JeasLee commited on
Commit
2ff932c
·
verified ·
1 Parent(s): b98539b

Delete RoboInterVLM_qwenvl25_3b/config.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. RoboInterVLM_qwenvl25_3b/config.json +0 -123
RoboInterVLM_qwenvl25_3b/config.json DELETED
@@ -1,123 +0,0 @@
1
- {
2
- "adaptive_gate": true,
3
- "architectures": [
4
- "Qwen2_5_VLForConditionalGeneration"
5
- ],
6
- "attention_dropout": 0.0,
7
- "bos_token_id": 151643,
8
- "decoder_sparse_step": 1,
9
- "eos_token_id": 151645,
10
- "hidden_act": "silu",
11
- "hidden_size": 2048,
12
- "image_token_id": 151655,
13
- "initializer_range": 0.02,
14
- "intermediate_size": 11008,
15
- "max_position_embeddings": 128000,
16
- "max_window_layers": 70,
17
- "mlp_only_layers": [],
18
- "model_type": "qwen2_5_vl",
19
- "moe_intermediate_size": 256,
20
- "norm_topk_prob": 0.01,
21
- "num_attention_heads": 16,
22
- "num_experts": 4,
23
- "num_experts_per_tok": 2,
24
- "num_hidden_layers": 36,
25
- "num_key_value_heads": 2,
26
- "output_router_logits": false,
27
- "rms_norm_eps": 1e-06,
28
- "rope_scaling": {
29
- "mrope_section": [
30
- 16,
31
- 24,
32
- 24
33
- ],
34
- "rope_type": "default",
35
- "type": "default"
36
- },
37
- "rope_theta": 1000000.0,
38
- "router_aux_loss_coef": 0.001,
39
- "sliding_window": 32768,
40
- "text_config": {
41
- "adaptive_gate": true,
42
- "architectures": [
43
- "Qwen2_5_VLForConditionalGeneration"
44
- ],
45
- "attention_dropout": 0.0,
46
- "bos_token_id": 151643,
47
- "decoder_sparse_step": 1,
48
- "eos_token_id": 151645,
49
- "hidden_act": "silu",
50
- "hidden_size": 2048,
51
- "image_token_id": null,
52
- "initializer_range": 0.02,
53
- "intermediate_size": 11008,
54
- "max_position_embeddings": 128000,
55
- "max_window_layers": 70,
56
- "mlp_only_layers": [],
57
- "model_type": "qwen2_5_vl_text",
58
- "moe_intermediate_size": 256,
59
- "norm_topk_prob": 0.01,
60
- "num_attention_heads": 16,
61
- "num_experts": 4,
62
- "num_experts_per_tok": 2,
63
- "num_hidden_layers": 36,
64
- "num_key_value_heads": 2,
65
- "output_router_logits": false,
66
- "rms_norm_eps": 1e-06,
67
- "rope_scaling": {
68
- "mrope_section": [
69
- 16,
70
- 24,
71
- 24
72
- ],
73
- "rope_type": "default",
74
- "type": "default"
75
- },
76
- "rope_theta": 1000000.0,
77
- "router_aux_loss_coef": 0.001,
78
- "sliding_window": 32768,
79
- "tie_word_embeddings": true,
80
- "torch_dtype": "bfloat16",
81
- "use_cache": true,
82
- "use_sliding_window": false,
83
- "video_token_id": null,
84
- "vision_end_token_id": 151653,
85
- "vision_start_token_id": 151652,
86
- "vision_token_id": 151654,
87
- "vocab_size": 151936
88
- },
89
- "torch_dtype": "bfloat16",
90
- "transformers_version": "4.51.3",
91
- "use_cache": true,
92
- "use_sliding_window": false,
93
- "video_token_id": 151656,
94
- "vision_config": {
95
- "depth": 32,
96
- "fullatt_block_indexes": [
97
- 7,
98
- 15,
99
- 23,
100
- 31
101
- ],
102
- "hidden_act": "silu",
103
- "hidden_size": 1280,
104
- "in_channels": 3,
105
- "in_chans": 3,
106
- "initializer_range": 0.02,
107
- "intermediate_size": 3420,
108
- "model_type": "qwen2_5_vl",
109
- "num_heads": 16,
110
- "out_hidden_size": 2048,
111
- "patch_size": 14,
112
- "spatial_merge_size": 2,
113
- "spatial_patch_size": 14,
114
- "temporal_patch_size": 2,
115
- "tokens_per_second": 2,
116
- "torch_dtype": "bfloat16",
117
- "window_size": 112
118
- },
119
- "vision_end_token_id": 151653,
120
- "vision_start_token_id": 151652,
121
- "vision_token_id": 151654,
122
- "vocab_size": 151936
123
- }