ruihanglai
commited on
Commit
•
5dc45e6
1
Parent(s):
fb32613
Update mlc-chat-config.json
Browse files- mlc-chat-config.json +83 -0
mlc-chat-config.json
ADDED
@@ -0,0 +1,83 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"version": "0.1.0",
|
3 |
+
"model_type": "gemma2",
|
4 |
+
"quantization": "q4f16_1",
|
5 |
+
"model_config": {
|
6 |
+
"hidden_size": 4608,
|
7 |
+
"intermediate_size": 36864,
|
8 |
+
"attention_bias": false,
|
9 |
+
"num_attention_heads": 32,
|
10 |
+
"num_key_value_heads": 16,
|
11 |
+
"head_dim": 128,
|
12 |
+
"num_hidden_layers": 46,
|
13 |
+
"rms_norm_eps": 1e-06,
|
14 |
+
"vocab_size": 256000,
|
15 |
+
"hidden_activation": "gelu_pytorch_tanh",
|
16 |
+
"position_embedding_base": 10000.0,
|
17 |
+
"context_window_size": 4096,
|
18 |
+
"prefill_chunk_size": 2048,
|
19 |
+
"tensor_parallel_shards": 1,
|
20 |
+
"max_batch_size": 80,
|
21 |
+
"attn_logit_softcapping": 50.0,
|
22 |
+
"final_logit_softcapping": 30.0,
|
23 |
+
"query_pre_attn_scalar": 144,
|
24 |
+
"sliding_window": 4096
|
25 |
+
},
|
26 |
+
"vocab_size": 256000,
|
27 |
+
"context_window_size": 4096,
|
28 |
+
"sliding_window_size": -1,
|
29 |
+
"prefill_chunk_size": 2048,
|
30 |
+
"attention_sink_size": -1,
|
31 |
+
"tensor_parallel_shards": 1,
|
32 |
+
"temperature": 1.0,
|
33 |
+
"presence_penalty": 0.0,
|
34 |
+
"frequency_penalty": 0.0,
|
35 |
+
"repetition_penalty": 1.0,
|
36 |
+
"top_p": 1.0,
|
37 |
+
"tokenizer_files": [
|
38 |
+
"tokenizer.model",
|
39 |
+
"tokenizer.json",
|
40 |
+
"tokenizer_config.json"
|
41 |
+
],
|
42 |
+
"tokenizer_info": {
|
43 |
+
"token_postproc_method": "byte_fallback",
|
44 |
+
"prepend_space_in_encode": false,
|
45 |
+
"strip_space_in_decode": false
|
46 |
+
},
|
47 |
+
"conv_template": {
|
48 |
+
"name": "gemma_instruction",
|
49 |
+
"system_template": "<bos>{system_message}",
|
50 |
+
"system_message": "",
|
51 |
+
"system_prefix_token_ids": [
|
52 |
+
2
|
53 |
+
],
|
54 |
+
"add_role_after_system_message": true,
|
55 |
+
"roles": {
|
56 |
+
"user": "<start_of_turn>user",
|
57 |
+
"assistant": "<start_of_turn>model"
|
58 |
+
},
|
59 |
+
"role_templates": {
|
60 |
+
"user": "{user_message}",
|
61 |
+
"assistant": "{assistant_message}",
|
62 |
+
"tool": "{tool_message}"
|
63 |
+
},
|
64 |
+
"messages": [],
|
65 |
+
"seps": [
|
66 |
+
"<end_of_turn>\n"
|
67 |
+
],
|
68 |
+
"role_content_sep": "\n",
|
69 |
+
"role_empty_sep": "\n",
|
70 |
+
"stop_str": [
|
71 |
+
"<end_of_turn>"
|
72 |
+
],
|
73 |
+
"stop_token_ids": [
|
74 |
+
1,
|
75 |
+
107
|
76 |
+
],
|
77 |
+
"function_string": "",
|
78 |
+
"use_function_calling": false
|
79 |
+
},
|
80 |
+
"pad_token_id": 0,
|
81 |
+
"bos_token_id": 2,
|
82 |
+
"eos_token_id": 1
|
83 |
+
}
|