uploaded model
Browse files- config.json +95 -0
config.json
ADDED
@@ -0,0 +1,95 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "Llama-3.1-Storm-8B-FP8-Dynamic",
|
3 |
+
"architectures": [
|
4 |
+
"LlamaForCausalLM"
|
5 |
+
],
|
6 |
+
"attention_bias": false,
|
7 |
+
"attention_dropout": 0.0,
|
8 |
+
"bos_token_id": 128000,
|
9 |
+
"compression_config": {
|
10 |
+
"config_groups": {
|
11 |
+
"group_0": {
|
12 |
+
"input_activations": {
|
13 |
+
"block_structure": null,
|
14 |
+
"dynamic": true,
|
15 |
+
"group_size": null,
|
16 |
+
"num_bits": 8,
|
17 |
+
"observer": "memoryless",
|
18 |
+
"observer_kwargs": {},
|
19 |
+
"strategy": "token",
|
20 |
+
"symmetric": true,
|
21 |
+
"type": "float"
|
22 |
+
},
|
23 |
+
"output_activations": null,
|
24 |
+
"targets": [
|
25 |
+
"Linear"
|
26 |
+
],
|
27 |
+
"weights": {
|
28 |
+
"block_structure": null,
|
29 |
+
"dynamic": false,
|
30 |
+
"group_size": null,
|
31 |
+
"num_bits": 8,
|
32 |
+
"observer": "minmax",
|
33 |
+
"observer_kwargs": {},
|
34 |
+
"strategy": "channel",
|
35 |
+
"symmetric": true,
|
36 |
+
"type": "float"
|
37 |
+
}
|
38 |
+
}
|
39 |
+
},
|
40 |
+
"format": "float-quantized",
|
41 |
+
"global_compression_ratio": 1.239290831149584,
|
42 |
+
"ignore": [
|
43 |
+
"lm_head"
|
44 |
+
],
|
45 |
+
"kv_cache_scheme": null,
|
46 |
+
"quant_method": "compressed-tensors",
|
47 |
+
"quantization_status": "frozen"
|
48 |
+
},
|
49 |
+
"eos_token_id": [
|
50 |
+
128001,
|
51 |
+
128008,
|
52 |
+
128009
|
53 |
+
],
|
54 |
+
"head_dim": 128,
|
55 |
+
"hidden_act": "silu",
|
56 |
+
"hidden_size": 4096,
|
57 |
+
"initializer_range": 0.02,
|
58 |
+
"intermediate_size": 14336,
|
59 |
+
"max_position_embeddings": 131072,
|
60 |
+
"mlp_bias": false,
|
61 |
+
"model_type": "llama",
|
62 |
+
"num_attention_heads": 32,
|
63 |
+
"num_hidden_layers": 32,
|
64 |
+
"num_key_value_heads": 8,
|
65 |
+
"pretraining_tp": 1,
|
66 |
+
"quantization_config": {
|
67 |
+
"_load_in_4bit": false,
|
68 |
+
"_load_in_8bit": true,
|
69 |
+
"bnb_4bit_compute_dtype": "float32",
|
70 |
+
"bnb_4bit_quant_storage": "uint8",
|
71 |
+
"bnb_4bit_quant_type": "fp4",
|
72 |
+
"bnb_4bit_use_double_quant": false,
|
73 |
+
"llm_int8_enable_fp32_cpu_offload": false,
|
74 |
+
"llm_int8_has_fp16_weight": false,
|
75 |
+
"llm_int8_skip_modules": null,
|
76 |
+
"llm_int8_threshold": 6.0,
|
77 |
+
"load_in_4bit": false,
|
78 |
+
"load_in_8bit": true,
|
79 |
+
"quant_method": "bitsandbytes"
|
80 |
+
},
|
81 |
+
"rms_norm_eps": 1e-05,
|
82 |
+
"rope_scaling": {
|
83 |
+
"factor": 8.0,
|
84 |
+
"high_freq_factor": 4.0,
|
85 |
+
"low_freq_factor": 1.0,
|
86 |
+
"original_max_position_embeddings": 8192,
|
87 |
+
"rope_type": "llama3"
|
88 |
+
},
|
89 |
+
"rope_theta": 500000.0,
|
90 |
+
"tie_word_embeddings": false,
|
91 |
+
"torch_dtype": "float16",
|
92 |
+
"transformers_version": "4.47.1",
|
93 |
+
"use_cache": true,
|
94 |
+
"vocab_size": 128256
|
95 |
+
}
|