": "model-00010-of-00012.safetensors",
|
| 575 |
+
"model.layers.26.attention.query_key_value.weight": "model-00010-of-00012.safetensors",
|
| 576 |
+
"model.layers.26.attention.query_layernorm.weight": "model-00010-of-00012.safetensors",
|
| 577 |
+
"model.layers.26.input_layernorm.weight": "model-00010-of-00012.safetensors",
|
| 578 |
+
"model.layers.26.mlp.gate.expert_bias": "model-00010-of-00012.safetensors",
|
| 579 |
+
"model.layers.26.mlp.gate.weight": "model-00010-of-00012.safetensors",
|
| 580 |
+
"model.layers.26.mlp.shared_experts.down_proj.biases": "model-00010-of-00012.safetensors",
|
| 581 |
+
"model.layers.26.mlp.shared_experts.down_proj.scales": "model-00010-of-00012.safetensors",
|
| 582 |
+
"model.layers.26.mlp.shared_experts.down_proj.weight": "model-00010-of-00012.safetensors",
|
| 583 |
+
"model.layers.26.mlp.shared_experts.gate_proj.biases": "model-00010-of-00012.safetensors",
|
| 584 |
+
"model.layers.26.mlp.shared_experts.gate_proj.scales": "model-00010-of-00012.safetensors",
|
| 585 |
+
"model.layers.26.mlp.shared_experts.gate_proj.weight": "model-00010-of-00012.safetensors",
|
| 586 |
+
"model.layers.26.mlp.shared_experts.up_proj.biases": "model-00010-of-00012.safetensors",
|
| 587 |
+
"model.layers.26.mlp.shared_experts.up_proj.scales": "model-00010-of-00012.safetensors",
|
| 588 |
+
"model.layers.26.mlp.shared_experts.up_proj.weight": "model-00010-of-00012.safetensors",
|
| 589 |
+
"model.layers.26.mlp.switch_mlp.down_proj.biases": "model-00010-of-00012.safetensors",
|
| 590 |
+
"model.layers.26.mlp.switch_mlp.down_proj.scales": "model-00010-of-00012.safetensors",
|
| 591 |
+
"model.layers.26.mlp.switch_mlp.down_proj.weight": "model-00010-of-00012.safetensors",
|
| 592 |
+
"model.layers.26.mlp.switch_mlp.gate_proj.biases": "model-00010-of-00012.safetensors",
|
| 593 |
+
"model.layers.26.mlp.switch_mlp.gate_proj.scales": "model-00010-of-00012.safetensors",
|
| 594 |
+
"model.layers.26.mlp.switch_mlp.gate_proj.weight": "model-00010-of-00012.safetensors",
|
| 595 |
+
"model.layers.26.mlp.switch_mlp.up_proj.biases": "model-00010-of-00012.safetensors",
|
| 596 |
+
"model.layers.26.mlp.switch_mlp.up_proj.scales": "model-00010-of-00012.safetensors",
|
| 597 |
+
"model.layers.26.mlp.switch_mlp.up_proj.weight": "model-00010-of-00012.safetensors",
|
| 598 |
+
"model.layers.26.post_attention_layernorm.weight": "model-00010-of-00012.safetensors",
|
| 599 |
+
"model.layers.27.attention.dense.biases": "model-00010-of-00012.safetensors",
|
| 600 |
+
"model.layers.27.attention.dense.scales": "model-00010-of-00012.safetensors",
|
| 601 |
+
"model.layers.27.attention.dense.weight": "model-00010-of-00012.safetensors",
|
| 602 |
+
"model.layers.27.attention.key_layernorm.weight": "model-00010-of-00012.safetensors",
|
| 603 |
+
"model.layers.27.attention.query_key_value.biases": "model-00010-of-00012.safetensors",
|
| 604 |
+
"model.layers.27.attention.query_key_value.scales": "model-00010-of-00012.safetensors",
|
| 605 |
+
"model.layers.27.attention.query_key_value.weight": "model-00010-of-00012.safetensors",
|
| 606 |
+
"model.layers.27.attention.query_layernorm.weight": "model-00010-of-00012.safetensors",
|
| 607 |
+
"model.layers.27.input_layernorm.weight": "model-00011-of-00012.safetensors",
|
| 608 |
+
"model.layers.27.mlp.gate.expert_bias": "model-00011-of-00012.safetensors",
|
| 609 |
+
"model.layers.27.mlp.gate.weight": "model-00011-of-00012.safetensors",
|
| 610 |
+
"model.layers.27.mlp.shared_experts.down_proj.biases": "model-00011-of-00012.safetensors",
|
| 611 |
+
"model.layers.27.mlp.shared_experts.down_proj.scales": "model-00011-of-00012.safetensors",
|
| 612 |
+
"model.layers.27.mlp.shared_experts.down_proj.weight": "model-00011-of-00012.safetensors",
|
| 613 |
+
"model.layers.27.mlp.shared_experts.gate_proj.biases": "model-00011-of-00012.safetensors",
|
| 614 |
+
"model.layers.27.mlp.shared_experts.gate_proj.scales": "model-00011-of-00012.safetensors",
|
| 615 |
+
"model.layers.27.mlp.shared_experts.gate_proj.weight": "model-00011-of-00012.safetensors",
|
| 616 |
+
"model.layers.27.mlp.shared_experts.up_proj.biases": "model-00011-of-00012.safetensors",
|
| 617 |
+
"model.layers.27.mlp.shared_experts.up_proj.scales": "model-00011-of-00012.safetensors",
|
| 618 |
+
"model.layers.27.mlp.shared_experts.up_proj.weight": "model-00011-of-00012.safetensors",
|
| 619 |
+
"model.layers.27.mlp.switch_mlp.down_proj.biases": "model-00011-of-00012.safetensors",
|
| 620 |
+
"model.layers.27.mlp.switch_mlp.down_proj.scales": "model-00011-of-00012.safetensors",
|
| 621 |
+
"model.layers.27.mlp.switch_mlp.down_proj.weight": "model-00011-of-00012.safetensors",
|
| 622 |
+
"model.layers.27.mlp.switch_mlp.gate_proj.biases": "model-00010-of-00012.safetensors",
|
| 623 |
+
"model.layers.27.mlp.switch_mlp.gate_proj.scales": "model-00010-of-00012.safetensors",
|
| 624 |
+
"model.layers.27.mlp.switch_mlp.gate_proj.weight": "model-00010-of-00012.safetensors",
|
| 625 |
+
"model.layers.27.mlp.switch_mlp.up_proj.biases": "model-00010-of-00012.safetensors",
|
| 626 |
+
"model.layers.27.mlp.switch_mlp.up_proj.scales": "model-00010-of-00012.safetensors",
|
| 627 |
+
"model.layers.27.mlp.switch_mlp.up_proj.weight": "model-00010-of-00012.safetensors",
|
| 628 |
+
"model.layers.27.post_attention_layernorm.weight": "model-00011-of-00012.safetensors",
|
| 629 |
+
"model.layers.28.attention.dense.biases": "model-00011-of-00012.safetensors",
|
| 630 |
+
"model.layers.28.attention.dense.scales": "model-00011-of-00012.safetensors",
|
| 631 |
+
"model.layers.28.attention.dense.weight": "model-00011-of-00012.safetensors",
|
| 632 |
+
"model.layers.28.attention.key_layernorm.weight": "model-00011-of-00012.safetensors",
|
| 633 |
+
"model.layers.28.attention.query_key_value.biases": "model-00011-of-00012.safetensors",
|
| 634 |
+
"model.layers.28.attention.query_key_value.scales": "model-00011-of-00012.safetensors",
|
| 635 |
+
"model.layers.28.attention.query_key_value.weight": "model-00011-of-00012.safetensors",
|
| 636 |
+
"model.layers.28.attention.query_layernorm.weight": "model-00011-of-00012.safetensors",
|
| 637 |
+
"model.layers.28.input_layernorm.weight": "model-00011-of-00012.safetensors",
|
| 638 |
+
"model.layers.28.mlp.gate.expert_bias": "model-00011-of-00012.safetensors",
|
| 639 |
+
"model.layers.28.mlp.gate.weight": "model-00011-of-00012.safetensors",
|
| 640 |
+
"model.layers.28.mlp.shared_experts.down_proj.biases": "model-00011-of-00012.safetensors",
|
| 641 |
+
"model.layers.28.mlp.shared_experts.down_proj.scales": "model-00011-of-00012.safetensors",
|
| 642 |
+
"model.layers.28.mlp.shared_experts.down_proj.weight": "model-00011-of-00012.safetensors",
|
| 643 |
+
"model.layers.28.mlp.shared_experts.gate_proj.biases": "model-00011-of-00012.safetensors",
|
| 644 |
+
"model.layers.28.mlp.shared_experts.gate_proj.scales": "model-00011-of-00012.safetensors",
|
| 645 |
+
"model.layers.28.mlp.shared_experts.gate_proj.weight": "model-00011-of-00012.safetensors",
|
| 646 |
+
"model.layers.28.mlp.shared_experts.up_proj.biases": "model-00011-of-00012.safetensors",
|
| 647 |
+
"model.layers.28.mlp.shared_experts.up_proj.scales": "model-00011-of-00012.safetensors",
|
| 648 |
+
"model.layers.28.mlp.shared_experts.up_proj.weight": "model-00011-of-00012.safetensors",
|
| 649 |
+
"model.layers.28.mlp.switch_mlp.down_proj.biases": "model-00011-of-00012.safetensors",
|
| 650 |
+
"model.layers.28.mlp.switch_mlp.down_proj.scales": "model-00011-of-00012.safetensors",
|
| 651 |
+
"model.layers.28.mlp.switch_mlp.down_proj.weight": "model-00011-of-00012.safetensors",
|
| 652 |
+
"model.layers.28.mlp.switch_mlp.gate_proj.biases": "model-00011-of-00012.safetensors",
|
| 653 |
+
"model.layers.28.mlp.switch_mlp.gate_proj.scales": "model-00011-of-00012.safetensors",
|
| 654 |
+
"model.layers.28.mlp.switch_mlp.gate_proj.weight": "model-00011-of-00012.safetensors",
|
| 655 |
+
"model.layers.28.mlp.switch_mlp.up_proj.biases": "model-00011-of-00012.safetensors",
|
| 656 |
+
"model.layers.28.mlp.switch_mlp.up_proj.scales": "model-00011-of-00012.safetensors",
|
| 657 |
+
"model.layers.28.mlp.switch_mlp.up_proj.weight": "model-00011-of-00012.safetensors",
|
| 658 |
+
"model.layers.28.post_attention_layernorm.weight": "model-00011-of-00012.safetensors",
|
| 659 |
+
"model.layers.29.attention.dense.biases": "model-00011-of-00012.safetensors",
|
| 660 |
+
"model.layers.29.attention.dense.scales": "model-00011-of-00012.safetensors",
|
| 661 |
+
"model.layers.29.attention.dense.weight": "model-00011-of-00012.safetensors",
|
| 662 |
+
"model.layers.29.attention.key_layernorm.weight": "model-00011-of-00012.safetensors",
|
| 663 |
+
"model.layers.29.attention.query_key_value.biases": "model-00011-of-00012.safetensors",
|
| 664 |
+
"model.layers.29.attention.query_key_value.scales": "model-00011-of-00012.safetensors",
|
| 665 |
+
"model.layers.29.attention.query_key_value.weight": "model-00011-of-00012.safetensors",
|
| 666 |
+
"model.layers.29.attention.query_layernorm.weight": "model-00011-of-00012.safetensors",
|
| 667 |
+
"model.layers.29.input_layernorm.weight": "model-00011-of-00012.safetensors",
|
| 668 |
+
"model.layers.29.mlp.gate.expert_bias": "model-00011-of-00012.safetensors",
|
| 669 |
+
"model.layers.29.mlp.gate.weight": "model-00011-of-00012.safetensors",
|
| 670 |
+
"model.layers.29.mlp.shared_experts.down_proj.biases": "model-00011-of-00012.safetensors",
|
| 671 |
+
"model.layers.29.mlp.shared_experts.down_proj.scales": "model-00011-of-00012.safetensors",
|
| 672 |
+
"model.layers.29.mlp.shared_experts.down_proj.weight": "model-00011-of-00012.safetensors",
|
| 673 |
+
"model.layers.29.mlp.shared_experts.gate_proj.biases": "model-00011-of-00012.safetensors",
|
| 674 |
+
"model.layers.29.mlp.shared_experts.gate_proj.scales": "model-00011-of-00012.safetensors",
|
| 675 |
+
"model.layers.29.mlp.shared_experts.gate_proj.weight": "model-00011-of-00012.safetensors",
|
| 676 |
+
"model.layers.29.mlp.shared_experts.up_proj.biases": "model-00011-of-00012.safetensors",
|
| 677 |
+
"model.layers.29.mlp.shared_experts.up_proj.scales": "model-00011-of-00012.safetensors",
|
| 678 |
+
"model.layers.29.mlp.shared_experts.up_proj.weight": "model-00011-of-00012.safetensors",
|
| 679 |
+
"model.layers.29.mlp.switch_mlp.down_proj.biases": "model-00011-of-00012.safetensors",
|
| 680 |
+
"model.layers.29.mlp.switch_mlp.down_proj.scales": "model-00011-of-00012.safetensors",
|
| 681 |
+
"model.layers.29.mlp.switch_mlp.down_proj.weight": "model-00011-of-00012.safetensors",
|
| 682 |
+
"model.layers.29.mlp.switch_mlp.gate_proj.biases": "model-00011-of-00012.safetensors",
|
| 683 |
+
"model.layers.29.mlp.switch_mlp.gate_proj.scales": "model-00011-of-00012.safetensors",
|
| 684 |
+
"model.layers.29.mlp.switch_mlp.gate_proj.weight": "model-00011-of-00012.safetensors",
|
| 685 |
+
"model.layers.29.mlp.switch_mlp.up_proj.biases": "model-00011-of-00012.safetensors",
|
| 686 |
+
"model.layers.29.mlp.switch_mlp.up_proj.scales": "model-00011-of-00012.safetensors",
|
| 687 |
+
"model.layers.29.mlp.switch_mlp.up_proj.weight": "model-00011-of-00012.safetensors",
|
| 688 |
+
"model.layers.29.post_attention_layernorm.weight": "model-00011-of-00012.safetensors",
|
| 689 |
+
"model.layers.3.attention.dense.biases": "model-00001-of-00012.safetensors",
|
| 690 |
+
"model.layers.3.attention.dense.scales": "model-00001-of-00012.safetensors",
|
| 691 |
+
"model.layers.3.attention.dense.weight": "model-00001-of-00012.safetensors",
|
| 692 |
+
"model.layers.3.attention.key_layernorm.weight": "model-00001-of-00012.safetensors",
|
| 693 |
+
"model.layers.3.attention.query_key_value.biases": "model-00001-of-00012.safetensors",
|
| 694 |
+
"model.layers.3.attention.query_key_value.scales": "model-00001-of-00012.safetensors",
|
| 695 |
+
"model.layers.3.attention.query_key_value.weight": "model-00001-of-00012.safetensors",
|
| 696 |
+
"model.layers.3.attention.query_layernorm.weight": "model-00001-of-00012.safetensors",
|
| 697 |
+
"model.layers.3.input_layernorm.weight": "model-00002-of-00012.safetensors",
|
| 698 |
+
"model.layers.3.mlp.gate.expert_bias": "model-00002-of-00012.safetensors",
|
| 699 |
+
"model.layers.3.mlp.gate.weight": "model-00002-of-00012.safetensors",
|
| 700 |
+
"model.layers.3.mlp.shared_experts.down_proj.biases": "model-00002-of-00012.safetensors",
|
| 701 |
+
"model.layers.3.mlp.shared_experts.down_proj.scales": "model-00002-of-00012.safetensors",
|
| 702 |
+
"model.layers.3.mlp.shared_experts.down_proj.weight": "model-00002-of-00012.safetensors",
|
| 703 |
+
"model.layers.3.mlp.shared_experts.gate_proj.biases": "model-00002-of-00012.safetensors",
|
| 704 |
+
"model.layers.3.mlp.shared_experts.gate_proj.scales": "model-00002-of-00012.safetensors",
|
| 705 |
+
"model.layers.3.mlp.shared_experts.gate_proj.weight": "model-00002-of-00012.safetensors",
|
| 706 |
+
"model.layers.3.mlp.shared_experts.up_proj.biases": "model-00002-of-00012.safetensors",
|
| 707 |
+
"model.layers.3.mlp.shared_experts.up_proj.scales": "model-00002-of-00012.safetensors",
|
| 708 |
+
"model.layers.3.mlp.shared_experts.up_proj.weight": "model-00002-of-00012.safetensors",
|
| 709 |
+
"model.layers.3.mlp.switch_mlp.down_proj.biases": "model-00002-of-00012.safetensors",
|
| 710 |
+
"model.layers.3.mlp.switch_mlp.down_proj.scales": "model-00002-of-00012.safetensors",
|
| 711 |
+
"model.layers.3.mlp.switch_mlp.down_proj.weight": "model-00002-of-00012.safetensors",
|
| 712 |
+
"model.layers.3.mlp.switch_mlp.gate_proj.biases": "model-00001-of-00012.safetensors",
|
| 713 |
+
"model.layers.3.mlp.switch_mlp.gate_proj.scales": "model-00001-of-00012.safetensors",
|
| 714 |
+
"model.layers.3.mlp.switch_mlp.gate_proj.weight": "model-00001-of-00012.safetensors",
|
| 715 |
+
"model.layers.3.mlp.switch_mlp.up_proj.biases": "model-00001-of-00012.safetensors",
|
| 716 |
+
"model.layers.3.mlp.switch_mlp.up_proj.scales": "model-00001-of-00012.safetensors",
|
| 717 |
+
"model.layers.3.mlp.switch_mlp.up_proj.weight": "model-00001-of-00012.safetensors",
|
| 718 |
+
"model.layers.3.post_attention_layernorm.weight": "model-00002-of-00012.safetensors",
|
| 719 |
+
"model.layers.30.attention.dense.biases": "model-00011-of-00012.safetensors",
|
| 720 |
+
"model.layers.30.attention.dense.scales": "model-00011-of-00012.safetensors",
|
| 721 |
+
"model.layers.30.attention.dense.weight": "model-00011-of-00012.safetensors",
|
| 722 |
+
"model.layers.30.attention.key_layernorm.weight": "model-00011-of-00012.safetensors",
|
| 723 |
+
"model.layers.30.attention.query_key_value.biases": "model-00011-of-00012.safetensors",
|
| 724 |
+
"model.layers.30.attention.query_key_value.scales": "model-00011-of-00012.safetensors",
|
| 725 |
+
"model.layers.30.attention.query_key_value.weight": "model-00011-of-00012.safetensors",
|
| 726 |
+
"model.layers.30.attention.query_layernorm.weight": "model-00011-of-00012.safetensors",
|
| 727 |
+
"model.layers.30.input_layernorm.weight": "model-00012-of-00012.safetensors",
|
| 728 |
+
"model.layers.30.mlp.gate.expert_bias": "model-00012-of-00012.safetensors",
|
| 729 |
+
"model.layers.30.mlp.gate.weight": "model-00012-of-00012.safetensors",
|
| 730 |
+
"model.layers.30.mlp.shared_experts.down_proj.biases": "model-00012-of-00012.safetensors",
|
| 731 |
+
"model.layers.30.mlp.shared_experts.down_proj.scales": "model-00012-of-00012.safetensors",
|
| 732 |
+
"model.layers.30.mlp.shared_experts.down_proj.weight": "model-00012-of-00012.safetensors",
|
| 733 |
+
"model.layers.30.mlp.shared_experts.gate_proj.biases": "model-00012-of-00012.safetensors",
|
| 734 |
+
"model.layers.30.mlp.shared_experts.gate_proj.scales": "model-00012-of-00012.safetensors",
|
| 735 |
+
"model.layers.30.mlp.shared_experts.gate_proj.weight": "model-00012-of-00012.safetensors",
|
| 736 |
+
"model.layers.30.mlp.shared_experts.up_proj.biases": "model-00012-of-00012.safetensors",
|
| 737 |
+
"model.layers.30.mlp.shared_experts.up_proj.scales": "model-00012-of-00012.safetensors",
|
| 738 |
+
"model.layers.30.mlp.shared_experts.up_proj.weight": "model-00012-of-00012.safetensors",
|
| 739 |
+
"model.layers.30.mlp.switch_mlp.down_proj.biases": "model-00012-of-00012.safetensors",
|
| 740 |
+
"model.layers.30.mlp.switch_mlp.down_proj.scales": "model-00012-of-00012.safetensors",
|
| 741 |
+
"model.layers.30.mlp.switch_mlp.down_proj.weight": "model-00012-of-00012.safetensors",
|
| 742 |
+
"model.layers.30.mlp.switch_mlp.gate_proj.biases": "model-00011-of-00012.safetensors",
|
| 743 |
+
"model.layers.30.mlp.switch_mlp.gate_proj.scales": "model-00011-of-00012.safetensors",
|
| 744 |
+
"model.layers.30.mlp.switch_mlp.gate_proj.weight": "model-00011-of-00012.safetensors",
|
| 745 |
+
"model.layers.30.mlp.switch_mlp.up_proj.biases": "model-00012-of-00012.safetensors",
|
| 746 |
+
"model.layers.30.mlp.switch_mlp.up_proj.scales": "model-00012-of-00012.safetensors",
|
| 747 |
+
"model.layers.30.mlp.switch_mlp.up_proj.weight": "model-00012-of-00012.safetensors",
|
| 748 |
+
"model.layers.30.post_attention_layernorm.weight": "model-00012-of-00012.safetensors",
|
| 749 |
+
"model.layers.31.attention.dense.biases": "model-00012-of-00012.safetensors",
|
| 750 |
+
"model.layers.31.attention.dense.scales": "model-00012-of-00012.safetensors",
|
| 751 |
+
"model.layers.31.attention.dense.weight": "model-00012-of-00012.safetensors",
|
| 752 |
+
"model.layers.31.attention.key_layernorm.weight": "model-00012-of-00012.safetensors",
|
| 753 |
+
"model.layers.31.attention.query_key_value.biases": "model-00012-of-00012.safetensors",
|
| 754 |
+
"model.layers.31.attention.query_key_value.scales": "model-00012-of-00012.safetensors",
|
| 755 |
+
"model.layers.31.attention.query_key_value.weight": "model-00012-of-00012.safetensors",
|
| 756 |
+
"model.layers.31.attention.query_layernorm.weight": "model-00012-of-00012.safetensors",
|
| 757 |
+
"model.layers.31.input_layernorm.weight": "model-00012-of-00012.safetensors",
|
| 758 |
+
"model.layers.31.mlp.gate.expert_bias": "model-00012-of-00012.safetensors",
|
| 759 |
+
"model.layers.31.mlp.gate.weight": "model-00012-of-00012.safetensors",
|
| 760 |
+
"model.layers.31.mlp.shared_experts.down_proj.biases": "model-00012-of-00012.safetensors",
|
| 761 |
+
"model.layers.31.mlp.shared_experts.down_proj.scales": "model-00012-of-00012.safetensors",
|
| 762 |
+
"model.layers.31.mlp.shared_experts.down_proj.weight": "model-00012-of-00012.safetensors",
|
| 763 |
+
"model.layers.31.mlp.shared_experts.gate_proj.biases": "model-00012-of-00012.safetensors",
|
| 764 |
+
"model.layers.31.mlp.shared_experts.gate_proj.scales": "model-00012-of-00012.safetensors",
|
| 765 |
+
"model.layers.31.mlp.shared_experts.gate_proj.weight": "model-00012-of-00012.safetensors",
|
| 766 |
+
"model.layers.31.mlp.shared_experts.up_proj.biases": "model-00012-of-00012.safetensors",
|
| 767 |
+
"model.layers.31.mlp.shared_experts.up_proj.scales": "model-00012-of-00012.safetensors",
|
| 768 |
+
"model.layers.31.mlp.shared_experts.up_proj.weight": "model-00012-of-00012.safetensors",
|
| 769 |
+
"model.layers.31.mlp.switch_mlp.down_proj.biases": "model-00012-of-00012.safetensors",
|
| 770 |
+
"model.layers.31.mlp.switch_mlp.down_proj.scales": "model-00012-of-00012.safetensors",
|
| 771 |
+
"model.layers.31.mlp.switch_mlp.down_proj.weight": "model-00012-of-00012.safetensors",
|
| 772 |
+
"model.layers.31.mlp.switch_mlp.gate_proj.biases": "model-00012-of-00012.safetensors",
|
| 773 |
+
"model.layers.31.mlp.switch_mlp.gate_proj.scales": "model-00012-of-00012.safetensors",
|
| 774 |
+
"model.layers.31.mlp.switch_mlp.gate_proj.weight": "model-00012-of-00012.safetensors",
|
| 775 |
+
"model.layers.31.mlp.switch_mlp.up_proj.biases": "model-00012-of-00012.safetensors",
|
| 776 |
+
"model.layers.31.mlp.switch_mlp.up_proj.scales": "model-00012-of-00012.safetensors",
|
| 777 |
+
"model.layers.31.mlp.switch_mlp.up_proj.weight": "model-00012-of-00012.safetensors",
|
| 778 |
+
"model.layers.31.post_attention_layernorm.weight": "model-00012-of-00012.safetensors",
|
| 779 |
+
"model.layers.4.attention.dense.biases": "model-00002-of-00012.safetensors",
|
| 780 |
+
"model.layers.4.attention.dense.scales": "model-00002-of-00012.safetensors",
|
| 781 |
+
"model.layers.4.attention.dense.weight": "model-00002-of-00012.safetensors",
|
| 782 |
+
"model.layers.4.attention.key_layernorm.weight": "model-00002-of-00012.safetensors",
|
| 783 |
+
"model.layers.4.attention.query_key_value.biases": "model-00002-of-00012.safetensors",
|
| 784 |
+
"model.layers.4.attention.query_key_value.scales": "model-00002-of-00012.safetensors",
|
| 785 |
+
"model.layers.4.attention.query_key_value.weight": "model-00002-of-00012.safetensors",
|
| 786 |
+
"model.layers.4.attention.query_layernorm.weight": "model-00002-of-00012.safetensors",
|
| 787 |
+
"model.layers.4.input_layernorm.weight": "model-00002-of-00012.safetensors",
|
| 788 |
+
"model.layers.4.mlp.gate.expert_bias": "model-00002-of-00012.safetensors",
|
| 789 |
+
"model.layers.4.mlp.gate.weight": "model-00002-of-00012.safetensors",
|
| 790 |
+
"model.layers.4.mlp.shared_experts.down_proj.biases": "model-00002-of-00012.safetensors",
|
| 791 |
+
"model.layers.4.mlp.shared_experts.down_proj.scales": "model-00002-of-00012.safetensors",
|
| 792 |
+
"model.layers.4.mlp.shared_experts.down_proj.weight": "model-00002-of-00012.safetensors",
|
| 793 |
+
"model.layers.4.mlp.shared_experts.gate_proj.biases": "model-00002-of-00012.safetensors",
|
| 794 |
+
"model.layers.4.mlp.shared_experts.gate_proj.scales": "model-00002-of-00012.safetensors",
|
| 795 |
+
"model.layers.4.mlp.shared_experts.gate_proj.weight": "model-00002-of-00012.safetensors",
|
| 796 |
+
"model.layers.4.mlp.shared_experts.up_proj.biases": "model-00002-of-00012.safetensors",
|
| 797 |
+
"model.layers.4.mlp.shared_experts.up_proj.scales": "model-00002-of-00012.safetensors",
|
| 798 |
+
"model.layers.4.mlp.shared_experts.up_proj.weight": "model-00002-of-00012.safetensors",
|
| 799 |
+
"model.layers.4.mlp.switch_mlp.down_proj.biases": "model-00002-of-00012.safetensors",
|
| 800 |
+
"model.layers.4.mlp.switch_mlp.down_proj.scales": "model-00002-of-00012.safetensors",
|
| 801 |
+
"model.layers.4.mlp.switch_mlp.down_proj.weight": "model-00002-of-00012.safetensors",
|
| 802 |
+
"model.layers.4.mlp.switch_mlp.gate_proj.biases": "model-00002-of-00012.safetensors",
|
| 803 |
+
"model.layers.4.mlp.switch_mlp.gate_proj.scales": "model-00002-of-00012.safetensors",
|
| 804 |
+
"model.layers.4.mlp.switch_mlp.gate_proj.weight": "model-00002-of-00012.safetensors",
|
| 805 |
+
"model.layers.4.mlp.switch_mlp.up_proj.biases": "model-00002-of-00012.safetensors",
|
| 806 |
+
"model.layers.4.mlp.switch_mlp.up_proj.scales": "model-00002-of-00012.safetensors",
|
| 807 |
+
"model.layers.4.mlp.switch_mlp.up_proj.weight": "model-00002-of-00012.safetensors",
|
| 808 |
+
"model.layers.4.post_attention_layernorm.weight": "model-00002-of-00012.safetensors",
|
| 809 |
+
"model.layers.5.attention.dense.biases": "model-00002-of-00012.safetensors",
|
| 810 |
+
"model.layers.5.attention.dense.scales": "model-00002-of-00012.safetensors",
|
| 811 |
+
"model.layers.5.attention.dense.weight": "model-00002-of-00012.safetensors",
|
| 812 |
+
"model.layers.5.attention.key_layernorm.weight": "model-00002-of-00012.safetensors",
|
| 813 |
+
"model.layers.5.attention.query_key_value.biases": "model-00002-of-00012.safetensors",
|
| 814 |
+
"model.layers.5.attention.query_key_value.scales": "model-00002-of-00012.safetensors",
|
| 815 |
+
"model.layers.5.attention.query_key_value.weight": "model-00002-of-00012.safetensors",
|
| 816 |
+
"model.layers.5.attention.query_layernorm.weight": "model-00002-of-00012.safetensors",
|
| 817 |
+
"model.layers.5.input_layernorm.weight": "model-00002-of-00012.safetensors",
|
| 818 |
+
"model.layers.5.mlp.gate.expert_bias": "model-00002-of-00012.safetensors",
|
| 819 |
+
"model.layers.5.mlp.gate.weight": "model-00002-of-00012.safetensors",
|
| 820 |
+
"model.layers.5.mlp.shared_experts.down_proj.biases": "model-00002-of-00012.safetensors",
|
| 821 |
+
"model.layers.5.mlp.shared_experts.down_proj.scales": "model-00002-of-00012.safetensors",
|
| 822 |
+
"model.layers.5.mlp.shared_experts.down_proj.weight": "model-00002-of-00012.safetensors",
|
| 823 |
+
"model.layers.5.mlp.shared_experts.gate_proj.biases": "model-00002-of-00012.safetensors",
|
| 824 |
+
"model.layers.5.mlp.shared_experts.gate_proj.scales": "model-00002-of-00012.safetensors",
|
| 825 |
+
"model.layers.5.mlp.shared_experts.gate_proj.weight": "model-00002-of-00012.safetensors",
|
| 826 |
+
"model.layers.5.mlp.shared_experts.up_proj.biases": "model-00002-of-00012.safetensors",
|
| 827 |
+
"model.layers.5.mlp.shared_experts.up_proj.scales": "model-00002-of-00012.safetensors",
|
| 828 |
+
"model.layers.5.mlp.shared_experts.up_proj.weight": "model-00002-of-00012.safetensors",
|
| 829 |
+
"model.layers.5.mlp.switch_mlp.down_proj.biases": "model-00002-of-00012.safetensors",
|
| 830 |
+
"model.layers.5.mlp.switch_mlp.down_proj.scales": "model-00002-of-00012.safetensors",
|
| 831 |
+
"model.layers.5.mlp.switch_mlp.down_proj.weight": "model-00002-of-00012.safetensors",
|
| 832 |
+
"model.layers.5.mlp.switch_mlp.gate_proj.biases": "model-00002-of-00012.safetensors",
|
| 833 |
+
"model.layers.5.mlp.switch_mlp.gate_proj.scales": "model-00002-of-00012.safetensors",
|
| 834 |
+
"model.layers.5.mlp.switch_mlp.gate_proj.weight": "model-00002-of-00012.safetensors",
|
| 835 |
+
"model.layers.5.mlp.switch_mlp.up_proj.biases": "model-00002-of-00012.safetensors",
|
| 836 |
+
"model.layers.5.mlp.switch_mlp.up_proj.scales": "model-00002-of-00012.safetensors",
|
| 837 |
+
"model.layers.5.mlp.switch_mlp.up_proj.weight": "model-00002-of-00012.safetensors",
|
| 838 |
+
"model.layers.5.post_attention_layernorm.weight": "model-00002-of-00012.safetensors",
|
| 839 |
+
"model.layers.6.attention.dense.biases": "model-00002-of-00012.safetensors",
|
| 840 |
+
"model.layers.6.attention.dense.scales": "model-00002-of-00012.safetensors",
|
| 841 |
+
"model.layers.6.attention.dense.weight": "model-00002-of-00012.safetensors",
|
| 842 |
+
"model.layers.6.attention.key_layernorm.weight": "model-00002-of-00012.safetensors",
|
| 843 |
+
"model.layers.6.attention.query_key_value.biases": "model-00002-of-00012.safetensors",
|
| 844 |
+
"model.layers.6.attention.query_key_value.scales": "model-00002-of-00012.safetensors",
|
| 845 |
+
"model.layers.6.attention.query_key_value.weight": "model-00002-of-00012.safetensors",
|
| 846 |
+
"model.layers.6.attention.query_layernorm.weight": "model-00002-of-00012.safetensors",
|
| 847 |
+
"model.layers.6.input_layernorm.weight": "model-00003-of-00012.safetensors",
|
| 848 |
+
"model.layers.6.mlp.gate.expert_bias": "model-00003-of-00012.safetensors",
|
| 849 |
+
"model.layers.6.mlp.gate.weight": "model-00003-of-00012.safetensors",
|
| 850 |
+
"model.layers.6.mlp.shared_experts.down_proj.biases": "model-00003-of-00012.safetensors",
|
| 851 |
+
"model.layers.6.mlp.shared_experts.down_proj.scales": "model-00003-of-00012.safetensors",
|
| 852 |
+
"model.layers.6.mlp.shared_experts.down_proj.weight": "model-00003-of-00012.safetensors",
|
| 853 |
+
"model.layers.6.mlp.shared_experts.gate_proj.biases": "model-00003-of-00012.safetensors",
|
| 854 |
+
"model.layers.6.mlp.shared_experts.gate_proj.scales": "model-00003-of-00012.safetensors",
|
| 855 |
+
"model.layers.6.mlp.shared_experts.gate_proj.weight": "model-00003-of-00012.safetensors",
|
| 856 |
+
"model.layers.6.mlp.shared_experts.up_proj.biases": "model-00003-of-00012.safetensors",
|
| 857 |
+
"model.layers.6.mlp.shared_experts.up_proj.scales": "model-00003-of-00012.safetensors",
|
| 858 |
+
"model.layers.6.mlp.shared_experts.up_proj.weight": "model-00003-of-00012.safetensors",
|
| 859 |
+
"model.layers.6.mlp.switch_mlp.down_proj.biases": "model-00003-of-00012.safetensors",
|
| 860 |
+
"model.layers.6.mlp.switch_mlp.down_proj.scales": "model-00003-of-00012.safetensors",
|
| 861 |
+
"model.layers.6.mlp.switch_mlp.down_proj.weight": "model-00003-of-00012.safetensors",
|
| 862 |
+
"model.layers.6.mlp.switch_mlp.gate_proj.biases": "model-00002-of-00012.safetensors",
|
| 863 |
+
"model.layers.6.mlp.switch_mlp.gate_proj.scales": "model-00002-of-00012.safetensors",
|
| 864 |
+
"model.layers.6.mlp.switch_mlp.gate_proj.weight": "model-00002-of-00012.safetensors",
|
| 865 |
+
"model.layers.6.mlp.switch_mlp.up_proj.biases": "model-00003-of-00012.safetensors",
|
| 866 |
+
"model.layers.6.mlp.switch_mlp.up_proj.scales": "model-00003-of-00012.safetensors",
|
| 867 |
+
"model.layers.6.mlp.switch_mlp.up_proj.weight": "model-00003-of-00012.safetensors",
|
| 868 |
+
"model.layers.6.post_attention_layernorm.weight": "model-00003-of-00012.safetensors",
|
| 869 |
+
"model.layers.7.attention.dense.biases": "model-00003-of-00012.safetensors",
|
| 870 |
+
"model.layers.7.attention.dense.scales": "model-00003-of-00012.safetensors",
|
| 871 |
+
"model.layers.7.attention.dense.weight": "model-00003-of-00012.safetensors",
|
| 872 |
+
"model.layers.7.attention.key_layernorm.weight": "model-00003-of-00012.safetensors",
|
| 873 |
+
"model.layers.7.attention.query_key_value.biases": "model-00003-of-00012.safetensors",
|
| 874 |
+
"model.layers.7.attention.query_key_value.scales": "model-00003-of-00012.safetensors",
|
| 875 |
+
"model.layers.7.attention.query_key_value.weight": "model-00003-of-00012.safetensors",
|
| 876 |
+
"model.layers.7.attention.query_layernorm.weight": "model-00003-of-00012.safetensors",
|
| 877 |
+
"model.layers.7.input_layernorm.weight": "model-00003-of-00012.safetensors",
|
| 878 |
+
"model.layers.7.mlp.gate.expert_bias": "model-00003-of-00012.safetensors",
|
| 879 |
+
"model.layers.7.mlp.gate.weight": "model-00003-of-00012.safetensors",
|
| 880 |
+
"model.layers.7.mlp.shared_experts.down_proj.biases": "model-00003-of-00012.safetensors",
|
| 881 |
+
"model.layers.7.mlp.shared_experts.down_proj.scales": "model-00003-of-00012.safetensors",
|
| 882 |
+
"model.layers.7.mlp.shared_experts.down_proj.weight": "model-00003-of-00012.safetensors",
|
| 883 |
+
"model.layers.7.mlp.shared_experts.gate_proj.biases": "model-00003-of-00012.safetensors",
|
| 884 |
+
"model.layers.7.mlp.shared_experts.gate_proj.scales": "model-00003-of-00012.safetensors",
|
| 885 |
+
"model.layers.7.mlp.shared_experts.gate_proj.weight": "model-00003-of-00012.safetensors",
|
| 886 |
+
"model.layers.7.mlp.shared_experts.up_proj.biases": "model-00003-of-00012.safetensors",
|
| 887 |
+
"model.layers.7.mlp.shared_experts.up_proj.scales": "model-00003-of-00012.safetensors",
|
| 888 |
+
"model.layers.7.mlp.shared_experts.up_proj.weight": "model-00003-of-00012.safetensors",
|
| 889 |
+
"model.layers.7.mlp.switch_mlp.down_proj.biases": "model-00003-of-00012.safetensors",
|
| 890 |
+
"model.layers.7.mlp.switch_mlp.down_proj.scales": "model-00003-of-00012.safetensors",
|
| 891 |
+
"model.layers.7.mlp.switch_mlp.down_proj.weight": "model-00003-of-00012.safetensors",
|
| 892 |
+
"model.layers.7.mlp.switch_mlp.gate_proj.biases": "model-00003-of-00012.safetensors",
|
| 893 |
+
"model.layers.7.mlp.switch_mlp.gate_proj.scales": "model-00003-of-00012.safetensors",
|
| 894 |
+
"model.layers.7.mlp.switch_mlp.gate_proj.weight": "model-00003-of-00012.safetensors",
|
| 895 |
+
"model.layers.7.mlp.switch_mlp.up_proj.biases": "model-00003-of-00012.safetensors",
|
| 896 |
+
"model.layers.7.mlp.switch_mlp.up_proj.scales": "model-00003-of-00012.safetensors",
|
| 897 |
+
"model.layers.7.mlp.switch_mlp.up_proj.weight": "model-00003-of-00012.safetensors",
|
| 898 |
+
"model.layers.7.post_attention_layernorm.weight": "model-00003-of-00012.safetensors",
|
| 899 |
+
"model.layers.8.attention.dense.biases": "model-00003-of-00012.safetensors",
|
| 900 |
+
"model.layers.8.attention.dense.scales": "model-00003-of-00012.safetensors",
|
| 901 |
+
"model.layers.8.attention.dense.weight": "model-00003-of-00012.safetensors",
|
| 902 |
+
"model.layers.8.attention.key_layernorm.weight": "model-00003-of-00012.safetensors",
|
| 903 |
+
"model.layers.8.attention.query_key_value.biases": "model-00003-of-00012.safetensors",
|
| 904 |
+
"model.layers.8.attention.query_key_value.scales": "model-00003-of-00012.safetensors",
|
| 905 |
+
"model.layers.8.attention.query_key_value.weight": "model-00003-of-00012.safetensors",
|
| 906 |
+
"model.layers.8.attention.query_layernorm.weight": "model-00003-of-00012.safetensors",
|
| 907 |
+
"model.layers.8.input_layernorm.weight": "model-00003-of-00012.safetensors",
|
| 908 |
+
"model.layers.8.mlp.gate.expert_bias": "model-00003-of-00012.safetensors",
|
| 909 |
+
"model.layers.8.mlp.gate.weight": "model-00003-of-00012.safetensors",
|
| 910 |
+
"model.layers.8.mlp.shared_experts.down_proj.biases": "model-00003-of-00012.safetensors",
|
| 911 |
+
"model.layers.8.mlp.shared_experts.down_proj.scales": "model-00003-of-00012.safetensors",
|
| 912 |
+
"model.layers.8.mlp.shared_experts.down_proj.weight": "model-00003-of-00012.safetensors",
|
| 913 |
+
"model.layers.8.mlp.shared_experts.gate_proj.biases": "model-00003-of-00012.safetensors",
|
| 914 |
+
"model.layers.8.mlp.shared_experts.gate_proj.scales": "model-00003-of-00012.safetensors",
|
| 915 |
+
"model.layers.8.mlp.shared_experts.gate_proj.weight": "model-00003-of-00012.safetensors",
|
| 916 |
+
"model.layers.8.mlp.shared_experts.up_proj.biases": "model-00003-of-00012.safetensors",
|
| 917 |
+
"model.layers.8.mlp.shared_experts.up_proj.scales": "model-00003-of-00012.safetensors",
|
| 918 |
+
"model.layers.8.mlp.shared_experts.up_proj.weight": "model-00003-of-00012.safetensors",
|
| 919 |
+
"model.layers.8.mlp.switch_mlp.down_proj.biases": "model-00003-of-00012.safetensors",
|
| 920 |
+
"model.layers.8.mlp.switch_mlp.down_proj.scales": "model-00003-of-00012.safetensors",
|
| 921 |
+
"model.layers.8.mlp.switch_mlp.down_proj.weight": "model-00003-of-00012.safetensors",
|
| 922 |
+
"model.layers.8.mlp.switch_mlp.gate_proj.biases": "model-00003-of-00012.safetensors",
|
| 923 |
+
"model.layers.8.mlp.switch_mlp.gate_proj.scales": "model-00003-of-00012.safetensors",
|
| 924 |
+
"model.layers.8.mlp.switch_mlp.gate_proj.weight": "model-00003-of-00012.safetensors",
|
| 925 |
+
"model.layers.8.mlp.switch_mlp.up_proj.biases": "model-00003-of-00012.safetensors",
|
| 926 |
+
"model.layers.8.mlp.switch_mlp.up_proj.scales": "model-00003-of-00012.safetensors",
|
| 927 |
+
"model.layers.8.mlp.switch_mlp.up_proj.weight": "model-00003-of-00012.safetensors",
|
| 928 |
+
"model.layers.8.post_attention_layernorm.weight": "model-00003-of-00012.safetensors",
|
| 929 |
+
"model.layers.9.attention.dense.biases": "model-00003-of-00012.safetensors",
|
| 930 |
+
"model.layers.9.attention.dense.scales": "model-00003-of-00012.safetensors",
|
| 931 |
+
"model.layers.9.attention.dense.weight": "model-00003-of-00012.safetensors",
|
| 932 |
+
"model.layers.9.attention.key_layernorm.weight": "model-00003-of-00012.safetensors",
|
| 933 |
+
"model.layers.9.attention.query_key_value.biases": "model-00003-of-00012.safetensors",
|
| 934 |
+
"model.layers.9.attention.query_key_value.scales": "model-00003-of-00012.safetensors",
|
| 935 |
+
"model.layers.9.attention.query_key_value.weight": "model-00003-of-00012.safetensors",
|
| 936 |
+
"model.layers.9.attention.query_layernorm.weight": "model-00003-of-00012.safetensors",
|
| 937 |
+
"model.layers.9.input_layernorm.weight": "model-00004-of-00012.safetensors",
|
| 938 |
+
"model.layers.9.mlp.gate.expert_bias": "model-00004-of-00012.safetensors",
|
| 939 |
+
"model.layers.9.mlp.gate.weight": "model-00004-of-00012.safetensors",
|
| 940 |
+
"model.layers.9.mlp.shared_experts.down_proj.biases": "model-00004-of-00012.safetensors",
|
| 941 |
+
"model.layers.9.mlp.shared_experts.down_proj.scales": "model-00004-of-00012.safetensors",
|
| 942 |
+
"model.layers.9.mlp.shared_experts.down_proj.weight": "model-00004-of-00012.safetensors",
|
| 943 |
+
"model.layers.9.mlp.shared_experts.gate_proj.biases": "model-00004-of-00012.safetensors",
|
| 944 |
+
"model.layers.9.mlp.shared_experts.gate_proj.scales": "model-00004-of-00012.safetensors",
|
| 945 |
+
"model.layers.9.mlp.shared_experts.gate_proj.weight": "model-00004-of-00012.safetensors",
|
| 946 |
+
"model.layers.9.mlp.shared_experts.up_proj.biases": "model-00004-of-00012.safetensors",
|
| 947 |
+
"model.layers.9.mlp.shared_experts.up_proj.scales": "model-00004-of-00012.safetensors",
|
| 948 |
+
"model.layers.9.mlp.shared_experts.up_proj.weight": "model-00004-of-00012.safetensors",
|
| 949 |
+
"model.layers.9.mlp.switch_mlp.down_proj.biases": "model-00004-of-00012.safetensors",
|
| 950 |
+
"model.layers.9.mlp.switch_mlp.down_proj.scales": "model-00004-of-00012.safetensors",
|
| 951 |
+
"model.layers.9.mlp.switch_mlp.down_proj.weight": "model-00004-of-00012.safetensors",
|
| 952 |
+
"model.layers.9.mlp.switch_mlp.gate_proj.biases": "model-00004-of-00012.safetensors",
|
| 953 |
+
"model.layers.9.mlp.switch_mlp.gate_proj.scales": "model-00004-of-00012.safetensors",
|
| 954 |
+
"model.layers.9.mlp.switch_mlp.gate_proj.weight": "model-00004-of-00012.safetensors",
|
| 955 |
+
"model.layers.9.mlp.switch_mlp.up_proj.biases": "model-00004-of-00012.safetensors",
|
| 956 |
+
"model.layers.9.mlp.switch_mlp.up_proj.scales": "model-00004-of-00012.safetensors",
|
| 957 |
+
"model.layers.9.mlp.switch_mlp.up_proj.weight": "model-00004-of-00012.safetensors",
|
| 958 |
+
"model.layers.9.post_attention_layernorm.weight": "model-00004-of-00012.safetensors",
|
| 959 |
+
"model.norm.weight": "model-00012-of-00012.safetensors",
|
| 960 |
+
"model.word_embeddings.biases": "model-00001-of-00012.safetensors",
|
| 961 |
+
"model.word_embeddings.scales": "model-00001-of-00012.safetensors",
|
| 962 |
+
"model.word_embeddings.weight": "model-00001-of-00012.safetensors"
|
| 963 |
+
}
|
| 964 |
+
}
|
modeling_llada2_moe.py
ADDED
|
@@ -0,0 +1,1621 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# coding=utf-8
|
| 2 |
+
# Copyright 2025 Antgroup and The HuggingFace Inc. team. All rights reserved.
|
| 3 |
+
#
|
| 4 |
+
# This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX
|
| 5 |
+
# and OPT implementations in this library. It has been modified from its
|
| 6 |
+
# original forms to accommodate minor architectural differences compared
|
| 7 |
+
# to GPT-NeoX and OPT used by the Meta AI team that trained the model.
|
| 8 |
+
#
|
| 9 |
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
| 10 |
+
# you may not use this file except in compliance with the License.
|
| 11 |
+
# You may obtain a copy of the License at
|
| 12 |
+
#
|
| 13 |
+
# http://www.apache.org/licenses/LICENSE-2.0
|
| 14 |
+
#
|
| 15 |
+
# Unless required by applicable law or agreed to in writing, software
|
| 16 |
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
| 17 |
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
| 18 |
+
# See the License for the specific language governing permissions and
|
| 19 |
+
# limitations under the License.
|
| 20 |
+
"""PyTorch LLaDA2MoE model."""
|
| 21 |
+
|
| 22 |
+
import math
|
| 23 |
+
import warnings
|
| 24 |
+
from typing import List, Optional, Tuple, Union
|
| 25 |
+
|
| 26 |
+
import torch
|
| 27 |
+
import torch.nn.functional as F
|
| 28 |
+
import torch.utils.checkpoint
|
| 29 |
+
from torch import nn
|
| 30 |
+
from torch.nn import CrossEntropyLoss
|
| 31 |
+
|
| 32 |
+
from transformers.activations import ACT2FN
|
| 33 |
+
from transformers.cache_utils import Cache, DynamicCache
|
| 34 |
+
from transformers.modeling_attn_mask_utils import (
|
| 35 |
+
AttentionMaskConverter,
|
| 36 |
+
_prepare_4d_attention_mask,
|
| 37 |
+
_prepare_4d_causal_attention_mask,
|
| 38 |
+
_prepare_4d_causal_attention_mask_for_sdpa,
|
| 39 |
+
)
|
| 40 |
+
from transformers.modeling_outputs import (
|
| 41 |
+
MoeModelOutputWithPast,
|
| 42 |
+
MoeCausalLMOutputWithPast,
|
| 43 |
+
)
|
| 44 |
+
from transformers.modeling_rope_utils import ROPE_INIT_FUNCTIONS, dynamic_rope_update
|
| 45 |
+
from transformers.modeling_utils import PreTrainedModel
|
| 46 |
+
from transformers.pytorch_utils import ALL_LAYERNORM_LAYERS, is_torch_greater_or_equal_than_1_13
|
| 47 |
+
from transformers.utils import (
|
| 48 |
+
add_start_docstrings,
|
| 49 |
+
add_start_docstrings_to_model_forward,
|
| 50 |
+
is_flash_attn_2_available,
|
| 51 |
+
is_flash_attn_greater_or_equal_2_10,
|
| 52 |
+
logging,
|
| 53 |
+
replace_return_docstrings,
|
| 54 |
+
)
|
| 55 |
+
from transformers.utils.import_utils import is_torch_fx_available
|
| 56 |
+
from .configuration_llada2_moe import LLaDA2MoeConfig
|
| 57 |
+
from transformers.generation.utils import GenerationMixin
|
| 58 |
+
|
| 59 |
+
|
| 60 |
+
if is_flash_attn_2_available():
|
| 61 |
+
from flash_attn import flash_attn_func, flash_attn_varlen_func
|
| 62 |
+
from flash_attn.bert_padding import index_first_axis, pad_input, unpad_input # noqa
|
| 63 |
+
|
| 64 |
+
|
| 65 |
+
# This makes `_prepare_4d_causal_attention_mask` a leaf function in the FX graph.
|
| 66 |
+
# It means that the function will not be traced through and simply appear as a node in the graph.
|
| 67 |
+
if is_torch_fx_available():
|
| 68 |
+
if not is_torch_greater_or_equal_than_1_13:
|
| 69 |
+
import torch.fx
|
| 70 |
+
|
| 71 |
+
_prepare_4d_causal_attention_mask = torch.fx.wrap(_prepare_4d_causal_attention_mask)
|
| 72 |
+
|
| 73 |
+
|
| 74 |
+
logger = logging.get_logger(__name__)
|
| 75 |
+
|
| 76 |
+
_CONFIG_FOR_DOC = "LLaDA2MoeConfig"
|
| 77 |
+
|
| 78 |
+
|
| 79 |
+
def _get_unpad_data(attention_mask):
|
| 80 |
+
seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
|
| 81 |
+
indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
|
| 82 |
+
max_seqlen_in_batch = seqlens_in_batch.max().item()
|
| 83 |
+
cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.torch.int32), (1, 0))
|
| 84 |
+
return (
|
| 85 |
+
indices,
|
| 86 |
+
cu_seqlens,
|
| 87 |
+
max_seqlen_in_batch,
|
| 88 |
+
)
|
| 89 |
+
|
| 90 |
+
|
| 91 |
+
class LLaDA2MoeRMSNorm(nn.Module):
|
| 92 |
+
def __init__(self, hidden_size, eps=1e-6):
|
| 93 |
+
"""
|
| 94 |
+
LLaDA2MoeRMSNorm is equivalent to T5LayerNorm
|
| 95 |
+
"""
|
| 96 |
+
super().__init__()
|
| 97 |
+
self.weight = nn.Parameter(torch.ones(hidden_size))
|
| 98 |
+
self.variance_epsilon = eps
|
| 99 |
+
|
| 100 |
+
def forward(self, hidden_states):
|
| 101 |
+
input_dtype = hidden_states.dtype
|
| 102 |
+
hidden_states = hidden_states.to(torch.float32)
|
| 103 |
+
variance = hidden_states.pow(2).mean(-1, keepdim=True)
|
| 104 |
+
hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)
|
| 105 |
+
return self.weight * hidden_states.to(input_dtype)
|
| 106 |
+
|
| 107 |
+
|
| 108 |
+
ALL_LAYERNORM_LAYERS.append(LLaDA2MoeRMSNorm)
|
| 109 |
+
|
| 110 |
+
|
| 111 |
+
class LLaDA2MoeRotaryEmbedding(nn.Module):
|
| 112 |
+
def __init__(self, config: LLaDA2MoeConfig, device=None):
|
| 113 |
+
super().__init__()
|
| 114 |
+
# BC: "rope_type" was originally "type"
|
| 115 |
+
if hasattr(config, "rope_scaling") and config.rope_scaling is not None:
|
| 116 |
+
self.rope_type = config.rope_scaling.get("rope_type", config.rope_scaling.get("type"))
|
| 117 |
+
else:
|
| 118 |
+
self.rope_type = "default"
|
| 119 |
+
self.max_seq_len_cached = config.max_position_embeddings
|
| 120 |
+
self.original_max_seq_len = config.max_position_embeddings
|
| 121 |
+
|
| 122 |
+
self.config = config
|
| 123 |
+
self.rope_init_fn = ROPE_INIT_FUNCTIONS[self.rope_type]
|
| 124 |
+
|
| 125 |
+
inv_freq, self.attention_scaling = self.rope_init_fn(self.config, device)
|
| 126 |
+
self.register_buffer("inv_freq", inv_freq, persistent=False)
|
| 127 |
+
self.original_inv_freq = self.inv_freq
|
| 128 |
+
|
| 129 |
+
@torch.no_grad()
|
| 130 |
+
@dynamic_rope_update # power user: used with advanced RoPE types (e.g. dynamic rope)
|
| 131 |
+
def forward(self, x, position_ids):
|
| 132 |
+
inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1).to(x.device)
|
| 133 |
+
position_ids_expanded = position_ids[:, None, :].float()
|
| 134 |
+
|
| 135 |
+
device_type = x.device.type if isinstance(x.device.type, str) and x.device.type != "mps" else "cpu"
|
| 136 |
+
with torch.autocast(device_type=device_type, enabled=False): # Force float32
|
| 137 |
+
freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).transpose(1, 2)
|
| 138 |
+
emb = torch.cat((freqs, freqs), dim=-1)
|
| 139 |
+
cos = emb.cos() * self.attention_scaling
|
| 140 |
+
sin = emb.sin() * self.attention_scaling
|
| 141 |
+
|
| 142 |
+
return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)
|
| 143 |
+
|
| 144 |
+
|
| 145 |
+
# Copied from transformers.models.llama.modeling_llama.rotate_half
|
| 146 |
+
def rotate_half(x):
|
| 147 |
+
"""Rotates half the hidden dims of the input."""
|
| 148 |
+
x1 = x[..., : x.shape[-1] // 2]
|
| 149 |
+
x2 = x[..., x.shape[-1] // 2 :]
|
| 150 |
+
return torch.cat((-x2, x1), dim=-1)
|
| 151 |
+
|
| 152 |
+
|
| 153 |
+
# Copied from transformers.models.llama.modeling_llama.apply_rotary_pos_emb
|
| 154 |
+
def apply_rotary_pos_emb(q, k, cos, sin, position_ids, unsqueeze_dim=1):
|
| 155 |
+
"""Applies Rotary Position Embedding to the query and key tensors.
|
| 156 |
+
|
| 157 |
+
Args:
|
| 158 |
+
q (`torch.Tensor`): The query tensor.
|
| 159 |
+
k (`torch.Tensor`): The key tensor.
|
| 160 |
+
cos (`torch.Tensor`): The cosine part of the rotary embedding.
|
| 161 |
+
sin (`torch.Tensor`): The sine part of the rotary embedding.
|
| 162 |
+
position_ids (`torch.Tensor`):
|
| 163 |
+
The position indices of the tokens corresponding to the query and key tensors. For example, this can be
|
| 164 |
+
used to pass offsetted position ids when working with a KV-cache.
|
| 165 |
+
unsqueeze_dim (`int`, *optional*, defaults to 1):
|
| 166 |
+
The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
|
| 167 |
+
sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
|
| 168 |
+
that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
|
| 169 |
+
k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
|
| 170 |
+
cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
|
| 171 |
+
the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
|
| 172 |
+
Returns:
|
| 173 |
+
`tuple(torch.Tensor)` comprising the query and key tensors rotated using the Rotary Position Embedding.
|
| 174 |
+
"""
|
| 175 |
+
cos = cos.unsqueeze(unsqueeze_dim)
|
| 176 |
+
sin = sin.unsqueeze(unsqueeze_dim)
|
| 177 |
+
|
| 178 |
+
# Keep half or full tensor for later concatenation
|
| 179 |
+
rotary_dim = cos.shape[-1]
|
| 180 |
+
q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]
|
| 181 |
+
k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]
|
| 182 |
+
|
| 183 |
+
# Apply rotary embeddings on the first half or full tensor
|
| 184 |
+
q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)
|
| 185 |
+
k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)
|
| 186 |
+
|
| 187 |
+
# Concatenate back to full shape
|
| 188 |
+
q_embed = torch.cat([q_embed, q_pass], dim=-1)
|
| 189 |
+
k_embed = torch.cat([k_embed, k_pass], dim=-1)
|
| 190 |
+
return q_embed, k_embed
|
| 191 |
+
|
| 192 |
+
|
| 193 |
+
class LLaDA2MoeMLP(nn.Module):
|
| 194 |
+
def __init__(self, config: LLaDA2MoeConfig, intermediate_size: int):
|
| 195 |
+
super().__init__()
|
| 196 |
+
self.config = config
|
| 197 |
+
self.hidden_size = config.hidden_size
|
| 198 |
+
self.intermediate_size = intermediate_size
|
| 199 |
+
|
| 200 |
+
self.gate_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
|
| 201 |
+
self.up_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
|
| 202 |
+
self.down_proj = nn.Linear(self.intermediate_size, self.hidden_size, bias=False)
|
| 203 |
+
self.act_fn = ACT2FN[config.hidden_act]
|
| 204 |
+
|
| 205 |
+
def forward(self, x):
|
| 206 |
+
return self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))
|
| 207 |
+
|
| 208 |
+
|
| 209 |
+
class LLaDA2MoeGate(nn.Module):
|
| 210 |
+
def __init__(self, config):
|
| 211 |
+
super().__init__()
|
| 212 |
+
self.config = config
|
| 213 |
+
self.top_k = config.num_experts_per_tok
|
| 214 |
+
self.num_experts = config.num_experts
|
| 215 |
+
|
| 216 |
+
self.n_group = config.n_group
|
| 217 |
+
self.topk_group = config.topk_group
|
| 218 |
+
|
| 219 |
+
# topk selection algorithm
|
| 220 |
+
self.gating_dim = config.hidden_size
|
| 221 |
+
self.weight = nn.Parameter(torch.empty((self.num_experts, self.gating_dim)))
|
| 222 |
+
self.routed_scaling_factor = config.routed_scaling_factor
|
| 223 |
+
|
| 224 |
+
self.register_buffer("expert_bias", torch.zeros((self.num_experts)))
|
| 225 |
+
self.reset_parameters()
|
| 226 |
+
|
| 227 |
+
def reset_parameters(self) -> None:
|
| 228 |
+
import torch.nn.init as init
|
| 229 |
+
|
| 230 |
+
init.kaiming_uniform_(self.weight, a=math.sqrt(5))
|
| 231 |
+
|
| 232 |
+
def group_limited_topk(
|
| 233 |
+
self,
|
| 234 |
+
scores: torch.Tensor,
|
| 235 |
+
):
|
| 236 |
+
num_tokens, _ = scores.size()
|
| 237 |
+
# Organize the experts into groups
|
| 238 |
+
group_scores = scores.view(num_tokens, self.n_group, -1).topk(2, dim=-1)[0].sum(dim=-1)
|
| 239 |
+
group_idx = torch.topk(group_scores, k=self.topk_group, dim=-1, sorted=False)[1]
|
| 240 |
+
group_mask = torch.zeros_like(group_scores)
|
| 241 |
+
group_mask.scatter_(1, group_idx, 1)
|
| 242 |
+
|
| 243 |
+
# Mask the experts based on selection groups
|
| 244 |
+
score_mask = (
|
| 245 |
+
group_mask.unsqueeze(-1)
|
| 246 |
+
.expand(num_tokens, self.n_group, self.num_experts // self.n_group)
|
| 247 |
+
.reshape(num_tokens, -1)
|
| 248 |
+
)
|
| 249 |
+
|
| 250 |
+
masked_scores = scores.masked_fill(~score_mask.bool(), float('-inf'))
|
| 251 |
+
probs, top_indices = torch.topk(masked_scores, k=self.top_k, dim=-1)
|
| 252 |
+
|
| 253 |
+
return probs, top_indices
|
| 254 |
+
|
| 255 |
+
def forward(self, hidden_states):
|
| 256 |
+
# compute gating score
|
| 257 |
+
hidden_states = hidden_states.view(-1, hidden_states.shape[-1])
|
| 258 |
+
logits = F.linear(hidden_states.type(torch.float32), self.weight.type(torch.float32))
|
| 259 |
+
|
| 260 |
+
scores = torch.sigmoid(logits.float()).type_as(logits)
|
| 261 |
+
|
| 262 |
+
scores_for_routing = scores + self.expert_bias
|
| 263 |
+
_, topk_idx = self.group_limited_topk(scores_for_routing)
|
| 264 |
+
|
| 265 |
+
scores = torch.gather(scores, dim=1, index=topk_idx).type_as(logits)
|
| 266 |
+
|
| 267 |
+
topk_weight = scores / (scores.sum(dim=-1, keepdim=True) + 1e-20) if self.top_k > 1 else scores
|
| 268 |
+
topk_weight = topk_weight * self.routed_scaling_factor
|
| 269 |
+
|
| 270 |
+
return topk_idx, topk_weight, logits
|
| 271 |
+
|
| 272 |
+
|
| 273 |
+
class LLaDA2MoeSparseMoeBlock(nn.Module):
|
| 274 |
+
"""
|
| 275 |
+
A mixed expert module containing shared experts.
|
| 276 |
+
"""
|
| 277 |
+
|
| 278 |
+
def __init__(self, config: LLaDA2MoeConfig):
|
| 279 |
+
super().__init__()
|
| 280 |
+
self.config = config
|
| 281 |
+
self.num_experts_per_tok = config.num_experts_per_tok
|
| 282 |
+
self._setup_experts()
|
| 283 |
+
self.gate = LLaDA2MoeGate(config)
|
| 284 |
+
if config.num_shared_experts is not None:
|
| 285 |
+
self.shared_experts = LLaDA2MoeMLP(
|
| 286 |
+
config=config, intermediate_size=config.moe_intermediate_size * config.num_shared_experts
|
| 287 |
+
)
|
| 288 |
+
|
| 289 |
+
def _setup_experts(self):
|
| 290 |
+
self.experts = nn.ModuleList(
|
| 291 |
+
[
|
| 292 |
+
LLaDA2MoeMLP(config=self.config, intermediate_size=self.config.moe_intermediate_size)
|
| 293 |
+
for _ in range(self.config.num_experts)
|
| 294 |
+
]
|
| 295 |
+
)
|
| 296 |
+
|
| 297 |
+
def forward(self, hidden_states):
|
| 298 |
+
identity = hidden_states
|
| 299 |
+
bsz, seq_len, h = hidden_states.shape
|
| 300 |
+
topk_idx, topk_weight, router_logits = self.gate(hidden_states)
|
| 301 |
+
hidden_states = hidden_states.view(-1, hidden_states.shape[-1])
|
| 302 |
+
flat_topk_idx = topk_idx.view(-1)
|
| 303 |
+
if self.training:
|
| 304 |
+
hidden_states = hidden_states.repeat_interleave(self.num_experts_per_tok, dim=0)
|
| 305 |
+
y = torch.empty_like(hidden_states)
|
| 306 |
+
for i, expert in enumerate(self.experts):
|
| 307 |
+
y[flat_topk_idx == i] = expert(hidden_states[flat_topk_idx == i])
|
| 308 |
+
y = (y.view(*topk_weight.shape, -1) * topk_weight.unsqueeze(-1)).sum(dim=1)
|
| 309 |
+
y = y.to(hidden_states.dtype).view(bsz, seq_len, h)
|
| 310 |
+
else:
|
| 311 |
+
y = self.moe_infer(hidden_states, topk_idx, topk_weight).view(bsz, seq_len, h)
|
| 312 |
+
if self.config.num_shared_experts is not None:
|
| 313 |
+
y = y + self.shared_experts(identity)
|
| 314 |
+
return y, (router_logits.view(bsz, seq_len, -1), topk_idx.view(bsz, seq_len, -1))
|
| 315 |
+
|
| 316 |
+
@torch.no_grad()
|
| 317 |
+
def moe_infer(self, x, topk_ids, topk_weight):
|
| 318 |
+
cnts = topk_ids.new_zeros((topk_ids.shape[0], len(self.experts)))
|
| 319 |
+
cnts.scatter_(1, topk_ids, 1)
|
| 320 |
+
tokens_per_expert = cnts.sum(dim=0)
|
| 321 |
+
idxs = topk_ids.view(-1).argsort()
|
| 322 |
+
sorted_tokens = x[idxs // topk_ids.shape[1]]
|
| 323 |
+
sorted_tokens_shape = sorted_tokens.shape
|
| 324 |
+
tokens_per_expert = tokens_per_expert.cpu().numpy()
|
| 325 |
+
outputs = []
|
| 326 |
+
start_idx = 0
|
| 327 |
+
for i, num_tokens in enumerate(tokens_per_expert):
|
| 328 |
+
end_idx = start_idx + num_tokens
|
| 329 |
+
if num_tokens == 0:
|
| 330 |
+
continue
|
| 331 |
+
expert = self.experts[i]
|
| 332 |
+
tokens_for_this_expert = sorted_tokens[start_idx:end_idx]
|
| 333 |
+
expert_out = expert(tokens_for_this_expert)
|
| 334 |
+
outputs.append(expert_out.to(x.device))
|
| 335 |
+
start_idx = end_idx
|
| 336 |
+
|
| 337 |
+
outs = torch.cat(outputs, dim=0) if len(outputs) else sorted_tokens.new_empty(0)
|
| 338 |
+
new_x = torch.empty_like(outs)
|
| 339 |
+
new_x[idxs] = outs
|
| 340 |
+
final_out = (
|
| 341 |
+
new_x.view(*topk_ids.shape, -1)
|
| 342 |
+
.type(topk_weight.dtype)
|
| 343 |
+
.mul_(topk_weight.unsqueeze(dim=-1))
|
| 344 |
+
.sum(dim=1)
|
| 345 |
+
.type(new_x.dtype)
|
| 346 |
+
)
|
| 347 |
+
return final_out
|
| 348 |
+
|
| 349 |
+
|
| 350 |
+
# Copied from transformers.models.llama.modeling_llama.repeat_kv
|
| 351 |
+
def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
|
| 352 |
+
"""
|
| 353 |
+
This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
|
| 354 |
+
num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
|
| 355 |
+
"""
|
| 356 |
+
batch, num_key_value_heads, slen, head_dim = hidden_states.shape
|
| 357 |
+
if n_rep == 1:
|
| 358 |
+
return hidden_states
|
| 359 |
+
hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)
|
| 360 |
+
return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)
|
| 361 |
+
|
| 362 |
+
|
| 363 |
+
# Copied from transformers.models.llama.modeling_llama.LlamaAttention with Llama->LLaDA2Moe
|
| 364 |
+
class LLaDA2MoeAttention(nn.Module):
|
| 365 |
+
"""Multi-headed attention from 'Attention Is All You Need' paper"""
|
| 366 |
+
|
| 367 |
+
def __init__(self, config: LLaDA2MoeConfig, layer_idx: Optional[int] = None):
|
| 368 |
+
super().__init__()
|
| 369 |
+
self.config = config
|
| 370 |
+
self.layer_idx = layer_idx
|
| 371 |
+
if layer_idx is None:
|
| 372 |
+
logger.warning_once(
|
| 373 |
+
f"Instantiating {self.__class__.__name__} without passing `layer_idx` is not recommended and will "
|
| 374 |
+
"to errors during the forward call, if caching is used. Please make sure to provide a `layer_idx` "
|
| 375 |
+
"when creating this class."
|
| 376 |
+
)
|
| 377 |
+
|
| 378 |
+
self.attention_dropout = config.attention_dropout
|
| 379 |
+
self.hidden_size = config.hidden_size
|
| 380 |
+
self.num_heads = config.num_attention_heads
|
| 381 |
+
self.head_dim = config.head_dim or self.hidden_size // self.num_heads
|
| 382 |
+
partial_rotary_factor = config.partial_rotary_factor if hasattr(config, "partial_rotary_factor") else 1.0
|
| 383 |
+
self.rope_dim = int(self.head_dim * partial_rotary_factor)
|
| 384 |
+
self.num_key_value_heads = config.num_key_value_heads
|
| 385 |
+
self.num_key_value_groups = self.num_heads // self.num_key_value_heads
|
| 386 |
+
self.max_position_embeddings = config.max_position_embeddings
|
| 387 |
+
self.rope_theta = config.rope_theta
|
| 388 |
+
self.is_causal = False
|
| 389 |
+
|
| 390 |
+
self.query_key_value = nn.Linear(
|
| 391 |
+
self.hidden_size,
|
| 392 |
+
(self.num_heads + 2 * self.num_key_value_heads) * self.head_dim,
|
| 393 |
+
bias=config.use_qkv_bias,
|
| 394 |
+
)
|
| 395 |
+
|
| 396 |
+
self.query_layernorm = LLaDA2MoeRMSNorm(self.head_dim, eps=config.rms_norm_eps)
|
| 397 |
+
self.key_layernorm = LLaDA2MoeRMSNorm(self.head_dim, eps=config.rms_norm_eps)
|
| 398 |
+
self.dense = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=config.use_bias)
|
| 399 |
+
|
| 400 |
+
def _shape(self, tensor: torch.Tensor, seq_len: int, bsz: int):
|
| 401 |
+
return tensor.view(bsz, seq_len, self.num_heads, self.head_dim).transpose(1, 2).contiguous()
|
| 402 |
+
|
| 403 |
+
def forward(
|
| 404 |
+
self,
|
| 405 |
+
hidden_states: torch.Tensor,
|
| 406 |
+
attention_mask: Optional[torch.Tensor] = None,
|
| 407 |
+
position_ids: Optional[torch.LongTensor] = None,
|
| 408 |
+
past_key_value: Optional[Cache] = None,
|
| 409 |
+
output_attentions: bool = False,
|
| 410 |
+
use_cache: bool = False,
|
| 411 |
+
position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None, # necessary, but kept here for BC
|
| 412 |
+
**kwargs,
|
| 413 |
+
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
|
| 414 |
+
if "padding_mask" in kwargs:
|
| 415 |
+
warnings.warn(
|
| 416 |
+
"Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use `attention_mask` instead.`"
|
| 417 |
+
)
|
| 418 |
+
|
| 419 |
+
bsz, q_len, _ = hidden_states.size()
|
| 420 |
+
|
| 421 |
+
qkv = self.query_key_value(hidden_states)
|
| 422 |
+
qkv = qkv.view(bsz, q_len, self.num_heads + 2 * self.num_key_value_heads, self.head_dim)
|
| 423 |
+
|
| 424 |
+
query_states, key_states, value_states = qkv.split(
|
| 425 |
+
[self.num_heads, self.num_key_value_heads, self.num_key_value_heads], dim=-2
|
| 426 |
+
)
|
| 427 |
+
query_states = query_states.transpose(1, 2)
|
| 428 |
+
key_states = key_states.transpose(1, 2)
|
| 429 |
+
value_states = value_states.transpose(1, 2)
|
| 430 |
+
|
| 431 |
+
query_states = self.query_layernorm(query_states)
|
| 432 |
+
key_states = self.key_layernorm(key_states)
|
| 433 |
+
|
| 434 |
+
kv_seq_len = key_states.shape[-2]
|
| 435 |
+
if past_key_value is not None:
|
| 436 |
+
if self.layer_idx is None:
|
| 437 |
+
raise ValueError(
|
| 438 |
+
f"The cache structure has changed since version v4.36. If you are using {self.__class__.__name__} "
|
| 439 |
+
"for auto-regressive decoding with k/v caching, please make sure to initialize the attention class "
|
| 440 |
+
"with a layer index."
|
| 441 |
+
)
|
| 442 |
+
kv_seq_len += past_key_value.get_usable_length(kv_seq_len, self.layer_idx)
|
| 443 |
+
cos, sin = position_embeddings
|
| 444 |
+
query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids)
|
| 445 |
+
|
| 446 |
+
if past_key_value is not None:
|
| 447 |
+
cache_kwargs = {"sin": sin, "cos": cos} # Specific to RoPE models
|
| 448 |
+
key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
|
| 449 |
+
|
| 450 |
+
key_states = repeat_kv(key_states, self.num_key_value_groups)
|
| 451 |
+
value_states = repeat_kv(value_states, self.num_key_value_groups)
|
| 452 |
+
|
| 453 |
+
attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) / math.sqrt(self.head_dim)
|
| 454 |
+
|
| 455 |
+
if attn_weights.size() != (bsz, self.num_heads, q_len, kv_seq_len):
|
| 456 |
+
raise ValueError(
|
| 457 |
+
f"Attention weights should be of size {(bsz, self.num_heads, q_len, kv_seq_len)}, but is"
|
| 458 |
+
f" {attn_weights.size()}"
|
| 459 |
+
)
|
| 460 |
+
# attention_mask = None
|
| 461 |
+
if attention_mask is not None:
|
| 462 |
+
if attention_mask.size() != (bsz, 1, q_len, kv_seq_len):
|
| 463 |
+
raise ValueError(
|
| 464 |
+
f"Attention mask should be of size {(bsz, 1, q_len, kv_seq_len)}, but is {attention_mask.size()}"
|
| 465 |
+
)
|
| 466 |
+
attn_weights = attn_weights + attention_mask
|
| 467 |
+
|
| 468 |
+
# upcast attention to fp32
|
| 469 |
+
attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query_states.dtype)
|
| 470 |
+
attn_weights = nn.functional.dropout(attn_weights, p=self.attention_dropout, training=self.training)
|
| 471 |
+
attn_output = torch.matmul(attn_weights, value_states)
|
| 472 |
+
|
| 473 |
+
if attn_output.size() != (bsz, self.num_heads, q_len, self.head_dim):
|
| 474 |
+
raise ValueError(
|
| 475 |
+
f"`attn_output` should be of size {(bsz, self.num_heads, q_len, self.head_dim)}, but is"
|
| 476 |
+
f" {attn_output.size()}"
|
| 477 |
+
)
|
| 478 |
+
|
| 479 |
+
attn_output = attn_output.transpose(1, 2).contiguous()
|
| 480 |
+
|
| 481 |
+
attn_output = attn_output.reshape(bsz, q_len, -1)
|
| 482 |
+
|
| 483 |
+
attn_output = self.dense(attn_output)
|
| 484 |
+
|
| 485 |
+
if not output_attentions:
|
| 486 |
+
attn_weights = None
|
| 487 |
+
|
| 488 |
+
return attn_output, attn_weights, past_key_value
|
| 489 |
+
|
| 490 |
+
|
| 491 |
+
# Copied from transformers.models.llama.modeling_llama.LlamaFlashAttention2 with Llama->LLaDA2Moe
|
| 492 |
+
class LLaDA2MoeFlashAttention2(LLaDA2MoeAttention):
|
| 493 |
+
"""
|
| 494 |
+
LLaDA2Moe flash attention module. This module inherits from `LLaDA2MoeAttention` as the weights of the module stays
|
| 495 |
+
untouched. The only required change would be on the forward pass where it needs to correctly call the public API of
|
| 496 |
+
flash attention and deal with padding tokens in case the input contains any of them.
|
| 497 |
+
"""
|
| 498 |
+
|
| 499 |
+
def __init__(self, *args, **kwargs):
|
| 500 |
+
super().__init__(*args, **kwargs)
|
| 501 |
+
|
| 502 |
+
# TODO: Should be removed once Flash Attention for RoCm is bumped to 2.1.
|
| 503 |
+
# flash_attn<2.1 generates top-left aligned causal mask, while what is needed here is bottom-right alignement, that was made default for flash_attn>=2.1. This attribute is used to handle this difference. Reference: https://github.com/Dao-AILab/flash-attention/releases/tag/v2.1.0.
|
| 504 |
+
# Beware that with flash_attn<2.1, using q_seqlen != k_seqlen (except for the case q_seqlen == 1) produces a wrong mask (top-left).
|
| 505 |
+
self._flash_attn_uses_top_left_mask = not is_flash_attn_greater_or_equal_2_10()
|
| 506 |
+
|
| 507 |
+
def forward(
|
| 508 |
+
self,
|
| 509 |
+
hidden_states: torch.Tensor,
|
| 510 |
+
attention_mask: Optional[torch.LongTensor] = None,
|
| 511 |
+
position_ids: Optional[torch.LongTensor] = None,
|
| 512 |
+
past_key_value: Optional[Cache] = None,
|
| 513 |
+
output_attentions: bool = False,
|
| 514 |
+
use_cache: bool = False,
|
| 515 |
+
position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None, # necessary, but kept here for BC
|
| 516 |
+
**kwargs,
|
| 517 |
+
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
|
| 518 |
+
# LLaDA2MoeFlashAttention2 attention does not support output_attentions
|
| 519 |
+
if "padding_mask" in kwargs:
|
| 520 |
+
warnings.warn(
|
| 521 |
+
"Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use `attention_mask` instead.`"
|
| 522 |
+
)
|
| 523 |
+
|
| 524 |
+
# overwrite attention_mask with padding_mask
|
| 525 |
+
attention_mask = kwargs.pop("padding_mask")
|
| 526 |
+
|
| 527 |
+
output_attentions = False
|
| 528 |
+
|
| 529 |
+
bsz, q_len, _ = hidden_states.size()
|
| 530 |
+
|
| 531 |
+
# Flash attention requires the input to have the shape
|
| 532 |
+
# batch_size x seq_length x head_dim x hidden_dim
|
| 533 |
+
# therefore we just need to keep the original shape
|
| 534 |
+
|
| 535 |
+
qkv = self.query_key_value(hidden_states)
|
| 536 |
+
qkv = qkv.view(bsz, q_len, self.num_heads + 2 * self.num_key_value_heads, self.head_dim)
|
| 537 |
+
|
| 538 |
+
query_states, key_states, value_states = qkv.split(
|
| 539 |
+
[self.num_heads, self.num_key_value_heads, self.num_key_value_heads], dim=-2
|
| 540 |
+
)
|
| 541 |
+
query_states = query_states.transpose(1, 2)
|
| 542 |
+
key_states = key_states.transpose(1, 2)
|
| 543 |
+
value_states = value_states.transpose(1, 2)
|
| 544 |
+
|
| 545 |
+
query_states = self.query_layernorm(query_states)
|
| 546 |
+
key_states = self.key_layernorm(key_states)
|
| 547 |
+
|
| 548 |
+
kv_seq_len = key_states.shape[-2]
|
| 549 |
+
if past_key_value is not None:
|
| 550 |
+
kv_seq_len += past_key_value.get_usable_length(kv_seq_len, self.layer_idx)
|
| 551 |
+
cos, sin = position_embeddings
|
| 552 |
+
query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids)
|
| 553 |
+
|
| 554 |
+
if past_key_value is not None:
|
| 555 |
+
cache_kwargs = {"sin": sin, "cos": cos} # Specific to RoPE models
|
| 556 |
+
key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
|
| 557 |
+
|
| 558 |
+
# TODO: These transpose are quite inefficient but Flash Attention requires the layout [batch_size, sequence_length, num_heads, head_dim]. We would need to refactor the KV cache
|
| 559 |
+
# to be able to avoid many of these transpose/reshape/view.
|
| 560 |
+
query_states = query_states.transpose(1, 2)
|
| 561 |
+
key_states = key_states.transpose(1, 2)
|
| 562 |
+
value_states = value_states.transpose(1, 2)
|
| 563 |
+
|
| 564 |
+
dropout_rate = self.attention_dropout if self.training else 0.0
|
| 565 |
+
|
| 566 |
+
# In PEFT, usually we cast the layer norms in float32 for training stability reasons
|
| 567 |
+
# therefore the input hidden states gets silently cast in float32. Hence, we need
|
| 568 |
+
# cast them back in the correct dtype just to be sure everything works as expected.
|
| 569 |
+
# This might slow down training & inference so it is recommended to not cast the LayerNorms
|
| 570 |
+
# in fp32. (LLaDA2MoeRMSNorm handles it correctly)
|
| 571 |
+
|
| 572 |
+
input_dtype = query_states.dtype
|
| 573 |
+
if input_dtype == torch.float32:
|
| 574 |
+
# Handle the case where the model is quantized
|
| 575 |
+
if hasattr(self.config, "_pre_quantization_dtype"):
|
| 576 |
+
target_dtype = self.config._pre_quantization_dtype
|
| 577 |
+
elif torch.is_autocast_enabled():
|
| 578 |
+
target_dtype = torch.get_autocast_gpu_dtype()
|
| 579 |
+
else:
|
| 580 |
+
target_dtype = self.query_key_value.weight.dtype
|
| 581 |
+
|
| 582 |
+
logger.warning_once(
|
| 583 |
+
f"The input hidden states seems to be silently casted in float32, this might be related to"
|
| 584 |
+
f" the fact you have upcasted embedding or layer norm layers in float32. We will cast back the input in"
|
| 585 |
+
f" {target_dtype}."
|
| 586 |
+
)
|
| 587 |
+
|
| 588 |
+
query_states = query_states.to(target_dtype)
|
| 589 |
+
key_states = key_states.to(target_dtype)
|
| 590 |
+
value_states = value_states.to(target_dtype)
|
| 591 |
+
|
| 592 |
+
attn_output = self._flash_attention_forward(
|
| 593 |
+
query_states, key_states, value_states, attention_mask, q_len, dropout=dropout_rate
|
| 594 |
+
)
|
| 595 |
+
|
| 596 |
+
attn_output = attn_output.reshape(bsz, q_len, -1).contiguous()
|
| 597 |
+
attn_output = self.dense(attn_output)
|
| 598 |
+
|
| 599 |
+
if not output_attentions:
|
| 600 |
+
attn_weights = None
|
| 601 |
+
|
| 602 |
+
return attn_output, attn_weights, past_key_value
|
| 603 |
+
|
| 604 |
+
def _flash_attention_forward(
|
| 605 |
+
self, query_states, key_states, value_states, attention_mask, query_length, dropout=0.0, softmax_scale=None
|
| 606 |
+
):
|
| 607 |
+
"""
|
| 608 |
+
Calls the forward method of Flash Attention - if the input hidden states contain at least one padding token
|
| 609 |
+
first unpad the input, then computes the attention scores and pad the final attention scores.
|
| 610 |
+
|
| 611 |
+
Args:
|
| 612 |
+
query_states (`torch.Tensor`):
|
| 613 |
+
Input query states to be passed to Flash Attention API
|
| 614 |
+
key_states (`torch.Tensor`):
|
| 615 |
+
Input key states to be passed to Flash Attention API
|
| 616 |
+
value_states (`torch.Tensor`):
|
| 617 |
+
Input value states to be passed to Flash Attention API
|
| 618 |
+
attention_mask (`torch.Tensor`):
|
| 619 |
+
The padding mask - corresponds to a tensor of size `(batch_size, seq_len)` where 0 stands for the
|
| 620 |
+
position of padding tokens and 1 for the position of non-padding tokens.
|
| 621 |
+
dropout (`int`, *optional*):
|
| 622 |
+
Attention dropout
|
| 623 |
+
softmax_scale (`float`, *optional*):
|
| 624 |
+
The scaling of QK^T before applying softmax. Default to 1 / sqrt(head_dim)
|
| 625 |
+
query_length (`int`):
|
| 626 |
+
The length of the query sequence in terms of tokens. This represents the number of tokens in the
|
| 627 |
+
`query_states` tensor along the sequence dimension. It is used to determine the effective sequence
|
| 628 |
+
length for attention computations.
|
| 629 |
+
"""
|
| 630 |
+
if not self._flash_attn_uses_top_left_mask:
|
| 631 |
+
causal = self.is_causal
|
| 632 |
+
else:
|
| 633 |
+
# TODO: Remove the `query_length != 1` check once Flash Attention for RoCm is bumped to 2.1. For details, please see the comment in LLaDA2MoeFlashAttention2 __init__.
|
| 634 |
+
causal = self.is_causal and query_length != 1
|
| 635 |
+
|
| 636 |
+
# attention_mask = None
|
| 637 |
+
# Contains at least one padding token in the sequence
|
| 638 |
+
if attention_mask is not None:
|
| 639 |
+
batch_size = query_states.shape[0]
|
| 640 |
+
query_states, key_states, value_states, indices_q, cu_seq_lens, max_seq_lens = self._upad_input(
|
| 641 |
+
query_states, key_states, value_states, attention_mask, query_length
|
| 642 |
+
)
|
| 643 |
+
|
| 644 |
+
cu_seqlens_q, cu_seqlens_k = cu_seq_lens
|
| 645 |
+
max_seqlen_in_batch_q, max_seqlen_in_batch_k = max_seq_lens
|
| 646 |
+
|
| 647 |
+
attn_output_unpad = flash_attn_varlen_func(
|
| 648 |
+
query_states,
|
| 649 |
+
key_states,
|
| 650 |
+
value_states,
|
| 651 |
+
cu_seqlens_q=cu_seqlens_q,
|
| 652 |
+
cu_seqlens_k=cu_seqlens_k,
|
| 653 |
+
max_seqlen_q=max_seqlen_in_batch_q,
|
| 654 |
+
max_seqlen_k=max_seqlen_in_batch_k,
|
| 655 |
+
dropout_p=dropout,
|
| 656 |
+
softmax_scale=softmax_scale,
|
| 657 |
+
causal=causal,
|
| 658 |
+
)
|
| 659 |
+
|
| 660 |
+
attn_output = pad_input(attn_output_unpad, indices_q, batch_size, query_length)
|
| 661 |
+
else:
|
| 662 |
+
attn_output = flash_attn_func(
|
| 663 |
+
query_states, key_states, value_states, dropout, softmax_scale=softmax_scale, causal=causal
|
| 664 |
+
)
|
| 665 |
+
|
| 666 |
+
return attn_output
|
| 667 |
+
|
| 668 |
+
def _upad_input(self, query_layer, key_layer, value_layer, attention_mask, query_length):
|
| 669 |
+
indices_k, cu_seqlens_k, max_seqlen_in_batch_k = _get_unpad_data(attention_mask)
|
| 670 |
+
batch_size, kv_seq_len, num_key_value_heads, head_dim = key_layer.shape
|
| 671 |
+
|
| 672 |
+
key_layer = index_first_axis(
|
| 673 |
+
key_layer.reshape(batch_size * kv_seq_len, num_key_value_heads, head_dim), indices_k
|
| 674 |
+
)
|
| 675 |
+
value_layer = index_first_axis(
|
| 676 |
+
value_layer.reshape(batch_size * kv_seq_len, num_key_value_heads, head_dim), indices_k
|
| 677 |
+
)
|
| 678 |
+
if query_length == kv_seq_len:
|
| 679 |
+
query_layer = index_first_axis(
|
| 680 |
+
query_layer.reshape(batch_size * kv_seq_len, self.num_heads, head_dim), indices_k
|
| 681 |
+
)
|
| 682 |
+
cu_seqlens_q = cu_seqlens_k
|
| 683 |
+
max_seqlen_in_batch_q = max_seqlen_in_batch_k
|
| 684 |
+
indices_q = indices_k
|
| 685 |
+
elif query_length == 1:
|
| 686 |
+
max_seqlen_in_batch_q = 1
|
| 687 |
+
cu_seqlens_q = torch.arange(
|
| 688 |
+
batch_size + 1, dtype=torch.int32, device=query_layer.device
|
| 689 |
+
) # There is a memcpy here, that is very bad.
|
| 690 |
+
indices_q = cu_seqlens_q[:-1]
|
| 691 |
+
query_layer = query_layer.squeeze(1)
|
| 692 |
+
else:
|
| 693 |
+
# The -q_len: slice assumes left padding.
|
| 694 |
+
attention_mask = attention_mask[:, -query_length:]
|
| 695 |
+
query_layer, indices_q, cu_seqlens_q, max_seqlen_in_batch_q = unpad_input(query_layer, attention_mask)
|
| 696 |
+
|
| 697 |
+
return (
|
| 698 |
+
query_layer,
|
| 699 |
+
key_layer,
|
| 700 |
+
value_layer,
|
| 701 |
+
indices_q,
|
| 702 |
+
(cu_seqlens_q, cu_seqlens_k),
|
| 703 |
+
(max_seqlen_in_batch_q, max_seqlen_in_batch_k),
|
| 704 |
+
)
|
| 705 |
+
|
| 706 |
+
|
| 707 |
+
# Copied from transformers.models.llama.modeling_llama.LlamaSdpaAttention with Llama->LLaDA2Moe
|
| 708 |
+
class LLaDA2MoeSdpaAttention(LLaDA2MoeAttention):
|
| 709 |
+
"""
|
| 710 |
+
LLaDA2Moe attention module using torch.nn.functional.scaled_dot_product_attention. This module inherits from
|
| 711 |
+
`LLaDA2MoeAttention` as the weights of the module stays untouched. The only changes are on the forward pass to adapt to
|
| 712 |
+
SDPA API.
|
| 713 |
+
"""
|
| 714 |
+
|
| 715 |
+
# Adapted from LLaDA2MoeAttention.forward
|
| 716 |
+
def forward(
|
| 717 |
+
self,
|
| 718 |
+
hidden_states: torch.Tensor,
|
| 719 |
+
attention_mask: Optional[torch.Tensor] = None,
|
| 720 |
+
position_ids: Optional[torch.LongTensor] = None,
|
| 721 |
+
past_key_value: Optional[Cache] = None,
|
| 722 |
+
output_attentions: bool = False,
|
| 723 |
+
use_cache: bool = False,
|
| 724 |
+
position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None, # necessary, but kept here for BC
|
| 725 |
+
**kwargs,
|
| 726 |
+
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
|
| 727 |
+
if output_attentions:
|
| 728 |
+
# TODO: Improve this warning with e.g. `model.config.attn_implementation = "manual"` once this is implemented.
|
| 729 |
+
logger.warning_once(
|
| 730 |
+
"LLaDA2MoeModel is using LLaDA2MoeSdpaAttention, but `torch.nn.functional.scaled_dot_product_attention` does not support `output_attentions=True`. Falling back to the manual attention implementation, "
|
| 731 |
+
'but specifying the manual implementation will be required from Transformers version v5.0.0 onwards. This warning can be removed using the argument `attn_implementation="eager"` when loading the model.'
|
| 732 |
+
)
|
| 733 |
+
return super().forward(
|
| 734 |
+
hidden_states=hidden_states,
|
| 735 |
+
attention_mask=attention_mask,
|
| 736 |
+
position_ids=position_ids,
|
| 737 |
+
past_key_value=past_key_value,
|
| 738 |
+
output_attentions=output_attentions,
|
| 739 |
+
use_cache=use_cache,
|
| 740 |
+
)
|
| 741 |
+
|
| 742 |
+
bsz, q_len, _ = hidden_states.size()
|
| 743 |
+
|
| 744 |
+
qkv = self.query_key_value(hidden_states)
|
| 745 |
+
qkv = qkv.view(bsz, q_len, self.num_heads + 2 * self.num_key_value_heads, self.head_dim)
|
| 746 |
+
|
| 747 |
+
query_states, key_states, value_states = qkv.split(
|
| 748 |
+
[self.num_heads, self.num_key_value_heads, self.num_key_value_heads], dim=-2
|
| 749 |
+
)
|
| 750 |
+
query_states = query_states.transpose(1, 2)
|
| 751 |
+
key_states = key_states.transpose(1, 2)
|
| 752 |
+
value_states = value_states.transpose(1, 2)
|
| 753 |
+
|
| 754 |
+
query_states = self.query_layernorm(query_states)
|
| 755 |
+
key_states = self.key_layernorm(key_states)
|
| 756 |
+
|
| 757 |
+
kv_seq_len = key_states.shape[-2]
|
| 758 |
+
if past_key_value is not None:
|
| 759 |
+
kv_seq_len += past_key_value.get_usable_length(kv_seq_len, self.layer_idx)
|
| 760 |
+
cos, sin = position_embeddings
|
| 761 |
+
|
| 762 |
+
query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids)
|
| 763 |
+
|
| 764 |
+
if past_key_value is not None:
|
| 765 |
+
cache_kwargs = {"sin": sin, "cos": cos} # Specific to RoPE models
|
| 766 |
+
key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
|
| 767 |
+
|
| 768 |
+
key_states = repeat_kv(key_states, self.num_key_value_groups)
|
| 769 |
+
value_states = repeat_kv(value_states, self.num_key_value_groups)
|
| 770 |
+
|
| 771 |
+
# attention_mask = None
|
| 772 |
+
if attention_mask is not None:
|
| 773 |
+
if attention_mask.size() != (bsz, 1, q_len, kv_seq_len):
|
| 774 |
+
raise ValueError(
|
| 775 |
+
f"Attention mask should be of size {(bsz, 1, q_len, kv_seq_len)}, but is {attention_mask.size()}"
|
| 776 |
+
)
|
| 777 |
+
|
| 778 |
+
# SDPA with memory-efficient backend is currently (torch==2.1.2) bugged with non-contiguous inputs with custom attn_mask,
|
| 779 |
+
# Reference: https://github.com/pytorch/pytorch/issues/112577.
|
| 780 |
+
if query_states.device.type == "cuda" and attention_mask is not None:
|
| 781 |
+
query_states = query_states.contiguous()
|
| 782 |
+
key_states = key_states.contiguous()
|
| 783 |
+
value_states = value_states.contiguous()
|
| 784 |
+
|
| 785 |
+
attn_output = torch.nn.functional.scaled_dot_product_attention(
|
| 786 |
+
query_states,
|
| 787 |
+
key_states,
|
| 788 |
+
value_states,
|
| 789 |
+
attn_mask=attention_mask,
|
| 790 |
+
dropout_p=self.attention_dropout if self.training else 0.0,
|
| 791 |
+
# The q_len > 1 is necessary to match with AttentionMaskConverter.to_causal_4d that does not create a causal mask in case q_len == 1.
|
| 792 |
+
is_causal=self.is_causal and attention_mask is None and q_len > 1,
|
| 793 |
+
)
|
| 794 |
+
|
| 795 |
+
attn_output = attn_output.transpose(1, 2).contiguous()
|
| 796 |
+
attn_output = attn_output.reshape(bsz, q_len, -1)
|
| 797 |
+
|
| 798 |
+
attn_output = self.dense(attn_output)
|
| 799 |
+
|
| 800 |
+
return attn_output, None, past_key_value
|
| 801 |
+
|
| 802 |
+
|
| 803 |
+
ATTENTION_CLASSES = {
|
| 804 |
+
"eager": LLaDA2MoeAttention,
|
| 805 |
+
"flash_attention_2": LLaDA2MoeFlashAttention2,
|
| 806 |
+
"sdpa": LLaDA2MoeSdpaAttention,
|
| 807 |
+
}
|
| 808 |
+
|
| 809 |
+
|
| 810 |
+
class LLaDA2MoeDecoderLayer(nn.Module):
|
| 811 |
+
def __init__(self, config: LLaDA2MoeConfig, layer_idx: int):
|
| 812 |
+
super().__init__()
|
| 813 |
+
self.hidden_size = config.hidden_size
|
| 814 |
+
|
| 815 |
+
self.attention = ATTENTION_CLASSES[config._attn_implementation](config=config, layer_idx=layer_idx)
|
| 816 |
+
|
| 817 |
+
self.mlp = (
|
| 818 |
+
LLaDA2MoeSparseMoeBlock(config)
|
| 819 |
+
if (config.num_experts is not None and layer_idx >= config.first_k_dense_replace)
|
| 820 |
+
else LLaDA2MoeMLP(config=config, intermediate_size=config.intermediate_size)
|
| 821 |
+
)
|
| 822 |
+
self.input_layernorm = LLaDA2MoeRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
|
| 823 |
+
self.post_attention_layernorm = LLaDA2MoeRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
|
| 824 |
+
|
| 825 |
+
def forward(
|
| 826 |
+
self,
|
| 827 |
+
hidden_states: torch.Tensor,
|
| 828 |
+
attention_mask: Optional[torch.Tensor] = None,
|
| 829 |
+
position_ids: Optional[torch.LongTensor] = None,
|
| 830 |
+
past_key_value: Optional[Tuple[torch.Tensor]] = None,
|
| 831 |
+
output_attentions: Optional[bool] = False,
|
| 832 |
+
output_router_logits: Optional[bool] = False,
|
| 833 |
+
use_cache: Optional[bool] = False,
|
| 834 |
+
position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None, # necessary, but kept here for BC
|
| 835 |
+
**kwargs,
|
| 836 |
+
) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
|
| 837 |
+
"""
|
| 838 |
+
Args:
|
| 839 |
+
hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
|
| 840 |
+
attention_mask (`torch.FloatTensor`, *optional*):
|
| 841 |
+
attention mask of size `(batch_size, sequence_length)` if flash attention is used or `(batch_size, 1,
|
| 842 |
+
query_sequence_length, key_sequence_length)` if default attention is used.
|
| 843 |
+
position_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
|
| 844 |
+
Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0,
|
| 845 |
+
config.n_positions - 1]`.
|
| 846 |
+
past_key_value (`Tuple(torch.FloatTensor)`, *optional*):
|
| 847 |
+
cached past key and value projection states
|
| 848 |
+
output_attentions (`bool`, *optional*):
|
| 849 |
+
Whether to return the attentions tensors of all attention layers. See `attentions` under
|
| 850 |
+
returned tensors for more detail.
|
| 851 |
+
output_router_logits (`bool`, *optional*):
|
| 852 |
+
Whether or not to return the logits of all the routers. They are useful for computing the router loss,
|
| 853 |
+
and should not be returned during inference.
|
| 854 |
+
use_cache (`bool`, *optional*):
|
| 855 |
+
If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding
|
| 856 |
+
(see `past_key_values`).
|
| 857 |
+
"""
|
| 858 |
+
if "padding_mask" in kwargs:
|
| 859 |
+
warnings.warn(
|
| 860 |
+
"Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use `attention_mask` instead.`"
|
| 861 |
+
)
|
| 862 |
+
residual = hidden_states
|
| 863 |
+
|
| 864 |
+
hidden_states = self.input_layernorm(hidden_states)
|
| 865 |
+
|
| 866 |
+
# Self Attention
|
| 867 |
+
hidden_states, self_attn_weights, present_key_value = self.attention(
|
| 868 |
+
hidden_states=hidden_states,
|
| 869 |
+
attention_mask=attention_mask,
|
| 870 |
+
position_ids=position_ids,
|
| 871 |
+
past_key_value=past_key_value,
|
| 872 |
+
output_attentions=output_attentions,
|
| 873 |
+
position_embeddings=position_embeddings,
|
| 874 |
+
use_cache=use_cache,
|
| 875 |
+
)
|
| 876 |
+
hidden_states = residual + hidden_states
|
| 877 |
+
|
| 878 |
+
# Fully Connected
|
| 879 |
+
residual = hidden_states
|
| 880 |
+
hidden_states = self.post_attention_layernorm(hidden_states)
|
| 881 |
+
hidden_states = self.mlp(hidden_states)
|
| 882 |
+
if isinstance(hidden_states, tuple):
|
| 883 |
+
hidden_states, router_logits = hidden_states
|
| 884 |
+
else:
|
| 885 |
+
router_logits = None
|
| 886 |
+
hidden_states = residual + hidden_states.to(residual.device)
|
| 887 |
+
|
| 888 |
+
outputs = (hidden_states,)
|
| 889 |
+
|
| 890 |
+
if output_attentions:
|
| 891 |
+
outputs += (self_attn_weights,)
|
| 892 |
+
|
| 893 |
+
if use_cache:
|
| 894 |
+
outputs += (present_key_value,)
|
| 895 |
+
|
| 896 |
+
if output_router_logits:
|
| 897 |
+
outputs += (router_logits,)
|
| 898 |
+
|
| 899 |
+
return outputs
|
| 900 |
+
|
| 901 |
+
|
| 902 |
+
LLADA2MOE_START_DOCSTRING = r"""
|
| 903 |
+
This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the
|
| 904 |
+
library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
|
| 905 |
+
etc.)
|
| 906 |
+
|
| 907 |
+
This model is also a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass.
|
| 908 |
+
Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage
|
| 909 |
+
and behavior.
|
| 910 |
+
|
| 911 |
+
Parameters:
|
| 912 |
+
config ([`LLaDA2MoeConfig`]):
|
| 913 |
+
Model configuration class with all the parameters of the model. Initializing with a config file does not
|
| 914 |
+
load the weights associated with the model, only the configuration. Check out the
|
| 915 |
+
[`~PreTrainedModel.from_pretrained`] method to load the model weights.
|
| 916 |
+
"""
|
| 917 |
+
|
| 918 |
+
|
| 919 |
+
@add_start_docstrings(
|
| 920 |
+
"The bare LLaDA2Moe Model outputting raw hidden-states without any specific head on top.",
|
| 921 |
+
LLADA2MOE_START_DOCSTRING,
|
| 922 |
+
)
|
| 923 |
+
class LLaDA2MoePreTrainedModel(PreTrainedModel):
|
| 924 |
+
config_class = LLaDA2MoeConfig
|
| 925 |
+
base_model_prefix = "model"
|
| 926 |
+
supports_gradient_checkpointing = True
|
| 927 |
+
_no_split_modules = ["LLaDA2MoeDecoderLayer"]
|
| 928 |
+
_skip_keys_device_placement = "past_key_values"
|
| 929 |
+
_supports_flash_attn_2 = True
|
| 930 |
+
_supports_sdpa = True
|
| 931 |
+
_supports_cache_class = True
|
| 932 |
+
|
| 933 |
+
def _init_weights(self, module):
|
| 934 |
+
std = self.config.initializer_range
|
| 935 |
+
if isinstance(module, nn.Linear):
|
| 936 |
+
module.weight.data.normal_(mean=0.0, std=std)
|
| 937 |
+
if module.bias is not None:
|
| 938 |
+
module.bias.data.zero_()
|
| 939 |
+
elif isinstance(module, nn.Embedding):
|
| 940 |
+
module.weight.data.normal_(mean=0.0, std=std)
|
| 941 |
+
if module.padding_idx is not None:
|
| 942 |
+
module.weight.data[module.padding_idx].zero_()
|
| 943 |
+
|
| 944 |
+
|
| 945 |
+
LLADA2MOE_INPUTS_DOCSTRING = r"""
|
| 946 |
+
Args:
|
| 947 |
+
input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
|
| 948 |
+
Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
|
| 949 |
+
it.
|
| 950 |
+
|
| 951 |
+
Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
|
| 952 |
+
[`PreTrainedTokenizer.__call__`] for details.
|
| 953 |
+
|
| 954 |
+
[What are input IDs?](../glossary#input-ids)
|
| 955 |
+
attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
|
| 956 |
+
Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
|
| 957 |
+
|
| 958 |
+
- 1 for tokens that are **not masked**,
|
| 959 |
+
- 0 for tokens that are **masked**.
|
| 960 |
+
|
| 961 |
+
[What are attention masks?](../glossary#attention-mask)
|
| 962 |
+
|
| 963 |
+
Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
|
| 964 |
+
[`PreTrainedTokenizer.__call__`] for details.
|
| 965 |
+
|
| 966 |
+
If `past_key_values` is used, optionally only the last `input_ids` have to be input (see
|
| 967 |
+
`past_key_values`).
|
| 968 |
+
|
| 969 |
+
If you want to change padding behavior, you should read [`modeling_opt._prepare_decoder_attention_mask`]
|
| 970 |
+
and modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more
|
| 971 |
+
information on the default strategy.
|
| 972 |
+
|
| 973 |
+
- 1 indicates the head is **not masked**,
|
| 974 |
+
- 0 indicates the head is **masked**.
|
| 975 |
+
position_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
|
| 976 |
+
Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0,
|
| 977 |
+
config.n_positions - 1]`.
|
| 978 |
+
|
| 979 |
+
[What are position IDs?](../glossary#position-ids)
|
| 980 |
+
past_key_values (`Cache` or `tuple(tuple(torch.FloatTensor))`, *optional*):
|
| 981 |
+
Pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention
|
| 982 |
+
blocks) that can be used to speed up sequential decoding. This typically consists in the `past_key_values`
|
| 983 |
+
returned by the model at a previous stage of decoding, when `use_cache=True` or `config.use_cache=True`.
|
| 984 |
+
|
| 985 |
+
Two formats are allowed:
|
| 986 |
+
- a [`~cache_utils.Cache`] instance;
|
| 987 |
+
- Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of
|
| 988 |
+
shape `(batch_size, num_heads, sequence_length, embed_size_per_head)`). This is also known as the legacy
|
| 989 |
+
cache format.
|
| 990 |
+
|
| 991 |
+
The model will output the same cache format that is fed as input. If no `past_key_values` are passed, the
|
| 992 |
+
legacy cache format will be returned.
|
| 993 |
+
|
| 994 |
+
If `past_key_values` are used, the user can optionally input only the last `input_ids` (those that don't
|
| 995 |
+
have their past key value states given to this model) of shape `(batch_size, 1)` instead of all `input_ids`
|
| 996 |
+
of shape `(batch_size, sequence_length)`.
|
| 997 |
+
inputs_embeds (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
|
| 998 |
+
Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation. This
|
| 999 |
+
is useful if you want more control over how to convert `input_ids` indices into associated vectors than the
|
| 1000 |
+
model's internal embedding lookup matrix.
|
| 1001 |
+
use_cache (`bool`, *optional*):
|
| 1002 |
+
If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see
|
| 1003 |
+
`past_key_values`).
|
| 1004 |
+
output_attentions (`bool`, *optional*):
|
| 1005 |
+
Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
|
| 1006 |
+
tensors for more detail.
|
| 1007 |
+
output_hidden_states (`bool`, *optional*):
|
| 1008 |
+
Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
|
| 1009 |
+
more detail.
|
| 1010 |
+
return_dict (`bool`, *optional*):
|
| 1011 |
+
Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
|
| 1012 |
+
"""
|
| 1013 |
+
|
| 1014 |
+
|
| 1015 |
+
@add_start_docstrings(
|
| 1016 |
+
"The bare LLaDA2Moe Model outputting raw hidden-states without any specific head on top.",
|
| 1017 |
+
LLADA2MOE_START_DOCSTRING,
|
| 1018 |
+
)
|
| 1019 |
+
class LLaDA2MoeModel(LLaDA2MoePreTrainedModel):
|
| 1020 |
+
"""
|
| 1021 |
+
Transformer decoder consisting of *config.num_hidden_layers* layers. Each layer is a [`LLaDA2MoeDecoderLayer`]
|
| 1022 |
+
|
| 1023 |
+
Args:
|
| 1024 |
+
config: LLaDA2MoeConfig
|
| 1025 |
+
"""
|
| 1026 |
+
|
| 1027 |
+
def __init__(self, config: LLaDA2MoeConfig):
|
| 1028 |
+
super().__init__(config)
|
| 1029 |
+
self.padding_idx = config.pad_token_id
|
| 1030 |
+
self.vocab_size = config.vocab_size
|
| 1031 |
+
|
| 1032 |
+
self.word_embeddings = nn.Embedding(config.vocab_size, config.hidden_size, self.padding_idx)
|
| 1033 |
+
self.layers = nn.ModuleList(
|
| 1034 |
+
[LLaDA2MoeDecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
|
| 1035 |
+
)
|
| 1036 |
+
self._use_sdpa = config._attn_implementation == "sdpa"
|
| 1037 |
+
self._use_flash_attention_2 = config._attn_implementation == "flash_attention_2"
|
| 1038 |
+
self.norm = LLaDA2MoeRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
|
| 1039 |
+
self.rotary_emb = LLaDA2MoeRotaryEmbedding(config=config)
|
| 1040 |
+
self.gradient_checkpointing = False
|
| 1041 |
+
# Initialize weights and apply final processing
|
| 1042 |
+
self.post_init()
|
| 1043 |
+
|
| 1044 |
+
def get_input_embeddings(self):
|
| 1045 |
+
return self.word_embeddings
|
| 1046 |
+
|
| 1047 |
+
def set_input_embeddings(self, value):
|
| 1048 |
+
self.word_embeddings = value
|
| 1049 |
+
|
| 1050 |
+
@add_start_docstrings_to_model_forward(LLADA2MOE_INPUTS_DOCSTRING)
|
| 1051 |
+
def forward(
|
| 1052 |
+
self,
|
| 1053 |
+
input_ids: torch.LongTensor = None,
|
| 1054 |
+
attention_mask: Optional[torch.Tensor] = None,
|
| 1055 |
+
position_ids: Optional[torch.LongTensor] = None,
|
| 1056 |
+
past_key_values: Optional[List[torch.FloatTensor]] = None,
|
| 1057 |
+
inputs_embeds: Optional[torch.FloatTensor] = None,
|
| 1058 |
+
use_cache: Optional[bool] = None,
|
| 1059 |
+
output_attentions: Optional[bool] = None,
|
| 1060 |
+
output_hidden_states: Optional[bool] = None,
|
| 1061 |
+
output_router_logits: Optional[bool] = None,
|
| 1062 |
+
return_dict: Optional[bool] = None,
|
| 1063 |
+
**kwargs,
|
| 1064 |
+
) -> Union[Tuple, MoeModelOutputWithPast]:
|
| 1065 |
+
output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
|
| 1066 |
+
output_hidden_states = (
|
| 1067 |
+
output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
|
| 1068 |
+
)
|
| 1069 |
+
output_router_logits = (
|
| 1070 |
+
output_router_logits if output_router_logits is not None else self.config.output_router_logits
|
| 1071 |
+
)
|
| 1072 |
+
use_cache = use_cache if use_cache is not None else self.config.use_cache
|
| 1073 |
+
|
| 1074 |
+
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
|
| 1075 |
+
|
| 1076 |
+
# retrieve input_ids and inputs_embeds
|
| 1077 |
+
if input_ids is not None and inputs_embeds is not None:
|
| 1078 |
+
raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
|
| 1079 |
+
elif input_ids is not None:
|
| 1080 |
+
batch_size, seq_length = input_ids.shape[:2]
|
| 1081 |
+
elif inputs_embeds is not None:
|
| 1082 |
+
batch_size, seq_length = inputs_embeds.shape[:2]
|
| 1083 |
+
else:
|
| 1084 |
+
raise ValueError("You have to specify either input_ids or inputs_embeds")
|
| 1085 |
+
|
| 1086 |
+
if self.gradient_checkpointing and self.training:
|
| 1087 |
+
if use_cache:
|
| 1088 |
+
logger.warning_once(
|
| 1089 |
+
"`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`transformers."
|
| 1090 |
+
)
|
| 1091 |
+
use_cache = False
|
| 1092 |
+
|
| 1093 |
+
past_key_values_length = 0
|
| 1094 |
+
if use_cache:
|
| 1095 |
+
use_legacy_cache = not isinstance(past_key_values, Cache)
|
| 1096 |
+
if use_legacy_cache:
|
| 1097 |
+
past_key_values = DynamicCache.from_legacy_cache(past_key_values)
|
| 1098 |
+
past_key_values_length = past_key_values.get_usable_length(seq_length)
|
| 1099 |
+
|
| 1100 |
+
if position_ids is None:
|
| 1101 |
+
device = input_ids.device if input_ids is not None else inputs_embeds.device
|
| 1102 |
+
position_ids = torch.arange(
|
| 1103 |
+
past_key_values_length, seq_length + past_key_values_length, dtype=torch.long, device=device
|
| 1104 |
+
)
|
| 1105 |
+
position_ids = position_ids.unsqueeze(0)
|
| 1106 |
+
|
| 1107 |
+
if inputs_embeds is None:
|
| 1108 |
+
inputs_embeds = self.word_embeddings(input_ids)
|
| 1109 |
+
|
| 1110 |
+
# TODO flash attention 2 can not support custom attention mask
|
| 1111 |
+
# if self._use_flash_attention_2:
|
| 1112 |
+
# # 2d mask is passed through the layers
|
| 1113 |
+
# attention_mask = attention_mask if (attention_mask is not None and 0 in attention_mask) else None
|
| 1114 |
+
if self._use_sdpa and not output_attentions:
|
| 1115 |
+
# output_attentions=True can not be supported when using SDPA, and we fall back on
|
| 1116 |
+
# the manual implementation that requires a 4D causal mask in all cases.
|
| 1117 |
+
attention_mask = _prepare_4d_causal_attention_mask_for_sdpa(
|
| 1118 |
+
attention_mask,
|
| 1119 |
+
(batch_size, seq_length),
|
| 1120 |
+
inputs_embeds,
|
| 1121 |
+
past_key_values_length,
|
| 1122 |
+
)
|
| 1123 |
+
else:
|
| 1124 |
+
# 4d mask is passed through the layers
|
| 1125 |
+
attention_mask = _prepare_4d_causal_attention_mask(
|
| 1126 |
+
attention_mask, (batch_size, seq_length), inputs_embeds, past_key_values_length
|
| 1127 |
+
)
|
| 1128 |
+
|
| 1129 |
+
# embed positions
|
| 1130 |
+
hidden_states = inputs_embeds
|
| 1131 |
+
|
| 1132 |
+
# create position embeddings to be shared across the decoder layers
|
| 1133 |
+
position_embeddings = self.rotary_emb(hidden_states, position_ids)
|
| 1134 |
+
|
| 1135 |
+
# decoder layers
|
| 1136 |
+
all_hidden_states = () if output_hidden_states else None
|
| 1137 |
+
all_self_attns = () if output_attentions else None
|
| 1138 |
+
all_router_logits = () if output_router_logits else None
|
| 1139 |
+
next_decoder_cache = None
|
| 1140 |
+
|
| 1141 |
+
for decoder_layer in self.layers:
|
| 1142 |
+
if output_hidden_states:
|
| 1143 |
+
all_hidden_states += (hidden_states,)
|
| 1144 |
+
|
| 1145 |
+
if self.gradient_checkpointing and self.training:
|
| 1146 |
+
layer_outputs = self._gradient_checkpointing_func(
|
| 1147 |
+
decoder_layer.__call__,
|
| 1148 |
+
hidden_states,
|
| 1149 |
+
attention_mask,
|
| 1150 |
+
position_ids,
|
| 1151 |
+
past_key_values,
|
| 1152 |
+
output_attentions,
|
| 1153 |
+
output_router_logits,
|
| 1154 |
+
use_cache,
|
| 1155 |
+
position_embeddings,
|
| 1156 |
+
)
|
| 1157 |
+
else:
|
| 1158 |
+
layer_outputs = decoder_layer(
|
| 1159 |
+
hidden_states,
|
| 1160 |
+
attention_mask=attention_mask,
|
| 1161 |
+
position_ids=position_ids,
|
| 1162 |
+
past_key_value=past_key_values,
|
| 1163 |
+
output_attentions=output_attentions,
|
| 1164 |
+
output_router_logits=output_router_logits,
|
| 1165 |
+
use_cache=use_cache,
|
| 1166 |
+
position_embeddings=position_embeddings,
|
| 1167 |
+
)
|
| 1168 |
+
hidden_states = layer_outputs[0]
|
| 1169 |
+
|
| 1170 |
+
if use_cache:
|
| 1171 |
+
next_decoder_cache = layer_outputs[2 if output_attentions else 1]
|
| 1172 |
+
|
| 1173 |
+
if output_attentions:
|
| 1174 |
+
all_self_attns += (layer_outputs[1],)
|
| 1175 |
+
|
| 1176 |
+
if output_router_logits and layer_outputs[-1] is not None:
|
| 1177 |
+
all_router_logits += (layer_outputs[-1],)
|
| 1178 |
+
|
| 1179 |
+
hidden_states = self.norm(hidden_states)
|
| 1180 |
+
|
| 1181 |
+
# add hidden states from the last decoder layer
|
| 1182 |
+
if output_hidden_states:
|
| 1183 |
+
all_hidden_states += (hidden_states,)
|
| 1184 |
+
|
| 1185 |
+
next_cache = None
|
| 1186 |
+
if use_cache:
|
| 1187 |
+
next_cache = next_decoder_cache.to_legacy_cache() if use_legacy_cache else next_decoder_cache
|
| 1188 |
+
if not return_dict:
|
| 1189 |
+
return tuple(
|
| 1190 |
+
v
|
| 1191 |
+
for v in [hidden_states, next_cache, all_hidden_states, all_self_attns, all_router_logits]
|
| 1192 |
+
if v is not None
|
| 1193 |
+
)
|
| 1194 |
+
return MoeModelOutputWithPast(
|
| 1195 |
+
last_hidden_state=hidden_states,
|
| 1196 |
+
past_key_values=next_cache,
|
| 1197 |
+
hidden_states=all_hidden_states,
|
| 1198 |
+
attentions=all_self_attns,
|
| 1199 |
+
router_logits=all_router_logits,
|
| 1200 |
+
)
|
| 1201 |
+
|
| 1202 |
+
|
| 1203 |
+
class LLaDA2MoeModelLM(LLaDA2MoePreTrainedModel, GenerationMixin):
|
| 1204 |
+
_tied_weights_keys = ["lm_head.weight"]
|
| 1205 |
+
|
| 1206 |
+
def __init__(self, config: LLaDA2MoeConfig):
|
| 1207 |
+
super().__init__(config)
|
| 1208 |
+
self.model = LLaDA2MoeModel(config)
|
| 1209 |
+
self.vocab_size = config.vocab_size
|
| 1210 |
+
self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
|
| 1211 |
+
|
| 1212 |
+
# Initialize weights and apply final processing
|
| 1213 |
+
self.post_init()
|
| 1214 |
+
|
| 1215 |
+
def get_input_embeddings(self):
|
| 1216 |
+
return self.model.word_embeddings
|
| 1217 |
+
|
| 1218 |
+
def set_input_embeddings(self, value):
|
| 1219 |
+
self.model.word_embeddings = value
|
| 1220 |
+
|
| 1221 |
+
def get_output_embeddings(self):
|
| 1222 |
+
return self.lm_head
|
| 1223 |
+
|
| 1224 |
+
def set_output_embeddings(self, new_embeddings):
|
| 1225 |
+
self.lm_head = new_embeddings
|
| 1226 |
+
|
| 1227 |
+
def set_decoder(self, decoder):
|
| 1228 |
+
self.model = decoder
|
| 1229 |
+
|
| 1230 |
+
def get_decoder(self):
|
| 1231 |
+
return self.model
|
| 1232 |
+
|
| 1233 |
+
@add_start_docstrings_to_model_forward(LLADA2MOE_INPUTS_DOCSTRING)
|
| 1234 |
+
@replace_return_docstrings(output_type=MoeCausalLMOutputWithPast, config_class=_CONFIG_FOR_DOC)
|
| 1235 |
+
def forward(
|
| 1236 |
+
self,
|
| 1237 |
+
input_ids: torch.LongTensor = None,
|
| 1238 |
+
attention_mask: Optional[torch.Tensor] = None,
|
| 1239 |
+
position_ids: Optional[torch.LongTensor] = None,
|
| 1240 |
+
past_key_values: Optional[List[torch.FloatTensor]] = None,
|
| 1241 |
+
inputs_embeds: Optional[torch.FloatTensor] = None,
|
| 1242 |
+
labels: Optional[torch.LongTensor] = None,
|
| 1243 |
+
use_cache: Optional[bool] = None,
|
| 1244 |
+
output_attentions: Optional[bool] = None,
|
| 1245 |
+
output_hidden_states: Optional[bool] = None,
|
| 1246 |
+
output_router_logits: Optional[bool] = None,
|
| 1247 |
+
return_dict: Optional[bool] = None,
|
| 1248 |
+
**kwargs,
|
| 1249 |
+
) -> Union[Tuple, MoeCausalLMOutputWithPast]:
|
| 1250 |
+
r"""
|
| 1251 |
+
Args:
|
| 1252 |
+
labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
|
| 1253 |
+
Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
|
| 1254 |
+
config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
|
| 1255 |
+
(masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`.
|
| 1256 |
+
|
| 1257 |
+
Returns:
|
| 1258 |
+
|
| 1259 |
+
Example:
|
| 1260 |
+
|
| 1261 |
+
```python
|
| 1262 |
+
>>> from transformers import AutoTokenizer
|
| 1263 |
+
|
| 1264 |
+
>>> model = LLaDA2MoeForCausalLM.from_pretrained(PATH_TO_CONVERTED_WEIGHTS)
|
| 1265 |
+
>>> tokenizer = AutoTokenizer.from_pretrained(PATH_TO_CONVERTED_TOKENIZER)
|
| 1266 |
+
|
| 1267 |
+
>>> prompt = "Hey, are you conscious? Can you talk to me?"
|
| 1268 |
+
>>> inputs = tokenizer(prompt, return_tensors="pt")
|
| 1269 |
+
|
| 1270 |
+
>>> # Generate
|
| 1271 |
+
>>> generate_ids = model.generate(inputs.input_ids, max_length=30)
|
| 1272 |
+
>>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
|
| 1273 |
+
"Hey, are you conscious? Can you talk to me?\nI'm not conscious, but I can talk to you."
|
| 1274 |
+
```"""
|
| 1275 |
+
output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
|
| 1276 |
+
output_hidden_states = (
|
| 1277 |
+
output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
|
| 1278 |
+
)
|
| 1279 |
+
output_router_logits = (
|
| 1280 |
+
output_router_logits if output_router_logits is not None else self.config.output_router_logits
|
| 1281 |
+
)
|
| 1282 |
+
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
|
| 1283 |
+
# decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
|
| 1284 |
+
outputs = self.model(
|
| 1285 |
+
input_ids=input_ids,
|
| 1286 |
+
attention_mask=attention_mask,
|
| 1287 |
+
position_ids=position_ids,
|
| 1288 |
+
past_key_values=past_key_values,
|
| 1289 |
+
inputs_embeds=inputs_embeds,
|
| 1290 |
+
use_cache=use_cache,
|
| 1291 |
+
output_attentions=output_attentions,
|
| 1292 |
+
output_hidden_states=output_hidden_states,
|
| 1293 |
+
output_router_logits=output_router_logits,
|
| 1294 |
+
return_dict=return_dict,
|
| 1295 |
+
**kwargs,
|
| 1296 |
+
)
|
| 1297 |
+
|
| 1298 |
+
hidden_states = outputs[0]
|
| 1299 |
+
|
| 1300 |
+
logits = self.lm_head(hidden_states)
|
| 1301 |
+
logits = logits.float()
|
| 1302 |
+
|
| 1303 |
+
loss = None
|
| 1304 |
+
aux_loss = None
|
| 1305 |
+
|
| 1306 |
+
if labels is not None:
|
| 1307 |
+
# Shift so that tokens < n predict n
|
| 1308 |
+
shift_logits = logits[..., :-1, :].contiguous()
|
| 1309 |
+
shift_labels = labels[..., 1:].contiguous()
|
| 1310 |
+
# Flatten the tokens
|
| 1311 |
+
loss_fct = CrossEntropyLoss()
|
| 1312 |
+
shift_logits = shift_logits.view(-1, self.config.vocab_size)
|
| 1313 |
+
shift_labels = shift_labels.view(-1)
|
| 1314 |
+
# Enable model parallelism
|
| 1315 |
+
shift_labels = shift_labels.to(shift_logits.device)
|
| 1316 |
+
loss = loss_fct(shift_logits, shift_labels)
|
| 1317 |
+
|
| 1318 |
+
if not return_dict:
|
| 1319 |
+
output = (logits,) + outputs[1:]
|
| 1320 |
+
if output_router_logits:
|
| 1321 |
+
output = (aux_loss,) + output
|
| 1322 |
+
return (loss,) + output if loss is not None else output
|
| 1323 |
+
|
| 1324 |
+
return MoeCausalLMOutputWithPast(
|
| 1325 |
+
loss=loss,
|
| 1326 |
+
aux_loss=aux_loss,
|
| 1327 |
+
logits=logits,
|
| 1328 |
+
past_key_values=outputs.past_key_values,
|
| 1329 |
+
hidden_states=outputs.hidden_states,
|
| 1330 |
+
attentions=outputs.attentions,
|
| 1331 |
+
router_logits=outputs.router_logits,
|
| 1332 |
+
)
|
| 1333 |
+
|
| 1334 |
+
def prepare_inputs_for_generation(
|
| 1335 |
+
self, input_ids, past_key_values=None, attention_mask=None, inputs_embeds=None, token_type_ids=None, **kwargs
|
| 1336 |
+
):
|
| 1337 |
+
if past_key_values is not None:
|
| 1338 |
+
if isinstance(past_key_values, Cache):
|
| 1339 |
+
cache_length = past_key_values.get_seq_length()
|
| 1340 |
+
past_length = past_key_values.seen_tokens
|
| 1341 |
+
max_cache_length = (
|
| 1342 |
+
past_key_values.get_max_length()
|
| 1343 |
+
if hasattr(past_key_values, "get_max_length")
|
| 1344 |
+
else past_key_values.get_max_cache_shape()
|
| 1345 |
+
)
|
| 1346 |
+
else:
|
| 1347 |
+
cache_length = past_length = past_key_values[0][0].shape[2]
|
| 1348 |
+
max_cache_length = None
|
| 1349 |
+
|
| 1350 |
+
# Keep only the unprocessed tokens:
|
| 1351 |
+
# 1 - If the length of the attention_mask exceeds the length of input_ids, then we are in a setting where
|
| 1352 |
+
# some of the inputs are exclusivelly passed as part of the cache (e.g. when passing input_embeds as input)
|
| 1353 |
+
if attention_mask is not None and attention_mask.shape[1] > input_ids.shape[1]:
|
| 1354 |
+
input_ids = input_ids[:, -(attention_mask.shape[1] - past_length) :]
|
| 1355 |
+
# 2 - If the past_length is smaller than input_ids', then input_ids holds all input tokens. We can discard
|
| 1356 |
+
# input_ids based on the past_length.
|
| 1357 |
+
elif past_length < input_ids.shape[1]:
|
| 1358 |
+
input_ids = input_ids[:, past_length:]
|
| 1359 |
+
# 3 - Otherwise (past_length >= input_ids.shape[1]), let's assume input_ids only has unprocessed tokens.
|
| 1360 |
+
|
| 1361 |
+
# If we are about to go beyond the maximum cache length, we need to crop the input attention mask.
|
| 1362 |
+
if (
|
| 1363 |
+
max_cache_length is not None
|
| 1364 |
+
and attention_mask is not None
|
| 1365 |
+
and cache_length + input_ids.shape[1] > max_cache_length
|
| 1366 |
+
):
|
| 1367 |
+
attention_mask = attention_mask[:, -max_cache_length:]
|
| 1368 |
+
|
| 1369 |
+
position_ids = kwargs.get("position_ids", None)
|
| 1370 |
+
if attention_mask is not None and position_ids is None:
|
| 1371 |
+
# create position_ids on the fly for batch generation
|
| 1372 |
+
position_ids = attention_mask.long().cumsum(-1) - 1
|
| 1373 |
+
position_ids.masked_fill_(attention_mask == 0, 1)
|
| 1374 |
+
if past_key_values:
|
| 1375 |
+
position_ids = position_ids[:, -input_ids.shape[1] :]
|
| 1376 |
+
|
| 1377 |
+
# if `inputs_embeds` are passed, we only want to use them in the 1st generation step
|
| 1378 |
+
if inputs_embeds is not None and past_key_values is None:
|
| 1379 |
+
model_inputs = {"inputs_embeds": inputs_embeds}
|
| 1380 |
+
else:
|
| 1381 |
+
model_inputs = {"input_ids": input_ids}
|
| 1382 |
+
|
| 1383 |
+
model_inputs.update(
|
| 1384 |
+
{
|
| 1385 |
+
"position_ids": position_ids,
|
| 1386 |
+
"past_key_values": past_key_values,
|
| 1387 |
+
"use_cache": kwargs.get("use_cache"),
|
| 1388 |
+
"attention_mask": attention_mask,
|
| 1389 |
+
}
|
| 1390 |
+
)
|
| 1391 |
+
return model_inputs
|
| 1392 |
+
|
| 1393 |
+
@staticmethod
|
| 1394 |
+
def _reorder_cache(past_key_values, beam_idx):
|
| 1395 |
+
reordered_past = ()
|
| 1396 |
+
for layer_past in past_key_values:
|
| 1397 |
+
reordered_past += (
|
| 1398 |
+
tuple(past_state.index_select(0, beam_idx.to(past_state.device)) for past_state in layer_past),
|
| 1399 |
+
)
|
| 1400 |
+
return reordered_past
|
| 1401 |
+
|
| 1402 |
+
@staticmethod
|
| 1403 |
+
def _top_k_logits(logits, k):
|
| 1404 |
+
if k is None or k <= 0:
|
| 1405 |
+
return logits
|
| 1406 |
+
else:
|
| 1407 |
+
values, _ = torch.topk(logits, k)
|
| 1408 |
+
min_values = values[..., -1, None]
|
| 1409 |
+
return torch.where(
|
| 1410 |
+
logits < min_values, torch.full_like(logits, float("-inf")), logits
|
| 1411 |
+
)
|
| 1412 |
+
|
| 1413 |
+
@staticmethod
|
| 1414 |
+
def _top_p_logits(logits, p):
|
| 1415 |
+
if p is None or p >= 1.0:
|
| 1416 |
+
return logits
|
| 1417 |
+
sorted_logits, sorted_indices = torch.sort(logits, descending=True)
|
| 1418 |
+
cumulative_probs = torch.cumsum(F.softmax(sorted_logits, dim=-1), dim=-1)
|
| 1419 |
+
sorted_mask = cumulative_probs > p
|
| 1420 |
+
sorted_mask[..., 1:] = sorted_mask[..., :-1].clone()
|
| 1421 |
+
sorted_mask[..., 0] = False
|
| 1422 |
+
mask_indices = torch.scatter(
|
| 1423 |
+
torch.full_like(logits, False, dtype=torch.bool),
|
| 1424 |
+
-1,
|
| 1425 |
+
sorted_indices,
|
| 1426 |
+
sorted_mask,
|
| 1427 |
+
)
|
| 1428 |
+
return logits.masked_fill(mask_indices, float("-inf"))
|
| 1429 |
+
|
| 1430 |
+
def _sample_with_temperature_topk_topp(self, logits, temperature=1.0, top_k=0, top_p=1.0):
|
| 1431 |
+
orig_shape = logits.shape[:-1]
|
| 1432 |
+
vocab_size = logits.shape[-1]
|
| 1433 |
+
logits = logits.reshape(-1, vocab_size)
|
| 1434 |
+
if temperature > 0 and temperature != 1.0:
|
| 1435 |
+
logits = logits / temperature
|
| 1436 |
+
logits = self._top_k_logits(logits, top_k)
|
| 1437 |
+
logits = self._top_p_logits(logits, top_p)
|
| 1438 |
+
probs = F.softmax(logits, dim=-1)
|
| 1439 |
+
token = torch.multinomial(probs, num_samples=1)
|
| 1440 |
+
token_prob = torch.gather(probs, -1, token)
|
| 1441 |
+
return token.view(*orig_shape), token_prob.view(*orig_shape)
|
| 1442 |
+
|
| 1443 |
+
@staticmethod
|
| 1444 |
+
def _get_num_transfer_tokens(block_length, steps):
|
| 1445 |
+
if steps == 0:
|
| 1446 |
+
return torch.tensor([], dtype=torch.int64)
|
| 1447 |
+
base = block_length // steps
|
| 1448 |
+
remainder = block_length % steps
|
| 1449 |
+
num_transfer_tokens = torch.full((steps,), base, dtype=torch.int64)
|
| 1450 |
+
num_transfer_tokens[:remainder] += 1
|
| 1451 |
+
return num_transfer_tokens
|
| 1452 |
+
|
| 1453 |
+
@torch.no_grad()
|
| 1454 |
+
def generate(
|
| 1455 |
+
self,
|
| 1456 |
+
inputs: Optional[torch.Tensor] = None,
|
| 1457 |
+
temperature: int = 0.0,
|
| 1458 |
+
block_length: int = 32,
|
| 1459 |
+
steps: int = 32,
|
| 1460 |
+
gen_length: int = 2048,
|
| 1461 |
+
top_p: Optional[int] = None,
|
| 1462 |
+
top_k: Optional[int] = None,
|
| 1463 |
+
eos_early_stop: bool = False,
|
| 1464 |
+
minimal_topk: int = 1,
|
| 1465 |
+
threshold: float = 0.95,
|
| 1466 |
+
eos_id: int = 156892,
|
| 1467 |
+
mask_id: int = 156895,
|
| 1468 |
+
):
|
| 1469 |
+
r"""
|
| 1470 |
+
Generates tokens using a block-wise, iterative refinement strategy.
|
| 1471 |
+
|
| 1472 |
+
This method operates differently from standard autoregressive generation. It first creates a template of the
|
| 1473 |
+
full desired length, filled with a special `mask_id`. It then processes this template in segments (`blocks`)
|
| 1474 |
+
and iteratively "denoises" or "refines" the `mask_id` tokens into actual tokens over a series of `steps` for
|
| 1475 |
+
each block. A custom block-diagonal causal attention mask ensures that generation within a block can attend to
|
| 1476 |
+
all previous blocks but not future ones.
|
| 1477 |
+
|
| 1478 |
+
<Tip warning={true}>
|
| 1479 |
+
|
| 1480 |
+
This is a specialized generation method. The quality and speed of the output are highly dependent on the interplay
|
| 1481 |
+
between `block_length`, `steps`, and `threshold`. It aims to achieve faster generation through parallel
|
| 1482 |
+
decoding within blocks, which is a departure from the token-by-token generation of standard `.generate()` methods.
|
| 1483 |
+
|
| 1484 |
+
</Tip>
|
| 1485 |
+
|
| 1486 |
+
Parameters:
|
| 1487 |
+
inputs (`torch.Tensor`):
|
| 1488 |
+
The token sequence used as a prompt for the generation.
|
| 1489 |
+
temperature (`float`, *optional*, defaults to 0.0):
|
| 1490 |
+
The value used to module the next token probabilities. A value of 0.0 corresponds to greedy decoding.
|
| 1491 |
+
block_length (`int`, *optional*, defaults to 32):
|
| 1492 |
+
The size of each generation block. The model generates text in parallel within these blocks. This is a
|
| 1493 |
+
key parameter for controlling the granularity of the generation process.
|
| 1494 |
+
steps (`int`, *optional*, defaults to 32):
|
| 1495 |
+
The number of iterative refinement (or "denoising") steps to perform for each block. Within each block,
|
| 1496 |
+
the model will try to replace `mask_id` tokens with real tokens for this many iterations.
|
| 1497 |
+
gen_length (`int`, *optional*, defaults to 2048):
|
| 1498 |
+
The maximum number of tokens to generate, excluding the prompt.
|
| 1499 |
+
top_p (`float`, *optional*):
|
| 1500 |
+
If set to a float value between 0 and 1, only the most probable tokens with probabilities that add up to
|
| 1501 |
+
`top_p` or higher are kept for generation (nucleus sampling).
|
| 1502 |
+
top_k (`int`, *optional*):
|
| 1503 |
+
The number of highest probability vocabulary tokens to keep for top-k-filtering.
|
| 1504 |
+
eos_early_stop (`bool`, *optional*, defaults to `False`):
|
| 1505 |
+
If `True`, generation will stop as soon as a valid End-Of-Sequence token is generated and confirmed,
|
| 1506 |
+
even if `gen_length` has not been reached.
|
| 1507 |
+
minimal_topk (`int`, *optional*, defaults to 1):
|
| 1508 |
+
A parameter used to dynamically adjust the number of refinement `steps`. The effective number of steps
|
| 1509 |
+
is capped at `gen_length // minimal_topk`.
|
| 1510 |
+
threshold (`float`, *optional*, defaults to 0.95):
|
| 1511 |
+
The confidence probability threshold for accepting a sampled token. During each refinement step, a
|
| 1512 |
+
sampled token is only kept if its probability is above this threshold. If not enough tokens meet the
|
| 1513 |
+
threshold, the ones with the highest confidence are chosen.
|
| 1514 |
+
eos_id (`int`, *optional*, defaults to 156892):
|
| 1515 |
+
The token ID for the end-of-sequence token. Used for `eos_early_stop`.
|
| 1516 |
+
mask_id (`int`, *optional*, defaults to 156895):
|
| 1517 |
+
The token ID used as a placeholder for tokens that are yet to be generated. This is central to the
|
| 1518 |
+
iterative refinement algorithm.
|
| 1519 |
+
|
| 1520 |
+
Return:
|
| 1521 |
+
`torch.Tensor`: A string containing the generated token IDs, starting
|
| 1522 |
+
after the prompt and stopping at the first `eos_id` or `gen_length`.
|
| 1523 |
+
"""
|
| 1524 |
+
steps = min(steps, gen_length // minimal_topk)
|
| 1525 |
+
input_ids = inputs.to(self.device)
|
| 1526 |
+
|
| 1527 |
+
prompt_length = input_ids.shape[1]
|
| 1528 |
+
num_blocks = (prompt_length + gen_length + block_length - 1) // block_length
|
| 1529 |
+
total_length = num_blocks * block_length
|
| 1530 |
+
|
| 1531 |
+
block_mask = torch.tril(torch.ones(num_blocks, num_blocks, device=self.device))
|
| 1532 |
+
block_diffusion_attention_mask = (
|
| 1533 |
+
block_mask.repeat_interleave(block_length, dim=0)
|
| 1534 |
+
.repeat_interleave(block_length, dim=1)
|
| 1535 |
+
.unsqueeze(0)
|
| 1536 |
+
.unsqueeze(0)
|
| 1537 |
+
).bool()
|
| 1538 |
+
block_diffusion_attention_mask = torch.where(
|
| 1539 |
+
block_diffusion_attention_mask, 0.0, float("-inf")
|
| 1540 |
+
).to(torch.bfloat16)
|
| 1541 |
+
|
| 1542 |
+
position_ids = torch.arange(total_length, device=self.device).unsqueeze(0)
|
| 1543 |
+
x = torch.full((1, total_length), mask_id, dtype=torch.long, device=self.device)
|
| 1544 |
+
x[:, :prompt_length] = input_ids.clone()
|
| 1545 |
+
|
| 1546 |
+
prompt_index_full = torch.zeros_like(x, dtype=torch.bool)
|
| 1547 |
+
prompt_index_full[:, :prompt_length] = True
|
| 1548 |
+
|
| 1549 |
+
prefill_blocks = prompt_length // block_length
|
| 1550 |
+
|
| 1551 |
+
denoising_steps_per_block = steps
|
| 1552 |
+
num_transfer_tokens_schedule = self._get_num_transfer_tokens(
|
| 1553 |
+
block_length, denoising_steps_per_block
|
| 1554 |
+
)
|
| 1555 |
+
for num_block in range(prefill_blocks, num_blocks):
|
| 1556 |
+
current_window_end = (num_block + 1) * block_length
|
| 1557 |
+
cur_x = x[:, :current_window_end]
|
| 1558 |
+
cur_attn_mask = block_diffusion_attention_mask[
|
| 1559 |
+
:, :, :current_window_end, :current_window_end
|
| 1560 |
+
]
|
| 1561 |
+
cur_position_ids = position_ids[:, :current_window_end]
|
| 1562 |
+
|
| 1563 |
+
for step in range(denoising_steps_per_block):
|
| 1564 |
+
active_block_mask = cur_x[:, -block_length:] == mask_id
|
| 1565 |
+
if active_block_mask.sum() == 0:
|
| 1566 |
+
break
|
| 1567 |
+
|
| 1568 |
+
logits = self.forward(
|
| 1569 |
+
cur_x,
|
| 1570 |
+
attention_mask=cur_attn_mask,
|
| 1571 |
+
position_ids=cur_position_ids,
|
| 1572 |
+
).logits
|
| 1573 |
+
|
| 1574 |
+
active_logits = logits[:, -block_length:, :]
|
| 1575 |
+
x0, x0_p = self._sample_with_temperature_topk_topp(
|
| 1576 |
+
active_logits, temperature=temperature, top_k=top_k, top_p=top_p
|
| 1577 |
+
)
|
| 1578 |
+
|
| 1579 |
+
num_to_transfer = num_transfer_tokens_schedule[step].item()
|
| 1580 |
+
transfer_index = torch.zeros_like(x0, dtype=torch.bool)
|
| 1581 |
+
|
| 1582 |
+
confidence = torch.where(active_block_mask, x0_p, -torch.inf)
|
| 1583 |
+
high_conf_mask = confidence[0] > threshold
|
| 1584 |
+
num_high_confidence = high_conf_mask.sum().item()
|
| 1585 |
+
|
| 1586 |
+
if num_high_confidence >= num_to_transfer:
|
| 1587 |
+
transfer_index[0] = high_conf_mask
|
| 1588 |
+
else:
|
| 1589 |
+
_, idx = torch.topk(
|
| 1590 |
+
confidence[0],
|
| 1591 |
+
k=min(num_to_transfer, active_block_mask.sum().item()),
|
| 1592 |
+
)
|
| 1593 |
+
transfer_index[0, idx] = True
|
| 1594 |
+
|
| 1595 |
+
if transfer_index.any():
|
| 1596 |
+
cur_x[:, -block_length:][transfer_index] = x0[transfer_index]
|
| 1597 |
+
if eos_early_stop and (x0[transfer_index] == eos_id).any():
|
| 1598 |
+
eos_pos_in_x = (cur_x[0] == eos_id).nonzero(as_tuple=True)
|
| 1599 |
+
if len(eos_pos_in_x[0]) > 0:
|
| 1600 |
+
eos_pos = eos_pos_in_x[0][0].item()
|
| 1601 |
+
if (cur_x[0, prompt_length:eos_pos] != mask_id).all():
|
| 1602 |
+
final_x = x[:, :total_length][:, : eos_pos + 1]
|
| 1603 |
+
return final_x
|
| 1604 |
+
|
| 1605 |
+
x[:, :current_window_end] = cur_x
|
| 1606 |
+
if (
|
| 1607 |
+
eos_id is not None
|
| 1608 |
+
and (x[0, prompt_length:current_window_end] == eos_id).any()
|
| 1609 |
+
):
|
| 1610 |
+
break
|
| 1611 |
+
|
| 1612 |
+
generated_answer = x[:, : prompt_length + gen_length]
|
| 1613 |
+
|
| 1614 |
+
mask_positions = (generated_answer[0][input_ids.shape[1] :] == eos_id).nonzero(
|
| 1615 |
+
as_tuple=True
|
| 1616 |
+
)[0]
|
| 1617 |
+
if len(mask_positions) > 0:
|
| 1618 |
+
first_mask_position = mask_positions[0].item()
|
| 1619 |
+
else:
|
| 1620 |
+
first_mask_position = gen_length
|
| 1621 |
+
return generated_answer[:, input_ids.shape[1] : input_ids.shape[1] + first_mask_position + 1]
|
special_tokens_map.json
ADDED
|
@@ -0,0 +1,37 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token": {
|
| 3 |
+
"content": "<|startoftext|>",
|
| 4 |
+
"lstrip": false,
|
| 5 |
+
"normalized": false,
|
| 6 |
+
"rstrip": false,
|
| 7 |
+
"single_word": false
|
| 8 |
+
},
|
| 9 |
+
"cls_token": {
|
| 10 |
+
"content": "[CLS]",
|
| 11 |
+
"lstrip": false,
|
| 12 |
+
"normalized": false,
|
| 13 |
+
"rstrip": false,
|
| 14 |
+
"single_word": false
|
| 15 |
+
},
|
| 16 |
+
"eos_token": {
|
| 17 |
+
"content": "<|endoftext|>",
|
| 18 |
+
"lstrip": false,
|
| 19 |
+
"normalized": false,
|
| 20 |
+
"rstrip": false,
|
| 21 |
+
"single_word": false
|
| 22 |
+
},
|
| 23 |
+
"mask_token": {
|
| 24 |
+
"content": "<|mask|>",
|
| 25 |
+
"lstrip": false,
|
| 26 |
+
"normalized": false,
|
| 27 |
+
"rstrip": false,
|
| 28 |
+
"single_word": false
|
| 29 |
+
},
|
| 30 |
+
"pad_token": {
|
| 31 |
+
"content": "<|endoftext|>",
|
| 32 |
+
"lstrip": false,
|
| 33 |
+
"normalized": false,
|
| 34 |
+
"rstrip": false,
|
| 35 |
+
"single_word": false
|
| 36 |
+
}
|
| 37 |
+
}
|
tokenizer.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:878eb900ea4a42da7e54bc18ad38352bec005bbfd81bdf76a5ecf1aad1093a6c
|
| 3 |
+
size 12205801
|
tokenizer_config.json
ADDED
|
@@ -0,0 +1,2115 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_bos_token": false,
|
| 3 |
+
"add_eos_token": false,
|
| 4 |
+
"added_tokens_decoder": {
|
| 5 |
+
"156891": {
|
| 6 |
+
"content": "<|startoftext|>",
|
| 7 |
+
"lstrip": false,
|
| 8 |
+
"normalized": false,
|
| 9 |
+
"rstrip": false,
|
| 10 |
+
"single_word": false,
|
| 11 |
+
"special": true
|
| 12 |
+
},
|
| 13 |
+
"156892": {
|
| 14 |
+
"content": "<|endoftext|>",
|
| 15 |
+
"lstrip": false,
|
| 16 |
+
"normalized": false,
|
| 17 |
+
"rstrip": false,
|
| 18 |
+
"single_word": false,
|
| 19 |
+
"special": true
|
| 20 |
+
},
|
| 21 |
+
"156893": {
|
| 22 |
+
"content": "[CLS]",
|
| 23 |
+
"lstrip": false,
|
| 24 |
+
"normalized": false,
|
| 25 |
+
"rstrip": false,
|
| 26 |
+
"single_word": false,
|
| 27 |
+
"special": true
|
| 28 |
+
},
|
| 29 |
+
"156894": {
|
| 30 |
+
"content": "[gMASK]",
|
| 31 |
+
"lstrip": false,
|
| 32 |
+
"normalized": false,
|
| 33 |
+
"rstrip": false,
|
| 34 |
+
"single_word": false,
|
| 35 |
+
"special": true
|
| 36 |
+
},
|
| 37 |
+
"156895": {
|
| 38 |
+
"content": "<|mask|>",
|
| 39 |
+
"lstrip": false,
|
| 40 |
+
"normalized": false,
|
| 41 |
+
"rstrip": false,
|
| 42 |
+
"single_word": false,
|
| 43 |
+
"special": true
|
| 44 |
+
},
|
| 45 |
+
"156896": {
|
| 46 |
+
"content": "<tool_call>",
|
| 47 |
+
"lstrip": false,
|
| 48 |
+
"normalized": false,
|
| 49 |
+
"rstrip": false,
|
| 50 |
+
"single_word": false,
|
| 51 |
+
"special": true
|
| 52 |
+
},
|
| 53 |
+
"156897": {
|
| 54 |
+
"content": "</tool_call>",
|
| 55 |
+
"lstrip": false,
|
| 56 |
+
"normalized": false,
|
| 57 |
+
"rstrip": false,
|
| 58 |
+
"single_word": false,
|
| 59 |
+
"special": true
|
| 60 |
+
},
|
| 61 |
+
"156898": {
|
| 62 |
+
"content": "<tool_response>",
|
| 63 |
+
"lstrip": false,
|
| 64 |
+
"normalized": false,
|
| 65 |
+
"rstrip": false,
|
| 66 |
+
"single_word": false,
|
| 67 |
+
"special": true
|
| 68 |
+
},
|
| 69 |
+
"156899": {
|
| 70 |
+
"content": "</tool_response>",
|
| 71 |
+
"lstrip": false,
|
| 72 |
+
"normalized": false,
|
| 73 |
+
"rstrip": false,
|
| 74 |
+
"single_word": false,
|
| 75 |
+
"special": true
|
| 76 |
+
},
|
| 77 |
+
"156900": {
|
| 78 |
+
"content": "<|role_end|>",
|
| 79 |
+
"lstrip": false,
|
| 80 |
+
"normalized": false,
|
| 81 |
+
"rstrip": false,
|
| 82 |
+
"single_word": false,
|
| 83 |
+
"special": true
|
| 84 |
+
},
|
| 85 |
+
"156901": {
|
| 86 |
+
"content": "<|reserved_token_6|>",
|
| 87 |
+
"lstrip": false,
|
| 88 |
+
"normalized": false,
|
| 89 |
+
"rstrip": false,
|
| 90 |
+
"single_word": false,
|
| 91 |
+
"special": true
|
| 92 |
+
},
|
| 93 |
+
"156902": {
|
| 94 |
+
"content": "<|reserved_token_7|>",
|
| 95 |
+
"lstrip": false,
|
| 96 |
+
"normalized": false,
|
| 97 |
+
"rstrip": false,
|
| 98 |
+
"single_word": false,
|
| 99 |
+
"special": true
|
| 100 |
+
},
|
| 101 |
+
"156903": {
|
| 102 |
+
"content": "<|reserved_token_8|>",
|
| 103 |
+
"lstrip": false,
|
| 104 |
+
"normalized": false,
|
| 105 |
+
"rstrip": false,
|
| 106 |
+
"single_word": false,
|
| 107 |
+
"special": true
|
| 108 |
+
},
|
| 109 |
+
"156904": {
|
| 110 |
+
"content": "<|reserved_token_9|>",
|
| 111 |
+
"lstrip": false,
|
| 112 |
+
"normalized": false,
|
| 113 |
+
"rstrip": false,
|
| 114 |
+
"single_word": false,
|
| 115 |
+
"special": true
|
| 116 |
+
},
|
| 117 |
+
"156905": {
|
| 118 |
+
"content": "<|reserved_token_10|>",
|
| 119 |
+
"lstrip": false,
|
| 120 |
+
"normalized": false,
|
| 121 |
+
"rstrip": false,
|
| 122 |
+
"single_word": false,
|
| 123 |
+
"special": true
|
| 124 |
+
},
|
| 125 |
+
"156906": {
|
| 126 |
+
"content": "<|reserved_token_11|>",
|
| 127 |
+
"lstrip": false,
|
| 128 |
+
"normalized": false,
|
| 129 |
+
"rstrip": false,
|
| 130 |
+
"single_word": false,
|
| 131 |
+
"special": true
|
| 132 |
+
},
|
| 133 |
+
"156907": {
|
| 134 |
+
"content": "<|reserved_token_12|>",
|
| 135 |
+
"lstrip": false,
|
| 136 |
+
"normalized": false,
|
| 137 |
+
"rstrip": false,
|
| 138 |
+
"single_word": false,
|
| 139 |
+
"special": true
|
| 140 |
+
},
|
| 141 |
+
"156908": {
|
| 142 |
+
"content": "<|reserved_token_13|>",
|
| 143 |
+
"lstrip": false,
|
| 144 |
+
"normalized": false,
|
| 145 |
+
"rstrip": false,
|
| 146 |
+
"single_word": false,
|
| 147 |
+
"special": true
|
| 148 |
+
},
|
| 149 |
+
"156909": {
|
| 150 |
+
"content": "<|reserved_token_14|>",
|
| 151 |
+
"lstrip": false,
|
| 152 |
+
"normalized": false,
|
| 153 |
+
"rstrip": false,
|
| 154 |
+
"single_word": false,
|
| 155 |
+
"special": true
|
| 156 |
+
},
|
| 157 |
+
"156910": {
|
| 158 |
+
"content": "<|reserved_token_15|>",
|
| 159 |
+
"lstrip": false,
|
| 160 |
+
"normalized": false,
|
| 161 |
+
"rstrip": false,
|
| 162 |
+
"single_word": false,
|
| 163 |
+
"special": true
|
| 164 |
+
},
|
| 165 |
+
"156911": {
|
| 166 |
+
"content": "<|reserved_token_16|>",
|
| 167 |
+
"lstrip": false,
|
| 168 |
+
"normalized": false,
|
| 169 |
+
"rstrip": false,
|
| 170 |
+
"single_word": false,
|
| 171 |
+
"special": true
|
| 172 |
+
},
|
| 173 |
+
"156912": {
|
| 174 |
+
"content": "<|reserved_token_17|>",
|
| 175 |
+
"lstrip": false,
|
| 176 |
+
"normalized": false,
|
| 177 |
+
"rstrip": false,
|
| 178 |
+
"single_word": false,
|
| 179 |
+
"special": true
|
| 180 |
+
},
|
| 181 |
+
"156913": {
|
| 182 |
+
"content": "<|reserved_token_18|>",
|
| 183 |
+
"lstrip": false,
|
| 184 |
+
"normalized": false,
|
| 185 |
+
"rstrip": false,
|
| 186 |
+
"single_word": false,
|
| 187 |
+
"special": true
|
| 188 |
+
},
|
| 189 |
+
"156914": {
|
| 190 |
+
"content": "<|reserved_token_19|>",
|
| 191 |
+
"lstrip": false,
|
| 192 |
+
"normalized": false,
|
| 193 |
+
"rstrip": false,
|
| 194 |
+
"single_word": false,
|
| 195 |
+
"special": true
|
| 196 |
+
},
|
| 197 |
+
"156915": {
|
| 198 |
+
"content": "<|reserved_token_20|>",
|
| 199 |
+
"lstrip": false,
|
| 200 |
+
"normalized": false,
|
| 201 |
+
"rstrip": false,
|
| 202 |
+
"single_word": false,
|
| 203 |
+
"special": true
|
| 204 |
+
},
|
| 205 |
+
"156916": {
|
| 206 |
+
"content": "<|reserved_token_21|>",
|
| 207 |
+
"lstrip": false,
|
| 208 |
+
"normalized": false,
|
| 209 |
+
"rstrip": false,
|
| 210 |
+
"single_word": false,
|
| 211 |
+
"special": true
|
| 212 |
+
},
|
| 213 |
+
"156917": {
|
| 214 |
+
"content": "<|reserved_token_22|>",
|
| 215 |
+
"lstrip": false,
|
| 216 |
+
"normalized": false,
|
| 217 |
+
"rstrip": false,
|
| 218 |
+
"single_word": false,
|
| 219 |
+
"special": true
|
| 220 |
+
},
|
| 221 |
+
"156918": {
|
| 222 |
+
"content": "<|reserved_token_23|>",
|
| 223 |
+
"lstrip": false,
|
| 224 |
+
"normalized": false,
|
| 225 |
+
"rstrip": false,
|
| 226 |
+
"single_word": false,
|
| 227 |
+
"special": true
|
| 228 |
+
},
|
| 229 |
+
"156919": {
|
| 230 |
+
"content": "<|reserved_token_24|>",
|
| 231 |
+
"lstrip": false,
|
| 232 |
+
"normalized": false,
|
| 233 |
+
"rstrip": false,
|
| 234 |
+
"single_word": false,
|
| 235 |
+
"special": true
|
| 236 |
+
},
|
| 237 |
+
"156920": {
|
| 238 |
+
"content": "<|reserved_token_25|>",
|
| 239 |
+
"lstrip": false,
|
| 240 |
+
"normalized": false,
|
| 241 |
+
"rstrip": false,
|
| 242 |
+
"single_word": false,
|
| 243 |
+
"special": true
|
| 244 |
+
},
|
| 245 |
+
"156921": {
|
| 246 |
+
"content": "<|reserved_token_26|>",
|
| 247 |
+
"lstrip": false,
|
| 248 |
+
"normalized": false,
|
| 249 |
+
"rstrip": false,
|
| 250 |
+
"single_word": false,
|
| 251 |
+
"special": true
|
| 252 |
+
},
|
| 253 |
+
"156922": {
|
| 254 |
+
"content": "<|reserved_token_27|>",
|
| 255 |
+
"lstrip": false,
|
| 256 |
+
"normalized": false,
|
| 257 |
+
"rstrip": false,
|
| 258 |
+
"single_word": false,
|
| 259 |
+
"special": true
|
| 260 |
+
},
|
| 261 |
+
"156923": {
|
| 262 |
+
"content": "<|reserved_token_28|>",
|
| 263 |
+
"lstrip": false,
|
| 264 |
+
"normalized": false,
|
| 265 |
+
"rstrip": false,
|
| 266 |
+
"single_word": false,
|
| 267 |
+
"special": true
|
| 268 |
+
},
|
| 269 |
+
"156924": {
|
| 270 |
+
"content": "<|reserved_token_29|>",
|
| 271 |
+
"lstrip": false,
|
| 272 |
+
"normalized": false,
|
| 273 |
+
"rstrip": false,
|
| 274 |
+
"single_word": false,
|
| 275 |
+
"special": true
|
| 276 |
+
},
|
| 277 |
+
"156925": {
|
| 278 |
+
"content": "<|reserved_token_30|>",
|
| 279 |
+
"lstrip": false,
|
| 280 |
+
"normalized": false,
|
| 281 |
+
"rstrip": false,
|
| 282 |
+
"single_word": false,
|
| 283 |
+
"special": true
|
| 284 |
+
},
|
| 285 |
+
"156926": {
|
| 286 |
+
"content": "<|reserved_token_31|>",
|
| 287 |
+
"lstrip": false,
|
| 288 |
+
"normalized": false,
|
| 289 |
+
"rstrip": false,
|
| 290 |
+
"single_word": false,
|
| 291 |
+
"special": true
|
| 292 |
+
},
|
| 293 |
+
"156927": {
|
| 294 |
+
"content": "<|reserved_token_32|>",
|
| 295 |
+
"lstrip": false,
|
| 296 |
+
"normalized": false,
|
| 297 |
+
"rstrip": false,
|
| 298 |
+
"single_word": false,
|
| 299 |
+
"special": true
|
| 300 |
+
},
|
| 301 |
+
"156928": {
|
| 302 |
+
"content": "<|reserved_token_33|>",
|
| 303 |
+
"lstrip": false,
|
| 304 |
+
"normalized": false,
|
| 305 |
+
"rstrip": false,
|
| 306 |
+
"single_word": false,
|
| 307 |
+
"special": true
|
| 308 |
+
},
|
| 309 |
+
"156929": {
|
| 310 |
+
"content": "<|reserved_token_34|>",
|
| 311 |
+
"lstrip": false,
|
| 312 |
+
"normalized": false,
|
| 313 |
+
"rstrip": false,
|
| 314 |
+
"single_word": false,
|
| 315 |
+
"special": true
|
| 316 |
+
},
|
| 317 |
+
"156930": {
|
| 318 |
+
"content": "<|reserved_token_35|>",
|
| 319 |
+
"lstrip": false,
|
| 320 |
+
"normalized": false,
|
| 321 |
+
"rstrip": false,
|
| 322 |
+
"single_word": false,
|
| 323 |
+
"special": true
|
| 324 |
+
},
|
| 325 |
+
"156931": {
|
| 326 |
+
"content": "<|reserved_token_36|>",
|
| 327 |
+
"lstrip": false,
|
| 328 |
+
"normalized": false,
|
| 329 |
+
"rstrip": false,
|
| 330 |
+
"single_word": false,
|
| 331 |
+
"special": true
|
| 332 |
+
},
|
| 333 |
+
"156932": {
|
| 334 |
+
"content": "<|reserved_token_37|>",
|
| 335 |
+
"lstrip": false,
|
| 336 |
+
"normalized": false,
|
| 337 |
+
"rstrip": false,
|
| 338 |
+
"single_word": false,
|
| 339 |
+
"special": true
|
| 340 |
+
},
|
| 341 |
+
"156933": {
|
| 342 |
+
"content": "<|reserved_token_38|>",
|
| 343 |
+
"lstrip": false,
|
| 344 |
+
"normalized": false,
|
| 345 |
+
"rstrip": false,
|
| 346 |
+
"single_word": false,
|
| 347 |
+
"special": true
|
| 348 |
+
},
|
| 349 |
+
"156934": {
|
| 350 |
+
"content": "<|reserved_token_39|>",
|
| 351 |
+
"lstrip": false,
|
| 352 |
+
"normalized": false,
|
| 353 |
+
"rstrip": false,
|
| 354 |
+
"single_word": false,
|
| 355 |
+
"special": true
|
| 356 |
+
},
|
| 357 |
+
"156935": {
|
| 358 |
+
"content": "<|reserved_token_40|>",
|
| 359 |
+
"lstrip": false,
|
| 360 |
+
"normalized": false,
|
| 361 |
+
"rstrip": false,
|
| 362 |
+
"single_word": false,
|
| 363 |
+
"special": true
|
| 364 |
+
},
|
| 365 |
+
"156936": {
|
| 366 |
+
"content": "<|reserved_token_41|>",
|
| 367 |
+
"lstrip": false,
|
| 368 |
+
"normalized": false,
|
| 369 |
+
"rstrip": false,
|
| 370 |
+
"single_word": false,
|
| 371 |
+
"special": true
|
| 372 |
+
},
|
| 373 |
+
"156937": {
|
| 374 |
+
"content": "<|reserved_token_42|>",
|
| 375 |
+
"lstrip": false,
|
| 376 |
+
"normalized": false,
|
| 377 |
+
"rstrip": false,
|
| 378 |
+
"single_word": false,
|
| 379 |
+
"special": true
|
| 380 |
+
},
|
| 381 |
+
"156938": {
|
| 382 |
+
"content": "<|reserved_token_43|>",
|
| 383 |
+
"lstrip": false,
|
| 384 |
+
"normalized": false,
|
| 385 |
+
"rstrip": false,
|
| 386 |
+
"single_word": false,
|
| 387 |
+
"special": true
|
| 388 |
+
},
|
| 389 |
+
"156939": {
|
| 390 |
+
"content": "<|reserved_token_44|>",
|
| 391 |
+
"lstrip": false,
|
| 392 |
+
"normalized": false,
|
| 393 |
+
"rstrip": false,
|
| 394 |
+
"single_word": false,
|
| 395 |
+
"special": true
|
| 396 |
+
},
|
| 397 |
+
"156940": {
|
| 398 |
+
"content": "<|reserved_token_45|>",
|
| 399 |
+
"lstrip": false,
|
| 400 |
+
"normalized": false,
|
| 401 |
+
"rstrip": false,
|
| 402 |
+
"single_word": false,
|
| 403 |
+
"special": true
|
| 404 |
+
},
|
| 405 |
+
"156941": {
|
| 406 |
+
"content": "<|reserved_token_46|>",
|
| 407 |
+
"lstrip": false,
|
| 408 |
+
"normalized": false,
|
| 409 |
+
"rstrip": false,
|
| 410 |
+
"single_word": false,
|
| 411 |
+
"special": true
|
| 412 |
+
},
|
| 413 |
+
"156942": {
|
| 414 |
+
"content": "<|reserved_token_47|>",
|
| 415 |
+
"lstrip": false,
|
| 416 |
+
"normalized": false,
|
| 417 |
+
"rstrip": false,
|
| 418 |
+
"single_word": false,
|
| 419 |
+
"special": true
|
| 420 |
+
},
|
| 421 |
+
"156943": {
|
| 422 |
+
"content": "<|reserved_token_48|>",
|
| 423 |
+
"lstrip": false,
|
| 424 |
+
"normalized": false,
|
| 425 |
+
"rstrip": false,
|
| 426 |
+
"single_word": false,
|
| 427 |
+
"special": true
|
| 428 |
+
},
|
| 429 |
+
"156944": {
|
| 430 |
+
"content": "<|reserved_token_49|>",
|
| 431 |
+
"lstrip": false,
|
| 432 |
+
"normalized": false,
|
| 433 |
+
"rstrip": false,
|
| 434 |
+
"single_word": false,
|
| 435 |
+
"special": true
|
| 436 |
+
},
|
| 437 |
+
"156945": {
|
| 438 |
+
"content": "<|reserved_token_50|>",
|
| 439 |
+
"lstrip": false,
|
| 440 |
+
"normalized": false,
|
| 441 |
+
"rstrip": false,
|
| 442 |
+
"single_word": false,
|
| 443 |
+
"special": true
|
| 444 |
+
},
|
| 445 |
+
"156946": {
|
| 446 |
+
"content": "<|reserved_token_51|>",
|
| 447 |
+
"lstrip": false,
|
| 448 |
+
"normalized": false,
|
| 449 |
+
"rstrip": false,
|
| 450 |
+
"single_word": false,
|
| 451 |
+
"special": true
|
| 452 |
+
},
|
| 453 |
+
"156947": {
|
| 454 |
+
"content": "<|reserved_token_52|>",
|
| 455 |
+
"lstrip": false,
|
| 456 |
+
"normalized": false,
|
| 457 |
+
"rstrip": false,
|
| 458 |
+
"single_word": false,
|
| 459 |
+
"special": true
|
| 460 |
+
},
|
| 461 |
+
"156948": {
|
| 462 |
+
"content": "<|reserved_token_53|>",
|
| 463 |
+
"lstrip": false,
|
| 464 |
+
"normalized": false,
|
| 465 |
+
"rstrip": false,
|
| 466 |
+
"single_word": false,
|
| 467 |
+
"special": true
|
| 468 |
+
},
|
| 469 |
+
"156949": {
|
| 470 |
+
"content": "<|reserved_token_54|>",
|
| 471 |
+
"lstrip": false,
|
| 472 |
+
"normalized": false,
|
| 473 |
+
"rstrip": false,
|
| 474 |
+
"single_word": false,
|
| 475 |
+
"special": true
|
| 476 |
+
},
|
| 477 |
+
"156950": {
|
| 478 |
+
"content": "<|reserved_token_55|>",
|
| 479 |
+
"lstrip": false,
|
| 480 |
+
"normalized": false,
|
| 481 |
+
"rstrip": false,
|
| 482 |
+
"single_word": false,
|
| 483 |
+
"special": true
|
| 484 |
+
},
|
| 485 |
+
"156951": {
|
| 486 |
+
"content": "<|reserved_token_56|>",
|
| 487 |
+
"lstrip": false,
|
| 488 |
+
"normalized": false,
|
| 489 |
+
"rstrip": false,
|
| 490 |
+
"single_word": false,
|
| 491 |
+
"special": true
|
| 492 |
+
},
|
| 493 |
+
"156952": {
|
| 494 |
+
"content": "<|reserved_token_57|>",
|
| 495 |
+
"lstrip": false,
|
| 496 |
+
"normalized": false,
|
| 497 |
+
"rstrip": false,
|
| 498 |
+
"single_word": false,
|
| 499 |
+
"special": true
|
| 500 |
+
},
|
| 501 |
+
"156953": {
|
| 502 |
+
"content": "<|reserved_token_58|>",
|
| 503 |
+
"lstrip": false,
|
| 504 |
+
"normalized": false,
|
| 505 |
+
"rstrip": false,
|
| 506 |
+
"single_word": false,
|
| 507 |
+
"special": true
|
| 508 |
+
},
|
| 509 |
+
"156954": {
|
| 510 |
+
"content": "<|reserved_token_59|>",
|
| 511 |
+
"lstrip": false,
|
| 512 |
+
"normalized": false,
|
| 513 |
+
"rstrip": false,
|
| 514 |
+
"single_word": false,
|
| 515 |
+
"special": true
|
| 516 |
+
},
|
| 517 |
+
"156955": {
|
| 518 |
+
"content": "<|reserved_token_60|>",
|
| 519 |
+
"lstrip": false,
|
| 520 |
+
"normalized": false,
|
| 521 |
+
"rstrip": false,
|
| 522 |
+
"single_word": false,
|
| 523 |
+
"special": true
|
| 524 |
+
},
|
| 525 |
+
"156956": {
|
| 526 |
+
"content": "<|reserved_token_61|>",
|
| 527 |
+
"lstrip": false,
|
| 528 |
+
"normalized": false,
|
| 529 |
+
"rstrip": false,
|
| 530 |
+
"single_word": false,
|
| 531 |
+
"special": true
|
| 532 |
+
},
|
| 533 |
+
"156957": {
|
| 534 |
+
"content": "<|reserved_token_62|>",
|
| 535 |
+
"lstrip": false,
|
| 536 |
+
"normalized": false,
|
| 537 |
+
"rstrip": false,
|
| 538 |
+
"single_word": false,
|
| 539 |
+
"special": true
|
| 540 |
+
},
|
| 541 |
+
"156958": {
|
| 542 |
+
"content": "<|reserved_token_63|>",
|
| 543 |
+
"lstrip": false,
|
| 544 |
+
"normalized": false,
|
| 545 |
+
"rstrip": false,
|
| 546 |
+
"single_word": false,
|
| 547 |
+
"special": true
|
| 548 |
+
},
|
| 549 |
+
"156959": {
|
| 550 |
+
"content": "<|reserved_token_64|>",
|
| 551 |
+
"lstrip": false,
|
| 552 |
+
"normalized": false,
|
| 553 |
+
"rstrip": false,
|
| 554 |
+
"single_word": false,
|
| 555 |
+
"special": true
|
| 556 |
+
},
|
| 557 |
+
"156960": {
|
| 558 |
+
"content": "<|reserved_token_65|>",
|
| 559 |
+
"lstrip": false,
|
| 560 |
+
"normalized": false,
|
| 561 |
+
"rstrip": false,
|
| 562 |
+
"single_word": false,
|
| 563 |
+
"special": true
|
| 564 |
+
},
|
| 565 |
+
"156961": {
|
| 566 |
+
"content": "<|reserved_token_66|>",
|
| 567 |
+
"lstrip": false,
|
| 568 |
+
"normalized": false,
|
| 569 |
+
"rstrip": false,
|
| 570 |
+
"single_word": false,
|
| 571 |
+
"special": true
|
| 572 |
+
},
|
| 573 |
+
"156962": {
|
| 574 |
+
"content": "<|reserved_token_67|>",
|
| 575 |
+
"lstrip": false,
|
| 576 |
+
"normalized": false,
|
| 577 |
+
"rstrip": false,
|
| 578 |
+
"single_word": false,
|
| 579 |
+
"special": true
|
| 580 |
+
},
|
| 581 |
+
"156963": {
|
| 582 |
+
"content": "<|reserved_token_68|>",
|
| 583 |
+
"lstrip": false,
|
| 584 |
+
"normalized": false,
|
| 585 |
+
"rstrip": false,
|
| 586 |
+
"single_word": false,
|
| 587 |
+
"special": true
|
| 588 |
+
},
|
| 589 |
+
"156964": {
|
| 590 |
+
"content": "<|reserved_token_69|>",
|
| 591 |
+
"lstrip": false,
|
| 592 |
+
"normalized": false,
|
| 593 |
+
"rstrip": false,
|
| 594 |
+
"single_word": false,
|
| 595 |
+
"special": true
|
| 596 |
+
},
|
| 597 |
+
"156965": {
|
| 598 |
+
"content": "<|reserved_token_70|>",
|
| 599 |
+
"lstrip": false,
|
| 600 |
+
"normalized": false,
|
| 601 |
+
"rstrip": false,
|
| 602 |
+
"single_word": false,
|
| 603 |
+
"special": true
|
| 604 |
+
},
|
| 605 |
+
"156966": {
|
| 606 |
+
"content": "<|reserved_token_71|>",
|
| 607 |
+
"lstrip": false,
|
| 608 |
+
"normalized": false,
|
| 609 |
+
"rstrip": false,
|
| 610 |
+
"single_word": false,
|
| 611 |
+
"special": true
|
| 612 |
+
},
|
| 613 |
+
"156967": {
|
| 614 |
+
"content": "<|reserved_token_72|>",
|
| 615 |
+
"lstrip": false,
|
| 616 |
+
"normalized": false,
|
| 617 |
+
"rstrip": false,
|
| 618 |
+
"single_word": false,
|
| 619 |
+
"special": true
|
| 620 |
+
},
|
| 621 |
+
"156968": {
|
| 622 |
+
"content": "<|reserved_token_73|>",
|
| 623 |
+
"lstrip": false,
|
| 624 |
+
"normalized": false,
|
| 625 |
+
"rstrip": false,
|
| 626 |
+
"single_word": false,
|
| 627 |
+
"special": true
|
| 628 |
+
},
|
| 629 |
+
"156969": {
|
| 630 |
+
"content": "<|reserved_token_74|>",
|
| 631 |
+
"lstrip": false,
|
| 632 |
+
"normalized": false,
|
| 633 |
+
"rstrip": false,
|
| 634 |
+
"single_word": false,
|
| 635 |
+
"special": true
|
| 636 |
+
},
|
| 637 |
+
"156970": {
|
| 638 |
+
"content": "<|reserved_token_75|>",
|
| 639 |
+
"lstrip": false,
|
| 640 |
+
"normalized": false,
|
| 641 |
+
"rstrip": false,
|
| 642 |
+
"single_word": false,
|
| 643 |
+
"special": true
|
| 644 |
+
},
|
| 645 |
+
"156971": {
|
| 646 |
+
"content": "<|reserved_token_76|>",
|
| 647 |
+
"lstrip": false,
|
| 648 |
+
"normalized": false,
|
| 649 |
+
"rstrip": false,
|
| 650 |
+
"single_word": false,
|
| 651 |
+
"special": true
|
| 652 |
+
},
|
| 653 |
+
"156972": {
|
| 654 |
+
"content": "<|reserved_token_77|>",
|
| 655 |
+
"lstrip": false,
|
| 656 |
+
"normalized": false,
|
| 657 |
+
"rstrip": false,
|
| 658 |
+
"single_word": false,
|
| 659 |
+
"special": true
|
| 660 |
+
},
|
| 661 |
+
"156973": {
|
| 662 |
+
"content": "<|reserved_token_78|>",
|
| 663 |
+
"lstrip": false,
|
| 664 |
+
"normalized": false,
|
| 665 |
+
"rstrip": false,
|
| 666 |
+
"single_word": false,
|
| 667 |
+
"special": true
|
| 668 |
+
},
|
| 669 |
+
"156974": {
|
| 670 |
+
"content": "<|reserved_token_79|>",
|
| 671 |
+
"lstrip": false,
|
| 672 |
+
"normalized": false,
|
| 673 |
+
"rstrip": false,
|
| 674 |
+
"single_word": false,
|
| 675 |
+
"special": true
|
| 676 |
+
},
|
| 677 |
+
"156975": {
|
| 678 |
+
"content": "<|reserved_token_80|>",
|
| 679 |
+
"lstrip": false,
|
| 680 |
+
"normalized": false,
|
| 681 |
+
"rstrip": false,
|
| 682 |
+
"single_word": false,
|
| 683 |
+
"special": true
|
| 684 |
+
},
|
| 685 |
+
"156976": {
|
| 686 |
+
"content": "<|reserved_token_81|>",
|
| 687 |
+
"lstrip": false,
|
| 688 |
+
"normalized": false,
|
| 689 |
+
"rstrip": false,
|
| 690 |
+
"single_word": false,
|
| 691 |
+
"special": true
|
| 692 |
+
},
|
| 693 |
+
"156977": {
|
| 694 |
+
"content": "<|reserved_token_82|>",
|
| 695 |
+
"lstrip": false,
|
| 696 |
+
"normalized": false,
|
| 697 |
+
"rstrip": false,
|
| 698 |
+
"single_word": false,
|
| 699 |
+
"special": true
|
| 700 |
+
},
|
| 701 |
+
"156978": {
|
| 702 |
+
"content": "<|reserved_token_83|>",
|
| 703 |
+
"lstrip": false,
|
| 704 |
+
"normalized": false,
|
| 705 |
+
"rstrip": false,
|
| 706 |
+
"single_word": false,
|
| 707 |
+
"special": true
|
| 708 |
+
},
|
| 709 |
+
"156979": {
|
| 710 |
+
"content": "<|reserved_token_84|>",
|
| 711 |
+
"lstrip": false,
|
| 712 |
+
"normalized": false,
|
| 713 |
+
"rstrip": false,
|
| 714 |
+
"single_word": false,
|
| 715 |
+
"special": true
|
| 716 |
+
},
|
| 717 |
+
"156980": {
|
| 718 |
+
"content": "<|reserved_token_85|>",
|
| 719 |
+
"lstrip": false,
|
| 720 |
+
"normalized": false,
|
| 721 |
+
"rstrip": false,
|
| 722 |
+
"single_word": false,
|
| 723 |
+
"special": true
|
| 724 |
+
},
|
| 725 |
+
"156981": {
|
| 726 |
+
"content": "<|reserved_token_86|>",
|
| 727 |
+
"lstrip": false,
|
| 728 |
+
"normalized": false,
|
| 729 |
+
"rstrip": false,
|
| 730 |
+
"single_word": false,
|
| 731 |
+
"special": true
|
| 732 |
+
},
|
| 733 |
+
"156982": {
|
| 734 |
+
"content": "<|reserved_token_87|>",
|
| 735 |
+
"lstrip": false,
|
| 736 |
+
"normalized": false,
|
| 737 |
+
"rstrip": false,
|
| 738 |
+
"single_word": false,
|
| 739 |
+
"special": true
|
| 740 |
+
},
|
| 741 |
+
"156983": {
|
| 742 |
+
"content": "<|reserved_token_88|>",
|
| 743 |
+
"lstrip": false,
|
| 744 |
+
"normalized": false,
|
| 745 |
+
"rstrip": false,
|
| 746 |
+
"single_word": false,
|
| 747 |
+
"special": true
|
| 748 |
+
},
|
| 749 |
+
"156984": {
|
| 750 |
+
"content": "<|reserved_token_89|>",
|
| 751 |
+
"lstrip": false,
|
| 752 |
+
"normalized": false,
|
| 753 |
+
"rstrip": false,
|
| 754 |
+
"single_word": false,
|
| 755 |
+
"special": true
|
| 756 |
+
},
|
| 757 |
+
"156985": {
|
| 758 |
+
"content": "<|reserved_token_90|>",
|
| 759 |
+
"lstrip": false,
|
| 760 |
+
"normalized": false,
|
| 761 |
+
"rstrip": false,
|
| 762 |
+
"single_word": false,
|
| 763 |
+
"special": true
|
| 764 |
+
},
|
| 765 |
+
"156986": {
|
| 766 |
+
"content": "<|reserved_token_91|>",
|
| 767 |
+
"lstrip": false,
|
| 768 |
+
"normalized": false,
|
| 769 |
+
"rstrip": false,
|
| 770 |
+
"single_word": false,
|
| 771 |
+
"special": true
|
| 772 |
+
},
|
| 773 |
+
"156987": {
|
| 774 |
+
"content": "<|reserved_token_92|>",
|
| 775 |
+
"lstrip": false,
|
| 776 |
+
"normalized": false,
|
| 777 |
+
"rstrip": false,
|
| 778 |
+
"single_word": false,
|
| 779 |
+
"special": true
|
| 780 |
+
},
|
| 781 |
+
"156988": {
|
| 782 |
+
"content": "<|reserved_token_93|>",
|
| 783 |
+
"lstrip": false,
|
| 784 |
+
"normalized": false,
|
| 785 |
+
"rstrip": false,
|
| 786 |
+
"single_word": false,
|
| 787 |
+
"special": true
|
| 788 |
+
},
|
| 789 |
+
"156989": {
|
| 790 |
+
"content": "<|reserved_token_94|>",
|
| 791 |
+
"lstrip": false,
|
| 792 |
+
"normalized": false,
|
| 793 |
+
"rstrip": false,
|
| 794 |
+
"single_word": false,
|
| 795 |
+
"special": true
|
| 796 |
+
},
|
| 797 |
+
"156990": {
|
| 798 |
+
"content": "<|reserved_token_95|>",
|
| 799 |
+
"lstrip": false,
|
| 800 |
+
"normalized": false,
|
| 801 |
+
"rstrip": false,
|
| 802 |
+
"single_word": false,
|
| 803 |
+
"special": true
|
| 804 |
+
},
|
| 805 |
+
"156991": {
|
| 806 |
+
"content": "<|reserved_token_96|>",
|
| 807 |
+
"lstrip": false,
|
| 808 |
+
"normalized": false,
|
| 809 |
+
"rstrip": false,
|
| 810 |
+
"single_word": false,
|
| 811 |
+
"special": true
|
| 812 |
+
},
|
| 813 |
+
"156992": {
|
| 814 |
+
"content": "<|reserved_token_97|>",
|
| 815 |
+
"lstrip": false,
|
| 816 |
+
"normalized": false,
|
| 817 |
+
"rstrip": false,
|
| 818 |
+
"single_word": false,
|
| 819 |
+
"special": true
|
| 820 |
+
},
|
| 821 |
+
"156993": {
|
| 822 |
+
"content": "<|reserved_token_98|>",
|
| 823 |
+
"lstrip": false,
|
| 824 |
+
"normalized": false,
|
| 825 |
+
"rstrip": false,
|
| 826 |
+
"single_word": false,
|
| 827 |
+
"special": true
|
| 828 |
+
},
|
| 829 |
+
"156994": {
|
| 830 |
+
"content": "<|reserved_token_99|>",
|
| 831 |
+
"lstrip": false,
|
| 832 |
+
"normalized": false,
|
| 833 |
+
"rstrip": false,
|
| 834 |
+
"single_word": false,
|
| 835 |
+
"special": true
|
| 836 |
+
},
|
| 837 |
+
"156995": {
|
| 838 |
+
"content": "<|reserved_token_100|>",
|
| 839 |
+
"lstrip": false,
|
| 840 |
+
"normalized": false,
|
| 841 |
+
"rstrip": false,
|
| 842 |
+
"single_word": false,
|
| 843 |
+
"special": true
|
| 844 |
+
},
|
| 845 |
+
"156996": {
|
| 846 |
+
"content": "<|reserved_token_101|>",
|
| 847 |
+
"lstrip": false,
|
| 848 |
+
"normalized": false,
|
| 849 |
+
"rstrip": false,
|
| 850 |
+
"single_word": false,
|
| 851 |
+
"special": true
|
| 852 |
+
},
|
| 853 |
+
"156997": {
|
| 854 |
+
"content": "<|reserved_token_102|>",
|
| 855 |
+
"lstrip": false,
|
| 856 |
+
"normalized": false,
|
| 857 |
+
"rstrip": false,
|
| 858 |
+
"single_word": false,
|
| 859 |
+
"special": true
|
| 860 |
+
},
|
| 861 |
+
"156998": {
|
| 862 |
+
"content": "<|reserved_token_103|>",
|
| 863 |
+
"lstrip": false,
|
| 864 |
+
"normalized": false,
|
| 865 |
+
"rstrip": false,
|
| 866 |
+
"single_word": false,
|
| 867 |
+
"special": true
|
| 868 |
+
},
|
| 869 |
+
"156999": {
|
| 870 |
+
"content": "<|reserved_token_104|>",
|
| 871 |
+
"lstrip": false,
|
| 872 |
+
"normalized": false,
|
| 873 |
+
"rstrip": false,
|
| 874 |
+
"single_word": false,
|
| 875 |
+
"special": true
|
| 876 |
+
},
|
| 877 |
+
"157000": {
|
| 878 |
+
"content": "<|reserved_token_105|>",
|
| 879 |
+
"lstrip": false,
|
| 880 |
+
"normalized": false,
|
| 881 |
+
"rstrip": false,
|
| 882 |
+
"single_word": false,
|
| 883 |
+
"special": true
|
| 884 |
+
},
|
| 885 |
+
"157001": {
|
| 886 |
+
"content": "<|reserved_token_106|>",
|
| 887 |
+
"lstrip": false,
|
| 888 |
+
"normalized": false,
|
| 889 |
+
"rstrip": false,
|
| 890 |
+
"single_word": false,
|
| 891 |
+
"special": true
|
| 892 |
+
},
|
| 893 |
+
"157002": {
|
| 894 |
+
"content": "<|reserved_token_107|>",
|
| 895 |
+
"lstrip": false,
|
| 896 |
+
"normalized": false,
|
| 897 |
+
"rstrip": false,
|
| 898 |
+
"single_word": false,
|
| 899 |
+
"special": true
|
| 900 |
+
},
|
| 901 |
+
"157003": {
|
| 902 |
+
"content": "<|reserved_token_108|>",
|
| 903 |
+
"lstrip": false,
|
| 904 |
+
"normalized": false,
|
| 905 |
+
"rstrip": false,
|
| 906 |
+
"single_word": false,
|
| 907 |
+
"special": true
|
| 908 |
+
},
|
| 909 |
+
"157004": {
|
| 910 |
+
"content": "<|reserved_token_109|>",
|
| 911 |
+
"lstrip": false,
|
| 912 |
+
"normalized": false,
|
| 913 |
+
"rstrip": false,
|
| 914 |
+
"single_word": false,
|
| 915 |
+
"special": true
|
| 916 |
+
},
|
| 917 |
+
"157005": {
|
| 918 |
+
"content": "<|reserved_token_110|>",
|
| 919 |
+
"lstrip": false,
|
| 920 |
+
"normalized": false,
|
| 921 |
+
"rstrip": false,
|
| 922 |
+
"single_word": false,
|
| 923 |
+
"special": true
|
| 924 |
+
},
|
| 925 |
+
"157006": {
|
| 926 |
+
"content": "<|reserved_token_111|>",
|
| 927 |
+
"lstrip": false,
|
| 928 |
+
"normalized": false,
|
| 929 |
+
"rstrip": false,
|
| 930 |
+
"single_word": false,
|
| 931 |
+
"special": true
|
| 932 |
+
},
|
| 933 |
+
"157007": {
|
| 934 |
+
"content": "<|reserved_token_112|>",
|
| 935 |
+
"lstrip": false,
|
| 936 |
+
"normalized": false,
|
| 937 |
+
"rstrip": false,
|
| 938 |
+
"single_word": false,
|
| 939 |
+
"special": true
|
| 940 |
+
},
|
| 941 |
+
"157008": {
|
| 942 |
+
"content": "<|reserved_token_113|>",
|
| 943 |
+
"lstrip": false,
|
| 944 |
+
"normalized": false,
|
| 945 |
+
"rstrip": false,
|
| 946 |
+
"single_word": false,
|
| 947 |
+
"special": true
|
| 948 |
+
},
|
| 949 |
+
"157009": {
|
| 950 |
+
"content": "<|reserved_token_114|>",
|
| 951 |
+
"lstrip": false,
|
| 952 |
+
"normalized": false,
|
| 953 |
+
"rstrip": false,
|
| 954 |
+
"single_word": false,
|
| 955 |
+
"special": true
|
| 956 |
+
},
|
| 957 |
+
"157010": {
|
| 958 |
+
"content": "<|reserved_token_115|>",
|
| 959 |
+
"lstrip": false,
|
| 960 |
+
"normalized": false,
|
| 961 |
+
"rstrip": false,
|
| 962 |
+
"single_word": false,
|
| 963 |
+
"special": true
|
| 964 |
+
},
|
| 965 |
+
"157011": {
|
| 966 |
+
"content": "<|reserved_token_116|>",
|
| 967 |
+
"lstrip": false,
|
| 968 |
+
"normalized": false,
|
| 969 |
+
"rstrip": false,
|
| 970 |
+
"single_word": false,
|
| 971 |
+
"special": true
|
| 972 |
+
},
|
| 973 |
+
"157012": {
|
| 974 |
+
"content": "<|reserved_token_117|>",
|
| 975 |
+
"lstrip": false,
|
| 976 |
+
"normalized": false,
|
| 977 |
+
"rstrip": false,
|
| 978 |
+
"single_word": false,
|
| 979 |
+
"special": true
|
| 980 |
+
},
|
| 981 |
+
"157013": {
|
| 982 |
+
"content": "<|reserved_token_118|>",
|
| 983 |
+
"lstrip": false,
|
| 984 |
+
"normalized": false,
|
| 985 |
+
"rstrip": false,
|
| 986 |
+
"single_word": false,
|
| 987 |
+
"special": true
|
| 988 |
+
},
|
| 989 |
+
"157014": {
|
| 990 |
+
"content": "<|reserved_token_119|>",
|
| 991 |
+
"lstrip": false,
|
| 992 |
+
"normalized": false,
|
| 993 |
+
"rstrip": false,
|
| 994 |
+
"single_word": false,
|
| 995 |
+
"special": true
|
| 996 |
+
},
|
| 997 |
+
"157015": {
|
| 998 |
+
"content": "<|reserved_token_120|>",
|
| 999 |
+
"lstrip": false,
|
| 1000 |
+
"normalized": false,
|
| 1001 |
+
"rstrip": false,
|
| 1002 |
+
"single_word": false,
|
| 1003 |
+
"special": true
|
| 1004 |
+
},
|
| 1005 |
+
"157016": {
|
| 1006 |
+
"content": "<|reserved_token_121|>",
|
| 1007 |
+
"lstrip": false,
|
| 1008 |
+
"normalized": false,
|
| 1009 |
+
"rstrip": false,
|
| 1010 |
+
"single_word": false,
|
| 1011 |
+
"special": true
|
| 1012 |
+
},
|
| 1013 |
+
"157017": {
|
| 1014 |
+
"content": "<|reserved_token_122|>",
|
| 1015 |
+
"lstrip": false,
|
| 1016 |
+
"normalized": false,
|
| 1017 |
+
"rstrip": false,
|
| 1018 |
+
"single_word": false,
|
| 1019 |
+
"special": true
|
| 1020 |
+
},
|
| 1021 |
+
"157018": {
|
| 1022 |
+
"content": "<|reserved_token_123|>",
|
| 1023 |
+
"lstrip": false,
|
| 1024 |
+
"normalized": false,
|
| 1025 |
+
"rstrip": false,
|
| 1026 |
+
"single_word": false,
|
| 1027 |
+
"special": true
|
| 1028 |
+
},
|
| 1029 |
+
"157019": {
|
| 1030 |
+
"content": "<|reserved_token_124|>",
|
| 1031 |
+
"lstrip": false,
|
| 1032 |
+
"normalized": false,
|
| 1033 |
+
"rstrip": false,
|
| 1034 |
+
"single_word": false,
|
| 1035 |
+
"special": true
|
| 1036 |
+
},
|
| 1037 |
+
"157020": {
|
| 1038 |
+
"content": "<|reserved_token_125|>",
|
| 1039 |
+
"lstrip": false,
|
| 1040 |
+
"normalized": false,
|
| 1041 |
+
"rstrip": false,
|
| 1042 |
+
"single_word": false,
|
| 1043 |
+
"special": true
|
| 1044 |
+
},
|
| 1045 |
+
"157021": {
|
| 1046 |
+
"content": "<|reserved_token_126|>",
|
| 1047 |
+
"lstrip": false,
|
| 1048 |
+
"normalized": false,
|
| 1049 |
+
"rstrip": false,
|
| 1050 |
+
"single_word": false,
|
| 1051 |
+
"special": true
|
| 1052 |
+
},
|
| 1053 |
+
"157022": {
|
| 1054 |
+
"content": "<|reserved_token_127|>",
|
| 1055 |
+
"lstrip": false,
|
| 1056 |
+
"normalized": false,
|
| 1057 |
+
"rstrip": false,
|
| 1058 |
+
"single_word": false,
|
| 1059 |
+
"special": true
|
| 1060 |
+
},
|
| 1061 |
+
"157023": {
|
| 1062 |
+
"content": "<|reserved_token_128|>",
|
| 1063 |
+
"lstrip": false,
|
| 1064 |
+
"normalized": false,
|
| 1065 |
+
"rstrip": false,
|
| 1066 |
+
"single_word": false,
|
| 1067 |
+
"special": true
|
| 1068 |
+
},
|
| 1069 |
+
"157024": {
|
| 1070 |
+
"content": "<|reserved_token_129|>",
|
| 1071 |
+
"lstrip": false,
|
| 1072 |
+
"normalized": false,
|
| 1073 |
+
"rstrip": false,
|
| 1074 |
+
"single_word": false,
|
| 1075 |
+
"special": true
|
| 1076 |
+
},
|
| 1077 |
+
"157025": {
|
| 1078 |
+
"content": "<|reserved_token_130|>",
|
| 1079 |
+
"lstrip": false,
|
| 1080 |
+
"normalized": false,
|
| 1081 |
+
"rstrip": false,
|
| 1082 |
+
"single_word": false,
|
| 1083 |
+
"special": true
|
| 1084 |
+
},
|
| 1085 |
+
"157026": {
|
| 1086 |
+
"content": "<|reserved_token_131|>",
|
| 1087 |
+
"lstrip": false,
|
| 1088 |
+
"normalized": false,
|
| 1089 |
+
"rstrip": false,
|
| 1090 |
+
"single_word": false,
|
| 1091 |
+
"special": true
|
| 1092 |
+
},
|
| 1093 |
+
"157027": {
|
| 1094 |
+
"content": "<|reserved_token_132|>",
|
| 1095 |
+
"lstrip": false,
|
| 1096 |
+
"normalized": false,
|
| 1097 |
+
"rstrip": false,
|
| 1098 |
+
"single_word": false,
|
| 1099 |
+
"special": true
|
| 1100 |
+
},
|
| 1101 |
+
"157028": {
|
| 1102 |
+
"content": "<|reserved_token_133|>",
|
| 1103 |
+
"lstrip": false,
|
| 1104 |
+
"normalized": false,
|
| 1105 |
+
"rstrip": false,
|
| 1106 |
+
"single_word": false,
|
| 1107 |
+
"special": true
|
| 1108 |
+
},
|
| 1109 |
+
"157029": {
|
| 1110 |
+
"content": "<|reserved_token_134|>",
|
| 1111 |
+
"lstrip": false,
|
| 1112 |
+
"normalized": false,
|
| 1113 |
+
"rstrip": false,
|
| 1114 |
+
"single_word": false,
|
| 1115 |
+
"special": true
|
| 1116 |
+
},
|
| 1117 |
+
"157030": {
|
| 1118 |
+
"content": "<|reserved_token_135|>",
|
| 1119 |
+
"lstrip": false,
|
| 1120 |
+
"normalized": false,
|
| 1121 |
+
"rstrip": false,
|
| 1122 |
+
"single_word": false,
|
| 1123 |
+
"special": true
|
| 1124 |
+
},
|
| 1125 |
+
"157031": {
|
| 1126 |
+
"content": "<|reserved_token_136|>",
|
| 1127 |
+
"lstrip": false,
|
| 1128 |
+
"normalized": false,
|
| 1129 |
+
"rstrip": false,
|
| 1130 |
+
"single_word": false,
|
| 1131 |
+
"special": true
|
| 1132 |
+
},
|
| 1133 |
+
"157032": {
|
| 1134 |
+
"content": "<|reserved_token_137|>",
|
| 1135 |
+
"lstrip": false,
|
| 1136 |
+
"normalized": false,
|
| 1137 |
+
"rstrip": false,
|
| 1138 |
+
"single_word": false,
|
| 1139 |
+
"special": true
|
| 1140 |
+
},
|
| 1141 |
+
"157033": {
|
| 1142 |
+
"content": "<|reserved_token_138|>",
|
| 1143 |
+
"lstrip": false,
|
| 1144 |
+
"normalized": false,
|
| 1145 |
+
"rstrip": false,
|
| 1146 |
+
"single_word": false,
|
| 1147 |
+
"special": true
|
| 1148 |
+
},
|
| 1149 |
+
"157034": {
|
| 1150 |
+
"content": "<|reserved_token_139|>",
|
| 1151 |
+
"lstrip": false,
|
| 1152 |
+
"normalized": false,
|
| 1153 |
+
"rstrip": false,
|
| 1154 |
+
"single_word": false,
|
| 1155 |
+
"special": true
|
| 1156 |
+
},
|
| 1157 |
+
"157035": {
|
| 1158 |
+
"content": "<|reserved_token_140|>",
|
| 1159 |
+
"lstrip": false,
|
| 1160 |
+
"normalized": false,
|
| 1161 |
+
"rstrip": false,
|
| 1162 |
+
"single_word": false,
|
| 1163 |
+
"special": true
|
| 1164 |
+
},
|
| 1165 |
+
"157036": {
|
| 1166 |
+
"content": "<|reserved_token_141|>",
|
| 1167 |
+
"lstrip": false,
|
| 1168 |
+
"normalized": false,
|
| 1169 |
+
"rstrip": false,
|
| 1170 |
+
"single_word": false,
|
| 1171 |
+
"special": true
|
| 1172 |
+
},
|
| 1173 |
+
"157037": {
|
| 1174 |
+
"content": "<|reserved_token_142|>",
|
| 1175 |
+
"lstrip": false,
|
| 1176 |
+
"normalized": false,
|
| 1177 |
+
"rstrip": false,
|
| 1178 |
+
"single_word": false,
|
| 1179 |
+
"special": true
|
| 1180 |
+
},
|
| 1181 |
+
"157038": {
|
| 1182 |
+
"content": "<|reserved_token_143|>",
|
| 1183 |
+
"lstrip": false,
|
| 1184 |
+
"normalized": false,
|
| 1185 |
+
"rstrip": false,
|
| 1186 |
+
"single_word": false,
|
| 1187 |
+
"special": true
|
| 1188 |
+
},
|
| 1189 |
+
"157039": {
|
| 1190 |
+
"content": "<|reserved_token_144|>",
|
| 1191 |
+
"lstrip": false,
|
| 1192 |
+
"normalized": false,
|
| 1193 |
+
"rstrip": false,
|
| 1194 |
+
"single_word": false,
|
| 1195 |
+
"special": true
|
| 1196 |
+
},
|
| 1197 |
+
"157040": {
|
| 1198 |
+
"content": "<|reserved_token_145|>",
|
| 1199 |
+
"lstrip": false,
|
| 1200 |
+
"normalized": false,
|
| 1201 |
+
"rstrip": false,
|
| 1202 |
+
"single_word": false,
|
| 1203 |
+
"special": true
|
| 1204 |
+
},
|
| 1205 |
+
"157041": {
|
| 1206 |
+
"content": "<|reserved_token_146|>",
|
| 1207 |
+
"lstrip": false,
|
| 1208 |
+
"normalized": false,
|
| 1209 |
+
"rstrip": false,
|
| 1210 |
+
"single_word": false,
|
| 1211 |
+
"special": true
|
| 1212 |
+
},
|
| 1213 |
+
"157042": {
|
| 1214 |
+
"content": "<|reserved_token_147|>",
|
| 1215 |
+
"lstrip": false,
|
| 1216 |
+
"normalized": false,
|
| 1217 |
+
"rstrip": false,
|
| 1218 |
+
"single_word": false,
|
| 1219 |
+
"special": true
|
| 1220 |
+
},
|
| 1221 |
+
"157043": {
|
| 1222 |
+
"content": "<|reserved_token_148|>",
|
| 1223 |
+
"lstrip": false,
|
| 1224 |
+
"normalized": false,
|
| 1225 |
+
"rstrip": false,
|
| 1226 |
+
"single_word": false,
|
| 1227 |
+
"special": true
|
| 1228 |
+
},
|
| 1229 |
+
"157044": {
|
| 1230 |
+
"content": "<|reserved_token_149|>",
|
| 1231 |
+
"lstrip": false,
|
| 1232 |
+
"normalized": false,
|
| 1233 |
+
"rstrip": false,
|
| 1234 |
+
"single_word": false,
|
| 1235 |
+
"special": true
|
| 1236 |
+
},
|
| 1237 |
+
"157045": {
|
| 1238 |
+
"content": "<|reserved_token_150|>",
|
| 1239 |
+
"lstrip": false,
|
| 1240 |
+
"normalized": false,
|
| 1241 |
+
"rstrip": false,
|
| 1242 |
+
"single_word": false,
|
| 1243 |
+
"special": true
|
| 1244 |
+
},
|
| 1245 |
+
"157046": {
|
| 1246 |
+
"content": "<|reserved_token_151|>",
|
| 1247 |
+
"lstrip": false,
|
| 1248 |
+
"normalized": false,
|
| 1249 |
+
"rstrip": false,
|
| 1250 |
+
"single_word": false,
|
| 1251 |
+
"special": true
|
| 1252 |
+
},
|
| 1253 |
+
"157047": {
|
| 1254 |
+
"content": "<|reserved_token_152|>",
|
| 1255 |
+
"lstrip": false,
|
| 1256 |
+
"normalized": false,
|
| 1257 |
+
"rstrip": false,
|
| 1258 |
+
"single_word": false,
|
| 1259 |
+
"special": true
|
| 1260 |
+
},
|
| 1261 |
+
"157048": {
|
| 1262 |
+
"content": "<|reserved_token_153|>",
|
| 1263 |
+
"lstrip": false,
|
| 1264 |
+
"normalized": false,
|
| 1265 |
+
"rstrip": false,
|
| 1266 |
+
"single_word": false,
|
| 1267 |
+
"special": true
|
| 1268 |
+
},
|
| 1269 |
+
"157049": {
|
| 1270 |
+
"content": "<|reserved_token_154|>",
|
| 1271 |
+
"lstrip": false,
|
| 1272 |
+
"normalized": false,
|
| 1273 |
+
"rstrip": false,
|
| 1274 |
+
"single_word": false,
|
| 1275 |
+
"special": true
|
| 1276 |
+
},
|
| 1277 |
+
"157050": {
|
| 1278 |
+
"content": "<|reserved_token_155|>",
|
| 1279 |
+
"lstrip": false,
|
| 1280 |
+
"normalized": false,
|
| 1281 |
+
"rstrip": false,
|
| 1282 |
+
"single_word": false,
|
| 1283 |
+
"special": true
|
| 1284 |
+
},
|
| 1285 |
+
"157051": {
|
| 1286 |
+
"content": "<|reserved_token_156|>",
|
| 1287 |
+
"lstrip": false,
|
| 1288 |
+
"normalized": false,
|
| 1289 |
+
"rstrip": false,
|
| 1290 |
+
"single_word": false,
|
| 1291 |
+
"special": true
|
| 1292 |
+
},
|
| 1293 |
+
"157052": {
|
| 1294 |
+
"content": "<|reserved_token_157|>",
|
| 1295 |
+
"lstrip": false,
|
| 1296 |
+
"normalized": false,
|
| 1297 |
+
"rstrip": false,
|
| 1298 |
+
"single_word": false,
|
| 1299 |
+
"special": true
|
| 1300 |
+
},
|
| 1301 |
+
"157053": {
|
| 1302 |
+
"content": "<|reserved_token_158|>",
|
| 1303 |
+
"lstrip": false,
|
| 1304 |
+
"normalized": false,
|
| 1305 |
+
"rstrip": false,
|
| 1306 |
+
"single_word": false,
|
| 1307 |
+
"special": true
|
| 1308 |
+
},
|
| 1309 |
+
"157054": {
|
| 1310 |
+
"content": "<|reserved_token_159|>",
|
| 1311 |
+
"lstrip": false,
|
| 1312 |
+
"normalized": false,
|
| 1313 |
+
"rstrip": false,
|
| 1314 |
+
"single_word": false,
|
| 1315 |
+
"special": true
|
| 1316 |
+
},
|
| 1317 |
+
"157055": {
|
| 1318 |
+
"content": "<|reserved_token_160|>",
|
| 1319 |
+
"lstrip": false,
|
| 1320 |
+
"normalized": false,
|
| 1321 |
+
"rstrip": false,
|
| 1322 |
+
"single_word": false,
|
| 1323 |
+
"special": true
|
| 1324 |
+
},
|
| 1325 |
+
"157056": {
|
| 1326 |
+
"content": "<|reserved_token_161|>",
|
| 1327 |
+
"lstrip": false,
|
| 1328 |
+
"normalized": false,
|
| 1329 |
+
"rstrip": false,
|
| 1330 |
+
"single_word": false,
|
| 1331 |
+
"special": true
|
| 1332 |
+
},
|
| 1333 |
+
"157057": {
|
| 1334 |
+
"content": "<|reserved_token_162|>",
|
| 1335 |
+
"lstrip": false,
|
| 1336 |
+
"normalized": false,
|
| 1337 |
+
"rstrip": false,
|
| 1338 |
+
"single_word": false,
|
| 1339 |
+
"special": true
|
| 1340 |
+
},
|
| 1341 |
+
"157058": {
|
| 1342 |
+
"content": "<|reserved_token_163|>",
|
| 1343 |
+
"lstrip": false,
|
| 1344 |
+
"normalized": false,
|
| 1345 |
+
"rstrip": false,
|
| 1346 |
+
"single_word": false,
|
| 1347 |
+
"special": true
|
| 1348 |
+
},
|
| 1349 |
+
"157059": {
|
| 1350 |
+
"content": "<|reserved_token_164|>",
|
| 1351 |
+
"lstrip": false,
|
| 1352 |
+
"normalized": false,
|
| 1353 |
+
"rstrip": false,
|
| 1354 |
+
"single_word": false,
|
| 1355 |
+
"special": true
|
| 1356 |
+
},
|
| 1357 |
+
"157060": {
|
| 1358 |
+
"content": "<|reserved_token_165|>",
|
| 1359 |
+
"lstrip": false,
|
| 1360 |
+
"normalized": false,
|
| 1361 |
+
"rstrip": false,
|
| 1362 |
+
"single_word": false,
|
| 1363 |
+
"special": true
|
| 1364 |
+
},
|
| 1365 |
+
"157061": {
|
| 1366 |
+
"content": "<|reserved_token_166|>",
|
| 1367 |
+
"lstrip": false,
|
| 1368 |
+
"normalized": false,
|
| 1369 |
+
"rstrip": false,
|
| 1370 |
+
"single_word": false,
|
| 1371 |
+
"special": true
|
| 1372 |
+
},
|
| 1373 |
+
"157062": {
|
| 1374 |
+
"content": "<|reserved_token_167|>",
|
| 1375 |
+
"lstrip": false,
|
| 1376 |
+
"normalized": false,
|
| 1377 |
+
"rstrip": false,
|
| 1378 |
+
"single_word": false,
|
| 1379 |
+
"special": true
|
| 1380 |
+
},
|
| 1381 |
+
"157063": {
|
| 1382 |
+
"content": "<|reserved_token_168|>",
|
| 1383 |
+
"lstrip": false,
|
| 1384 |
+
"normalized": false,
|
| 1385 |
+
"rstrip": false,
|
| 1386 |
+
"single_word": false,
|
| 1387 |
+
"special": true
|
| 1388 |
+
},
|
| 1389 |
+
"157064": {
|
| 1390 |
+
"content": "<|reserved_token_169|>",
|
| 1391 |
+
"lstrip": false,
|
| 1392 |
+
"normalized": false,
|
| 1393 |
+
"rstrip": false,
|
| 1394 |
+
"single_word": false,
|
| 1395 |
+
"special": true
|
| 1396 |
+
},
|
| 1397 |
+
"157065": {
|
| 1398 |
+
"content": "<|reserved_token_170|>",
|
| 1399 |
+
"lstrip": false,
|
| 1400 |
+
"normalized": false,
|
| 1401 |
+
"rstrip": false,
|
| 1402 |
+
"single_word": false,
|
| 1403 |
+
"special": true
|
| 1404 |
+
},
|
| 1405 |
+
"157066": {
|
| 1406 |
+
"content": "<|reserved_token_171|>",
|
| 1407 |
+
"lstrip": false,
|
| 1408 |
+
"normalized": false,
|
| 1409 |
+
"rstrip": false,
|
| 1410 |
+
"single_word": false,
|
| 1411 |
+
"special": true
|
| 1412 |
+
},
|
| 1413 |
+
"157067": {
|
| 1414 |
+
"content": "<|reserved_token_172|>",
|
| 1415 |
+
"lstrip": false,
|
| 1416 |
+
"normalized": false,
|
| 1417 |
+
"rstrip": false,
|
| 1418 |
+
"single_word": false,
|
| 1419 |
+
"special": true
|
| 1420 |
+
},
|
| 1421 |
+
"157068": {
|
| 1422 |
+
"content": "<|reserved_token_173|>",
|
| 1423 |
+
"lstrip": false,
|
| 1424 |
+
"normalized": false,
|
| 1425 |
+
"rstrip": false,
|
| 1426 |
+
"single_word": false,
|
| 1427 |
+
"special": true
|
| 1428 |
+
},
|
| 1429 |
+
"157069": {
|
| 1430 |
+
"content": "<|reserved_token_174|>",
|
| 1431 |
+
"lstrip": false,
|
| 1432 |
+
"normalized": false,
|
| 1433 |
+
"rstrip": false,
|
| 1434 |
+
"single_word": false,
|
| 1435 |
+
"special": true
|
| 1436 |
+
},
|
| 1437 |
+
"157070": {
|
| 1438 |
+
"content": "<|reserved_token_175|>",
|
| 1439 |
+
"lstrip": false,
|
| 1440 |
+
"normalized": false,
|
| 1441 |
+
"rstrip": false,
|
| 1442 |
+
"single_word": false,
|
| 1443 |
+
"special": true
|
| 1444 |
+
},
|
| 1445 |
+
"157071": {
|
| 1446 |
+
"content": "<|reserved_token_176|>",
|
| 1447 |
+
"lstrip": false,
|
| 1448 |
+
"normalized": false,
|
| 1449 |
+
"rstrip": false,
|
| 1450 |
+
"single_word": false,
|
| 1451 |
+
"special": true
|
| 1452 |
+
},
|
| 1453 |
+
"157072": {
|
| 1454 |
+
"content": "<|reserved_token_177|>",
|
| 1455 |
+
"lstrip": false,
|
| 1456 |
+
"normalized": false,
|
| 1457 |
+
"rstrip": false,
|
| 1458 |
+
"single_word": false,
|
| 1459 |
+
"special": true
|
| 1460 |
+
},
|
| 1461 |
+
"157073": {
|
| 1462 |
+
"content": "<|reserved_token_178|>",
|
| 1463 |
+
"lstrip": false,
|
| 1464 |
+
"normalized": false,
|
| 1465 |
+
"rstrip": false,
|
| 1466 |
+
"single_word": false,
|
| 1467 |
+
"special": true
|
| 1468 |
+
},
|
| 1469 |
+
"157074": {
|
| 1470 |
+
"content": "<|reserved_token_179|>",
|
| 1471 |
+
"lstrip": false,
|
| 1472 |
+
"normalized": false,
|
| 1473 |
+
"rstrip": false,
|
| 1474 |
+
"single_word": false,
|
| 1475 |
+
"special": true
|
| 1476 |
+
},
|
| 1477 |
+
"157075": {
|
| 1478 |
+
"content": "<|reserved_token_180|>",
|
| 1479 |
+
"lstrip": false,
|
| 1480 |
+
"normalized": false,
|
| 1481 |
+
"rstrip": false,
|
| 1482 |
+
"single_word": false,
|
| 1483 |
+
"special": true
|
| 1484 |
+
},
|
| 1485 |
+
"157076": {
|
| 1486 |
+
"content": "<|reserved_token_181|>",
|
| 1487 |
+
"lstrip": false,
|
| 1488 |
+
"normalized": false,
|
| 1489 |
+
"rstrip": false,
|
| 1490 |
+
"single_word": false,
|
| 1491 |
+
"special": true
|
| 1492 |
+
},
|
| 1493 |
+
"157077": {
|
| 1494 |
+
"content": "<|reserved_token_182|>",
|
| 1495 |
+
"lstrip": false,
|
| 1496 |
+
"normalized": false,
|
| 1497 |
+
"rstrip": false,
|
| 1498 |
+
"single_word": false,
|
| 1499 |
+
"special": true
|
| 1500 |
+
},
|
| 1501 |
+
"157078": {
|
| 1502 |
+
"content": "<|reserved_token_183|>",
|
| 1503 |
+
"lstrip": false,
|
| 1504 |
+
"normalized": false,
|
| 1505 |
+
"rstrip": false,
|
| 1506 |
+
"single_word": false,
|
| 1507 |
+
"special": true
|
| 1508 |
+
},
|
| 1509 |
+
"157079": {
|
| 1510 |
+
"content": "<|reserved_token_184|>",
|
| 1511 |
+
"lstrip": false,
|
| 1512 |
+
"normalized": false,
|
| 1513 |
+
"rstrip": false,
|
| 1514 |
+
"single_word": false,
|
| 1515 |
+
"special": true
|
| 1516 |
+
},
|
| 1517 |
+
"157080": {
|
| 1518 |
+
"content": "<|reserved_token_185|>",
|
| 1519 |
+
"lstrip": false,
|
| 1520 |
+
"normalized": false,
|
| 1521 |
+
"rstrip": false,
|
| 1522 |
+
"single_word": false,
|
| 1523 |
+
"special": true
|
| 1524 |
+
},
|
| 1525 |
+
"157081": {
|
| 1526 |
+
"content": "<|reserved_token_186|>",
|
| 1527 |
+
"lstrip": false,
|
| 1528 |
+
"normalized": false,
|
| 1529 |
+
"rstrip": false,
|
| 1530 |
+
"single_word": false,
|
| 1531 |
+
"special": true
|
| 1532 |
+
},
|
| 1533 |
+
"157082": {
|
| 1534 |
+
"content": "<|reserved_token_187|>",
|
| 1535 |
+
"lstrip": false,
|
| 1536 |
+
"normalized": false,
|
| 1537 |
+
"rstrip": false,
|
| 1538 |
+
"single_word": false,
|
| 1539 |
+
"special": true
|
| 1540 |
+
},
|
| 1541 |
+
"157083": {
|
| 1542 |
+
"content": "<|reserved_token_188|>",
|
| 1543 |
+
"lstrip": false,
|
| 1544 |
+
"normalized": false,
|
| 1545 |
+
"rstrip": false,
|
| 1546 |
+
"single_word": false,
|
| 1547 |
+
"special": true
|
| 1548 |
+
},
|
| 1549 |
+
"157084": {
|
| 1550 |
+
"content": "<|reserved_token_189|>",
|
| 1551 |
+
"lstrip": false,
|
| 1552 |
+
"normalized": false,
|
| 1553 |
+
"rstrip": false,
|
| 1554 |
+
"single_word": false,
|
| 1555 |
+
"special": true
|
| 1556 |
+
},
|
| 1557 |
+
"157085": {
|
| 1558 |
+
"content": "<|reserved_token_190|>",
|
| 1559 |
+
"lstrip": false,
|
| 1560 |
+
"normalized": false,
|
| 1561 |
+
"rstrip": false,
|
| 1562 |
+
"single_word": false,
|
| 1563 |
+
"special": true
|
| 1564 |
+
},
|
| 1565 |
+
"157086": {
|
| 1566 |
+
"content": "<|reserved_token_191|>",
|
| 1567 |
+
"lstrip": false,
|
| 1568 |
+
"normalized": false,
|
| 1569 |
+
"rstrip": false,
|
| 1570 |
+
"single_word": false,
|
| 1571 |
+
"special": true
|
| 1572 |
+
},
|
| 1573 |
+
"157087": {
|
| 1574 |
+
"content": "<|reserved_token_192|>",
|
| 1575 |
+
"lstrip": false,
|
| 1576 |
+
"normalized": false,
|
| 1577 |
+
"rstrip": false,
|
| 1578 |
+
"single_word": false,
|
| 1579 |
+
"special": true
|
| 1580 |
+
},
|
| 1581 |
+
"157088": {
|
| 1582 |
+
"content": "<|reserved_token_193|>",
|
| 1583 |
+
"lstrip": false,
|
| 1584 |
+
"normalized": false,
|
| 1585 |
+
"rstrip": false,
|
| 1586 |
+
"single_word": false,
|
| 1587 |
+
"special": true
|
| 1588 |
+
},
|
| 1589 |
+
"157089": {
|
| 1590 |
+
"content": "<|reserved_token_194|>",
|
| 1591 |
+
"lstrip": false,
|
| 1592 |
+
"normalized": false,
|
| 1593 |
+
"rstrip": false,
|
| 1594 |
+
"single_word": false,
|
| 1595 |
+
"special": true
|
| 1596 |
+
},
|
| 1597 |
+
"157090": {
|
| 1598 |
+
"content": "<|reserved_token_195|>",
|
| 1599 |
+
"lstrip": false,
|
| 1600 |
+
"normalized": false,
|
| 1601 |
+
"rstrip": false,
|
| 1602 |
+
"single_word": false,
|
| 1603 |
+
"special": true
|
| 1604 |
+
},
|
| 1605 |
+
"157091": {
|
| 1606 |
+
"content": "<|reserved_token_196|>",
|
| 1607 |
+
"lstrip": false,
|
| 1608 |
+
"normalized": false,
|
| 1609 |
+
"rstrip": false,
|
| 1610 |
+
"single_word": false,
|
| 1611 |
+
"special": true
|
| 1612 |
+
},
|
| 1613 |
+
"157092": {
|
| 1614 |
+
"content": "<|reserved_token_197|>",
|
| 1615 |
+
"lstrip": false,
|
| 1616 |
+
"normalized": false,
|
| 1617 |
+
"rstrip": false,
|
| 1618 |
+
"single_word": false,
|
| 1619 |
+
"special": true
|
| 1620 |
+
},
|
| 1621 |
+
"157093": {
|
| 1622 |
+
"content": "<|reserved_token_198|>",
|
| 1623 |
+
"lstrip": false,
|
| 1624 |
+
"normalized": false,
|
| 1625 |
+
"rstrip": false,
|
| 1626 |
+
"single_word": false,
|
| 1627 |
+
"special": true
|
| 1628 |
+
},
|
| 1629 |
+
"157094": {
|
| 1630 |
+
"content": "<|reserved_token_199|>",
|
| 1631 |
+
"lstrip": false,
|
| 1632 |
+
"normalized": false,
|
| 1633 |
+
"rstrip": false,
|
| 1634 |
+
"single_word": false,
|
| 1635 |
+
"special": true
|
| 1636 |
+
},
|
| 1637 |
+
"157095": {
|
| 1638 |
+
"content": "<|reserved_token_200|>",
|
| 1639 |
+
"lstrip": false,
|
| 1640 |
+
"normalized": false,
|
| 1641 |
+
"rstrip": false,
|
| 1642 |
+
"single_word": false,
|
| 1643 |
+
"special": true
|
| 1644 |
+
},
|
| 1645 |
+
"157096": {
|
| 1646 |
+
"content": "<|reserved_token_201|>",
|
| 1647 |
+
"lstrip": false,
|
| 1648 |
+
"normalized": false,
|
| 1649 |
+
"rstrip": false,
|
| 1650 |
+
"single_word": false,
|
| 1651 |
+
"special": true
|
| 1652 |
+
},
|
| 1653 |
+
"157097": {
|
| 1654 |
+
"content": "<|reserved_token_202|>",
|
| 1655 |
+
"lstrip": false,
|
| 1656 |
+
"normalized": false,
|
| 1657 |
+
"rstrip": false,
|
| 1658 |
+
"single_word": false,
|
| 1659 |
+
"special": true
|
| 1660 |
+
},
|
| 1661 |
+
"157098": {
|
| 1662 |
+
"content": "<|reserved_token_203|>",
|
| 1663 |
+
"lstrip": false,
|
| 1664 |
+
"normalized": false,
|
| 1665 |
+
"rstrip": false,
|
| 1666 |
+
"single_word": false,
|
| 1667 |
+
"special": true
|
| 1668 |
+
},
|
| 1669 |
+
"157099": {
|
| 1670 |
+
"content": "<|reserved_token_204|>",
|
| 1671 |
+
"lstrip": false,
|
| 1672 |
+
"normalized": false,
|
| 1673 |
+
"rstrip": false,
|
| 1674 |
+
"single_word": false,
|
| 1675 |
+
"special": true
|
| 1676 |
+
},
|
| 1677 |
+
"157100": {
|
| 1678 |
+
"content": "<|reserved_token_205|>",
|
| 1679 |
+
"lstrip": false,
|
| 1680 |
+
"normalized": false,
|
| 1681 |
+
"rstrip": false,
|
| 1682 |
+
"single_word": false,
|
| 1683 |
+
"special": true
|
| 1684 |
+
},
|
| 1685 |
+
"157101": {
|
| 1686 |
+
"content": "<|reserved_token_206|>",
|
| 1687 |
+
"lstrip": false,
|
| 1688 |
+
"normalized": false,
|
| 1689 |
+
"rstrip": false,
|
| 1690 |
+
"single_word": false,
|
| 1691 |
+
"special": true
|
| 1692 |
+
},
|
| 1693 |
+
"157102": {
|
| 1694 |
+
"content": "<|reserved_token_207|>",
|
| 1695 |
+
"lstrip": false,
|
| 1696 |
+
"normalized": false,
|
| 1697 |
+
"rstrip": false,
|
| 1698 |
+
"single_word": false,
|
| 1699 |
+
"special": true
|
| 1700 |
+
},
|
| 1701 |
+
"157103": {
|
| 1702 |
+
"content": "<|reserved_token_208|>",
|
| 1703 |
+
"lstrip": false,
|
| 1704 |
+
"normalized": false,
|
| 1705 |
+
"rstrip": false,
|
| 1706 |
+
"single_word": false,
|
| 1707 |
+
"special": true
|
| 1708 |
+
},
|
| 1709 |
+
"157104": {
|
| 1710 |
+
"content": "<|reserved_token_209|>",
|
| 1711 |
+
"lstrip": false,
|
| 1712 |
+
"normalized": false,
|
| 1713 |
+
"rstrip": false,
|
| 1714 |
+
"single_word": false,
|
| 1715 |
+
"special": true
|
| 1716 |
+
},
|
| 1717 |
+
"157105": {
|
| 1718 |
+
"content": "<|reserved_token_210|>",
|
| 1719 |
+
"lstrip": false,
|
| 1720 |
+
"normalized": false,
|
| 1721 |
+
"rstrip": false,
|
| 1722 |
+
"single_word": false,
|
| 1723 |
+
"special": true
|
| 1724 |
+
},
|
| 1725 |
+
"157106": {
|
| 1726 |
+
"content": "<|reserved_token_211|>",
|
| 1727 |
+
"lstrip": false,
|
| 1728 |
+
"normalized": false,
|
| 1729 |
+
"rstrip": false,
|
| 1730 |
+
"single_word": false,
|
| 1731 |
+
"special": true
|
| 1732 |
+
},
|
| 1733 |
+
"157107": {
|
| 1734 |
+
"content": "<|reserved_token_212|>",
|
| 1735 |
+
"lstrip": false,
|
| 1736 |
+
"normalized": false,
|
| 1737 |
+
"rstrip": false,
|
| 1738 |
+
"single_word": false,
|
| 1739 |
+
"special": true
|
| 1740 |
+
},
|
| 1741 |
+
"157108": {
|
| 1742 |
+
"content": "<|reserved_token_213|>",
|
| 1743 |
+
"lstrip": false,
|
| 1744 |
+
"normalized": false,
|
| 1745 |
+
"rstrip": false,
|
| 1746 |
+
"single_word": false,
|
| 1747 |
+
"special": true
|
| 1748 |
+
},
|
| 1749 |
+
"157109": {
|
| 1750 |
+
"content": "<|reserved_token_214|>",
|
| 1751 |
+
"lstrip": false,
|
| 1752 |
+
"normalized": false,
|
| 1753 |
+
"rstrip": false,
|
| 1754 |
+
"single_word": false,
|
| 1755 |
+
"special": true
|
| 1756 |
+
},
|
| 1757 |
+
"157110": {
|
| 1758 |
+
"content": "<|reserved_token_215|>",
|
| 1759 |
+
"lstrip": false,
|
| 1760 |
+
"normalized": false,
|
| 1761 |
+
"rstrip": false,
|
| 1762 |
+
"single_word": false,
|
| 1763 |
+
"special": true
|
| 1764 |
+
},
|
| 1765 |
+
"157111": {
|
| 1766 |
+
"content": "<|reserved_token_216|>",
|
| 1767 |
+
"lstrip": false,
|
| 1768 |
+
"normalized": false,
|
| 1769 |
+
"rstrip": false,
|
| 1770 |
+
"single_word": false,
|
| 1771 |
+
"special": true
|
| 1772 |
+
},
|
| 1773 |
+
"157112": {
|
| 1774 |
+
"content": "<|reserved_token_217|>",
|
| 1775 |
+
"lstrip": false,
|
| 1776 |
+
"normalized": false,
|
| 1777 |
+
"rstrip": false,
|
| 1778 |
+
"single_word": false,
|
| 1779 |
+
"special": true
|
| 1780 |
+
},
|
| 1781 |
+
"157113": {
|
| 1782 |
+
"content": "<|reserved_token_218|>",
|
| 1783 |
+
"lstrip": false,
|
| 1784 |
+
"normalized": false,
|
| 1785 |
+
"rstrip": false,
|
| 1786 |
+
"single_word": false,
|
| 1787 |
+
"special": true
|
| 1788 |
+
},
|
| 1789 |
+
"157114": {
|
| 1790 |
+
"content": "<|reserved_token_219|>",
|
| 1791 |
+
"lstrip": false,
|
| 1792 |
+
"normalized": false,
|
| 1793 |
+
"rstrip": false,
|
| 1794 |
+
"single_word": false,
|
| 1795 |
+
"special": true
|
| 1796 |
+
},
|
| 1797 |
+
"157115": {
|
| 1798 |
+
"content": "<|reserved_token_220|>",
|
| 1799 |
+
"lstrip": false,
|
| 1800 |
+
"normalized": false,
|
| 1801 |
+
"rstrip": false,
|
| 1802 |
+
"single_word": false,
|
| 1803 |
+
"special": true
|
| 1804 |
+
},
|
| 1805 |
+
"157116": {
|
| 1806 |
+
"content": "<|reserved_token_221|>",
|
| 1807 |
+
"lstrip": false,
|
| 1808 |
+
"normalized": false,
|
| 1809 |
+
"rstrip": false,
|
| 1810 |
+
"single_word": false,
|
| 1811 |
+
"special": true
|
| 1812 |
+
},
|
| 1813 |
+
"157117": {
|
| 1814 |
+
"content": "<|reserved_token_222|>",
|
| 1815 |
+
"lstrip": false,
|
| 1816 |
+
"normalized": false,
|
| 1817 |
+
"rstrip": false,
|
| 1818 |
+
"single_word": false,
|
| 1819 |
+
"special": true
|
| 1820 |
+
},
|
| 1821 |
+
"157118": {
|
| 1822 |
+
"content": "<|reserved_token_223|>",
|
| 1823 |
+
"lstrip": false,
|
| 1824 |
+
"normalized": false,
|
| 1825 |
+
"rstrip": false,
|
| 1826 |
+
"single_word": false,
|
| 1827 |
+
"special": true
|
| 1828 |
+
},
|
| 1829 |
+
"157119": {
|
| 1830 |
+
"content": "<|reserved_token_224|>",
|
| 1831 |
+
"lstrip": false,
|
| 1832 |
+
"normalized": false,
|
| 1833 |
+
"rstrip": false,
|
| 1834 |
+
"single_word": false,
|
| 1835 |
+
"special": true
|
| 1836 |
+
},
|
| 1837 |
+
"157120": {
|
| 1838 |
+
"content": "<|reserved_token_225|>",
|
| 1839 |
+
"lstrip": false,
|
| 1840 |
+
"normalized": false,
|
| 1841 |
+
"rstrip": false,
|
| 1842 |
+
"single_word": false,
|
| 1843 |
+
"special": true
|
| 1844 |
+
},
|
| 1845 |
+
"157121": {
|
| 1846 |
+
"content": "<|reserved_token_226|>",
|
| 1847 |
+
"lstrip": false,
|
| 1848 |
+
"normalized": false,
|
| 1849 |
+
"rstrip": false,
|
| 1850 |
+
"single_word": false,
|
| 1851 |
+
"special": true
|
| 1852 |
+
},
|
| 1853 |
+
"157122": {
|
| 1854 |
+
"content": "<|reserved_token_227|>",
|
| 1855 |
+
"lstrip": false,
|
| 1856 |
+
"normalized": false,
|
| 1857 |
+
"rstrip": false,
|
| 1858 |
+
"single_word": false,
|
| 1859 |
+
"special": true
|
| 1860 |
+
},
|
| 1861 |
+
"157123": {
|
| 1862 |
+
"content": "<|reserved_token_228|>",
|
| 1863 |
+
"lstrip": false,
|
| 1864 |
+
"normalized": false,
|
| 1865 |
+
"rstrip": false,
|
| 1866 |
+
"single_word": false,
|
| 1867 |
+
"special": true
|
| 1868 |
+
},
|
| 1869 |
+
"157124": {
|
| 1870 |
+
"content": "<|reserved_token_229|>",
|
| 1871 |
+
"lstrip": false,
|
| 1872 |
+
"normalized": false,
|
| 1873 |
+
"rstrip": false,
|
| 1874 |
+
"single_word": false,
|
| 1875 |
+
"special": true
|
| 1876 |
+
},
|
| 1877 |
+
"157125": {
|
| 1878 |
+
"content": "<|reserved_token_230|>",
|
| 1879 |
+
"lstrip": false,
|
| 1880 |
+
"normalized": false,
|
| 1881 |
+
"rstrip": false,
|
| 1882 |
+
"single_word": false,
|
| 1883 |
+
"special": true
|
| 1884 |
+
},
|
| 1885 |
+
"157126": {
|
| 1886 |
+
"content": "<|reserved_token_231|>",
|
| 1887 |
+
"lstrip": false,
|
| 1888 |
+
"normalized": false,
|
| 1889 |
+
"rstrip": false,
|
| 1890 |
+
"single_word": false,
|
| 1891 |
+
"special": true
|
| 1892 |
+
},
|
| 1893 |
+
"157127": {
|
| 1894 |
+
"content": "<|reserved_token_232|>",
|
| 1895 |
+
"lstrip": false,
|
| 1896 |
+
"normalized": false,
|
| 1897 |
+
"rstrip": false,
|
| 1898 |
+
"single_word": false,
|
| 1899 |
+
"special": true
|
| 1900 |
+
},
|
| 1901 |
+
"157128": {
|
| 1902 |
+
"content": "<|reserved_token_233|>",
|
| 1903 |
+
"lstrip": false,
|
| 1904 |
+
"normalized": false,
|
| 1905 |
+
"rstrip": false,
|
| 1906 |
+
"single_word": false,
|
| 1907 |
+
"special": true
|
| 1908 |
+
},
|
| 1909 |
+
"157129": {
|
| 1910 |
+
"content": "<|reserved_token_234|>",
|
| 1911 |
+
"lstrip": false,
|
| 1912 |
+
"normalized": false,
|
| 1913 |
+
"rstrip": false,
|
| 1914 |
+
"single_word": false,
|
| 1915 |
+
"special": true
|
| 1916 |
+
},
|
| 1917 |
+
"157130": {
|
| 1918 |
+
"content": "<|reserved_token_235|>",
|
| 1919 |
+
"lstrip": false,
|
| 1920 |
+
"normalized": false,
|
| 1921 |
+
"rstrip": false,
|
| 1922 |
+
"single_word": false,
|
| 1923 |
+
"special": true
|
| 1924 |
+
},
|
| 1925 |
+
"157131": {
|
| 1926 |
+
"content": "<|reserved_token_236|>",
|
| 1927 |
+
"lstrip": false,
|
| 1928 |
+
"normalized": false,
|
| 1929 |
+
"rstrip": false,
|
| 1930 |
+
"single_word": false,
|
| 1931 |
+
"special": true
|
| 1932 |
+
},
|
| 1933 |
+
"157132": {
|
| 1934 |
+
"content": "<|reserved_token_237|>",
|
| 1935 |
+
"lstrip": false,
|
| 1936 |
+
"normalized": false,
|
| 1937 |
+
"rstrip": false,
|
| 1938 |
+
"single_word": false,
|
| 1939 |
+
"special": true
|
| 1940 |
+
},
|
| 1941 |
+
"157133": {
|
| 1942 |
+
"content": "<|reserved_token_238|>",
|
| 1943 |
+
"lstrip": false,
|
| 1944 |
+
"normalized": false,
|
| 1945 |
+
"rstrip": false,
|
| 1946 |
+
"single_word": false,
|
| 1947 |
+
"special": true
|
| 1948 |
+
},
|
| 1949 |
+
"157134": {
|
| 1950 |
+
"content": "<|reserved_token_239|>",
|
| 1951 |
+
"lstrip": false,
|
| 1952 |
+
"normalized": false,
|
| 1953 |
+
"rstrip": false,
|
| 1954 |
+
"single_word": false,
|
| 1955 |
+
"special": true
|
| 1956 |
+
},
|
| 1957 |
+
"157135": {
|
| 1958 |
+
"content": "<|reserved_token_240|>",
|
| 1959 |
+
"lstrip": false,
|
| 1960 |
+
"normalized": false,
|
| 1961 |
+
"rstrip": false,
|
| 1962 |
+
"single_word": false,
|
| 1963 |
+
"special": true
|
| 1964 |
+
},
|
| 1965 |
+
"157136": {
|
| 1966 |
+
"content": "<|reserved_token_241|>",
|
| 1967 |
+
"lstrip": false,
|
| 1968 |
+
"normalized": false,
|
| 1969 |
+
"rstrip": false,
|
| 1970 |
+
"single_word": false,
|
| 1971 |
+
"special": true
|
| 1972 |
+
},
|
| 1973 |
+
"157137": {
|
| 1974 |
+
"content": "<|reserved_token_242|>",
|
| 1975 |
+
"lstrip": false,
|
| 1976 |
+
"normalized": false,
|
| 1977 |
+
"rstrip": false,
|
| 1978 |
+
"single_word": false,
|
| 1979 |
+
"special": true
|
| 1980 |
+
},
|
| 1981 |
+
"157138": {
|
| 1982 |
+
"content": "<|reserved_token_243|>",
|
| 1983 |
+
"lstrip": false,
|
| 1984 |
+
"normalized": false,
|
| 1985 |
+
"rstrip": false,
|
| 1986 |
+
"single_word": false,
|
| 1987 |
+
"special": true
|
| 1988 |
+
},
|
| 1989 |
+
"157139": {
|
| 1990 |
+
"content": "<|reserved_token_244|>",
|
| 1991 |
+
"lstrip": false,
|
| 1992 |
+
"normalized": false,
|
| 1993 |
+
"rstrip": false,
|
| 1994 |
+
"single_word": false,
|
| 1995 |
+
"special": true
|
| 1996 |
+
},
|
| 1997 |
+
"157140": {
|
| 1998 |
+
"content": "<|reserved_token_245|>",
|
| 1999 |
+
"lstrip": false,
|
| 2000 |
+
"normalized": false,
|
| 2001 |
+
"rstrip": false,
|
| 2002 |
+
"single_word": false,
|
| 2003 |
+
"special": true
|
| 2004 |
+
},
|
| 2005 |
+
"157141": {
|
| 2006 |
+
"content": "<|reserved_token_246|>",
|
| 2007 |
+
"lstrip": false,
|
| 2008 |
+
"normalized": false,
|
| 2009 |
+
"rstrip": false,
|
| 2010 |
+
"single_word": false,
|
| 2011 |
+
"special": true
|
| 2012 |
+
},
|
| 2013 |
+
"157142": {
|
| 2014 |
+
"content": "<|reserved_token_247|>",
|
| 2015 |
+
"lstrip": false,
|
| 2016 |
+
"normalized": false,
|
| 2017 |
+
"rstrip": false,
|
| 2018 |
+
"single_word": false,
|
| 2019 |
+
"special": true
|
| 2020 |
+
},
|
| 2021 |
+
"157143": {
|
| 2022 |
+
"content": "<|reserved_token_248|>",
|
| 2023 |
+
"lstrip": false,
|
| 2024 |
+
"normalized": false,
|
| 2025 |
+
"rstrip": false,
|
| 2026 |
+
"single_word": false,
|
| 2027 |
+
"special": true
|
| 2028 |
+
},
|
| 2029 |
+
"157144": {
|
| 2030 |
+
"content": "<|reserved_token_249|>",
|
| 2031 |
+
"lstrip": false,
|
| 2032 |
+
"normalized": false,
|
| 2033 |
+
"rstrip": false,
|
| 2034 |
+
"single_word": false,
|
| 2035 |
+
"special": true
|
| 2036 |
+
},
|
| 2037 |
+
"157145": {
|
| 2038 |
+
"content": "<|reserved_token_250|>",
|
| 2039 |
+
"lstrip": false,
|
| 2040 |
+
"normalized": false,
|
| 2041 |
+
"rstrip": false,
|
| 2042 |
+
"single_word": false,
|
| 2043 |
+
"special": true
|
| 2044 |
+
},
|
| 2045 |
+
"157146": {
|
| 2046 |
+
"content": "<|reserved_token_251|>",
|
| 2047 |
+
"lstrip": false,
|
| 2048 |
+
"normalized": false,
|
| 2049 |
+
"rstrip": false,
|
| 2050 |
+
"single_word": false,
|
| 2051 |
+
"special": true
|
| 2052 |
+
},
|
| 2053 |
+
"157147": {
|
| 2054 |
+
"content": "<|reserved_token_252|>",
|
| 2055 |
+
"lstrip": false,
|
| 2056 |
+
"normalized": false,
|
| 2057 |
+
"rstrip": false,
|
| 2058 |
+
"single_word": false,
|
| 2059 |
+
"special": true
|
| 2060 |
+
},
|
| 2061 |
+
"157148": {
|
| 2062 |
+
"content": "<|reserved_token_253|>",
|
| 2063 |
+
"lstrip": false,
|
| 2064 |
+
"normalized": false,
|
| 2065 |
+
"rstrip": false,
|
| 2066 |
+
"single_word": false,
|
| 2067 |
+
"special": true
|
| 2068 |
+
},
|
| 2069 |
+
"157149": {
|
| 2070 |
+
"content": "<|reserved_token_254|>",
|
| 2071 |
+
"lstrip": false,
|
| 2072 |
+
"normalized": false,
|
| 2073 |
+
"rstrip": false,
|
| 2074 |
+
"single_word": false,
|
| 2075 |
+
"special": true
|
| 2076 |
+
},
|
| 2077 |
+
"157150": {
|
| 2078 |
+
"content": "<|reserved_token_255|>",
|
| 2079 |
+
"lstrip": false,
|
| 2080 |
+
"normalized": false,
|
| 2081 |
+
"rstrip": false,
|
| 2082 |
+
"single_word": false,
|
| 2083 |
+
"special": true
|
| 2084 |
+
},
|
| 2085 |
+
"157151": {
|
| 2086 |
+
"content": "<role>",
|
| 2087 |
+
"lstrip": false,
|
| 2088 |
+
"normalized": false,
|
| 2089 |
+
"rstrip": false,
|
| 2090 |
+
"single_word": false,
|
| 2091 |
+
"special": true
|
| 2092 |
+
},
|
| 2093 |
+
"157152": {
|
| 2094 |
+
"content": "</role>",
|
| 2095 |
+
"lstrip": false,
|
| 2096 |
+
"normalized": false,
|
| 2097 |
+
"rstrip": false,
|
| 2098 |
+
"single_word": false,
|
| 2099 |
+
"special": true
|
| 2100 |
+
}
|
| 2101 |
+
},
|
| 2102 |
+
"bos_token": "<|startoftext|>",
|
| 2103 |
+
"clean_up_tokenization_spaces": false,
|
| 2104 |
+
"cls_token": "[CLS]",
|
| 2105 |
+
"eos_token": "<|endoftext|>",
|
| 2106 |
+
"extra_special_tokens": {},
|
| 2107 |
+
"fast_tokenizer": true,
|
| 2108 |
+
"gmask_token": "[gMASK]",
|
| 2109 |
+
"mask_token": "<|mask|>",
|
| 2110 |
+
"merges_file": null,
|
| 2111 |
+
"model_max_length": 8192,
|
| 2112 |
+
"pad_token": "<|endoftext|>",
|
| 2113 |
+
"tokenizer_class": "PreTrainedTokenizerFast",
|
| 2114 |
+
"trust_remote_code": true
|
| 2115 |
+
}
|