danielhanchen commited on
Commit
4b7ec94
·
verified ·
1 Parent(s): c0975ca

Add files using upload-large-folder tool

Browse files
config.json CHANGED
@@ -30,72 +30,74 @@
30
  "multi_modal_projector",
31
  "merger",
32
  "modality_projection",
33
- "model.layers.8.mlp",
 
 
34
  "model.layers.15.mlp",
35
  "visual.blocks.31.mlp",
36
- "visual.blocks.29.attn",
37
- "visual.blocks.24.attn",
38
- "visual.blocks.21.attn",
39
- "visual.blocks.31.attn",
40
  "visual.blocks.27.attn",
41
- "visual.blocks.26.attn",
42
  "visual.blocks.28.attn",
43
- "visual.blocks.30.attn",
44
- "visual.blocks.20.attn",
45
- "visual.blocks.25.attn",
46
  "visual.merger.mlp",
47
  "visual.blocks.30.mlp",
48
- "model.layers.6.mlp",
 
 
 
 
 
49
  "visual.blocks.22.attn",
 
 
 
 
50
  "visual.blocks.18.attn",
51
  "visual.blocks.26.mlp",
52
- "visual.blocks.29.mlp",
53
  "visual.blocks.27.mlp",
54
- "visual.blocks.19.attn",
55
- "visual.blocks.25.mlp",
56
  "visual.blocks.20.mlp",
 
57
  "visual.blocks.19.mlp",
 
58
  "visual.blocks.23.mlp",
 
59
  "visual.blocks.23.attn",
60
- "visual.blocks.17.attn",
61
- "visual.blocks.24.mlp",
62
- "visual.blocks.28.mlp",
63
- "visual.blocks.10.attn",
64
- "visual.blocks.18.mlp",
65
  "visual.blocks.22.mlp",
66
- "visual.blocks.11.mlp",
67
- "visual.blocks.14.attn",
68
  "visual.blocks.12.attn",
69
- "visual.blocks.21.mlp",
70
  "visual.blocks.9.attn",
 
71
  "visual.blocks.16.attn",
72
- "visual.blocks.11.attn",
 
 
 
 
 
73
  "visual.blocks.9.mlp",
 
 
 
74
  "visual.blocks.4.mlp",
75
- "visual.blocks.13.attn",
76
- "visual.blocks.6.mlp",
77
- "visual.blocks.10.mlp",
78
  "visual.blocks.16.mlp",
79
- "visual.blocks.8.attn",
80
- "visual.blocks.2.attn",
81
- "visual.blocks.5.mlp",
82
- "visual.blocks.15.mlp",
83
- "visual.blocks.2.mlp",
84
  "visual.blocks.8.mlp",
85
- "visual.blocks.6.attn",
 
 
86
  "visual.blocks.7.mlp",
87
- "visual.blocks.12.mlp",
88
- "visual.blocks.3.mlp",
89
  "visual.blocks.13.mlp",
90
- "visual.blocks.15.attn",
91
- "visual.blocks.0.attn",
 
92
  "visual.blocks.1.attn",
93
- "visual.blocks.3.attn",
94
- "visual.blocks.1.mlp",
95
  "visual.blocks.5.attn",
 
 
96
  "visual.blocks.4.attn",
97
- "visual.blocks.14.mlp",
98
  "visual.blocks.7.attn",
 
 
99
  "visual.blocks.0.mlp",
100
  "visual.blocks.17.mlp"
101
  ],
@@ -118,7 +120,7 @@
118
  "sliding_window": 32768,
119
  "tie_word_embeddings": false,
120
  "torch_dtype": "bfloat16",
121
- "transformers_version": "4.51.0.dev0",
122
  "unsloth_fixed": true,
123
  "use_cache": true,
124
  "use_sliding_window": false,
 
30
  "multi_modal_projector",
31
  "merger",
32
  "modality_projection",
33
+ "model.layers.10.mlp",
34
+ "model.layers.7.mlp",
35
+ "model.layers.22.mlp",
36
  "model.layers.15.mlp",
37
  "visual.blocks.31.mlp",
 
 
 
 
38
  "visual.blocks.27.attn",
 
39
  "visual.blocks.28.attn",
40
+ "visual.blocks.31.attn",
 
 
41
  "visual.merger.mlp",
42
  "visual.blocks.30.mlp",
43
+ "visual.blocks.29.attn",
44
+ "visual.blocks.26.attn",
45
+ "visual.blocks.24.attn",
46
+ "visual.blocks.30.attn",
47
+ "visual.blocks.21.attn",
48
+ "visual.blocks.25.attn",
49
  "visual.blocks.22.attn",
50
+ "visual.blocks.29.mlp",
51
+ "visual.blocks.19.attn",
52
+ "visual.blocks.20.attn",
53
+ "model.layers.6.mlp",
54
  "visual.blocks.18.attn",
55
  "visual.blocks.26.mlp",
 
56
  "visual.blocks.27.mlp",
 
 
57
  "visual.blocks.20.mlp",
58
+ "visual.blocks.18.mlp",
59
  "visual.blocks.19.mlp",
60
+ "visual.blocks.28.mlp",
61
  "visual.blocks.23.mlp",
62
+ "visual.blocks.25.mlp",
63
  "visual.blocks.23.attn",
 
 
 
 
 
64
  "visual.blocks.22.mlp",
65
+ "visual.blocks.17.attn",
 
66
  "visual.blocks.12.attn",
67
+ "visual.blocks.11.attn",
68
  "visual.blocks.9.attn",
69
+ "visual.blocks.21.mlp",
70
  "visual.blocks.16.attn",
71
+ "visual.blocks.14.attn",
72
+ "visual.blocks.10.mlp",
73
+ "visual.blocks.10.attn",
74
+ "visual.blocks.13.attn",
75
+ "visual.blocks.24.mlp",
76
+ "visual.blocks.11.mlp",
77
  "visual.blocks.9.mlp",
78
+ "visual.blocks.6.attn",
79
+ "visual.blocks.5.mlp",
80
+ "visual.blocks.12.mlp",
81
  "visual.blocks.4.mlp",
 
 
 
82
  "visual.blocks.16.mlp",
 
 
 
 
 
83
  "visual.blocks.8.mlp",
84
+ "visual.blocks.6.mlp",
85
+ "visual.blocks.2.attn",
86
+ "visual.blocks.1.mlp",
87
  "visual.blocks.7.mlp",
88
+ "visual.blocks.8.attn",
 
89
  "visual.blocks.13.mlp",
90
+ "visual.blocks.3.mlp",
91
+ "visual.blocks.15.mlp",
92
+ "visual.blocks.2.mlp",
93
  "visual.blocks.1.attn",
 
 
94
  "visual.blocks.5.attn",
95
+ "visual.blocks.0.attn",
96
+ "visual.blocks.15.attn",
97
  "visual.blocks.4.attn",
 
98
  "visual.blocks.7.attn",
99
+ "visual.blocks.3.attn",
100
+ "visual.blocks.14.mlp",
101
  "visual.blocks.0.mlp",
102
  "visual.blocks.17.mlp"
103
  ],
 
120
  "sliding_window": 32768,
121
  "tie_word_embeddings": false,
122
  "torch_dtype": "bfloat16",
123
+ "transformers_version": "4.50.1",
124
  "unsloth_fixed": true,
125
  "use_cache": true,
126
  "use_sliding_window": false,
generation_config.json CHANGED
@@ -9,5 +9,5 @@
9
  "pad_token_id": 151654,
10
  "repetition_penalty": 1.05,
11
  "temperature": 1e-06,
12
- "transformers_version": "4.51.0.dev0"
13
  }
 
9
  "pad_token_id": 151654,
10
  "repetition_penalty": 1.05,
11
  "temperature": 1e-06,
12
+ "transformers_version": "4.50.1"
13
  }
model-00002-of-00005.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:1bcbd8320b7662d0de082c66854b22e374607dad8b39fdded60baea49e9c7add
3
- size 4951256662
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:64f31c761976fdd3dba7117d1d8dddbda583cba145359980b73b04eb785c0170
3
+ size 4991807831
model-00003-of-00005.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:13cf5219c93b0b3de10a578e74dd06935e5e99139eed70d58008935489652610
3
- size 4958587465
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bbaa6846f566db7ad67a1caff12f4f8eab7b1cd9ce496d6a7a3b6214426079c2
3
+ size 4993690414
model-00004-of-00005.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:c7ffad638286c5a1725360cc42e0c831349a05e1dd1e1bee205a9ee252747a35
3
- size 4999136404
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7f1fd7b300eb6bd4b34e7f78c83faad6baaae561852651cb8ac70318af128fc5
3
+ size 4958587451
model-00005-of-00005.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:af9e7bbb3ff227428ea6a0df5e1502f3553b530d4774ad5e66641b932f69046e
3
- size 2815049538
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e4469c46443a245744089c31dfd38d43e7417a514799b62e72424d2037d9f0d3
3
+ size 4040475063
model.safetensors.index.json CHANGED
The diff for this file is too large to render. See raw diff