Update config.json
Browse files- config.json +86 -86
config.json
CHANGED
@@ -1,86 +1,86 @@
|
|
1 |
-
{
|
2 |
-
"architectures": [
|
3 |
-
"Qwen2_5_VLForConditionalGeneration"
|
4 |
-
],
|
5 |
-
"attention_dropout": 0.0,
|
6 |
-
"eos_token_id":
|
7 |
-
"hidden_act": "silu",
|
8 |
-
"hidden_size": 2048,
|
9 |
-
"image_token_id": 151655,
|
10 |
-
"initializer_range": 0.02,
|
11 |
-
"intermediate_size": 11008,
|
12 |
-
"max_position_embeddings": 128000,
|
13 |
-
"max_window_layers": 70,
|
14 |
-
"model_type": "qwen2_5_vl",
|
15 |
-
"num_attention_heads": 16,
|
16 |
-
"num_hidden_layers": 36,
|
17 |
-
"num_key_value_heads": 2,
|
18 |
-
"pad_token_id": 151654,
|
19 |
-
"quantization_config": {
|
20 |
-
"_load_in_4bit": true,
|
21 |
-
"_load_in_8bit": false,
|
22 |
-
"bnb_4bit_compute_dtype": "bfloat16",
|
23 |
-
"bnb_4bit_quant_storage": "uint8",
|
24 |
-
"bnb_4bit_quant_type": "nf4",
|
25 |
-
"bnb_4bit_use_double_quant": true,
|
26 |
-
"llm_int8_enable_fp32_cpu_offload": false,
|
27 |
-
"llm_int8_has_fp16_weight": false,
|
28 |
-
"llm_int8_skip_modules": [
|
29 |
-
"lm_head",
|
30 |
-
"multi_modal_projector",
|
31 |
-
"merger",
|
32 |
-
"modality_projection"
|
33 |
-
],
|
34 |
-
"llm_int8_threshold": 6.0,
|
35 |
-
"load_in_4bit": true,
|
36 |
-
"load_in_8bit": false,
|
37 |
-
"quant_method": "bitsandbytes"
|
38 |
-
},
|
39 |
-
"rms_norm_eps": 1e-06,
|
40 |
-
"rope_scaling": {
|
41 |
-
"mrope_section": [
|
42 |
-
16,
|
43 |
-
24,
|
44 |
-
24
|
45 |
-
],
|
46 |
-
"rope_type": "default",
|
47 |
-
"type": "default"
|
48 |
-
},
|
49 |
-
"rope_theta": 1000000.0,
|
50 |
-
"sliding_window": 32768,
|
51 |
-
"tie_word_embeddings": true,
|
52 |
-
"torch_dtype": "bfloat16",
|
53 |
-
"transformers_version": "4.51.3",
|
54 |
-
"unsloth_fixed": true,
|
55 |
-
"use_cache": true,
|
56 |
-
"use_sliding_window": false,
|
57 |
-
"video_token_id": 151656,
|
58 |
-
"vision_config": {
|
59 |
-
"depth": 32,
|
60 |
-
"fullatt_block_indexes": [
|
61 |
-
7,
|
62 |
-
15,
|
63 |
-
23,
|
64 |
-
31
|
65 |
-
],
|
66 |
-
"hidden_act": "silu",
|
67 |
-
"hidden_size": 1280,
|
68 |
-
"in_channels": 3,
|
69 |
-
"in_chans": 3,
|
70 |
-
"intermediate_size": 3420,
|
71 |
-
"model_type": "qwen2_5_vl",
|
72 |
-
"num_heads": 16,
|
73 |
-
"out_hidden_size": 2048,
|
74 |
-
"patch_size": 14,
|
75 |
-
"spatial_merge_size": 2,
|
76 |
-
"spatial_patch_size": 14,
|
77 |
-
"temporal_patch_size": 2,
|
78 |
-
"tokens_per_second": 2,
|
79 |
-
"torch_dtype": "bfloat16",
|
80 |
-
"window_size": 112
|
81 |
-
},
|
82 |
-
"vision_end_token_id": 151653,
|
83 |
-
"vision_start_token_id": 151652,
|
84 |
-
"vision_token_id": 151654,
|
85 |
-
"vocab_size": 151936
|
86 |
-
}
|
|
|
1 |
+
{
|
2 |
+
"architectures": [
|
3 |
+
"Qwen2_5_VLForConditionalGeneration"
|
4 |
+
],
|
5 |
+
"attention_dropout": 0.0,
|
6 |
+
"eos_token_id": 151643,
|
7 |
+
"hidden_act": "silu",
|
8 |
+
"hidden_size": 2048,
|
9 |
+
"image_token_id": 151655,
|
10 |
+
"initializer_range": 0.02,
|
11 |
+
"intermediate_size": 11008,
|
12 |
+
"max_position_embeddings": 128000,
|
13 |
+
"max_window_layers": 70,
|
14 |
+
"model_type": "qwen2_5_vl",
|
15 |
+
"num_attention_heads": 16,
|
16 |
+
"num_hidden_layers": 36,
|
17 |
+
"num_key_value_heads": 2,
|
18 |
+
"pad_token_id": 151654,
|
19 |
+
"quantization_config": {
|
20 |
+
"_load_in_4bit": true,
|
21 |
+
"_load_in_8bit": false,
|
22 |
+
"bnb_4bit_compute_dtype": "bfloat16",
|
23 |
+
"bnb_4bit_quant_storage": "uint8",
|
24 |
+
"bnb_4bit_quant_type": "nf4",
|
25 |
+
"bnb_4bit_use_double_quant": true,
|
26 |
+
"llm_int8_enable_fp32_cpu_offload": false,
|
27 |
+
"llm_int8_has_fp16_weight": false,
|
28 |
+
"llm_int8_skip_modules": [
|
29 |
+
"lm_head",
|
30 |
+
"multi_modal_projector",
|
31 |
+
"merger",
|
32 |
+
"modality_projection"
|
33 |
+
],
|
34 |
+
"llm_int8_threshold": 6.0,
|
35 |
+
"load_in_4bit": true,
|
36 |
+
"load_in_8bit": false,
|
37 |
+
"quant_method": "bitsandbytes"
|
38 |
+
},
|
39 |
+
"rms_norm_eps": 1e-06,
|
40 |
+
"rope_scaling": {
|
41 |
+
"mrope_section": [
|
42 |
+
16,
|
43 |
+
24,
|
44 |
+
24
|
45 |
+
],
|
46 |
+
"rope_type": "default",
|
47 |
+
"type": "default"
|
48 |
+
},
|
49 |
+
"rope_theta": 1000000.0,
|
50 |
+
"sliding_window": 32768,
|
51 |
+
"tie_word_embeddings": true,
|
52 |
+
"torch_dtype": "bfloat16",
|
53 |
+
"transformers_version": "4.51.3",
|
54 |
+
"unsloth_fixed": true,
|
55 |
+
"use_cache": true,
|
56 |
+
"use_sliding_window": false,
|
57 |
+
"video_token_id": 151656,
|
58 |
+
"vision_config": {
|
59 |
+
"depth": 32,
|
60 |
+
"fullatt_block_indexes": [
|
61 |
+
7,
|
62 |
+
15,
|
63 |
+
23,
|
64 |
+
31
|
65 |
+
],
|
66 |
+
"hidden_act": "silu",
|
67 |
+
"hidden_size": 1280,
|
68 |
+
"in_channels": 3,
|
69 |
+
"in_chans": 3,
|
70 |
+
"intermediate_size": 3420,
|
71 |
+
"model_type": "qwen2_5_vl",
|
72 |
+
"num_heads": 16,
|
73 |
+
"out_hidden_size": 2048,
|
74 |
+
"patch_size": 14,
|
75 |
+
"spatial_merge_size": 2,
|
76 |
+
"spatial_patch_size": 14,
|
77 |
+
"temporal_patch_size": 2,
|
78 |
+
"tokens_per_second": 2,
|
79 |
+
"torch_dtype": "bfloat16",
|
80 |
+
"window_size": 112
|
81 |
+
},
|
82 |
+
"vision_end_token_id": 151653,
|
83 |
+
"vision_start_token_id": 151652,
|
84 |
+
"vision_token_id": 151654,
|
85 |
+
"vocab_size": 151936
|
86 |
+
}
|