CserDu123 commited on
Commit
c6202c0
·
verified ·
1 Parent(s): 3dcadf8

Upload sft-017-qwen3-vl-30b-a3b-full-tuning-8k-ckpt-738/config.json with huggingface_hub

Browse files
sft-017-qwen3-vl-30b-a3b-full-tuning-8k-ckpt-738/config.json ADDED
@@ -0,0 +1,74 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen3VLMoeForConditionalGeneration"
4
+ ],
5
+ "dtype": "bfloat16",
6
+ "eos_token_id": 151645,
7
+ "image_token_id": 151655,
8
+ "model_type": "qwen3_vl_moe",
9
+ "pad_token_id": 151643,
10
+ "text_config": {
11
+ "attention_bias": false,
12
+ "attention_dropout": 0.0,
13
+ "bos_token_id": 151643,
14
+ "decoder_sparse_step": 1,
15
+ "dtype": "bfloat16",
16
+ "eos_token_id": 151645,
17
+ "head_dim": 128,
18
+ "hidden_act": "silu",
19
+ "hidden_size": 2048,
20
+ "initializer_range": 0.02,
21
+ "intermediate_size": 6144,
22
+ "max_position_embeddings": 262144,
23
+ "mlp_only_layers": [],
24
+ "model_type": "qwen3_vl_moe_text",
25
+ "moe_intermediate_size": 768,
26
+ "norm_topk_prob": true,
27
+ "num_attention_heads": 32,
28
+ "num_experts": 128,
29
+ "num_experts_per_tok": 8,
30
+ "num_hidden_layers": 48,
31
+ "num_key_value_heads": 4,
32
+ "rms_norm_eps": 1e-06,
33
+ "rope_scaling": {
34
+ "mrope_interleaved": true,
35
+ "mrope_section": [
36
+ 24,
37
+ 20,
38
+ 20
39
+ ],
40
+ "rope_type": "default"
41
+ },
42
+ "rope_theta": 5000000,
43
+ "router_aux_loss_coef": 0.001,
44
+ "use_cache": true,
45
+ "vocab_size": 151936
46
+ },
47
+ "tie_word_embeddings": false,
48
+ "transformers_version": "4.57.3",
49
+ "use_cache": false,
50
+ "video_token_id": 151656,
51
+ "vision_config": {
52
+ "deepstack_visual_indexes": [
53
+ 8,
54
+ 16,
55
+ 24
56
+ ],
57
+ "depth": 27,
58
+ "dtype": "bfloat16",
59
+ "hidden_act": "gelu_pytorch_tanh",
60
+ "hidden_size": 1152,
61
+ "in_channels": 3,
62
+ "initializer_range": 0.02,
63
+ "intermediate_size": 4304,
64
+ "model_type": "qwen3_vl_moe",
65
+ "num_heads": 16,
66
+ "num_position_embeddings": 2304,
67
+ "out_hidden_size": 2048,
68
+ "patch_size": 16,
69
+ "spatial_merge_size": 2,
70
+ "temporal_patch_size": 2
71
+ },
72
+ "vision_end_token_id": 151653,
73
+ "vision_start_token_id": 151652
74
+ }