testdummyvt commited on
Commit
80f40a1
·
verified ·
1 Parent(s): 0c04262

Upload Qwen2_5_VLForConditionalGeneration

Browse files
Files changed (2) hide show
  1. config.json +3 -45
  2. generation_config.json +1 -1
config.json CHANGED
@@ -17,7 +17,6 @@
17
  "num_hidden_layers": 36,
18
  "num_key_value_heads": 2,
19
  "quantization_config": {
20
- "include_input_output_embeddings": false,
21
  "modules_to_not_convert": null,
22
  "quant_method": "torchao",
23
  "quant_type": {
@@ -39,8 +38,7 @@
39
  "_version": 1
40
  }
41
  },
42
- "quant_type_kwargs": {},
43
- "untie_embedding_weights": false
44
  },
45
  "rms_norm_eps": 1e-06,
46
  "rope_scaling": {
@@ -54,48 +52,9 @@
54
  },
55
  "rope_theta": 1000000.0,
56
  "sliding_window": 32768,
57
- "text_config": {
58
- "architectures": [
59
- "Qwen2_5_VLForConditionalGeneration"
60
- ],
61
- "attention_dropout": 0.0,
62
- "bos_token_id": 151643,
63
- "eos_token_id": 151645,
64
- "hidden_act": "silu",
65
- "hidden_size": 2048,
66
- "image_token_id": null,
67
- "initializer_range": 0.02,
68
- "intermediate_size": 11008,
69
- "max_position_embeddings": 128000,
70
- "max_window_layers": 70,
71
- "model_type": "qwen2_5_vl_text",
72
- "num_attention_heads": 16,
73
- "num_hidden_layers": 36,
74
- "num_key_value_heads": 2,
75
- "rms_norm_eps": 1e-06,
76
- "rope_scaling": {
77
- "mrope_section": [
78
- 16,
79
- 24,
80
- 24
81
- ],
82
- "rope_type": "default",
83
- "type": "default"
84
- },
85
- "rope_theta": 1000000.0,
86
- "sliding_window": 32768,
87
- "tie_word_embeddings": true,
88
- "torch_dtype": "bfloat16",
89
- "use_cache": true,
90
- "use_sliding_window": false,
91
- "video_token_id": null,
92
- "vision_end_token_id": 151653,
93
- "vision_start_token_id": 151652,
94
- "vision_token_id": 151654,
95
- "vocab_size": 151936
96
- },
97
  "torch_dtype": "bfloat16",
98
- "transformers_version": "4.52.4",
99
  "use_cache": true,
100
  "use_sliding_window": false,
101
  "video_token_id": 151656,
@@ -111,7 +70,6 @@
111
  "hidden_size": 1280,
112
  "in_channels": 3,
113
  "in_chans": 3,
114
- "initializer_range": 0.02,
115
  "intermediate_size": 3420,
116
  "model_type": "qwen2_5_vl",
117
  "num_heads": 16,
 
17
  "num_hidden_layers": 36,
18
  "num_key_value_heads": 2,
19
  "quantization_config": {
 
20
  "modules_to_not_convert": null,
21
  "quant_method": "torchao",
22
  "quant_type": {
 
38
  "_version": 1
39
  }
40
  },
41
+ "quant_type_kwargs": {}
 
42
  },
43
  "rms_norm_eps": 1e-06,
44
  "rope_scaling": {
 
52
  },
53
  "rope_theta": 1000000.0,
54
  "sliding_window": 32768,
55
+ "tie_word_embeddings": true,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
56
  "torch_dtype": "bfloat16",
57
+ "transformers_version": "4.51.3",
58
  "use_cache": true,
59
  "use_sliding_window": false,
60
  "video_token_id": 151656,
 
70
  "hidden_size": 1280,
71
  "in_channels": 3,
72
  "in_chans": 3,
 
73
  "intermediate_size": 3420,
74
  "model_type": "qwen2_5_vl",
75
  "num_heads": 16,
generation_config.json CHANGED
@@ -8,5 +8,5 @@
8
  "pad_token_id": 151643,
9
  "repetition_penalty": 1.05,
10
  "temperature": 1e-06,
11
- "transformers_version": "4.52.4"
12
  }
 
8
  "pad_token_id": 151643,
9
  "repetition_penalty": 1.05,
10
  "temperature": 1e-06,
11
+ "transformers_version": "4.51.3"
12
  }