text_encoder/config.json
1.5 KB · 67 lines · json Raw
1 {
2 "architectures": [
3 "Qwen3VLModel"
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": 128,
15 "hidden_act": "silu",
16 "hidden_size": 2560,
17 "initializer_range": 0.02,
18 "intermediate_size": 9728,
19 "max_position_embeddings": 262144,
20 "model_type": "qwen3_vl_text",
21 "num_attention_heads": 32,
22 "num_hidden_layers": 36,
23 "num_key_value_heads": 8,
24 "pad_token_id": null,
25 "rms_norm_eps": 1e-06,
26 "rope_parameters": {
27 "mrope_interleaved": true,
28 "mrope_section": [
29 24,
30 20,
31 20
32 ],
33 "rope_theta": 5000000,
34 "rope_type": "default"
35 },
36 "tie_word_embeddings": true,
37 "use_cache": true,
38 "vocab_size": 151936
39 },
40 "tie_word_embeddings": true,
41 "transformers_version": "5.2.0",
42 "video_token_id": 151656,
43 "vision_config": {
44 "deepstack_visual_indexes": [
45 5,
46 11,
47 17
48 ],
49 "depth": 24,
50 "dtype": "bfloat16",
51 "hidden_act": "gelu_pytorch_tanh",
52 "hidden_size": 1024,
53 "in_channels": 3,
54 "initializer_range": 0.02,
55 "intermediate_size": 4096,
56 "model_type": "qwen3_vl",
57 "num_heads": 16,
58 "num_position_embeddings": 2304,
59 "out_hidden_size": 2560,
60 "patch_size": 16,
61 "spatial_merge_size": 2,
62 "temporal_patch_size": 2
63 },
64 "vision_end_token_id": 151653,
65 "vision_start_token_id": 151652
66 }
67