Hugging Face's logo

Journey9ni
/
Spatial_LLM

llava
Model card Files Community
Journey9ni commited on
Commit
5d69b94
·
verified ·
1 Parent(s): 18390d6

Upload folder using huggingface_hub

Browse files
Files changed (7) hide show
  1. 20_23/README.md +15 -0
  2. 20_23/adapter_config.json +215 -0
  3. 20_23/adapter_model.bin +3 -0
  4. 20_23/config.json +231 -0
  5. 20_23/generation_config.json +14 -0
  6. 20_23/non_lora_trainables.bin +3 -0
  7. 20_23/trainer_state.json +0 -0
20_23/README.md ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: peft
3
+ ---
4
+ ## Training procedure
5
+
6
+ ### Framework versions
7
+
8
+ - PEFT 0.4.0
9
+ - PEFT 0.4.0
10
+
11
+ - PEFT 0.4.0
12
+ rk versions
13
+
14
+
15
+ - PEFT 0.4.0
20_23/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.9.mlp.up_proj",
18
+ "model.layers.22.mlp.gate_proj",
19
+ "model.layers.15.mlp.up_proj",
20
+ "model.layers.23.mlp.gate_proj",
21
+ "model.layers.26.self_attn.v_proj",
22
+ "model.layers.2.mlp.up_proj",
23
+ "model.layers.22.mlp.down_proj",
24
+ "model.layers.3.self_attn.v_proj",
25
+ "model.layers.27.mlp.up_proj",
26
+ "model.layers.25.mlp.up_proj",
27
+ "model.layers.25.self_attn.k_proj",
28
+ "model.layers.19.self_attn.k_proj",
29
+ "model.layers.15.self_attn.o_proj",
30
+ "model.layers.16.self_attn.v_proj",
31
+ "model.layers.17.self_attn.q_proj",
32
+ "model.layers.10.mlp.down_proj",
33
+ "model.layers.22.mlp.up_proj",
34
+ "model.layers.14.mlp.down_proj",
35
+ "model.layers.17.mlp.up_proj",
36
+ "model.layers.17.self_attn.k_proj",
37
+ "model.layers.11.mlp.down_proj",
38
+ "model.layers.7.mlp.up_proj",
39
+ "model.layers.7.self_attn.o_proj",
40
+ "model.layers.6.mlp.up_proj",
41
+ "model.layers.24.self_attn.v_proj",
42
+ "model.layers.12.self_attn.k_proj",
43
+ "model.layers.5.mlp.up_proj",
44
+ "model.layers.25.self_attn.q_proj",
45
+ "model.layers.11.self_attn.k_proj",
46
+ "model.layers.0.self_attn.k_proj",
47
+ "model.layers.2.self_attn.o_proj",
48
+ "model.layers.14.self_attn.v_proj",
49
+ "model.layers.23.self_attn.v_proj",
50
+ "model.layers.6.self_attn.v_proj",
51
+ "model.layers.7.self_attn.k_proj",
52
+ "model.layers.16.mlp.down_proj",
53
+ "model.layers.12.mlp.down_proj",
54
+ "model.layers.24.mlp.gate_proj",
55
+ "model.layers.5.self_attn.q_proj",
56
+ "model.layers.11.mlp.gate_proj",
57
+ "model.layers.10.mlp.up_proj",
58
+ "model.layers.19.mlp.gate_proj",
59
+ "model.layers.25.mlp.gate_proj",
60
+ "model.layers.6.mlp.gate_proj",
61
+ "model.layers.18.self_attn.k_proj",
62
+ "model.layers.4.self_attn.k_proj",
63
+ "model.layers.20.self_attn.v_proj",
64
+ "model.layers.6.self_attn.q_proj",
65
+ "model.layers.4.self_attn.q_proj",
66
+ "model.layers.23.mlp.up_proj",
67
+ "model.layers.25.mlp.down_proj",
68
+ "model.layers.25.self_attn.o_proj",
69
+ "model.layers.18.self_attn.q_proj",
70
+ "model.layers.5.self_attn.o_proj",
71
+ "model.layers.2.self_attn.k_proj",
72
+ "model.layers.5.self_attn.v_proj",
73
+ "model.layers.0.mlp.down_proj",
74
+ "model.layers.19.self_attn.v_proj",
75
+ "model.layers.19.mlp.up_proj",
76
+ "model.layers.12.mlp.gate_proj",
77
+ "model.layers.9.self_attn.v_proj",
78
+ "model.layers.1.self_attn.q_proj",
79
+ "model.layers.16.self_attn.q_proj",
80
+ "model.layers.15.self_attn.k_proj",
81
+ "model.layers.26.mlp.gate_proj",
82
+ "model.layers.0.self_attn.o_proj",
83
+ "model.layers.1.mlp.down_proj",
84
+ "model.layers.23.self_attn.o_proj",
85
+ "model.layers.3.self_attn.k_proj",
86
+ "model.layers.10.self_attn.k_proj",
87
+ "model.layers.22.self_attn.v_proj",
88
+ "model.layers.14.self_attn.k_proj",
89
+ "model.layers.7.self_attn.q_proj",
90
+ "model.layers.9.self_attn.o_proj",
91
+ "model.layers.20.mlp.down_proj",
92
+ "model.layers.27.mlp.gate_proj",
93
+ "model.layers.21.self_attn.v_proj",
94
+ "model.layers.11.self_attn.v_proj",
95
+ "model.layers.27.self_attn.q_proj",
96
+ "model.layers.15.mlp.gate_proj",
97
+ "model.layers.2.self_attn.v_proj",
98
+ "model.layers.13.self_attn.q_proj",
99
+ "model.layers.1.self_attn.v_proj",
100
+ "model.layers.3.mlp.gate_proj",
101
+ "model.layers.4.self_attn.v_proj",
102
+ "model.layers.13.self_attn.k_proj",
103
+ "model.layers.21.self_attn.q_proj",
104
+ "model.layers.22.self_attn.k_proj",
105
+ "model.layers.2.mlp.down_proj",
106
+ "model.layers.18.self_attn.o_proj",
107
+ "model.layers.26.mlp.up_proj",
108
+ "model.layers.21.mlp.up_proj",
109
+ "model.layers.12.self_attn.o_proj",
110
+ "model.layers.22.self_attn.q_proj",
111
+ "model.layers.3.mlp.down_proj",
112
+ "model.layers.21.self_attn.o_proj",
113
+ "model.layers.10.mlp.gate_proj",
114
+ "model.layers.10.self_attn.q_proj",
115
+ "model.layers.15.self_attn.v_proj",
116
+ "model.layers.8.self_attn.o_proj",
117
+ "model.layers.16.self_attn.k_proj",
118
+ "model.layers.17.mlp.gate_proj",
119
+ "model.layers.24.mlp.up_proj",
120
+ "model.layers.14.mlp.gate_proj",
121
+ "model.layers.9.mlp.down_proj",
122
+ "model.layers.21.mlp.gate_proj",
123
+ "model.layers.16.mlp.up_proj",
124
+ "model.layers.3.self_attn.o_proj",
125
+ "model.layers.9.self_attn.q_proj",
126
+ "model.layers.17.self_attn.v_proj",
127
+ "model.layers.18.mlp.up_proj",
128
+ "model.layers.18.mlp.down_proj",
129
+ "model.layers.6.self_attn.o_proj",
130
+ "model.layers.9.mlp.gate_proj",
131
+ "model.layers.13.mlp.down_proj",
132
+ "model.layers.10.self_attn.v_proj",
133
+ "model.layers.21.mlp.down_proj",
134
+ "model.layers.26.self_attn.o_proj",
135
+ "model.layers.13.mlp.up_proj",
136
+ "model.layers.20.self_attn.k_proj",
137
+ "model.layers.27.self_attn.o_proj",
138
+ "model.layers.27.mlp.down_proj",
139
+ "model.layers.8.self_attn.k_proj",
140
+ "model.layers.12.self_attn.v_proj",
141
+ "model.layers.25.self_attn.v_proj",
142
+ "model.layers.5.mlp.gate_proj",
143
+ "model.layers.14.mlp.up_proj",
144
+ "model.layers.26.self_attn.k_proj",
145
+ "model.layers.15.mlp.down_proj",
146
+ "model.layers.1.mlp.up_proj",
147
+ "model.layers.2.self_attn.q_proj",
148
+ "model.layers.16.mlp.gate_proj",
149
+ "model.layers.18.self_attn.v_proj",
150
+ "model.layers.19.self_attn.o_proj",
151
+ "model.layers.27.self_attn.v_proj",
152
+ "model.layers.13.self_attn.o_proj",
153
+ "model.layers.14.self_attn.q_proj",
154
+ "model.layers.23.mlp.down_proj",
155
+ "model.layers.3.mlp.up_proj",
156
+ "model.layers.7.self_attn.v_proj",
157
+ "model.layers.23.self_attn.k_proj",
158
+ "model.layers.26.self_attn.q_proj",
159
+ "model.layers.4.self_attn.o_proj",
160
+ "model.layers.0.self_attn.v_proj",
161
+ "model.layers.8.mlp.gate_proj",
162
+ "model.layers.20.mlp.gate_proj",
163
+ "model.layers.11.self_attn.o_proj",
164
+ "model.layers.20.self_attn.o_proj",
165
+ "model.layers.14.self_attn.o_proj",
166
+ "model.layers.1.self_attn.k_proj",
167
+ "model.layers.10.self_attn.o_proj",
168
+ "model.layers.17.mlp.down_proj",
169
+ "model.layers.2.mlp.gate_proj",
170
+ "model.layers.19.mlp.down_proj",
171
+ "model.layers.20.self_attn.q_proj",
172
+ "model.layers.4.mlp.down_proj",
173
+ "model.layers.0.mlp.gate_proj",
174
+ "model.layers.6.self_attn.k_proj",
175
+ "model.layers.12.mlp.up_proj",
176
+ "model.layers.16.self_attn.o_proj",
177
+ "model.layers.22.self_attn.o_proj",
178
+ "model.layers.11.mlp.up_proj",
179
+ "model.layers.4.mlp.gate_proj",
180
+ "model.layers.1.self_attn.o_proj",
181
+ "model.layers.17.self_attn.o_proj",
182
+ "model.layers.20.mlp.up_proj",
183
+ "model.layers.13.self_attn.v_proj",
184
+ "model.layers.19.self_attn.q_proj",
185
+ "model.layers.24.self_attn.k_proj",
186
+ "model.layers.8.mlp.down_proj",
187
+ "model.layers.15.self_attn.q_proj",
188
+ "model.layers.11.self_attn.q_proj",
189
+ "model.layers.23.self_attn.q_proj",
190
+ "model.layers.13.mlp.gate_proj",
191
+ "model.layers.24.mlp.down_proj",
192
+ "model.layers.24.self_attn.q_proj",
193
+ "model.layers.7.mlp.down_proj",
194
+ "model.layers.27.self_attn.k_proj",
195
+ "model.layers.0.self_attn.q_proj",
196
+ "model.layers.3.self_attn.q_proj",
197
+ "model.layers.5.mlp.down_proj",
198
+ "model.layers.7.mlp.gate_proj",
199
+ "model.layers.0.mlp.up_proj",
200
+ "model.layers.26.mlp.down_proj",
201
+ "model.layers.5.self_attn.k_proj",
202
+ "model.layers.8.mlp.up_proj",
203
+ "model.layers.1.mlp.gate_proj",
204
+ "model.layers.24.self_attn.o_proj",
205
+ "model.layers.8.self_attn.q_proj",
206
+ "model.layers.21.self_attn.k_proj",
207
+ "model.layers.8.self_attn.v_proj",
208
+ "model.layers.6.mlp.down_proj",
209
+ "model.layers.18.mlp.gate_proj",
210
+ "model.layers.9.self_attn.k_proj",
211
+ "model.layers.4.mlp.up_proj",
212
+ "model.layers.12.self_attn.q_proj"
213
+ ],
214
+ "task_type": "CAUSAL_LM"
215
+ }
20_23/adapter_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8bb6f59ad321331a13b111701b3b43522650c5b6a69ce8b6bc0ce3e564aefad4
3
+ size 646059050
20_23/config.json ADDED
@@ -0,0 +1,231 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ "fusion_block": "mlp2x_gelu",
16
+ "hidden_act": "silu",
17
+ "hidden_size": 3584,
18
+ "ignore_index": -100,
19
+ "image_aspect_ratio": "anyres_max_9",
20
+ "image_crop_resolution": null,
21
+ "image_grid_pinpoints": [
22
+ [
23
+ 384,
24
+ 384
25
+ ],
26
+ [
27
+ 384,
28
+ 768
29
+ ],
30
+ [
31
+ 384,
32
+ 1152
33
+ ],
34
+ [
35
+ 384,
36
+ 1536
37
+ ],
38
+ [
39
+ 384,
40
+ 1920
41
+ ],
42
+ [
43
+ 384,
44
+ 2304
45
+ ],
46
+ [
47
+ 768,
48
+ 384
49
+ ],
50
+ [
51
+ 768,
52
+ 768
53
+ ],
54
+ [
55
+ 768,
56
+ 1152
57
+ ],
58
+ [
59
+ 768,
60
+ 1536
61
+ ],
62
+ [
63
+ 768,
64
+ 1920
65
+ ],
66
+ [
67
+ 768,
68
+ 2304
69
+ ],
70
+ [
71
+ 1152,
72
+ 384
73
+ ],
74
+ [
75
+ 1152,
76
+ 768
77
+ ],
78
+ [
79
+ 1152,
80
+ 1152
81
+ ],
82
+ [
83
+ 1152,
84
+ 1536
85
+ ],
86
+ [
87
+ 1152,
88
+ 1920
89
+ ],
90
+ [
91
+ 1152,
92
+ 2304
93
+ ],
94
+ [
95
+ 1536,
96
+ 384
97
+ ],
98
+ [
99
+ 1536,
100
+ 768
101
+ ],
102
+ [
103
+ 1536,
104
+ 1152
105
+ ],
106
+ [
107
+ 1536,
108
+ 1536
109
+ ],
110
+ [
111
+ 1536,
112
+ 1920
113
+ ],
114
+ [
115
+ 1536,
116
+ 2304
117
+ ],
118
+ [
119
+ 1920,
120
+ 384
121
+ ],
122
+ [
123
+ 1920,
124
+ 768
125
+ ],
126
+ [
127
+ 1920,
128
+ 1152
129
+ ],
130
+ [
131
+ 1920,
132
+ 1536
133
+ ],
134
+ [
135
+ 1920,
136
+ 1920
137
+ ],
138
+ [
139
+ 1920,
140
+ 2304
141
+ ],
142
+ [
143
+ 2304,
144
+ 384
145
+ ],
146
+ [
147
+ 2304,
148
+ 768
149
+ ],
150
+ [
151
+ 2304,
152
+ 1152
153
+ ],
154
+ [
155
+ 2304,
156
+ 1536
157
+ ],
158
+ [
159
+ 2304,
160
+ 1920
161
+ ],
162
+ [
163
+ 2304,
164
+ 2304
165
+ ]
166
+ ],
167
+ "image_split_resolution": null,
168
+ "image_token_index": 151646,
169
+ "initializer_range": 0.02,
170
+ "intermediate_size": 18944,
171
+ "max_position_embeddings": 32768,
172
+ "max_window_layers": 28,
173
+ "mm_hidden_size": 1152,
174
+ "mm_newline_position": "grid",
175
+ "mm_patch_merge_type": "spatial_unpad",
176
+ "mm_projector_lr": null,
177
+ "mm_projector_type": "mlp2x_gelu",
178
+ "mm_resampler_type": null,
179
+ "mm_spatial_pool_mode": "bilinear",
180
+ "mm_spatial_pool_stride": 2,
181
+ "mm_spatial_tower": "cut3r",
182
+ "mm_tunable_parts": "mm_vision_tower,mm_mlp_adapter,mm_language_model",
183
+ "mm_use_im_patch_token": false,
184
+ "mm_use_im_start_end": false,
185
+ "mm_vision_select_feature": "patch",
186
+ "mm_vision_select_layer": -2,
187
+ "mm_vision_tower": "google/siglip-so400m-patch14-384",
188
+ "mm_vision_tower_lr": null,
189
+ "model_type": "llava",
190
+ "num_attention_heads": 28,
191
+ "num_hidden_layers": 28,
192
+ "num_key_value_heads": 4,
193
+ "pos_skipping_range": 4096,
194
+ "projector_hidden_act": "gelu",
195
+ "rms_norm_eps": 1e-06,
196
+ "rope_scaling": null,
197
+ "rope_theta": 1000000.0,
198
+ "sliding_window": 131072,
199
+ "spatial_tower": "cut3r",
200
+ "text_config": {
201
+ "model_type": "llama"
202
+ },
203
+ "tie_word_embeddings": false,
204
+ "tokenizer_model_max_length": 32768,
205
+ "tokenizer_padding_side": "right",
206
+ "torch_dtype": "bfloat16",
207
+ "transformers_version": "4.40.0.dev0",
208
+ "tune_fusion_block": true,
209
+ "tune_mm_mlp_adapter": true,
210
+ "tune_mm_vision_resampler": false,
211
+ "tune_spatial_tower": false,
212
+ "unfreeze_mm_vision_tower": false,
213
+ "use_cache": true,
214
+ "use_mm_proj": true,
215
+ "use_pos_skipping": false,
216
+ "use_sliding_window": false,
217
+ "vision_config": {
218
+ "hidden_size": 1024,
219
+ "image_size": 336,
220
+ "intermediate_size": 4096,
221
+ "model_type": "clip_vision_model",
222
+ "num_attention_heads": 16,
223
+ "num_hidden_layers": 24,
224
+ "patch_size": 14,
225
+ "projection_dim": 768,
226
+ "vocab_size": 32000
227
+ },
228
+ "vision_feature_layer": -2,
229
+ "vision_feature_select_strategy": "default",
230
+ "vision_tower_pretrained": null
231
+ }
20_23/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
+ }
20_23/non_lora_trainables.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:603e675e44969a8d7613a7e1594afcb174df40a25b1a27129f9c945a36fb1d6f
3
+ size 65174928
20_23/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff