Hugging Face's logo

amd
/
Llama-3.2-3B-Instruct-awq-g128-int4-asym-bf16-onnx-ryzen-strix

ONNX
Model card Files Community
uday610 commited on
Commit
e42d25f
·
verified ·
1 Parent(s): a337849

Update genai_config.json

Browse files
Files changed (1) hide show
  1. genai_config.json +54 -55
genai_config.json CHANGED
@@ -1,56 +1,55 @@
1
- {
2
- "model": {
3
- "bos_token_id": 128000,
4
- "context_length": 131072,
5
- "decoder": {
6
- "session_options": {
7
- "log_id": "onnxruntime-genai",
8
- "custom_ops_library": "onnxruntime_vitis_ai_custom_ops.dll",
9
- "provider_options": [
10
- {
11
- "VitisAI": {
12
- "config_file": ".\\libs\\vaip_llm.json"
13
- }
14
- }
15
- ]
16
- },
17
- "filename": "model.onnx",
18
- "head_size": 128,
19
- "hidden_size": 3072,
20
- "inputs": {
21
- "input_ids": "input_ids",
22
- "attention_mask": "attention_mask",
23
- "past_key_names": "past_key_values.%d.key",
24
- "past_value_names": "past_key_values.%d.value"
25
- },
26
- "outputs": {
27
- "logits": "logits",
28
- "present_key_names": "present.%d.key",
29
- "present_value_names": "present.%d.value"
30
- },
31
- "num_attention_heads": 24,
32
- "num_hidden_layers": 28,
33
- "num_key_value_heads": 8
34
- },
35
- "eos_token_id": 128001,
36
- "pad_token_id": 128001,
37
- "type": "llama",
38
- "vocab_size": 128256
39
- },
40
- "search": {
41
- "diversity_penalty": 0.0,
42
- "do_sample": true,
43
- "early_stopping": true,
44
- "length_penalty": 1.0,
45
- "max_length": 131072,
46
- "min_length": 0,
47
- "no_repeat_ngram_size": 0,
48
- "num_beams": 1,
49
- "num_return_sequences": 1,
50
- "past_present_share_buffer": true,
51
- "repetition_penalty": 1.0,
52
- "temperature": 0.6,
53
- "top_k": 1,
54
- "top_p": 0.9
55
- }
56
  }
 
1
+ {
2
+ "model": {
3
+ "bos_token_id": 128000,
4
+ "context_length": 131072,
5
+ "decoder": {
6
+ "session_options": {
7
+ "log_id": "onnxruntime-genai",
8
+ "custom_ops_library": "onnxruntime_vitis_ai_custom_ops.dll",
9
+ "provider_options": [
10
+ {
11
+ "VitisAI": {
12
+ }
13
+ }
14
+ ]
15
+ },
16
+ "filename": "model.onnx",
17
+ "head_size": 128,
18
+ "hidden_size": 3072,
19
+ "inputs": {
20
+ "input_ids": "input_ids",
21
+ "attention_mask": "attention_mask",
22
+ "past_key_names": "past_key_values.%d.key",
23
+ "past_value_names": "past_key_values.%d.value"
24
+ },
25
+ "outputs": {
26
+ "logits": "logits",
27
+ "present_key_names": "present.%d.key",
28
+ "present_value_names": "present.%d.value"
29
+ },
30
+ "num_attention_heads": 24,
31
+ "num_hidden_layers": 28,
32
+ "num_key_value_heads": 8
33
+ },
34
+ "eos_token_id": 128001,
35
+ "pad_token_id": 128001,
36
+ "type": "llama",
37
+ "vocab_size": 128256
38
+ },
39
+ "search": {
40
+ "diversity_penalty": 0.0,
41
+ "do_sample": true,
42
+ "early_stopping": true,
43
+ "length_penalty": 1.0,
44
+ "max_length": 131072,
45
+ "min_length": 0,
46
+ "no_repeat_ngram_size": 0,
47
+ "num_beams": 1,
48
+ "num_return_sequences": 1,
49
+ "past_present_share_buffer": true,
50
+ "repetition_penalty": 1.0,
51
+ "temperature": 0.6,
52
+ "top_k": 1,
53
+ "top_p": 0.9
54
+ }
 
55
  }