Upload model weights and config

added_tokens.json

@@ -0,0 +1,13 @@
+{
+  "<|assistant|>": 32001,
+  "<|endoftext|>": 32000,
+  "<|end|>": 32007,
+  "<|placeholder1|>": 32002,
+  "<|placeholder2|>": 32003,
+  "<|placeholder3|>": 32004,
+  "<|placeholder4|>": 32005,
+  "<|placeholder5|>": 32008,
+  "<|placeholder6|>": 32009,
+  "<|system|>": 32006,
+  "<|user|>": 32010
+}

architecture.py

@@ -0,0 +1,241 @@
+# --- START OF FILE architecture.py ---
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from transformers import Phi3Config, Phi3ForCausalLM
+from typing import Optional, Dict, List
+
+# --- BUILDING BLOCK 1: Hierarchical VectorMemoryHead ---
+# This version is improved with a hierarchical memory system (L1/L2 cache)
+# to handle much longer contexts and a gated update mechanism for stability.
+class VectorMemoryHead(nn.Module):
+    def __init__(self, hidden_dim: int, num_memory_slots: int, num_heads: int, ff_dim: int,
+                 num_long_term_memory_slots: int = 0, # <-- NEW: Size of the L2 memory cache
+                 device=None, dtype=None):
+        super().__init__()
+        self.hidden_dim = hidden_dim
+        self.num_memory_slots = num_memory_slots # L1 cache size
+        self.num_long_term_memory_slots = num_long_term_memory_slots # L2 cache size
+
+        # --- L1 Working Memory Components (same as before) ---
+        encoder_layer = nn.TransformerEncoderLayer(
+            d_model=hidden_dim, nhead=num_heads, dim_feedforward=ff_dim, dropout=0.1, batch_first=True,
+            device=device, dtype=dtype
+        )
+        self.encoder = nn.TransformerEncoder(encoder_layer, num_layers=1)
+        self.memory_queries = nn.Parameter(torch.randn(1, num_memory_slots, hidden_dim, device=device, dtype=dtype))
+        self.memory_attention = nn.MultiheadAttention(
+            embed_dim=hidden_dim, num_heads=num_heads, dropout=0.1, batch_first=True,
+            device=device, dtype=dtype
+        )
+        self.memory_layernorm = nn.LayerNorm(hidden_dim, device=device, dtype=dtype)
+        self.decoder_attention = nn.MultiheadAttention(
+            embed_dim=hidden_dim, num_heads=num_heads, dropout=0.1, batch_first=True,
+            device=device, dtype=dtype
+        )
+        self.decoder_layernorm = nn.LayerNorm(hidden_dim, device=device, dtype=dtype)
+        self.decoder_ffn = nn.Sequential(
+            nn.Linear(hidden_dim, ff_dim, device=device, dtype=dtype),
+            nn.ReLU(),
+            nn.Linear(ff_dim, hidden_dim, device=device, dtype=dtype)
+        )
+
+        # --- NEW: L2 Long-Term Memory Components ---
+        self.use_long_term_memory = self.num_long_term_memory_slots > 0
+        if self.use_long_term_memory:
+            self.long_term_memory = nn.Parameter(
+                torch.zeros(1, self.num_long_term_memory_slots, hidden_dim, device=device, dtype=dtype)
+            )
+            # Gate for updating long-term memory (similar to GRU/LSTM gates)
+            self.memory_update_gate = nn.Sequential(
+                nn.Linear(2 * hidden_dim, hidden_dim, device=device, dtype=dtype),
+                nn.Sigmoid()
+            )
+            # Attention to read from L2 memory
+            self.ltm_retrieval_attention = nn.MultiheadAttention(
+                embed_dim=hidden_dim, num_heads=num_heads, dropout=0.1, batch_first=True,
+                device=device, dtype=dtype
+            )
+
+    def forward(self, memory_input_sequence: torch.Tensor):
+        batch_size = memory_input_sequence.shape[0]
+
+        # 1. Encode input sequence
+        encoded_vectors = self.encoder(memory_input_sequence)
+
+        # 2. Compress into L1 working memory
+        queries = self.memory_queries.expand(batch_size, -1, -1)
+        compressed_memory, _ = self.memory_attention(query=queries, key=encoded_vectors, value=encoded_vectors)
+        compressed_memory = self.memory_layernorm(compressed_memory + queries) # (B, num_memory_slots, D)
+
+        final_memory_context = compressed_memory
+
+        # --- NEW: Interact with L2 Long-Term Memory ---
+        if self.use_long_term_memory and self.long_term_memory.shape[0] == batch_size:
+            # 3a. Retrieve relevant context from L2 memory using L1 as query
+            retrieved_ltm, _ = self.ltm_retrieval_attention(
+                query=compressed_memory,
+                key=self.long_term_memory,
+                value=self.long_term_memory
+            )
+
+            # 3b. Gated update of the Long-Term Memory
+            # Average the L1 memory to get a summary vector for the update
+            l1_summary = compressed_memory.mean(dim=1)
+            ltm_summary = self.long_term_memory.mean(dim=1)
+            gate_input = torch.cat([l1_summary, ltm_summary], dim=-1)
+            update_gate = self.memory_update_gate(gate_input).unsqueeze(1) # (B, 1, D)
+
+            # Update LTM by blending new info from L1
+            self.long_term_memory.data = (update_gate * l1_summary.unsqueeze(1)) + ((1 - update_gate) * self.long_term_memory.data)
+
+            # Combine L1 and retrieved L2 context for the final output
+            final_memory_context = final_memory_context + retrieved_ltm
+
+        # 4. Decode from the final memory context to reconstruct original sequence
+        reconstructed, _ = self.decoder_attention(query=encoded_vectors, key=final_memory_context, value=final_memory_context)
+        reconstructed_vectors = self.decoder_layernorm(reconstructed + encoded_vectors)
+        reconstructed_vectors = self.decoder_ffn(reconstructed_vectors)
+
+        return compressed_memory, reconstructed_vectors
+
+# --- BUILDING BLOCK 2: The Custom Layer (With Per-Dataset Parameters and Refinement) ---
+class GCVectorMemoryLayer(nn.Module):
+    def __init__(self, original_layer: nn.Linear, global_input_dim: int,
+                 memory_dim: int, num_memory_slots: int, memory_num_heads: int,
+                 global_state_storage: Dict, dataset_keys: List[str]):
+        super().__init__()
+        self.input_dim = original_layer.in_features
+        self.output_dim = original_layer.out_features
+        self.memory_dim = memory_dim
+        self.global_state_storage = global_state_storage
+        self.dataset_keys = dataset_keys
+        self.linear = original_layer # Shared linear layer
+
+        device, dtype = self.linear.weight.device, self.linear.weight.dtype
+
+        # --- NEW: Per-dataset specialized parameters ---
+        self.local_state_projs = nn.ModuleDict()
+        self.global_state_projs = nn.ModuleDict()
+        self.memory_heads = nn.ModuleDict()
+        self.correction_heads = nn.ModuleDict()
+
+        for key in self.dataset_keys:
+            self.local_state_projs[key] = nn.Linear(self.input_dim, memory_dim, device=device, dtype=dtype)
+            self.global_state_projs[key] = nn.Linear(global_input_dim, memory_dim, device=device, dtype=dtype)
+            self.memory_heads[key] = VectorMemoryHead(
+                hidden_dim=memory_dim, num_memory_slots=num_memory_slots,
+                num_heads=memory_num_heads, ff_dim=memory_dim * 2,
+                num_long_term_memory_slots=32, # Enable L2 Cache
+                device=device, dtype=dtype
+            )
+            self.correction_heads[key] = nn.Linear(memory_dim, 2 * self.output_dim, device=device, dtype=dtype)
+
+        self.refinement_passes: int = 2 # Default to 2 passes for deeper refinement
+
+        self.last_corrected_activation: Optional[torch.Tensor] = None
+        self.last_additive_correction: Optional[torch.Tensor] = None
+        self.last_memory_input: Optional[torch.Tensor] = None
+        self.last_reconstructed_from_memory: Optional[torch.Tensor] = None
+
+    def forward(self, x: torch.Tensor):
+        base_output = self.linear(x)
+
+        # Determine which set of specialized parameters to use
+        dataset_key = self.global_state_storage.get('dataset_key')
+        if not dataset_key or 'embeds' not in self.global_state_storage or self.refinement_passes < 1:
+            return base_output
+
+        # Select the correct modules for the current context
+        local_state_proj = self.local_state_projs[dataset_key]
+        global_state_proj = self.global_state_projs[dataset_key]
+        memory_head = self.memory_heads[dataset_key]
+        correction_head = self.correction_heads[dataset_key]
+
+        global_embeds = self.global_state_storage['embeds']
+        if global_embeds.shape[1] != x.shape[1]: global_embeds = global_embeds[:, -x.shape[1]:, :]
+        B, S, _ = x.shape
+
+        # Ensure LTM is initialized with correct batch size for the specific memory head
+        if memory_head.use_long_term_memory and memory_head.long_term_memory.shape[0] != B:
+            memory_head.long_term_memory.data = memory_head.long_term_memory.data.expand(B, -1, -1)
+
+        with torch.no_grad(): # Use no_grad for the refinement loop as it's an inference-like process
+            proj_local = local_state_proj(x.detach())
+            proj_global = global_state_proj(global_embeds.detach())
+            memory_input = torch.stack([proj_global, proj_local], dim=2)
+            memory_input_flat = memory_input.view(B * S, 2, self.memory_dim)
+            compressed_mem_flat, _ = memory_head(memory_input_flat)
+            aggregated_thought = compressed_mem_flat.mean(dim=1).view(B, S, self.memory_dim)
+
+            # Iteratively refine the output using state-feedback
+            corrected_activation = base_output
+            current_thought = aggregated_thought
+            for _ in range(self.refinement_passes):
+                raw_correction = correction_head(current_thought)
+                gate, value = torch.chunk(raw_correction, 2, dim=-1)
+                corrected_activation = corrected_activation * torch.sigmoid(gate) + value
+                
+                current_thought_flat = current_thought.view(B * S, self.memory_dim)
+                refined_thought, _ = memory_head.decoder_attention(
+                    query=current_thought_flat.unsqueeze(1), key=compressed_mem_flat, value=compressed_mem_flat
+                )
+                refined_thought = memory_head.decoder_layernorm(refined_thought.squeeze(1) + current_thought_flat)
+                current_thought = refined_thought.view(B, S, self.memory_dim)
+
+        if self.training:
+            with torch.enable_grad():
+                proj_local_grad = local_state_proj(x)
+                proj_global_grad = global_state_proj(global_embeds)
+                memory_input_grad = torch.stack([proj_global_grad, proj_local_grad], dim=2)
+                memory_input_flat_grad = memory_input_grad.view(B * S, 2, self.memory_dim)
+                compressed_mem_flat_grad, recon_flat_grad = memory_head(memory_input_flat_grad)
+                aggregated_thought_grad = compressed_mem_flat_grad.mean(dim=1).view(B, S, self.memory_dim)
+
+                raw_correction_grad = correction_head(aggregated_thought_grad)
+                gate_grad, value_grad = torch.chunk(raw_correction_grad, 2, dim=-1)
+                final_activation = base_output * torch.sigmoid(gate_grad.to(x.dtype)) + value_grad.to(x.dtype)
+
+                self.last_corrected_activation = final_activation
+                self.last_additive_correction = value_grad
+                self.last_memory_input = memory_input_flat_grad
+                self.last_reconstructed_from_memory = recon_flat_grad
+                return final_activation
+        else:
+             return corrected_activation.to(x.dtype)
+
+# --- BUILDING BLOCK 3: The Full Custom Model Wrapper (for saving/loading) ---
+class Phi3WithVectorMemoryForCausalLM(Phi3ForCausalLM):
+    def __init__(self, config):
+        super().__init__(config)
+        self.global_state_storage = {}
+        # Target a central layer in the network for maximum impact
+        self.target_layer_path = "model.layers.15.mlp.gate_up_proj"
+
+        self.model.embed_tokens.register_forward_hook(
+            lambda module, input, output: self.global_state_storage.update({'embeds': output.detach()})
+        )
+        
+        # This logic is primarily for loading a pre-trained model.
+        # The training script handles the initial creation.
+        if hasattr(config, "dataset_keys") and config.dataset_keys:
+            try:
+                print(f"Re-initializing GCVectorMemoryLayer with dataset keys: {config.dataset_keys}")
+                original_layer = self.get_submodule(self.target_layer_path)
+                custom_layer = GCVectorMemoryLayer(
+                    original_layer=original_layer, global_input_dim=config.hidden_size,
+                    memory_dim=64,
+                    num_memory_slots=8,
+                    memory_num_heads=4,
+                    global_state_storage=self.global_state_storage,
+                    dataset_keys=config.dataset_keys # Use keys from config
+                )
+                parent_path = ".".join(self.target_layer_path.split('.')[:-1])
+                child_name = self.target_layer_path.split('.')[-1]
+                setattr(self.get_submodule(parent_path), child_name, custom_layer)
+                print(f"Successfully reloaded and replaced '{self.target_layer_path}' with specialized GCVectorMemoryLayer.")
+            except AttributeError:
+                print(f"Could not find target layer '{self.target_layer_path}' during reload. Model remains unmodified.")
+        else:
+             print("No 'dataset_keys' found in config. The custom layer will not be initialized.")

chat_template.jinja

@@ -0,0 +1,8 @@
+{% for message in messages %}{% if message['role'] == 'system' %}{{'<|system|>
+' + message['content'] + '<|end|>
+'}}{% elif message['role'] == 'user' %}{{'<|user|>
+' + message['content'] + '<|end|>
+'}}{% elif message['role'] == 'assistant' %}{{'<|assistant|>
+' + message['content'] + '<|end|>
+'}}{% endif %}{% endfor %}{% if add_generation_prompt %}{{ '<|assistant|>
+' }}{% else %}{{ eos_token }}{% endif %}

config.json

@@ -0,0 +1,41 @@
+{
+  "architectures": [
+    "Phi3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "auto_map": {
+    "AutoModelForCausalLM": "architecture.Phi3WithVectorMemoryForCausalLM"
+  },
+  "bos_token_id": 1,
+  "dataset_keys": [
+    "codenet_c",
+    "codenet_cpp",
+    "codenet_python",
+    "math",
+    "wikitext"
+  ],
+  "embd_pdrop": 0.0,
+  "eos_token_id": 32000,
+  "hidden_act": "silu",
+  "hidden_size": 3072,
+  "initializer_range": 0.02,
+  "intermediate_size": 8192,
+  "max_position_embeddings": 4096,
+  "model_type": "phi3",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 32,
+  "original_max_position_embeddings": 4096,
+  "pad_token_id": 32000,
+  "resid_pdrop": 0.0,
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": null,
+  "rope_theta": 10000.0,
+  "sliding_window": 2047,
+  "tie_word_embeddings": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.52.4",
+  "use_cache": true,
+  "vocab_size": 32064
+}

generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 1,
+  "eos_token_id": 32000,
+  "pad_token_id": 32000,
+  "transformers_version": "4.52.4"
+}

model-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b61786273d113f32062c6ea038774d4dbb7389c7695148083d350b6924ebd451
+size 4998865208

model-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f311787aa136e858556caa8543015161edcad85ba81b6a36072443d7fa73c87
+size 2669692552

model.safetensors.index.json

@@ -0,0 +1,412 @@
+{
+  "metadata": {
+    "total_size": 7668502144
+  },
+  "weight_map": {
+    "lm_head.weight": "model-00002-of-00002.safetensors",
+    "model.embed_tokens.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.correction_heads.codenet_c.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.correction_heads.codenet_c.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.correction_heads.codenet_cpp.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.correction_heads.codenet_cpp.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.correction_heads.codenet_python.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.correction_heads.codenet_python.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.correction_heads.math.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.correction_heads.math.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.correction_heads.wikitext.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.correction_heads.wikitext.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.global_state_projs.codenet_c.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.global_state_projs.codenet_c.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.global_state_projs.codenet_cpp.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.global_state_projs.codenet_cpp.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.global_state_projs.codenet_python.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.global_state_projs.codenet_python.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.global_state_projs.math.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.global_state_projs.math.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.global_state_projs.wikitext.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.global_state_projs.wikitext.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.linear.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.local_state_projs.codenet_c.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.local_state_projs.codenet_c.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.local_state_projs.codenet_cpp.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.local_state_projs.codenet_cpp.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.local_state_projs.codenet_python.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.local_state_projs.codenet_python.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.local_state_projs.math.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.local_state_projs.math.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.local_state_projs.wikitext.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.local_state_projs.wikitext.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.decoder_attention.in_proj_bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.decoder_attention.in_proj_weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.decoder_attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.decoder_attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.decoder_ffn.0.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.decoder_ffn.0.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.decoder_ffn.2.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.decoder_ffn.2.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.decoder_layernorm.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.decoder_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.encoder.layers.0.linear1.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.encoder.layers.0.linear1.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.encoder.layers.0.linear2.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.encoder.layers.0.linear2.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.encoder.layers.0.norm1.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.encoder.layers.0.norm1.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.encoder.layers.0.norm2.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.encoder.layers.0.norm2.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.encoder.layers.0.self_attn.in_proj_bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.encoder.layers.0.self_attn.in_proj_weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.encoder.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.encoder.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.long_term_memory": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.ltm_retrieval_attention.in_proj_bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.ltm_retrieval_attention.in_proj_weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.ltm_retrieval_attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.ltm_retrieval_attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.memory_attention.in_proj_bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.memory_attention.in_proj_weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.memory_attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.memory_attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.memory_layernorm.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.memory_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.memory_queries": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.memory_update_gate.0.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_c.memory_update_gate.0.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.decoder_attention.in_proj_bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.decoder_attention.in_proj_weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.decoder_attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.decoder_attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.decoder_ffn.0.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.decoder_ffn.0.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.decoder_ffn.2.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.decoder_ffn.2.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.decoder_layernorm.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.decoder_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.encoder.layers.0.linear1.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.encoder.layers.0.linear1.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.encoder.layers.0.linear2.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.encoder.layers.0.linear2.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.encoder.layers.0.norm1.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.encoder.layers.0.norm1.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.encoder.layers.0.norm2.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.encoder.layers.0.norm2.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.encoder.layers.0.self_attn.in_proj_bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.encoder.layers.0.self_attn.in_proj_weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.encoder.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.encoder.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.long_term_memory": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.ltm_retrieval_attention.in_proj_bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.ltm_retrieval_attention.in_proj_weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.ltm_retrieval_attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.ltm_retrieval_attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.memory_attention.in_proj_bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.memory_attention.in_proj_weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.memory_attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.memory_attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.memory_layernorm.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.memory_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.memory_queries": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.memory_update_gate.0.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_cpp.memory_update_gate.0.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.decoder_attention.in_proj_bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.decoder_attention.in_proj_weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.decoder_attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.decoder_attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.decoder_ffn.0.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.decoder_ffn.0.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.decoder_ffn.2.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.decoder_ffn.2.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.decoder_layernorm.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.decoder_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.encoder.layers.0.linear1.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.encoder.layers.0.linear1.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.encoder.layers.0.linear2.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.encoder.layers.0.linear2.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.encoder.layers.0.norm1.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.encoder.layers.0.norm1.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.encoder.layers.0.norm2.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.encoder.layers.0.norm2.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.encoder.layers.0.self_attn.in_proj_bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.encoder.layers.0.self_attn.in_proj_weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.encoder.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.encoder.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.long_term_memory": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.ltm_retrieval_attention.in_proj_bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.ltm_retrieval_attention.in_proj_weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.ltm_retrieval_attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.ltm_retrieval_attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.memory_attention.in_proj_bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.memory_attention.in_proj_weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.memory_attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.memory_attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.memory_layernorm.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.memory_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.memory_queries": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.memory_update_gate.0.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.codenet_python.memory_update_gate.0.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.decoder_attention.in_proj_bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.decoder_attention.in_proj_weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.decoder_attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.decoder_attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.decoder_ffn.0.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.decoder_ffn.0.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.decoder_ffn.2.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.decoder_ffn.2.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.decoder_layernorm.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.decoder_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.encoder.layers.0.linear1.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.encoder.layers.0.linear1.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.encoder.layers.0.linear2.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.encoder.layers.0.linear2.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.encoder.layers.0.norm1.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.encoder.layers.0.norm1.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.encoder.layers.0.norm2.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.encoder.layers.0.norm2.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.encoder.layers.0.self_attn.in_proj_bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.encoder.layers.0.self_attn.in_proj_weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.encoder.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.encoder.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.long_term_memory": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.ltm_retrieval_attention.in_proj_bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.ltm_retrieval_attention.in_proj_weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.ltm_retrieval_attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.ltm_retrieval_attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.memory_attention.in_proj_bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.memory_attention.in_proj_weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.memory_attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.memory_attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.memory_layernorm.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.memory_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.memory_queries": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.memory_update_gate.0.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.math.memory_update_gate.0.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.decoder_attention.in_proj_bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.decoder_attention.in_proj_weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.decoder_attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.decoder_attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.decoder_ffn.0.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.decoder_ffn.0.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.decoder_ffn.2.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.decoder_ffn.2.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.decoder_layernorm.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.decoder_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.encoder.layers.0.linear1.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.encoder.layers.0.linear1.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.encoder.layers.0.linear2.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.encoder.layers.0.linear2.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.encoder.layers.0.norm1.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.encoder.layers.0.norm1.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.encoder.layers.0.norm2.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.encoder.layers.0.norm2.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.encoder.layers.0.self_attn.in_proj_bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.encoder.layers.0.self_attn.in_proj_weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.encoder.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.encoder.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.long_term_memory": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.ltm_retrieval_attention.in_proj_bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.ltm_retrieval_attention.in_proj_weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.ltm_retrieval_attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.ltm_retrieval_attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.memory_attention.in_proj_bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.memory_attention.in_proj_weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.memory_attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.memory_attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.memory_layernorm.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.memory_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.memory_queries": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.memory_update_gate.0.bias": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.memory_heads.wikitext.memory_update_gate.0.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.20.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.20.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.20.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.20.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.20.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.20.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.21.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.mlp.gate_up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.21.self_attn.qkv_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.mlp.gate_up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.self_attn.qkv_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.mlp.gate_up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.self_attn.qkv_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.mlp.gate_up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.self_attn.qkv_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.mlp.gate_up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.self_attn.qkv_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.mlp.gate_up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.self_attn.qkv_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.mlp.gate_up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.self_attn.qkv_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.mlp.gate_up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.self_attn.qkv_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.mlp.gate_up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.self_attn.qkv_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.30.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.mlp.gate_up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.self_attn.qkv_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.mlp.gate_up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.self_attn.qkv_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.norm.weight": "model-00002-of-00002.safetensors"
+  }
+}

special_tokens_map.json

@@ -0,0 +1,30 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
+size 499723

tokenizer_config.json

@@ -0,0 +1,131 @@
+{
+  "add_bos_token": false,
+  "add_eos_token": false,
+  "add_prefix_space": null,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": false
+    },
+    "32000": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "32001": {
+      "content": "<|assistant|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32002": {
+      "content": "<|placeholder1|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32003": {
+      "content": "<|placeholder2|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32004": {
+      "content": "<|placeholder3|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32005": {
+      "content": "<|placeholder4|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32006": {
+      "content": "<|system|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32007": {
+      "content": "<|end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32008": {
+      "content": "<|placeholder5|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32009": {
+      "content": "<|placeholder6|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32010": {
+      "content": "<|user|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "extra_special_tokens": {},
+  "legacy": false,
+  "model_max_length": 4096,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "left",
+  "sp_model_kwargs": {},
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}