fozziethebeat commited on
Commit
0314bcc
1 Parent(s): 2b8a569

Upload LlavaLlamaForCausalLM

Browse files
config.json CHANGED
@@ -11,7 +11,6 @@
11
  "freeze_mm_vision_resampler": false,
12
  "hidden_act": "silu",
13
  "hidden_size": 4096,
14
- "ignore_index": -100,
15
  "image_aspect_ratio": "anyres",
16
  "image_crop_resolution": 224,
17
  "image_grid_pinpoints": [
@@ -37,7 +36,6 @@
37
  ]
38
  ],
39
  "image_split_resolution": 224,
40
- "image_token_index": 32000,
41
  "initializer_range": 0.02,
42
  "intermediate_size": 11008,
43
  "max_position_embeddings": 4096,
@@ -52,19 +50,15 @@
52
  "mm_vision_select_layer": -2,
53
  "mm_vision_tower": "openai/clip-vit-large-patch14-336",
54
  "mm_vision_tower_lr": 2e-06,
55
- "model_type": "llava",
56
  "num_attention_heads": 32,
57
  "num_hidden_layers": 32,
58
  "num_key_value_heads": 32,
59
- "pad_token_id": 32001,
60
  "pretraining_tp": 1,
61
- "projector_hidden_act": "gelu",
62
  "rms_norm_eps": 1e-05,
63
  "rope_scaling": null,
64
  "rope_theta": 10000.0,
65
- "text_config": {
66
- "model_type": "llama"
67
- },
68
  "tie_word_embeddings": false,
69
  "tokenizer_model_max_length": 4096,
70
  "tokenizer_padding_side": "right",
@@ -75,18 +69,5 @@
75
  "unfreeze_mm_vision_tower": true,
76
  "use_cache": true,
77
  "use_mm_proj": true,
78
- "vision_config": {
79
- "hidden_size": 1024,
80
- "image_size": 336,
81
- "intermediate_size": 4096,
82
- "model_type": "clip_vision_model",
83
- "num_attention_heads": 16,
84
- "num_hidden_layers": 24,
85
- "patch_size": 14,
86
- "projection_dim": 768,
87
- "vocab_size": 32000
88
- },
89
- "vision_feature_layer": -2,
90
- "vision_feature_select_strategy": "default",
91
  "vocab_size": 32000
92
  }
 
11
  "freeze_mm_vision_resampler": false,
12
  "hidden_act": "silu",
13
  "hidden_size": 4096,
 
14
  "image_aspect_ratio": "anyres",
15
  "image_crop_resolution": 224,
16
  "image_grid_pinpoints": [
 
36
  ]
37
  ],
38
  "image_split_resolution": 224,
 
39
  "initializer_range": 0.02,
40
  "intermediate_size": 11008,
41
  "max_position_embeddings": 4096,
 
50
  "mm_vision_select_layer": -2,
51
  "mm_vision_tower": "openai/clip-vit-large-patch14-336",
52
  "mm_vision_tower_lr": 2e-06,
53
+ "model_type": "llava_llama",
54
  "num_attention_heads": 32,
55
  "num_hidden_layers": 32,
56
  "num_key_value_heads": 32,
57
+ "pad_token_id": 0,
58
  "pretraining_tp": 1,
 
59
  "rms_norm_eps": 1e-05,
60
  "rope_scaling": null,
61
  "rope_theta": 10000.0,
 
 
 
62
  "tie_word_embeddings": false,
63
  "tokenizer_model_max_length": 4096,
64
  "tokenizer_padding_side": "right",
 
69
  "unfreeze_mm_vision_tower": true,
70
  "use_cache": true,
71
  "use_mm_proj": true,
 
 
 
 
 
 
 
 
 
 
 
 
 
72
  "vocab_size": 32000
73
  }
model-00001-of-00003.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:fb2f8c87745e0f04853e5d33c831435f2950176f20fd8268293ba87102280382
3
  size 4938993632
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:038a8cb8c9041477364e74f3931843825b6c4abd8edfdd70a3896a81e276ae50
3
  size 4938993632
model-00002-of-00003.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:d7849648d359fcae6a4f21ebc58f37a4a082ab6b5b359a470538dfdfac55b6a6
3
  size 4947390880
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f6ba49dbb3739a19c2fda1c500d0bab2fe0bbaed5cd70e878b0251e96fd11d02
3
  size 4947390880
model-00003-of-00003.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a142597a4b3908fcf66af6e3a96c4410cb962eca262a6ab83f4ae722903869d5
3
  size 4239524944
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c203edc06716c6841607e3d138880a143996a74e014604f132cedbe17ef029d0
3
  size 4239524944