Hugging Face's logo

Journey9ni
/
llava_video_7b_qwen2_lora_base

PEFT
llava
Model card Files
xet
Community
Journey9ni commited on
Commit
057ae4f
·
verified ·
1 Parent(s): 3ac7b40

Upload model files excluding checkpoints

Browse files
Files changed (7) hide show
  1. README.md +0 -0
  2. adapter_config.json +215 -0
  3. adapter_model.bin +3 -0
  4. config.json +228 -0
  5. generation_config.json +14 -0
  6. non_lora_trainables.bin +3 -0
  7. trainer_state.json +0 -0
README.md ADDED
Binary file (178 Bytes). View file
 
adapter_config.json ADDED
@@ -0,0 +1,215 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "auto_mapping": null,
3
+ "base_model_name_or_path": "checkpoints/LLaVA-Video-7B-Qwen2",
4
+ "bias": "none",
5
+ "fan_in_fan_out": false,
6
+ "inference_mode": true,
7
+ "init_lora_weights": true,
8
+ "layers_pattern": null,
9
+ "layers_to_transform": null,
10
+ "lora_alpha": 256,
11
+ "lora_dropout": 0.05,
12
+ "modules_to_save": null,
13
+ "peft_type": "LORA",
14
+ "r": 128,
15
+ "revision": null,
16
+ "target_modules": [
17
+ "model.layers.3.self_attn.k_proj",
18
+ "model.layers.4.mlp.down_proj",
19
+ "model.layers.22.mlp.gate_proj",
20
+ "model.layers.11.self_attn.q_proj",
21
+ "model.layers.14.mlp.gate_proj",
22
+ "model.layers.16.mlp.gate_proj",
23
+ "model.layers.8.mlp.down_proj",
24
+ "model.layers.25.mlp.up_proj",
25
+ "model.layers.2.self_attn.k_proj",
26
+ "model.layers.26.self_attn.o_proj",
27
+ "model.layers.26.mlp.up_proj",
28
+ "model.layers.3.mlp.up_proj",
29
+ "model.layers.16.mlp.up_proj",
30
+ "model.layers.20.mlp.gate_proj",
31
+ "model.layers.26.self_attn.k_proj",
32
+ "model.layers.4.self_attn.o_proj",
33
+ "model.layers.26.mlp.gate_proj",
34
+ "model.layers.2.mlp.down_proj",
35
+ "model.layers.10.mlp.gate_proj",
36
+ "model.layers.17.self_attn.o_proj",
37
+ "model.layers.19.self_attn.k_proj",
38
+ "model.layers.7.self_attn.o_proj",
39
+ "model.layers.10.self_attn.k_proj",
40
+ "model.layers.23.mlp.gate_proj",
41
+ "model.layers.11.self_attn.v_proj",
42
+ "model.layers.21.mlp.down_proj",
43
+ "model.layers.8.self_attn.k_proj",
44
+ "model.layers.2.mlp.gate_proj",
45
+ "model.layers.27.mlp.down_proj",
46
+ "model.layers.8.mlp.up_proj",
47
+ "model.layers.9.self_attn.v_proj",
48
+ "model.layers.7.mlp.gate_proj",
49
+ "model.layers.5.mlp.gate_proj",
50
+ "model.layers.15.self_attn.k_proj",
51
+ "model.layers.12.self_attn.k_proj",
52
+ "model.layers.4.self_attn.k_proj",
53
+ "model.layers.17.mlp.up_proj",
54
+ "model.layers.25.mlp.gate_proj",
55
+ "model.layers.8.self_attn.o_proj",
56
+ "model.layers.0.self_attn.o_proj",
57
+ "model.layers.6.self_attn.v_proj",
58
+ "model.layers.2.self_attn.q_proj",
59
+ "model.layers.13.mlp.down_proj",
60
+ "model.layers.1.self_attn.v_proj",
61
+ "model.layers.21.self_attn.q_proj",
62
+ "model.layers.5.self_attn.o_proj",
63
+ "model.layers.20.self_attn.k_proj",
64
+ "model.layers.6.mlp.gate_proj",
65
+ "model.layers.27.self_attn.q_proj",
66
+ "model.layers.13.mlp.up_proj",
67
+ "model.layers.14.self_attn.q_proj",
68
+ "model.layers.18.self_attn.q_proj",
69
+ "model.layers.19.self_attn.v_proj",
70
+ "model.layers.0.mlp.up_proj",
71
+ "model.layers.23.self_attn.v_proj",
72
+ "model.layers.13.mlp.gate_proj",
73
+ "model.layers.10.self_attn.o_proj",
74
+ "model.layers.6.mlp.up_proj",
75
+ "model.layers.11.self_attn.k_proj",
76
+ "model.layers.15.mlp.up_proj",
77
+ "model.layers.9.self_attn.k_proj",
78
+ "model.layers.20.self_attn.q_proj",
79
+ "model.layers.22.mlp.down_proj",
80
+ "model.layers.26.self_attn.v_proj",
81
+ "model.layers.26.mlp.down_proj",
82
+ "model.layers.18.self_attn.o_proj",
83
+ "model.layers.27.self_attn.v_proj",
84
+ "model.layers.15.self_attn.q_proj",
85
+ "model.layers.16.mlp.down_proj",
86
+ "model.layers.15.self_attn.o_proj",
87
+ "model.layers.2.self_attn.v_proj",
88
+ "model.layers.3.self_attn.v_proj",
89
+ "model.layers.3.self_attn.o_proj",
90
+ "model.layers.6.self_attn.k_proj",
91
+ "model.layers.17.mlp.gate_proj",
92
+ "model.layers.18.mlp.up_proj",
93
+ "model.layers.4.self_attn.q_proj",
94
+ "model.layers.20.self_attn.v_proj",
95
+ "model.layers.6.mlp.down_proj",
96
+ "model.layers.27.mlp.gate_proj",
97
+ "model.layers.17.self_attn.v_proj",
98
+ "model.layers.12.self_attn.q_proj",
99
+ "model.layers.1.mlp.gate_proj",
100
+ "model.layers.21.self_attn.k_proj",
101
+ "model.layers.15.self_attn.v_proj",
102
+ "model.layers.16.self_attn.k_proj",
103
+ "model.layers.27.self_attn.o_proj",
104
+ "model.layers.12.self_attn.o_proj",
105
+ "model.layers.18.self_attn.v_proj",
106
+ "model.layers.22.self_attn.o_proj",
107
+ "model.layers.5.self_attn.v_proj",
108
+ "model.layers.11.mlp.down_proj",
109
+ "model.layers.13.self_attn.q_proj",
110
+ "model.layers.21.self_attn.o_proj",
111
+ "model.layers.10.mlp.down_proj",
112
+ "model.layers.0.self_attn.q_proj",
113
+ "model.layers.13.self_attn.k_proj",
114
+ "model.layers.22.mlp.up_proj",
115
+ "model.layers.20.self_attn.o_proj",
116
+ "model.layers.23.self_attn.o_proj",
117
+ "model.layers.21.mlp.up_proj",
118
+ "model.layers.21.self_attn.v_proj",
119
+ "model.layers.5.self_attn.k_proj",
120
+ "model.layers.6.self_attn.o_proj",
121
+ "model.layers.9.mlp.up_proj",
122
+ "model.layers.8.mlp.gate_proj",
123
+ "model.layers.19.mlp.up_proj",
124
+ "model.layers.19.mlp.down_proj",
125
+ "model.layers.24.self_attn.q_proj",
126
+ "model.layers.23.mlp.up_proj",
127
+ "model.layers.24.self_attn.v_proj",
128
+ "model.layers.19.self_attn.q_proj",
129
+ "model.layers.9.mlp.gate_proj",
130
+ "model.layers.17.self_attn.k_proj",
131
+ "model.layers.25.self_attn.q_proj",
132
+ "model.layers.9.self_attn.o_proj",
133
+ "model.layers.3.mlp.gate_proj",
134
+ "model.layers.4.mlp.gate_proj",
135
+ "model.layers.27.mlp.up_proj",
136
+ "model.layers.3.self_attn.q_proj",
137
+ "model.layers.14.self_attn.o_proj",
138
+ "model.layers.3.mlp.down_proj",
139
+ "model.layers.8.self_attn.q_proj",
140
+ "model.layers.12.mlp.gate_proj",
141
+ "model.layers.7.self_attn.k_proj",
142
+ "model.layers.8.self_attn.v_proj",
143
+ "model.layers.20.mlp.up_proj",
144
+ "model.layers.22.self_attn.q_proj",
145
+ "model.layers.0.mlp.down_proj",
146
+ "model.layers.13.self_attn.o_proj",
147
+ "model.layers.10.mlp.up_proj",
148
+ "model.layers.25.self_attn.o_proj",
149
+ "model.layers.26.self_attn.q_proj",
150
+ "model.layers.12.self_attn.v_proj",
151
+ "model.layers.22.self_attn.k_proj",
152
+ "model.layers.10.self_attn.q_proj",
153
+ "model.layers.1.mlp.down_proj",
154
+ "model.layers.2.mlp.up_proj",
155
+ "model.layers.1.mlp.up_proj",
156
+ "model.layers.19.mlp.gate_proj",
157
+ "model.layers.0.self_attn.v_proj",
158
+ "model.layers.24.mlp.up_proj",
159
+ "model.layers.14.mlp.up_proj",
160
+ "model.layers.12.mlp.up_proj",
161
+ "model.layers.4.self_attn.v_proj",
162
+ "model.layers.25.self_attn.k_proj",
163
+ "model.layers.1.self_attn.q_proj",
164
+ "model.layers.23.mlp.down_proj",
165
+ "model.layers.5.mlp.up_proj",
166
+ "model.layers.7.self_attn.q_proj",
167
+ "model.layers.24.self_attn.k_proj",
168
+ "model.layers.14.mlp.down_proj",
169
+ "model.layers.24.mlp.gate_proj",
170
+ "model.layers.24.mlp.down_proj",
171
+ "model.layers.11.mlp.up_proj",
172
+ "model.layers.16.self_attn.v_proj",
173
+ "model.layers.15.mlp.down_proj",
174
+ "model.layers.2.self_attn.o_proj",
175
+ "model.layers.7.mlp.up_proj",
176
+ "model.layers.6.self_attn.q_proj",
177
+ "model.layers.0.mlp.gate_proj",
178
+ "model.layers.1.self_attn.k_proj",
179
+ "model.layers.11.mlp.gate_proj",
180
+ "model.layers.18.self_attn.k_proj",
181
+ "model.layers.21.mlp.gate_proj",
182
+ "model.layers.13.self_attn.v_proj",
183
+ "model.layers.10.self_attn.v_proj",
184
+ "model.layers.16.self_attn.o_proj",
185
+ "model.layers.17.self_attn.q_proj",
186
+ "model.layers.14.self_attn.v_proj",
187
+ "model.layers.12.mlp.down_proj",
188
+ "model.layers.25.self_attn.v_proj",
189
+ "model.layers.5.mlp.down_proj",
190
+ "model.layers.9.self_attn.q_proj",
191
+ "model.layers.0.self_attn.k_proj",
192
+ "model.layers.7.self_attn.v_proj",
193
+ "model.layers.11.self_attn.o_proj",
194
+ "model.layers.1.self_attn.o_proj",
195
+ "model.layers.24.self_attn.o_proj",
196
+ "model.layers.18.mlp.down_proj",
197
+ "model.layers.18.mlp.gate_proj",
198
+ "model.layers.9.mlp.down_proj",
199
+ "model.layers.20.mlp.down_proj",
200
+ "model.layers.5.self_attn.q_proj",
201
+ "model.layers.15.mlp.gate_proj",
202
+ "model.layers.7.mlp.down_proj",
203
+ "model.layers.19.self_attn.o_proj",
204
+ "model.layers.25.mlp.down_proj",
205
+ "model.layers.4.mlp.up_proj",
206
+ "model.layers.17.mlp.down_proj",
207
+ "model.layers.22.self_attn.v_proj",
208
+ "model.layers.23.self_attn.q_proj",
209
+ "model.layers.23.self_attn.k_proj",
210
+ "model.layers.16.self_attn.q_proj",
211
+ "model.layers.14.self_attn.k_proj",
212
+ "model.layers.27.self_attn.k_proj"
213
+ ],
214
+ "task_type": "CAUSAL_LM"
215
+ }
adapter_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:86239c7f571a168b191a01c88f5ca6e24b44a1ab7295d04d48ce866d0b844b99
3
+ size 646059050
config.json ADDED
@@ -0,0 +1,228 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "checkpoints/LLaVA-Video-7B-Qwen2",
3
+ "add_faster_video": false,
4
+ "add_time_instruction": true,
5
+ "architectures": [
6
+ "LlavaQwenForCausalLM"
7
+ ],
8
+ "attention_dropout": 0.0,
9
+ "bos_token_id": 151643,
10
+ "eos_token_id": 151645,
11
+ "faster_token_stride": 10,
12
+ "force_sample": true,
13
+ "freeze_mm_mlp_adapter": false,
14
+ "freeze_mm_vision_resampler": false,
15
+ "hidden_act": "silu",
16
+ "hidden_size": 3584,
17
+ "ignore_index": -100,
18
+ "image_aspect_ratio": "anyres_max_9",
19
+ "image_crop_resolution": null,
20
+ "image_grid_pinpoints": [
21
+ [
22
+ 384,
23
+ 384
24
+ ],
25
+ [
26
+ 384,
27
+ 768
28
+ ],
29
+ [
30
+ 384,
31
+ 1152
32
+ ],
33
+ [
34
+ 384,
35
+ 1536
36
+ ],
37
+ [
38
+ 384,
39
+ 1920
40
+ ],
41
+ [
42
+ 384,
43
+ 2304
44
+ ],
45
+ [
46
+ 768,
47
+ 384
48
+ ],
49
+ [
50
+ 768,
51
+ 768
52
+ ],
53
+ [
54
+ 768,
55
+ 1152
56
+ ],
57
+ [
58
+ 768,
59
+ 1536
60
+ ],
61
+ [
62
+ 768,
63
+ 1920
64
+ ],
65
+ [
66
+ 768,
67
+ 2304
68
+ ],
69
+ [
70
+ 1152,
71
+ 384
72
+ ],
73
+ [
74
+ 1152,
75
+ 768
76
+ ],
77
+ [
78
+ 1152,
79
+ 1152
80
+ ],
81
+ [
82
+ 1152,
83
+ 1536
84
+ ],
85
+ [
86
+ 1152,
87
+ 1920
88
+ ],
89
+ [
90
+ 1152,
91
+ 2304
92
+ ],
93
+ [
94
+ 1536,
95
+ 384
96
+ ],
97
+ [
98
+ 1536,
99
+ 768
100
+ ],
101
+ [
102
+ 1536,
103
+ 1152
104
+ ],
105
+ [
106
+ 1536,
107
+ 1536
108
+ ],
109
+ [
110
+ 1536,
111
+ 1920
112
+ ],
113
+ [
114
+ 1536,
115
+ 2304
116
+ ],
117
+ [
118
+ 1920,
119
+ 384
120
+ ],
121
+ [
122
+ 1920,
123
+ 768
124
+ ],
125
+ [
126
+ 1920,
127
+ 1152
128
+ ],
129
+ [
130
+ 1920,
131
+ 1536
132
+ ],
133
+ [
134
+ 1920,
135
+ 1920
136
+ ],
137
+ [
138
+ 1920,
139
+ 2304
140
+ ],
141
+ [
142
+ 2304,
143
+ 384
144
+ ],
145
+ [
146
+ 2304,
147
+ 768
148
+ ],
149
+ [
150
+ 2304,
151
+ 1152
152
+ ],
153
+ [
154
+ 2304,
155
+ 1536
156
+ ],
157
+ [
158
+ 2304,
159
+ 1920
160
+ ],
161
+ [
162
+ 2304,
163
+ 2304
164
+ ]
165
+ ],
166
+ "image_split_resolution": null,
167
+ "image_token_index": 151646,
168
+ "initializer_range": 0.02,
169
+ "intermediate_size": 18944,
170
+ "max_position_embeddings": 32768,
171
+ "max_window_layers": 28,
172
+ "mm_hidden_size": 1152,
173
+ "mm_newline_position": "grid",
174
+ "mm_patch_merge_type": "spatial_unpad",
175
+ "mm_projector_lr": null,
176
+ "mm_projector_type": "mlp2x_gelu",
177
+ "mm_resampler_type": null,
178
+ "mm_spatial_pool_mode": "bilinear",
179
+ "mm_spatial_pool_stride": 2,
180
+ "mm_tunable_parts": "mm_vision_tower,mm_mlp_adapter,mm_language_model",
181
+ "mm_use_im_patch_token": false,
182
+ "mm_use_im_start_end": false,
183
+ "mm_vision_select_feature": "patch",
184
+ "mm_vision_select_layer": -2,
185
+ "mm_vision_tower": "google/siglip-so400m-patch14-384",
186
+ "mm_vision_tower_lr": null,
187
+ "model_type": "llava",
188
+ "num_attention_heads": 28,
189
+ "num_hidden_layers": 28,
190
+ "num_key_value_heads": 4,
191
+ "pos_skipping_range": 4096,
192
+ "projector_hidden_act": "gelu",
193
+ "rms_norm_eps": 1e-06,
194
+ "rope_scaling": null,
195
+ "rope_theta": 1000000.0,
196
+ "sliding_window": 131072,
197
+ "text_config": {
198
+ "model_type": "llama"
199
+ },
200
+ "tie_word_embeddings": false,
201
+ "tokenizer_model_max_length": 32768,
202
+ "tokenizer_padding_side": "right",
203
+ "torch_dtype": "bfloat16",
204
+ "transformers_version": "4.40.0.dev0",
205
+ "tune_fusion_block": false,
206
+ "tune_mm_mlp_adapter": true,
207
+ "tune_mm_vision_resampler": false,
208
+ "tune_spatial_tower": false,
209
+ "unfreeze_mm_vision_tower": false,
210
+ "use_cache": true,
211
+ "use_mm_proj": true,
212
+ "use_pos_skipping": false,
213
+ "use_sliding_window": false,
214
+ "vision_config": {
215
+ "hidden_size": 1024,
216
+ "image_size": 336,
217
+ "intermediate_size": 4096,
218
+ "model_type": "clip_vision_model",
219
+ "num_attention_heads": 16,
220
+ "num_hidden_layers": 24,
221
+ "patch_size": 14,
222
+ "projection_dim": 768,
223
+ "vocab_size": 32000
224
+ },
225
+ "vision_feature_layer": -2,
226
+ "vision_feature_select_strategy": "default",
227
+ "vision_tower_pretrained": null
228
+ }
generation_config.json ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 151643,
3
+ "do_sample": true,
4
+ "eos_token_id": [
5
+ 151645,
6
+ 151643
7
+ ],
8
+ "pad_token_id": 151643,
9
+ "repetition_penalty": 1.05,
10
+ "temperature": 0.7,
11
+ "top_k": 20,
12
+ "top_p": 0.8,
13
+ "transformers_version": "4.40.0.dev0"
14
+ }
non_lora_trainables.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7d8b240180a8f405502ff22f76ff6d2513eab265badfce5dc855213ef5b41957
3
+ size 33964208
trainer_state.json ADDED
The diff for this file is too large to render. See raw diff