snake7gun commited on
Commit
5899b79
·
verified ·
1 Parent(s): 956de1a

Upload folder using huggingface_hub

Browse files
Files changed (3) hide show
  1. config.json +21 -20
  2. model copy.safetensors +3 -0
  3. model.safetensors +2 -2
config.json CHANGED
@@ -3,56 +3,57 @@
3
  "Qwen3VLForConditionalGeneration"
4
  ],
5
  "dtype": "float32",
6
- "image_token_id": 1001,
7
  "model_type": "qwen3_vl",
8
  "text_config": {
9
  "attention_bias": false,
10
  "attention_dropout": 0.0,
11
- "head_dim": 16,
12
  "hidden_act": "silu",
13
- "hidden_size": 64,
14
  "initializer_range": 0.02,
15
- "intermediate_size": 256,
16
- "max_position_embeddings": 1024,
17
  "model_type": "qwen3_vl_text",
18
- "num_attention_heads": 4,
19
  "num_hidden_layers": 2,
20
- "num_key_value_heads": 2,
21
  "rms_norm_eps": 1e-06,
22
  "rope_scaling": {
23
  "mrope_section": [
24
  8,
25
- 6,
26
- 6
27
  ],
28
  "rope_type": "default"
29
  },
30
- "rope_theta": 10000.0,
31
  "use_cache": true,
32
- "vocab_size": 1000
33
  },
34
  "tie_word_embeddings": false,
35
  "transformers_version": "4.57.0.dev0",
36
- "video_token_id": 1002,
37
  "vision_config": {
38
  "deepstack_visual_indexes": [
39
- 0,
40
- 1
 
41
  ],
42
- "depth": 2,
43
  "hidden_act": "gelu_pytorch_tanh",
44
  "hidden_size": 128,
45
  "in_channels": 3,
46
  "initializer_range": 0.02,
47
- "intermediate_size": 512,
48
  "model_type": "qwen3_vl",
49
  "num_heads": 4,
50
- "num_position_embeddings": 256,
51
- "out_hidden_size": 64,
52
  "patch_size": 16,
53
  "spatial_merge_size": 2,
54
  "temporal_patch_size": 2
55
  },
56
- "vision_end_token_id": 1004,
57
- "vision_start_token_id": 1003
58
  }
 
3
  "Qwen3VLForConditionalGeneration"
4
  ],
5
  "dtype": "float32",
6
+ "image_token_id": 151655,
7
  "model_type": "qwen3_vl",
8
  "text_config": {
9
  "attention_bias": false,
10
  "attention_dropout": 0.0,
11
+ "head_dim": 32,
12
  "hidden_act": "silu",
13
+ "hidden_size": 8,
14
  "initializer_range": 0.02,
15
+ "intermediate_size": 64,
16
+ "max_position_embeddings": 262144,
17
  "model_type": "qwen3_vl_text",
18
+ "num_attention_heads": 8,
19
  "num_hidden_layers": 2,
20
+ "num_key_value_heads": 4,
21
  "rms_norm_eps": 1e-06,
22
  "rope_scaling": {
23
  "mrope_section": [
24
  8,
25
+ 4,
26
+ 4
27
  ],
28
  "rope_type": "default"
29
  },
30
+ "rope_theta": 5000000,
31
  "use_cache": true,
32
+ "vocab_size": 151936
33
  },
34
  "tie_word_embeddings": false,
35
  "transformers_version": "4.57.0.dev0",
36
+ "video_token_id": 151656,
37
  "vision_config": {
38
  "deepstack_visual_indexes": [
39
+ 1,
40
+ 3,
41
+ 5
42
  ],
43
+ "depth": 6,
44
  "hidden_act": "gelu_pytorch_tanh",
45
  "hidden_size": 128,
46
  "in_channels": 3,
47
  "initializer_range": 0.02,
48
+ "intermediate_size": 64,
49
  "model_type": "qwen3_vl",
50
  "num_heads": 4,
51
+ "num_position_embeddings": 2304,
52
+ "out_hidden_size": 8,
53
  "patch_size": 16,
54
  "spatial_merge_size": 2,
55
  "temporal_patch_size": 2
56
  },
57
+ "vision_end_token_id": 151653,
58
+ "vision_start_token_id": 151652
59
  }
model copy.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:23896cc3671cb2e2bd291026dd8b058abf7d32568675e6775e29579e7a8d4666
3
+ size 9028288
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:d699b4ddd348288defffa40fb5d20a7f83291ecbfe697262608c1115b61a16ee
3
- size 7072080
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:20fc6cb4ed60246202d2496195acf2174ead73f90559ac9b943aa37ce2762fdd
3
+ size 18043024