snake7gun commited on
Commit
956de1a
·
verified ·
1 Parent(s): 2a5e44b

Upload folder using huggingface_hub

Browse files
Files changed (3) hide show
  1. config.json +21 -27
  2. generation_config.json +2 -11
  3. model.safetensors +2 -2
config.json CHANGED
@@ -2,63 +2,57 @@
2
  "architectures": [
3
  "Qwen3VLForConditionalGeneration"
4
  ],
5
- "dtype": "bfloat16",
6
- "image_token_id": 151655,
7
  "model_type": "qwen3_vl",
8
  "text_config": {
9
  "attention_bias": false,
10
  "attention_dropout": 0.0,
11
- "bos_token_id": 151643,
12
- "dtype": "bfloat16",
13
- "eos_token_id": 151645,
14
- "head_dim": 32,
15
  "hidden_act": "silu",
16
- "hidden_size": 8,
17
  "initializer_range": 0.02,
18
- "intermediate_size": 64,
19
- "max_position_embeddings": 262144,
20
  "model_type": "qwen3_vl_text",
21
- "moe_intermediate_size": 64,
22
- "num_attention_heads": 8,
23
  "num_hidden_layers": 2,
24
- "num_key_value_heads": 4,
25
  "rms_norm_eps": 1e-06,
26
  "rope_scaling": {
27
- "mrope_interleaved": true,
28
  "mrope_section": [
29
  8,
30
- 4,
31
- 4
32
  ],
33
  "rope_type": "default"
34
  },
35
- "rope_theta": 5000000,
36
  "use_cache": true,
37
- "vocab_size": 151936
38
  },
39
  "tie_word_embeddings": false,
40
  "transformers_version": "4.57.0.dev0",
41
- "video_token_id": 151656,
42
  "vision_config": {
43
  "deepstack_visual_indexes": [
44
- 1,
45
- 3,
46
- 5
47
  ],
48
- "depth": 6,
49
  "hidden_act": "gelu_pytorch_tanh",
50
  "hidden_size": 128,
51
  "in_channels": 3,
52
  "initializer_range": 0.02,
53
- "intermediate_size": 64,
54
  "model_type": "qwen3_vl",
55
  "num_heads": 4,
56
- "num_position_embeddings": 2304,
57
- "out_hidden_size": 8,
58
  "patch_size": 16,
59
  "spatial_merge_size": 2,
60
  "temporal_patch_size": 2
61
  },
62
- "vision_end_token_id": 151653,
63
- "vision_start_token_id": 151652
64
  }
 
2
  "architectures": [
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
  }
generation_config.json CHANGED
@@ -1,13 +1,4 @@
1
  {
2
- "bos_token_id": 151643,
3
- "do_sample": true,
4
- "eos_token_id": [
5
- 151645,
6
- 151643
7
- ],
8
- "pad_token_id": 151643,
9
- "top_k": 20,
10
- "top_p": 0.95,
11
- "transformers_version": "4.57.0.dev0",
12
- "trust_remote_code": true
13
  }
 
1
  {
2
+ "_from_model_config": true,
3
+ "transformers_version": "4.57.0.dev0"
 
 
 
 
 
 
 
 
 
4
  }
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:23896cc3671cb2e2bd291026dd8b058abf7d32568675e6775e29579e7a8d4666
3
- size 9028288
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d699b4ddd348288defffa40fb5d20a7f83291ecbfe697262608c1115b61a16ee
3
+ size 7072080