minhnguyent546 commited on
Commit
23206d2
·
verified ·
1 Parent(s): 1fa41da

Upload folder using huggingface_hub

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. scripts/.ipynb_checkpoints/ViViT-checkpoint.ipynb +829 -0
  2. scripts/.ipynb_checkpoints/VideoMAE-checkpoint.ipynb +713 -0
  3. scripts/.ipynb_checkpoints/run_self_influence-checkpoint.sh +13 -0
  4. scripts/.ipynb_checkpoints/soup_caformer_b36-checkpoint.sh +15 -0
  5. scripts/.ipynb_checkpoints/soup_caformer_m36-checkpoint.sh +17 -0
  6. scripts/.ipynb_checkpoints/soup_caformer_s36-checkpoint.sh +12 -0
  7. scripts/.ipynb_checkpoints/soup_coatnet_0-merged_4_15-checkpoint.sh +12 -0
  8. scripts/.ipynb_checkpoints/soup_coatnet_0_co_teaching-checkpoint.sh +12 -0
  9. scripts/.ipynb_checkpoints/soup_coatnet_0_random-checkpoint.sh +12 -0
  10. scripts/.ipynb_checkpoints/soup_coatnet_2-merged_4_15-checkpoint.sh +12 -0
  11. scripts/.ipynb_checkpoints/soup_coatnet_2_random-checkpoint.sh +12 -0
  12. scripts/.ipynb_checkpoints/soup_convnext_base-checkpoint.sh +15 -0
  13. scripts/.ipynb_checkpoints/soup_convnext_small-checkpoint.sh +15 -0
  14. scripts/.ipynb_checkpoints/soup_convnext_tiny-checkpoint.sh +15 -0
  15. scripts/.ipynb_checkpoints/soup_eva02_base-checkpoint.sh +13 -0
  16. scripts/.ipynb_checkpoints/soup_eva02_large-checkpoint.sh +16 -0
  17. scripts/.ipynb_checkpoints/soup_eva02_small-checkpoint.sh +16 -0
  18. scripts/.ipynb_checkpoints/soup_focalnet_base_srf-checkpoint.sh +15 -0
  19. scripts/.ipynb_checkpoints/soup_focalnet_small_lrf-checkpoint.sh +15 -0
  20. scripts/.ipynb_checkpoints/soup_focalnet_small_srf-checkpoint.sh +15 -0
  21. scripts/.ipynb_checkpoints/soup_hgnetv2_b6_ssld_stage2-checkpoint.sh +15 -0
  22. scripts/.ipynb_checkpoints/soup_maxvit_base-checkpoint.sh +15 -0
  23. scripts/.ipynb_checkpoints/soup_swin-checkpoint.sh +14 -0
  24. scripts/.ipynb_checkpoints/soup_tiny_vit_21m-checkpoint.sh +15 -0
  25. scripts/.ipynb_checkpoints/soup_vit_base-checkpoint.sh +15 -0
  26. scripts/.ipynb_checkpoints/soup_vit_base_laion2b-checkpoint.sh +15 -0
  27. scripts/.ipynb_checkpoints/test_caformer_b36-checkpoint.sh +10 -0
  28. scripts/.ipynb_checkpoints/test_caformer_m36-checkpoint.sh +12 -0
  29. scripts/.ipynb_checkpoints/test_caformer_s36-checkpoint.sh +10 -0
  30. scripts/.ipynb_checkpoints/test_coatnet_0-checkpoint.sh +10 -0
  31. scripts/.ipynb_checkpoints/test_coatnet_0-merged_4_15-checkpoint.sh +10 -0
  32. scripts/.ipynb_checkpoints/test_coatnet_0_co_teaching-checkpoint.sh +10 -0
  33. scripts/.ipynb_checkpoints/test_coatnet_0_random-checkpoint.sh +10 -0
  34. scripts/.ipynb_checkpoints/test_coatnet_2-merged_4_15-checkpoint.sh +10 -0
  35. scripts/.ipynb_checkpoints/test_coatnet_2_random-checkpoint.sh +10 -0
  36. scripts/.ipynb_checkpoints/test_convnext_base-checkpoint.sh +10 -0
  37. scripts/.ipynb_checkpoints/test_convnext_femto-checkpoint.sh +10 -0
  38. scripts/.ipynb_checkpoints/test_convnext_large-checkpoint.sh +10 -0
  39. scripts/.ipynb_checkpoints/test_convnext_nano-checkpoint.sh +10 -0
  40. scripts/.ipynb_checkpoints/test_convnext_pico-checkpoint.sh +10 -0
  41. scripts/.ipynb_checkpoints/test_convnext_small-checkpoint.sh +10 -0
  42. scripts/.ipynb_checkpoints/test_convnext_tiny-checkpoint.sh +10 -0
  43. scripts/.ipynb_checkpoints/test_convnext_v2_large-checkpoint.sh +10 -0
  44. scripts/.ipynb_checkpoints/test_convnext_xlarge-checkpoint.sh +10 -0
  45. scripts/.ipynb_checkpoints/test_convnext_xxlarge-checkpoint.sh +10 -0
  46. scripts/.ipynb_checkpoints/test_eva02_base-checkpoint.sh +10 -0
  47. scripts/.ipynb_checkpoints/test_eva02_large-checkpoint.sh +10 -0
  48. scripts/.ipynb_checkpoints/test_eva02_small-checkpoint.sh +10 -0
  49. scripts/.ipynb_checkpoints/test_focalnet_base_srf-checkpoint.sh +10 -0
  50. scripts/.ipynb_checkpoints/test_focalnet_small_lrf-checkpoint.sh +10 -0
scripts/.ipynb_checkpoints/ViViT-checkpoint.ipynb ADDED
@@ -0,0 +1,829 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "cells": [
3
+ {
4
+ "cell_type": "code",
5
+ "execution_count": 1,
6
+ "id": "18122027-63d4-45bc-a155-11d941da97b9",
7
+ "metadata": {},
8
+ "outputs": [],
9
+ "source": [
10
+ "DATASET_PATH = \"/media/khmt/HDD1/TQKhang/datasets/xarac\""
11
+ ]
12
+ },
13
+ {
14
+ "cell_type": "code",
15
+ "execution_count": 2,
16
+ "id": "bc5b5096-93dd-4661-bd32-e1c66a3facfc",
17
+ "metadata": {},
18
+ "outputs": [],
19
+ "source": [
20
+ "%load_ext autoreload\n",
21
+ "%autoreload 2"
22
+ ]
23
+ },
24
+ {
25
+ "cell_type": "code",
26
+ "execution_count": 3,
27
+ "id": "59965104-6b47-4bd4-ae52-9924a91feed3",
28
+ "metadata": {},
29
+ "outputs": [
30
+ {
31
+ "name": "stderr",
32
+ "output_type": "stream",
33
+ "text": [
34
+ "/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
35
+ " from .autonotebook import tqdm as notebook_tqdm\n"
36
+ ]
37
+ }
38
+ ],
39
+ "source": [
40
+ "from torch.optim import AdamW\n",
41
+ "from video_transformers import VideoModel\n",
42
+ "from video_transformers.backbones.transformers import TransformersBackbone\n",
43
+ "from video_transformers.backbones.timm import TimmBackbone\n",
44
+ "from video_transformers.data import VideoDataModule\n",
45
+ "from video_transformers.heads import LinearHead\n",
46
+ "from video_transformers.trainer import trainer_factory"
47
+ ]
48
+ },
49
+ {
50
+ "cell_type": "code",
51
+ "execution_count": 4,
52
+ "id": "2f5c18af-5a5a-4e6d-91dd-446123218792",
53
+ "metadata": {
54
+ "scrolled": true
55
+ },
56
+ "outputs": [
57
+ {
58
+ "name": "stderr",
59
+ "output_type": "stream",
60
+ "text": [
61
+ "Some weights of the model checkpoint at facebook/timesformer-base-finetuned-k400 were not used when initializing TimesformerModel: ['classifier.weight', 'classifier.bias']\n",
62
+ "- This IS expected if you are initializing TimesformerModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
63
+ "- This IS NOT expected if you are initializing TimesformerModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n",
64
+ "/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/site-packages/transformers/models/videomae/feature_extraction_videomae.py:31: FutureWarning: The class VideoMAEFeatureExtractor is deprecated and will be removed in version 5 of Transformers. Please use VideoMAEImageProcessor instead.\n",
65
+ " FutureWarning,\n"
66
+ ]
67
+ }
68
+ ],
69
+ "source": [
70
+ "backbone = TransformersBackbone(\"facebook/timesformer-base-finetuned-k400\", num_unfrozen_stages=1)"
71
+ ]
72
+ },
73
+ {
74
+ "cell_type": "code",
75
+ "execution_count": 5,
76
+ "id": "04eb1ed9-0497-48d0-a72e-7223939f257b",
77
+ "metadata": {},
78
+ "outputs": [],
79
+ "source": [
80
+ "datamodule = VideoDataModule(\n",
81
+ " train_root=\"/media/khmt/HDD1/TQKhang/datasets/xarac/train\",\n",
82
+ " val_root=\"/media/khmt/HDD1/TQKhang/datasets/xarac/val\",\n",
83
+ " test_root=\"/media/khmt/HDD1/TQKhang/datasets/xarac/val\",\n",
84
+ " batch_size=4,\n",
85
+ " num_workers=4,\n",
86
+ " num_timesteps=8,\n",
87
+ " preprocess_input_size=224,\n",
88
+ " preprocess_clip_duration=1,\n",
89
+ " preprocess_means=backbone.mean,\n",
90
+ " preprocess_stds=backbone.std,\n",
91
+ " preprocess_min_short_side=256,\n",
92
+ " preprocess_max_short_side=320,\n",
93
+ " preprocess_horizontal_flip_p=0.5,\n",
94
+ ")"
95
+ ]
96
+ },
97
+ {
98
+ "cell_type": "code",
99
+ "execution_count": 6,
100
+ "id": "12b51c8e-8420-456a-96fb-a185165c990a",
101
+ "metadata": {},
102
+ "outputs": [
103
+ {
104
+ "data": {
105
+ "text/plain": [
106
+ "2"
107
+ ]
108
+ },
109
+ "execution_count": 6,
110
+ "metadata": {},
111
+ "output_type": "execute_result"
112
+ }
113
+ ],
114
+ "source": [
115
+ "datamodule.num_classes"
116
+ ]
117
+ },
118
+ {
119
+ "cell_type": "code",
120
+ "execution_count": 7,
121
+ "id": "a6189aca-e28c-4147-8efc-fae604663d14",
122
+ "metadata": {},
123
+ "outputs": [],
124
+ "source": [
125
+ "head = LinearHead(hidden_size=backbone.num_features, num_classes=datamodule.num_classes)\n",
126
+ "model = VideoModel(backbone, head)"
127
+ ]
128
+ },
129
+ {
130
+ "cell_type": "code",
131
+ "execution_count": 8,
132
+ "id": "f3ed0a1e-d90c-49b1-a213-55815d0cdf15",
133
+ "metadata": {
134
+ "scrolled": true
135
+ },
136
+ "outputs": [
137
+ {
138
+ "data": {
139
+ "text/plain": [
140
+ "VideoModel(\n",
141
+ " (backbone): TransformersBackbone(\n",
142
+ " (model): TimesformerModel(\n",
143
+ " (embeddings): TimesformerEmbeddings(\n",
144
+ " (patch_embeddings): TimesformerPatchEmbeddings(\n",
145
+ " (projection): Conv2d(3, 768, kernel_size=(16, 16), stride=(16, 16))\n",
146
+ " )\n",
147
+ " (pos_drop): Dropout(p=0.0, inplace=False)\n",
148
+ " (time_drop): Dropout(p=0.0, inplace=False)\n",
149
+ " )\n",
150
+ " (encoder): TimesformerEncoder(\n",
151
+ " (layer): ModuleList(\n",
152
+ " (0): TimesformerLayer(\n",
153
+ " (drop_path): Identity()\n",
154
+ " (attention): TimeSformerAttention(\n",
155
+ " (attention): TimesformerSelfAttention(\n",
156
+ " (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
157
+ " (attn_drop): Dropout(p=0.0, inplace=False)\n",
158
+ " )\n",
159
+ " (output): TimesformerSelfOutput(\n",
160
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
161
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
162
+ " )\n",
163
+ " )\n",
164
+ " (intermediate): TimesformerIntermediate(\n",
165
+ " (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
166
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
167
+ " (intermediate_act_fn): GELUActivation()\n",
168
+ " )\n",
169
+ " (output): TimesformerOutput(\n",
170
+ " (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
171
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
172
+ " )\n",
173
+ " (layernorm_before): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
174
+ " (layernorm_after): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
175
+ " (temporal_layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
176
+ " (temporal_attention): TimeSformerAttention(\n",
177
+ " (attention): TimesformerSelfAttention(\n",
178
+ " (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
179
+ " (attn_drop): Dropout(p=0.0, inplace=False)\n",
180
+ " )\n",
181
+ " (output): TimesformerSelfOutput(\n",
182
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
183
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
184
+ " )\n",
185
+ " )\n",
186
+ " (temporal_dense): Linear(in_features=768, out_features=768, bias=True)\n",
187
+ " )\n",
188
+ " (1): TimesformerLayer(\n",
189
+ " (drop_path): Identity()\n",
190
+ " (attention): TimeSformerAttention(\n",
191
+ " (attention): TimesformerSelfAttention(\n",
192
+ " (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
193
+ " (attn_drop): Dropout(p=0.0, inplace=False)\n",
194
+ " )\n",
195
+ " (output): TimesformerSelfOutput(\n",
196
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
197
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
198
+ " )\n",
199
+ " )\n",
200
+ " (intermediate): TimesformerIntermediate(\n",
201
+ " (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
202
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
203
+ " (intermediate_act_fn): GELUActivation()\n",
204
+ " )\n",
205
+ " (output): TimesformerOutput(\n",
206
+ " (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
207
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
208
+ " )\n",
209
+ " (layernorm_before): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
210
+ " (layernorm_after): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
211
+ " (temporal_layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
212
+ " (temporal_attention): TimeSformerAttention(\n",
213
+ " (attention): TimesformerSelfAttention(\n",
214
+ " (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
215
+ " (attn_drop): Dropout(p=0.0, inplace=False)\n",
216
+ " )\n",
217
+ " (output): TimesformerSelfOutput(\n",
218
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
219
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
220
+ " )\n",
221
+ " )\n",
222
+ " (temporal_dense): Linear(in_features=768, out_features=768, bias=True)\n",
223
+ " )\n",
224
+ " (2): TimesformerLayer(\n",
225
+ " (drop_path): Identity()\n",
226
+ " (attention): TimeSformerAttention(\n",
227
+ " (attention): TimesformerSelfAttention(\n",
228
+ " (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
229
+ " (attn_drop): Dropout(p=0.0, inplace=False)\n",
230
+ " )\n",
231
+ " (output): TimesformerSelfOutput(\n",
232
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
233
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
234
+ " )\n",
235
+ " )\n",
236
+ " (intermediate): TimesformerIntermediate(\n",
237
+ " (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
238
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
239
+ " (intermediate_act_fn): GELUActivation()\n",
240
+ " )\n",
241
+ " (output): TimesformerOutput(\n",
242
+ " (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
243
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
244
+ " )\n",
245
+ " (layernorm_before): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
246
+ " (layernorm_after): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
247
+ " (temporal_layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
248
+ " (temporal_attention): TimeSformerAttention(\n",
249
+ " (attention): TimesformerSelfAttention(\n",
250
+ " (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
251
+ " (attn_drop): Dropout(p=0.0, inplace=False)\n",
252
+ " )\n",
253
+ " (output): TimesformerSelfOutput(\n",
254
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
255
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
256
+ " )\n",
257
+ " )\n",
258
+ " (temporal_dense): Linear(in_features=768, out_features=768, bias=True)\n",
259
+ " )\n",
260
+ " (3): TimesformerLayer(\n",
261
+ " (drop_path): Identity()\n",
262
+ " (attention): TimeSformerAttention(\n",
263
+ " (attention): TimesformerSelfAttention(\n",
264
+ " (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
265
+ " (attn_drop): Dropout(p=0.0, inplace=False)\n",
266
+ " )\n",
267
+ " (output): TimesformerSelfOutput(\n",
268
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
269
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
270
+ " )\n",
271
+ " )\n",
272
+ " (intermediate): TimesformerIntermediate(\n",
273
+ " (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
274
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
275
+ " (intermediate_act_fn): GELUActivation()\n",
276
+ " )\n",
277
+ " (output): TimesformerOutput(\n",
278
+ " (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
279
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
280
+ " )\n",
281
+ " (layernorm_before): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
282
+ " (layernorm_after): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
283
+ " (temporal_layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
284
+ " (temporal_attention): TimeSformerAttention(\n",
285
+ " (attention): TimesformerSelfAttention(\n",
286
+ " (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
287
+ " (attn_drop): Dropout(p=0.0, inplace=False)\n",
288
+ " )\n",
289
+ " (output): TimesformerSelfOutput(\n",
290
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
291
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
292
+ " )\n",
293
+ " )\n",
294
+ " (temporal_dense): Linear(in_features=768, out_features=768, bias=True)\n",
295
+ " )\n",
296
+ " (4): TimesformerLayer(\n",
297
+ " (drop_path): Identity()\n",
298
+ " (attention): TimeSformerAttention(\n",
299
+ " (attention): TimesformerSelfAttention(\n",
300
+ " (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
301
+ " (attn_drop): Dropout(p=0.0, inplace=False)\n",
302
+ " )\n",
303
+ " (output): TimesformerSelfOutput(\n",
304
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
305
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
306
+ " )\n",
307
+ " )\n",
308
+ " (intermediate): TimesformerIntermediate(\n",
309
+ " (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
310
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
311
+ " (intermediate_act_fn): GELUActivation()\n",
312
+ " )\n",
313
+ " (output): TimesformerOutput(\n",
314
+ " (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
315
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
316
+ " )\n",
317
+ " (layernorm_before): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
318
+ " (layernorm_after): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
319
+ " (temporal_layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
320
+ " (temporal_attention): TimeSformerAttention(\n",
321
+ " (attention): TimesformerSelfAttention(\n",
322
+ " (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
323
+ " (attn_drop): Dropout(p=0.0, inplace=False)\n",
324
+ " )\n",
325
+ " (output): TimesformerSelfOutput(\n",
326
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
327
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
328
+ " )\n",
329
+ " )\n",
330
+ " (temporal_dense): Linear(in_features=768, out_features=768, bias=True)\n",
331
+ " )\n",
332
+ " (5): TimesformerLayer(\n",
333
+ " (drop_path): Identity()\n",
334
+ " (attention): TimeSformerAttention(\n",
335
+ " (attention): TimesformerSelfAttention(\n",
336
+ " (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
337
+ " (attn_drop): Dropout(p=0.0, inplace=False)\n",
338
+ " )\n",
339
+ " (output): TimesformerSelfOutput(\n",
340
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
341
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
342
+ " )\n",
343
+ " )\n",
344
+ " (intermediate): TimesformerIntermediate(\n",
345
+ " (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
346
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
347
+ " (intermediate_act_fn): GELUActivation()\n",
348
+ " )\n",
349
+ " (output): TimesformerOutput(\n",
350
+ " (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
351
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
352
+ " )\n",
353
+ " (layernorm_before): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
354
+ " (layernorm_after): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
355
+ " (temporal_layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
356
+ " (temporal_attention): TimeSformerAttention(\n",
357
+ " (attention): TimesformerSelfAttention(\n",
358
+ " (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
359
+ " (attn_drop): Dropout(p=0.0, inplace=False)\n",
360
+ " )\n",
361
+ " (output): TimesformerSelfOutput(\n",
362
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
363
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
364
+ " )\n",
365
+ " )\n",
366
+ " (temporal_dense): Linear(in_features=768, out_features=768, bias=True)\n",
367
+ " )\n",
368
+ " (6): TimesformerLayer(\n",
369
+ " (drop_path): Identity()\n",
370
+ " (attention): TimeSformerAttention(\n",
371
+ " (attention): TimesformerSelfAttention(\n",
372
+ " (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
373
+ " (attn_drop): Dropout(p=0.0, inplace=False)\n",
374
+ " )\n",
375
+ " (output): TimesformerSelfOutput(\n",
376
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
377
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
378
+ " )\n",
379
+ " )\n",
380
+ " (intermediate): TimesformerIntermediate(\n",
381
+ " (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
382
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
383
+ " (intermediate_act_fn): GELUActivation()\n",
384
+ " )\n",
385
+ " (output): TimesformerOutput(\n",
386
+ " (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
387
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
388
+ " )\n",
389
+ " (layernorm_before): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
390
+ " (layernorm_after): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
391
+ " (temporal_layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
392
+ " (temporal_attention): TimeSformerAttention(\n",
393
+ " (attention): TimesformerSelfAttention(\n",
394
+ " (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
395
+ " (attn_drop): Dropout(p=0.0, inplace=False)\n",
396
+ " )\n",
397
+ " (output): TimesformerSelfOutput(\n",
398
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
399
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
400
+ " )\n",
401
+ " )\n",
402
+ " (temporal_dense): Linear(in_features=768, out_features=768, bias=True)\n",
403
+ " )\n",
404
+ " (7): TimesformerLayer(\n",
405
+ " (drop_path): Identity()\n",
406
+ " (attention): TimeSformerAttention(\n",
407
+ " (attention): TimesformerSelfAttention(\n",
408
+ " (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
409
+ " (attn_drop): Dropout(p=0.0, inplace=False)\n",
410
+ " )\n",
411
+ " (output): TimesformerSelfOutput(\n",
412
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
413
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
414
+ " )\n",
415
+ " )\n",
416
+ " (intermediate): TimesformerIntermediate(\n",
417
+ " (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
418
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
419
+ " (intermediate_act_fn): GELUActivation()\n",
420
+ " )\n",
421
+ " (output): TimesformerOutput(\n",
422
+ " (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
423
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
424
+ " )\n",
425
+ " (layernorm_before): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
426
+ " (layernorm_after): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
427
+ " (temporal_layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
428
+ " (temporal_attention): TimeSformerAttention(\n",
429
+ " (attention): TimesformerSelfAttention(\n",
430
+ " (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
431
+ " (attn_drop): Dropout(p=0.0, inplace=False)\n",
432
+ " )\n",
433
+ " (output): TimesformerSelfOutput(\n",
434
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
435
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
436
+ " )\n",
437
+ " )\n",
438
+ " (temporal_dense): Linear(in_features=768, out_features=768, bias=True)\n",
439
+ " )\n",
440
+ " (8): TimesformerLayer(\n",
441
+ " (drop_path): Identity()\n",
442
+ " (attention): TimeSformerAttention(\n",
443
+ " (attention): TimesformerSelfAttention(\n",
444
+ " (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
445
+ " (attn_drop): Dropout(p=0.0, inplace=False)\n",
446
+ " )\n",
447
+ " (output): TimesformerSelfOutput(\n",
448
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
449
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
450
+ " )\n",
451
+ " )\n",
452
+ " (intermediate): TimesformerIntermediate(\n",
453
+ " (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
454
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
455
+ " (intermediate_act_fn): GELUActivation()\n",
456
+ " )\n",
457
+ " (output): TimesformerOutput(\n",
458
+ " (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
459
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
460
+ " )\n",
461
+ " (layernorm_before): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
462
+ " (layernorm_after): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
463
+ " (temporal_layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
464
+ " (temporal_attention): TimeSformerAttention(\n",
465
+ " (attention): TimesformerSelfAttention(\n",
466
+ " (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
467
+ " (attn_drop): Dropout(p=0.0, inplace=False)\n",
468
+ " )\n",
469
+ " (output): TimesformerSelfOutput(\n",
470
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
471
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
472
+ " )\n",
473
+ " )\n",
474
+ " (temporal_dense): Linear(in_features=768, out_features=768, bias=True)\n",
475
+ " )\n",
476
+ " (9): TimesformerLayer(\n",
477
+ " (drop_path): Identity()\n",
478
+ " (attention): TimeSformerAttention(\n",
479
+ " (attention): TimesformerSelfAttention(\n",
480
+ " (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
481
+ " (attn_drop): Dropout(p=0.0, inplace=False)\n",
482
+ " )\n",
483
+ " (output): TimesformerSelfOutput(\n",
484
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
485
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
486
+ " )\n",
487
+ " )\n",
488
+ " (intermediate): TimesformerIntermediate(\n",
489
+ " (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
490
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
491
+ " (intermediate_act_fn): GELUActivation()\n",
492
+ " )\n",
493
+ " (output): TimesformerOutput(\n",
494
+ " (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
495
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
496
+ " )\n",
497
+ " (layernorm_before): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
498
+ " (layernorm_after): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
499
+ " (temporal_layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
500
+ " (temporal_attention): TimeSformerAttention(\n",
501
+ " (attention): TimesformerSelfAttention(\n",
502
+ " (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
503
+ " (attn_drop): Dropout(p=0.0, inplace=False)\n",
504
+ " )\n",
505
+ " (output): TimesformerSelfOutput(\n",
506
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
507
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
508
+ " )\n",
509
+ " )\n",
510
+ " (temporal_dense): Linear(in_features=768, out_features=768, bias=True)\n",
511
+ " )\n",
512
+ " (10): TimesformerLayer(\n",
513
+ " (drop_path): Identity()\n",
514
+ " (attention): TimeSformerAttention(\n",
515
+ " (attention): TimesformerSelfAttention(\n",
516
+ " (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
517
+ " (attn_drop): Dropout(p=0.0, inplace=False)\n",
518
+ " )\n",
519
+ " (output): TimesformerSelfOutput(\n",
520
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
521
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
522
+ " )\n",
523
+ " )\n",
524
+ " (intermediate): TimesformerIntermediate(\n",
525
+ " (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
526
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
527
+ " (intermediate_act_fn): GELUActivation()\n",
528
+ " )\n",
529
+ " (output): TimesformerOutput(\n",
530
+ " (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
531
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
532
+ " )\n",
533
+ " (layernorm_before): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
534
+ " (layernorm_after): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
535
+ " (temporal_layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
536
+ " (temporal_attention): TimeSformerAttention(\n",
537
+ " (attention): TimesformerSelfAttention(\n",
538
+ " (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
539
+ " (attn_drop): Dropout(p=0.0, inplace=False)\n",
540
+ " )\n",
541
+ " (output): TimesformerSelfOutput(\n",
542
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
543
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
544
+ " )\n",
545
+ " )\n",
546
+ " (temporal_dense): Linear(in_features=768, out_features=768, bias=True)\n",
547
+ " )\n",
548
+ " (11): TimesformerLayer(\n",
549
+ " (drop_path): Identity()\n",
550
+ " (attention): TimeSformerAttention(\n",
551
+ " (attention): TimesformerSelfAttention(\n",
552
+ " (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
553
+ " (attn_drop): Dropout(p=0.0, inplace=False)\n",
554
+ " )\n",
555
+ " (output): TimesformerSelfOutput(\n",
556
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
557
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
558
+ " )\n",
559
+ " )\n",
560
+ " (intermediate): TimesformerIntermediate(\n",
561
+ " (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
562
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
563
+ " (intermediate_act_fn): GELUActivation()\n",
564
+ " )\n",
565
+ " (output): TimesformerOutput(\n",
566
+ " (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
567
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
568
+ " )\n",
569
+ " (layernorm_before): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
570
+ " (layernorm_after): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
571
+ " (temporal_layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
572
+ " (temporal_attention): TimeSformerAttention(\n",
573
+ " (attention): TimesformerSelfAttention(\n",
574
+ " (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
575
+ " (attn_drop): Dropout(p=0.0, inplace=False)\n",
576
+ " )\n",
577
+ " (output): TimesformerSelfOutput(\n",
578
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
579
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
580
+ " )\n",
581
+ " )\n",
582
+ " (temporal_dense): Linear(in_features=768, out_features=768, bias=True)\n",
583
+ " )\n",
584
+ " )\n",
585
+ " )\n",
586
+ " (layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
587
+ " )\n",
588
+ " )\n",
589
+ " (head): LinearHead(\n",
590
+ " (linear): Linear(in_features=768, out_features=2, bias=True)\n",
591
+ " )\n",
592
+ ")"
593
+ ]
594
+ },
595
+ "execution_count": 8,
596
+ "metadata": {},
597
+ "output_type": "execute_result"
598
+ }
599
+ ],
600
+ "source": [
601
+ "model"
602
+ ]
603
+ },
604
+ {
605
+ "cell_type": "code",
606
+ "execution_count": 9,
607
+ "id": "a9d97e70-f539-4f34-8859-f69f1033a57e",
608
+ "metadata": {},
609
+ "outputs": [],
610
+ "source": [
611
+ "optimizer = AdamW(model.parameters(), lr=1e-4)\n",
612
+ "\n",
613
+ "Trainer = trainer_factory(\"single_label_classification\")\n",
614
+ "trainer = Trainer(datamodule, model, optimizer=optimizer, max_epochs=10, \n",
615
+ " mixed_precision=\"fp16\")"
616
+ ]
617
+ },
618
+ {
619
+ "cell_type": "code",
620
+ "execution_count": 10,
621
+ "id": "552054ca-55bd-4692-90ae-5f36b1c4d030",
622
+ "metadata": {
623
+ "scrolled": true
624
+ },
625
+ "outputs": [
626
+ {
627
+ "name": "stdout",
628
+ "output_type": "stream",
629
+ "text": [
630
+ "Trainable parameteres: 10045442\n",
631
+ "Total parameteres: 121260290\n"
632
+ ]
633
+ },
634
+ {
635
+ "name": "stderr",
636
+ "output_type": "stream",
637
+ "text": [
638
+ "Epoch 0 (Done) : 100%|█████████████████████████████████████████████████████████████████████████████| 168/168 [01:13<00:00, 2.30 batch/s, loss=0.6949, val/f1=0.488, train/f1=0.504]\n",
639
+ "Epoch 1 (Done) : 100%|█████████████████████████████████████████████████████████████████████████████| 168/168 [01:13<00:00, 2.29 batch/s, loss=0.5825, val/f1=0.738, train/f1=0.711]\n",
640
+ "Epoch 2 (Done) : 100%|█████████████████████████████████████████████████████████████████████████████| 168/168 [01:15<00:00, 2.23 batch/s, loss=0.4997, val/f1=0.755, train/f1=0.794]\n",
641
+ "Epoch 3 (Done) : 100%|█████████████████████████████████████████████████████████████████████████████| 168/168 [01:16<00:00, 2.20 batch/s, loss=0.5583, val/f1=0.744, train/f1=0.813]\n",
642
+ "Epoch 4 (Done) : 100%|█████████████████████████████████████████████████████████████████████████████| 168/168 [01:16<00:00, 2.20 batch/s, loss=0.5624, val/f1=0.711, train/f1=0.846]\n",
643
+ "Epoch 5 (Done) : 100%|█████████████████████████████████████████████████████████████████████████████| 168/168 [01:16<00:00, 2.18 batch/s, loss=0.5908, val/f1=0.735, train/f1=0.857]\n",
644
+ "Epoch 6 (Done) : 100%|█████████████████████████████████████████████████████████████████████████████| 168/168 [01:17<00:00, 2.17 batch/s, loss=0.5492, val/f1=0.743, train/f1=0.846]\n",
645
+ "Epoch 7 (Done) : 100%|█████████████████████████████████████████████████████████████████████████████| 168/168 [01:16<00:00, 2.19 batch/s, loss=0.5395, val/f1=0.752, train/f1=0.879]\n",
646
+ "Epoch 8 (Done) : 100%|█████████████████████████████████████████████████████████████████████████████| 168/168 [01:16<00:00, 2.21 batch/s, loss=0.5304, val/f1=0.741, train/f1=0.886]\n",
647
+ "Epoch 9 (Done) : 100%|█████████████████████████████████████████████████████████████████████████████| 168/168 [01:16<00:00, 2.20 batch/s, loss=0.5423, val/f1=0.758, train/f1=0.910]\n"
648
+ ]
649
+ }
650
+ ],
651
+ "source": [
652
+ "trainer.fit()"
653
+ ]
654
+ },
655
+ {
656
+ "cell_type": "code",
657
+ "execution_count": 11,
658
+ "id": "7f378e22-d85b-4714-981e-4b68c3fdeb73",
659
+ "metadata": {},
660
+ "outputs": [
661
+ {
662
+ "data": {
663
+ "text/plain": [
664
+ "['khongxarac', 'xarac']"
665
+ ]
666
+ },
667
+ "execution_count": 11,
668
+ "metadata": {},
669
+ "output_type": "execute_result"
670
+ }
671
+ ],
672
+ "source": [
673
+ "datamodule.labels"
674
+ ]
675
+ },
676
+ {
677
+ "cell_type": "code",
678
+ "execution_count": 12,
679
+ "id": "ef16f8e1-6ad1-446a-a817-b01d4259e60b",
680
+ "metadata": {},
681
+ "outputs": [
682
+ {
683
+ "data": {
684
+ "text/plain": [
685
+ "{'num_timesteps': 8,\n",
686
+ " 'input_size': 224,\n",
687
+ " 'means': [0.45, 0.45, 0.45],\n",
688
+ " 'stds': [0.225, 0.225, 0.225],\n",
689
+ " 'min_short_side': 256,\n",
690
+ " 'max_short_side': 320,\n",
691
+ " 'horizontal_flip_p': 0.5,\n",
692
+ " 'clip_duration': 1}"
693
+ ]
694
+ },
695
+ "execution_count": 12,
696
+ "metadata": {},
697
+ "output_type": "execute_result"
698
+ }
699
+ ],
700
+ "source": [
701
+ "datamodule.preprocessor_config"
702
+ ]
703
+ },
704
+ {
705
+ "cell_type": "code",
706
+ "execution_count": 13,
707
+ "id": "8e9c61dc-8c52-4280-8a02-becadafd304b",
708
+ "metadata": {},
709
+ "outputs": [],
710
+ "source": [
711
+ "import os, glob\n",
712
+ "from pathlib import Path\n",
713
+ "from tqdm import tqdm\n",
714
+ "from sklearn.metrics import accuracy_score, f1_score, classification_report, confusion_matrix"
715
+ ]
716
+ },
717
+ {
718
+ "cell_type": "code",
719
+ "execution_count": 14,
720
+ "id": "a215c096-f988-4cb8-a5bd-7d320cc0d86b",
721
+ "metadata": {},
722
+ "outputs": [],
723
+ "source": [
724
+ "test_loader = datamodule.test_dataloader\n",
725
+ "class_names = datamodule.labels"
726
+ ]
727
+ },
728
+ {
729
+ "cell_type": "code",
730
+ "execution_count": 15,
731
+ "id": "ba8fa1bc-3796-4076-ab46-ab1cbb44ce66",
732
+ "metadata": {},
733
+ "outputs": [
734
+ {
735
+ "name": "stderr",
736
+ "output_type": "stream",
737
+ "text": [
738
+ "Testing: 32it [00:12, 2.47it/s] \n"
739
+ ]
740
+ }
741
+ ],
742
+ "source": [
743
+ "import torch\n",
744
+ "model.eval()\n",
745
+ "y_true, y_pred = [], []\n",
746
+ "with torch.no_grad():\n",
747
+ " for batch in tqdm(test_loader, desc=\"Testing\"):\n",
748
+ " inputs = batch[\"video\"].to(\"cuda\")\n",
749
+ " labels = batch[\"label\"].to(\"cuda\")\n",
750
+ " outputs = model(inputs)\n",
751
+ " probabilities = torch.nn.functional.softmax(outputs, dim=1)\n",
752
+ " predictions = probabilities.argmax(dim=-1)\n",
753
+ " y_true.extend(labels.cpu().tolist())\n",
754
+ " y_pred.extend(predictions.cpu().tolist())"
755
+ ]
756
+ },
757
+ {
758
+ "cell_type": "code",
759
+ "execution_count": 16,
760
+ "id": "6fbcc06f-d230-4c46-9cd8-c2aa83aeca19",
761
+ "metadata": {},
762
+ "outputs": [
763
+ {
764
+ "name": "stdout",
765
+ "output_type": "stream",
766
+ "text": [
767
+ "Accuracy: 0.7692 | F1-macro: 0.7608\n",
768
+ " precision recall f1-score support\n",
769
+ "\n",
770
+ " khongxarac 0.79 0.82 0.81 68\n",
771
+ " xarac 0.74 0.69 0.72 49\n",
772
+ "\n",
773
+ " accuracy 0.77 117\n",
774
+ " macro avg 0.76 0.76 0.76 117\n",
775
+ "weighted avg 0.77 0.77 0.77 117\n",
776
+ "\n",
777
+ "Confusion matrix:\n",
778
+ " [[56 12]\n",
779
+ " [15 34]]\n"
780
+ ]
781
+ }
782
+ ],
783
+ "source": [
784
+ "acc = accuracy_score(y_true, y_pred)\n",
785
+ "f1m = f1_score(y_true, y_pred, average=\"macro\")\n",
786
+ "print(f\"Accuracy: {acc:.4f} | F1-macro: {f1m:.4f}\")\n",
787
+ "print(classification_report(y_true, y_pred, target_names=class_names))\n",
788
+ "print(\"Confusion matrix:\\n\", confusion_matrix(y_true, y_pred))"
789
+ ]
790
+ },
791
+ {
792
+ "cell_type": "code",
793
+ "execution_count": null,
794
+ "id": "0b78a3c0-72ef-4207-bb66-03a8a3b02782",
795
+ "metadata": {},
796
+ "outputs": [],
797
+ "source": []
798
+ },
799
+ {
800
+ "cell_type": "code",
801
+ "execution_count": null,
802
+ "id": "7acdbc58-f8c8-4841-84c5-78fbe837f4fb",
803
+ "metadata": {},
804
+ "outputs": [],
805
+ "source": []
806
+ }
807
+ ],
808
+ "metadata": {
809
+ "kernelspec": {
810
+ "display_name": "TQKhang-ViViT",
811
+ "language": "python",
812
+ "name": "tqkhang-vivit"
813
+ },
814
+ "language_info": {
815
+ "codemirror_mode": {
816
+ "name": "ipython",
817
+ "version": 3
818
+ },
819
+ "file_extension": ".py",
820
+ "mimetype": "text/x-python",
821
+ "name": "python",
822
+ "nbconvert_exporter": "python",
823
+ "pygments_lexer": "ipython3",
824
+ "version": "3.7.16"
825
+ }
826
+ },
827
+ "nbformat": 4,
828
+ "nbformat_minor": 5
829
+ }
scripts/.ipynb_checkpoints/VideoMAE-checkpoint.ipynb ADDED
@@ -0,0 +1,713 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "cells": [
3
+ {
4
+ "cell_type": "code",
5
+ "execution_count": 1,
6
+ "id": "18122027-63d4-45bc-a155-11d941da97b9",
7
+ "metadata": {},
8
+ "outputs": [],
9
+ "source": [
10
+ "DATASET_PATH = \"/media/khmt/HDD1/TQKhang/datasets/xarac\""
11
+ ]
12
+ },
13
+ {
14
+ "cell_type": "code",
15
+ "execution_count": 2,
16
+ "id": "bc5b5096-93dd-4661-bd32-e1c66a3facfc",
17
+ "metadata": {},
18
+ "outputs": [],
19
+ "source": [
20
+ "%load_ext autoreload\n",
21
+ "%autoreload 2"
22
+ ]
23
+ },
24
+ {
25
+ "cell_type": "code",
26
+ "execution_count": 3,
27
+ "id": "59965104-6b47-4bd4-ae52-9924a91feed3",
28
+ "metadata": {},
29
+ "outputs": [
30
+ {
31
+ "name": "stderr",
32
+ "output_type": "stream",
33
+ "text": [
34
+ "/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
35
+ " from .autonotebook import tqdm as notebook_tqdm\n"
36
+ ]
37
+ }
38
+ ],
39
+ "source": [
40
+ "from torch.optim import AdamW\n",
41
+ "from video_transformers import VideoModel\n",
42
+ "from video_transformers.backbones.transformers import TransformersBackbone\n",
43
+ "from video_transformers.backbones.timm import TimmBackbone\n",
44
+ "from video_transformers.data import VideoDataModule\n",
45
+ "from video_transformers.heads import LinearHead\n",
46
+ "from video_transformers.trainer import trainer_factory"
47
+ ]
48
+ },
49
+ {
50
+ "cell_type": "code",
51
+ "execution_count": 5,
52
+ "id": "2f5c18af-5a5a-4e6d-91dd-446123218792",
53
+ "metadata": {
54
+ "scrolled": true
55
+ },
56
+ "outputs": [
57
+ {
58
+ "name": "stderr",
59
+ "output_type": "stream",
60
+ "text": [
61
+ "Downloading: 22.9kB [00:00, 10.8MB/s]\n",
62
+ "Downloading: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 346M/346M [00:29<00:00, 11.9MB/s]\n",
63
+ "Some weights of the model checkpoint at MCG-NJU/videomae-base-finetuned-kinetics were not used when initializing VideoMAEModel: ['fc_norm.weight', 'fc_norm.bias', 'classifier.bias', 'classifier.weight']\n",
64
+ "- This IS expected if you are initializing VideoMAEModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
65
+ "- This IS NOT expected if you are initializing VideoMAEModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n",
66
+ "Downloading: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 271/271 [00:00<00:00, 220kB/s]\n",
67
+ "/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/site-packages/transformers/models/videomae/feature_extraction_videomae.py:31: FutureWarning: The class VideoMAEFeatureExtractor is deprecated and will be removed in version 5 of Transformers. Please use VideoMAEImageProcessor instead.\n",
68
+ " FutureWarning,\n"
69
+ ]
70
+ }
71
+ ],
72
+ "source": [
73
+ "backbone = TransformersBackbone(\"MCG-NJU/videomae-base-finetuned-kinetics\", num_unfrozen_stages=1)"
74
+ ]
75
+ },
76
+ {
77
+ "cell_type": "code",
78
+ "execution_count": 12,
79
+ "id": "04eb1ed9-0497-48d0-a72e-7223939f257b",
80
+ "metadata": {},
81
+ "outputs": [],
82
+ "source": [
83
+ "datamodule = VideoDataModule(\n",
84
+ " train_root=\"/media/khmt/HDD1/TQKhang/datasets/xarac/train\",\n",
85
+ " val_root=\"/media/khmt/HDD1/TQKhang/datasets/xarac/val\",\n",
86
+ " test_root=\"/media/khmt/HDD1/TQKhang/datasets/xarac/val\",\n",
87
+ " batch_size=4,\n",
88
+ " num_workers=4,\n",
89
+ " num_timesteps=16,\n",
90
+ " preprocess_input_size=224,\n",
91
+ " preprocess_clip_duration=1,\n",
92
+ " preprocess_means=backbone.mean,\n",
93
+ " preprocess_stds=backbone.std,\n",
94
+ " preprocess_min_short_side=256,\n",
95
+ " preprocess_max_short_side=320,\n",
96
+ " preprocess_horizontal_flip_p=0.5,\n",
97
+ ")"
98
+ ]
99
+ },
100
+ {
101
+ "cell_type": "code",
102
+ "execution_count": 13,
103
+ "id": "12b51c8e-8420-456a-96fb-a185165c990a",
104
+ "metadata": {},
105
+ "outputs": [
106
+ {
107
+ "data": {
108
+ "text/plain": [
109
+ "2"
110
+ ]
111
+ },
112
+ "execution_count": 13,
113
+ "metadata": {},
114
+ "output_type": "execute_result"
115
+ }
116
+ ],
117
+ "source": [
118
+ "datamodule.num_classes"
119
+ ]
120
+ },
121
+ {
122
+ "cell_type": "code",
123
+ "execution_count": 14,
124
+ "id": "a6189aca-e28c-4147-8efc-fae604663d14",
125
+ "metadata": {},
126
+ "outputs": [],
127
+ "source": [
128
+ "head = LinearHead(hidden_size=backbone.num_features, num_classes=datamodule.num_classes)\n",
129
+ "model = VideoModel(backbone, head)"
130
+ ]
131
+ },
132
+ {
133
+ "cell_type": "code",
134
+ "execution_count": 15,
135
+ "id": "f3ed0a1e-d90c-49b1-a213-55815d0cdf15",
136
+ "metadata": {
137
+ "scrolled": true
138
+ },
139
+ "outputs": [
140
+ {
141
+ "data": {
142
+ "text/plain": [
143
+ "VideoModel(\n",
144
+ " (backbone): TransformersBackbone(\n",
145
+ " (model): VideoMAEModel(\n",
146
+ " (embeddings): VideoMAEEmbeddings(\n",
147
+ " (patch_embeddings): VideoMAEPatchEmbeddings(\n",
148
+ " (projection): Conv3d(3, 768, kernel_size=(2, 16, 16), stride=(2, 16, 16))\n",
149
+ " )\n",
150
+ " )\n",
151
+ " (encoder): VideoMAEEncoder(\n",
152
+ " (layer): ModuleList(\n",
153
+ " (0): VideoMAELayer(\n",
154
+ " (attention): VideoMAEAttention(\n",
155
+ " (attention): VideoMAESelfAttention(\n",
156
+ " (query): Linear(in_features=768, out_features=768, bias=False)\n",
157
+ " (key): Linear(in_features=768, out_features=768, bias=False)\n",
158
+ " (value): Linear(in_features=768, out_features=768, bias=False)\n",
159
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
160
+ " )\n",
161
+ " (output): VideoMAESelfOutput(\n",
162
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
163
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
164
+ " )\n",
165
+ " )\n",
166
+ " (intermediate): VideoMAEIntermediate(\n",
167
+ " (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
168
+ " (intermediate_act_fn): GELUActivation()\n",
169
+ " )\n",
170
+ " (output): VideoMAEOutput(\n",
171
+ " (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
172
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
173
+ " )\n",
174
+ " (layernorm_before): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
175
+ " (layernorm_after): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
176
+ " )\n",
177
+ " (1): VideoMAELayer(\n",
178
+ " (attention): VideoMAEAttention(\n",
179
+ " (attention): VideoMAESelfAttention(\n",
180
+ " (query): Linear(in_features=768, out_features=768, bias=False)\n",
181
+ " (key): Linear(in_features=768, out_features=768, bias=False)\n",
182
+ " (value): Linear(in_features=768, out_features=768, bias=False)\n",
183
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
184
+ " )\n",
185
+ " (output): VideoMAESelfOutput(\n",
186
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
187
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
188
+ " )\n",
189
+ " )\n",
190
+ " (intermediate): VideoMAEIntermediate(\n",
191
+ " (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
192
+ " (intermediate_act_fn): GELUActivation()\n",
193
+ " )\n",
194
+ " (output): VideoMAEOutput(\n",
195
+ " (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
196
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
197
+ " )\n",
198
+ " (layernorm_before): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
199
+ " (layernorm_after): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
200
+ " )\n",
201
+ " (2): VideoMAELayer(\n",
202
+ " (attention): VideoMAEAttention(\n",
203
+ " (attention): VideoMAESelfAttention(\n",
204
+ " (query): Linear(in_features=768, out_features=768, bias=False)\n",
205
+ " (key): Linear(in_features=768, out_features=768, bias=False)\n",
206
+ " (value): Linear(in_features=768, out_features=768, bias=False)\n",
207
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
208
+ " )\n",
209
+ " (output): VideoMAESelfOutput(\n",
210
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
211
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
212
+ " )\n",
213
+ " )\n",
214
+ " (intermediate): VideoMAEIntermediate(\n",
215
+ " (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
216
+ " (intermediate_act_fn): GELUActivation()\n",
217
+ " )\n",
218
+ " (output): VideoMAEOutput(\n",
219
+ " (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
220
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
221
+ " )\n",
222
+ " (layernorm_before): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
223
+ " (layernorm_after): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
224
+ " )\n",
225
+ " (3): VideoMAELayer(\n",
226
+ " (attention): VideoMAEAttention(\n",
227
+ " (attention): VideoMAESelfAttention(\n",
228
+ " (query): Linear(in_features=768, out_features=768, bias=False)\n",
229
+ " (key): Linear(in_features=768, out_features=768, bias=False)\n",
230
+ " (value): Linear(in_features=768, out_features=768, bias=False)\n",
231
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
232
+ " )\n",
233
+ " (output): VideoMAESelfOutput(\n",
234
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
235
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
236
+ " )\n",
237
+ " )\n",
238
+ " (intermediate): VideoMAEIntermediate(\n",
239
+ " (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
240
+ " (intermediate_act_fn): GELUActivation()\n",
241
+ " )\n",
242
+ " (output): VideoMAEOutput(\n",
243
+ " (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
244
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
245
+ " )\n",
246
+ " (layernorm_before): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
247
+ " (layernorm_after): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
248
+ " )\n",
249
+ " (4): VideoMAELayer(\n",
250
+ " (attention): VideoMAEAttention(\n",
251
+ " (attention): VideoMAESelfAttention(\n",
252
+ " (query): Linear(in_features=768, out_features=768, bias=False)\n",
253
+ " (key): Linear(in_features=768, out_features=768, bias=False)\n",
254
+ " (value): Linear(in_features=768, out_features=768, bias=False)\n",
255
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
256
+ " )\n",
257
+ " (output): VideoMAESelfOutput(\n",
258
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
259
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
260
+ " )\n",
261
+ " )\n",
262
+ " (intermediate): VideoMAEIntermediate(\n",
263
+ " (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
264
+ " (intermediate_act_fn): GELUActivation()\n",
265
+ " )\n",
266
+ " (output): VideoMAEOutput(\n",
267
+ " (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
268
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
269
+ " )\n",
270
+ " (layernorm_before): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
271
+ " (layernorm_after): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
272
+ " )\n",
273
+ " (5): VideoMAELayer(\n",
274
+ " (attention): VideoMAEAttention(\n",
275
+ " (attention): VideoMAESelfAttention(\n",
276
+ " (query): Linear(in_features=768, out_features=768, bias=False)\n",
277
+ " (key): Linear(in_features=768, out_features=768, bias=False)\n",
278
+ " (value): Linear(in_features=768, out_features=768, bias=False)\n",
279
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
280
+ " )\n",
281
+ " (output): VideoMAESelfOutput(\n",
282
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
283
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
284
+ " )\n",
285
+ " )\n",
286
+ " (intermediate): VideoMAEIntermediate(\n",
287
+ " (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
288
+ " (intermediate_act_fn): GELUActivation()\n",
289
+ " )\n",
290
+ " (output): VideoMAEOutput(\n",
291
+ " (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
292
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
293
+ " )\n",
294
+ " (layernorm_before): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
295
+ " (layernorm_after): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
296
+ " )\n",
297
+ " (6): VideoMAELayer(\n",
298
+ " (attention): VideoMAEAttention(\n",
299
+ " (attention): VideoMAESelfAttention(\n",
300
+ " (query): Linear(in_features=768, out_features=768, bias=False)\n",
301
+ " (key): Linear(in_features=768, out_features=768, bias=False)\n",
302
+ " (value): Linear(in_features=768, out_features=768, bias=False)\n",
303
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
304
+ " )\n",
305
+ " (output): VideoMAESelfOutput(\n",
306
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
307
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
308
+ " )\n",
309
+ " )\n",
310
+ " (intermediate): VideoMAEIntermediate(\n",
311
+ " (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
312
+ " (intermediate_act_fn): GELUActivation()\n",
313
+ " )\n",
314
+ " (output): VideoMAEOutput(\n",
315
+ " (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
316
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
317
+ " )\n",
318
+ " (layernorm_before): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
319
+ " (layernorm_after): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
320
+ " )\n",
321
+ " (7): VideoMAELayer(\n",
322
+ " (attention): VideoMAEAttention(\n",
323
+ " (attention): VideoMAESelfAttention(\n",
324
+ " (query): Linear(in_features=768, out_features=768, bias=False)\n",
325
+ " (key): Linear(in_features=768, out_features=768, bias=False)\n",
326
+ " (value): Linear(in_features=768, out_features=768, bias=False)\n",
327
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
328
+ " )\n",
329
+ " (output): VideoMAESelfOutput(\n",
330
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
331
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
332
+ " )\n",
333
+ " )\n",
334
+ " (intermediate): VideoMAEIntermediate(\n",
335
+ " (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
336
+ " (intermediate_act_fn): GELUActivation()\n",
337
+ " )\n",
338
+ " (output): VideoMAEOutput(\n",
339
+ " (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
340
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
341
+ " )\n",
342
+ " (layernorm_before): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
343
+ " (layernorm_after): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
344
+ " )\n",
345
+ " (8): VideoMAELayer(\n",
346
+ " (attention): VideoMAEAttention(\n",
347
+ " (attention): VideoMAESelfAttention(\n",
348
+ " (query): Linear(in_features=768, out_features=768, bias=False)\n",
349
+ " (key): Linear(in_features=768, out_features=768, bias=False)\n",
350
+ " (value): Linear(in_features=768, out_features=768, bias=False)\n",
351
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
352
+ " )\n",
353
+ " (output): VideoMAESelfOutput(\n",
354
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
355
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
356
+ " )\n",
357
+ " )\n",
358
+ " (intermediate): VideoMAEIntermediate(\n",
359
+ " (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
360
+ " (intermediate_act_fn): GELUActivation()\n",
361
+ " )\n",
362
+ " (output): VideoMAEOutput(\n",
363
+ " (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
364
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
365
+ " )\n",
366
+ " (layernorm_before): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
367
+ " (layernorm_after): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
368
+ " )\n",
369
+ " (9): VideoMAELayer(\n",
370
+ " (attention): VideoMAEAttention(\n",
371
+ " (attention): VideoMAESelfAttention(\n",
372
+ " (query): Linear(in_features=768, out_features=768, bias=False)\n",
373
+ " (key): Linear(in_features=768, out_features=768, bias=False)\n",
374
+ " (value): Linear(in_features=768, out_features=768, bias=False)\n",
375
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
376
+ " )\n",
377
+ " (output): VideoMAESelfOutput(\n",
378
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
379
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
380
+ " )\n",
381
+ " )\n",
382
+ " (intermediate): VideoMAEIntermediate(\n",
383
+ " (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
384
+ " (intermediate_act_fn): GELUActivation()\n",
385
+ " )\n",
386
+ " (output): VideoMAEOutput(\n",
387
+ " (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
388
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
389
+ " )\n",
390
+ " (layernorm_before): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
391
+ " (layernorm_after): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
392
+ " )\n",
393
+ " (10): VideoMAELayer(\n",
394
+ " (attention): VideoMAEAttention(\n",
395
+ " (attention): VideoMAESelfAttention(\n",
396
+ " (query): Linear(in_features=768, out_features=768, bias=False)\n",
397
+ " (key): Linear(in_features=768, out_features=768, bias=False)\n",
398
+ " (value): Linear(in_features=768, out_features=768, bias=False)\n",
399
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
400
+ " )\n",
401
+ " (output): VideoMAESelfOutput(\n",
402
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
403
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
404
+ " )\n",
405
+ " )\n",
406
+ " (intermediate): VideoMAEIntermediate(\n",
407
+ " (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
408
+ " (intermediate_act_fn): GELUActivation()\n",
409
+ " )\n",
410
+ " (output): VideoMAEOutput(\n",
411
+ " (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
412
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
413
+ " )\n",
414
+ " (layernorm_before): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
415
+ " (layernorm_after): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
416
+ " )\n",
417
+ " (11): VideoMAELayer(\n",
418
+ " (attention): VideoMAEAttention(\n",
419
+ " (attention): VideoMAESelfAttention(\n",
420
+ " (query): Linear(in_features=768, out_features=768, bias=False)\n",
421
+ " (key): Linear(in_features=768, out_features=768, bias=False)\n",
422
+ " (value): Linear(in_features=768, out_features=768, bias=False)\n",
423
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
424
+ " )\n",
425
+ " (output): VideoMAESelfOutput(\n",
426
+ " (dense): Linear(in_features=768, out_features=768, bias=True)\n",
427
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
428
+ " )\n",
429
+ " )\n",
430
+ " (intermediate): VideoMAEIntermediate(\n",
431
+ " (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
432
+ " (intermediate_act_fn): GELUActivation()\n",
433
+ " )\n",
434
+ " (output): VideoMAEOutput(\n",
435
+ " (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
436
+ " (dropout): Dropout(p=0.0, inplace=False)\n",
437
+ " )\n",
438
+ " (layernorm_before): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
439
+ " (layernorm_after): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
440
+ " )\n",
441
+ " )\n",
442
+ " )\n",
443
+ " )\n",
444
+ " )\n",
445
+ " (head): LinearHead(\n",
446
+ " (linear): Linear(in_features=768, out_features=2, bias=True)\n",
447
+ " )\n",
448
+ ")"
449
+ ]
450
+ },
451
+ "execution_count": 15,
452
+ "metadata": {},
453
+ "output_type": "execute_result"
454
+ }
455
+ ],
456
+ "source": [
457
+ "model"
458
+ ]
459
+ },
460
+ {
461
+ "cell_type": "code",
462
+ "execution_count": 16,
463
+ "id": "a9d97e70-f539-4f34-8859-f69f1033a57e",
464
+ "metadata": {},
465
+ "outputs": [],
466
+ "source": [
467
+ "optimizer = AdamW(model.parameters(), lr=1e-4)\n",
468
+ "\n",
469
+ "Trainer = trainer_factory(\"single_label_classification\")\n",
470
+ "trainer = Trainer(datamodule, model, optimizer=optimizer, max_epochs=10, \n",
471
+ " mixed_precision=\"fp16\")"
472
+ ]
473
+ },
474
+ {
475
+ "cell_type": "code",
476
+ "execution_count": 17,
477
+ "id": "552054ca-55bd-4692-90ae-5f36b1c4d030",
478
+ "metadata": {
479
+ "scrolled": true
480
+ },
481
+ "outputs": [
482
+ {
483
+ "name": "stdout",
484
+ "output_type": "stream",
485
+ "text": [
486
+ "Trainable parameteres: 7088642\n",
487
+ "Total parameteres: 86227202\n"
488
+ ]
489
+ },
490
+ {
491
+ "name": "stderr",
492
+ "output_type": "stream",
493
+ "text": [
494
+ "Epoch 0 (Done) : 100%|███████████���█████████████████████████████████████████████████████████████████| 168/168 [01:27<00:00, 1.92 batch/s, loss=0.9388, val/f1=0.483, train/f1=0.471]\n",
495
+ "Epoch 1 (Done) : 100%|█████████████████████████████████████████████████████████████████████████████| 168/168 [01:27<00:00, 1.91 batch/s, loss=0.5735, val/f1=0.755, train/f1=0.696]\n",
496
+ "Epoch 2 (Done) : 100%|█████████████████████████████████████████████████████████████████████████████| 168/168 [01:28<00:00, 1.89 batch/s, loss=0.5928, val/f1=0.727, train/f1=0.802]\n",
497
+ "Epoch 3 (Done) : 100%|█████████████████████████████████████████████████████████████████████████████| 168/168 [01:29<00:00, 1.87 batch/s, loss=0.5903, val/f1=0.737, train/f1=0.822]\n",
498
+ "Epoch 4 (Done) : 100%|█████████████████████████████████████████████████████████████████████████████| 168/168 [01:30<00:00, 1.86 batch/s, loss=0.5391, val/f1=0.769, train/f1=0.855]\n",
499
+ "Epoch 5 (Done) : 100%|█████████████████████████████████████████████████████████████████████████████| 168/168 [01:30<00:00, 1.85 batch/s, loss=0.6027, val/f1=0.767, train/f1=0.884]\n",
500
+ "Epoch 6 (Done) : 100%|█████████████████████████████████████████████████████████████████████████████| 168/168 [01:30<00:00, 1.85 batch/s, loss=0.5935, val/f1=0.754, train/f1=0.887]\n",
501
+ "Epoch 7 (Done) : 100%|█████████████████████████████████████████████████████████████████████████████| 168/168 [01:29<00:00, 1.87 batch/s, loss=0.5759, val/f1=0.740, train/f1=0.892]\n",
502
+ "Epoch 8 (Done) : 100%|█████████████████████████████████████████████████████████████████████████████| 168/168 [01:29<00:00, 1.88 batch/s, loss=0.6144, val/f1=0.746, train/f1=0.921]\n",
503
+ "Epoch 9 (Done) : 100%|█████████████████████████████████████████████████████████████████████████████| 168/168 [01:30<00:00, 1.85 batch/s, loss=0.5750, val/f1=0.740, train/f1=0.923]\n"
504
+ ]
505
+ }
506
+ ],
507
+ "source": [
508
+ "trainer.fit()"
509
+ ]
510
+ },
511
+ {
512
+ "cell_type": "code",
513
+ "execution_count": 18,
514
+ "id": "7f378e22-d85b-4714-981e-4b68c3fdeb73",
515
+ "metadata": {},
516
+ "outputs": [
517
+ {
518
+ "data": {
519
+ "text/plain": [
520
+ "['khongxarac', 'xarac']"
521
+ ]
522
+ },
523
+ "execution_count": 18,
524
+ "metadata": {},
525
+ "output_type": "execute_result"
526
+ }
527
+ ],
528
+ "source": [
529
+ "datamodule.labels"
530
+ ]
531
+ },
532
+ {
533
+ "cell_type": "code",
534
+ "execution_count": 19,
535
+ "id": "ef16f8e1-6ad1-446a-a817-b01d4259e60b",
536
+ "metadata": {},
537
+ "outputs": [
538
+ {
539
+ "data": {
540
+ "text/plain": [
541
+ "{'num_timesteps': 16,\n",
542
+ " 'input_size': 224,\n",
543
+ " 'means': [0.485, 0.456, 0.406],\n",
544
+ " 'stds': [0.229, 0.224, 0.225],\n",
545
+ " 'min_short_side': 256,\n",
546
+ " 'max_short_side': 320,\n",
547
+ " 'horizontal_flip_p': 0.5,\n",
548
+ " 'clip_duration': 1}"
549
+ ]
550
+ },
551
+ "execution_count": 19,
552
+ "metadata": {},
553
+ "output_type": "execute_result"
554
+ }
555
+ ],
556
+ "source": [
557
+ "datamodule.preprocessor_config"
558
+ ]
559
+ },
560
+ {
561
+ "cell_type": "code",
562
+ "execution_count": 20,
563
+ "id": "8e9c61dc-8c52-4280-8a02-becadafd304b",
564
+ "metadata": {},
565
+ "outputs": [],
566
+ "source": [
567
+ "import os, glob\n",
568
+ "from pathlib import Path\n",
569
+ "from tqdm import tqdm\n",
570
+ "from sklearn.metrics import accuracy_score, f1_score, classification_report, confusion_matrix"
571
+ ]
572
+ },
573
+ {
574
+ "cell_type": "code",
575
+ "execution_count": 21,
576
+ "id": "a215c096-f988-4cb8-a5bd-7d320cc0d86b",
577
+ "metadata": {},
578
+ "outputs": [],
579
+ "source": [
580
+ "test_loader = datamodule.test_dataloader\n",
581
+ "class_names = datamodule.labels"
582
+ ]
583
+ },
584
+ {
585
+ "cell_type": "code",
586
+ "execution_count": 26,
587
+ "id": "ba8fa1bc-3796-4076-ab46-ab1cbb44ce66",
588
+ "metadata": {},
589
+ "outputs": [
590
+ {
591
+ "name": "stderr",
592
+ "output_type": "stream",
593
+ "text": [
594
+ "Testing: 0%| | 0/30 [00:00<?, ?it/s]Exception ignored in: <function _MultiProcessingDataLoaderIter.__del__ at 0x75ce96bc04d0>\n",
595
+ "Traceback (most recent call last):\n",
596
+ " File \"/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/site-packages/torch/utils/data/dataloader.py\", line 1358, in __del__\n",
597
+ " self._shutdown_workers()\n",
598
+ " File \"/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/site-packages/torch/utils/data/dataloader.py\", line 1341, in _shutdown_workers\n",
599
+ " if w.is_alive():\n",
600
+ " File \"/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/multiprocessing/process.py\", line 151, in is_alive\n",
601
+ " assert self._parent_pid == os.getpid(), 'can only test a child process'\n",
602
+ "AssertionError: can only test a child process\n",
603
+ "Exception ignored in: <function _MultiProcessingDataLoaderIter.__del__ at 0x75ce96bc04d0>\n",
604
+ "Traceback (most recent call last):\n",
605
+ " File \"/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/site-packages/torch/utils/data/dataloader.py\", line 1358, in __del__\n",
606
+ " self._shutdown_workers()\n",
607
+ " File \"/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/site-packages/torch/utils/data/dataloader.py\", line 1341, in _shutdown_workers\n",
608
+ " if w.is_alive():\n",
609
+ " File \"/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/multiprocessing/process.py\", line 151, in is_alive\n",
610
+ " assert self._parent_pid == os.getpid(), 'can only test a child process'\n",
611
+ "AssertionError: can only test a child process\n",
612
+ "Exception ignored in: <function _MultiProcessingDataLoaderIter.__del__ at 0x75ce96bc04d0>\n",
613
+ "Traceback (most recent call last):\n",
614
+ " File \"/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/site-packages/torch/utils/data/dataloader.py\", line 1358, in __del__\n",
615
+ " self._shutdown_workers()\n",
616
+ " File \"/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/site-packages/torch/utils/data/dataloader.py\", line 1341, in _shutdown_workers\n",
617
+ " if w.is_alive():\n",
618
+ " File \"/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/multiprocessing/process.py\", line 151, in is_alive\n",
619
+ " assert self._parent_pid == os.getpid(), 'can only test a child process'\n",
620
+ "AssertionError: can only test a child process\n",
621
+ "Exception ignored in: <function _MultiProcessingDataLoaderIter.__del__ at 0x75ce96bc04d0>\n",
622
+ "Traceback (most recent call last):\n",
623
+ " File \"/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/site-packages/torch/utils/data/dataloader.py\", line 1358, in __del__\n",
624
+ " self._shutdown_workers()\n",
625
+ " File \"/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/site-packages/torch/utils/data/dataloader.py\", line 1341, in _shutdown_workers\n",
626
+ " if w.is_alive():\n",
627
+ " File \"/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/multiprocessing/process.py\", line 151, in is_alive\n",
628
+ " assert self._parent_pid == os.getpid(), 'can only test a child process'\n",
629
+ "AssertionError: can only test a child process\n",
630
+ "Testing: 32it [00:15, 2.06it/s] \n"
631
+ ]
632
+ }
633
+ ],
634
+ "source": [
635
+ "import torch\n",
636
+ "model.eval()\n",
637
+ "y_true, y_pred = [], []\n",
638
+ "with torch.no_grad():\n",
639
+ " for batch in tqdm(test_loader, desc=\"Testing\"):\n",
640
+ " inputs = batch[\"video\"].to(\"cuda\")\n",
641
+ " labels = batch[\"label\"].to(\"cuda\")\n",
642
+ " outputs = model(inputs)\n",
643
+ " probabilities = torch.nn.functional.softmax(outputs, dim=1)\n",
644
+ " predictions = probabilities.argmax(dim=-1)\n",
645
+ " y_true.extend(labels.cpu().tolist())\n",
646
+ " y_pred.extend(predictions.cpu().tolist())"
647
+ ]
648
+ },
649
+ {
650
+ "cell_type": "code",
651
+ "execution_count": 27,
652
+ "id": "6fbcc06f-d230-4c46-9cd8-c2aa83aeca19",
653
+ "metadata": {},
654
+ "outputs": [
655
+ {
656
+ "name": "stdout",
657
+ "output_type": "stream",
658
+ "text": [
659
+ "Accuracy: 0.7607 | F1-macro: 0.7426\n",
660
+ " precision recall f1-score support\n",
661
+ "\n",
662
+ " khongxarac 0.75 0.88 0.81 68\n",
663
+ " xarac 0.78 0.59 0.67 49\n",
664
+ "\n",
665
+ " accuracy 0.76 117\n",
666
+ " macro avg 0.77 0.74 0.74 117\n",
667
+ "weighted avg 0.76 0.76 0.75 117\n",
668
+ "\n",
669
+ "Confusion matrix:\n",
670
+ " [[60 8]\n",
671
+ " [20 29]]\n"
672
+ ]
673
+ }
674
+ ],
675
+ "source": [
676
+ "acc = accuracy_score(y_true, y_pred)\n",
677
+ "f1m = f1_score(y_true, y_pred, average=\"macro\")\n",
678
+ "print(f\"Accuracy: {acc:.4f} | F1-macro: {f1m:.4f}\")\n",
679
+ "print(classification_report(y_true, y_pred, target_names=class_names))\n",
680
+ "print(\"Confusion matrix:\\n\", confusion_matrix(y_true, y_pred))"
681
+ ]
682
+ },
683
+ {
684
+ "cell_type": "code",
685
+ "execution_count": null,
686
+ "id": "0b78a3c0-72ef-4207-bb66-03a8a3b02782",
687
+ "metadata": {},
688
+ "outputs": [],
689
+ "source": []
690
+ }
691
+ ],
692
+ "metadata": {
693
+ "kernelspec": {
694
+ "display_name": "TQKhang-ViViT",
695
+ "language": "python",
696
+ "name": "tqkhang-vivit"
697
+ },
698
+ "language_info": {
699
+ "codemirror_mode": {
700
+ "name": "ipython",
701
+ "version": 3
702
+ },
703
+ "file_extension": ".py",
704
+ "mimetype": "text/x-python",
705
+ "name": "python",
706
+ "nbconvert_exporter": "python",
707
+ "pygments_lexer": "ipython3",
708
+ "version": "3.7.16"
709
+ }
710
+ },
711
+ "nbformat": 4,
712
+ "nbformat_minor": 5
713
+ }
scripts/.ipynb_checkpoints/run_self_influence-checkpoint.sh ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-01_16-22-06-caformer_s36/f1/model_epoch_16_f1_0.7191.pth'
4
+
5
+ python -m soups.self_influence \
6
+ --seed 42 \
7
+ --checkpoint_path "${CHECKPOINTS_DIR}" \
8
+ --device cpu \
9
+ --output_file self_influence_results/_test.json \
10
+ --model timm/caformer_s36.sail_in22k_ft_in1k \
11
+ --dataset_dir data/ich-split-renamed \
12
+ --eval_batch_size 16 \
13
+ --num_workers 8
scripts/.ipynb_checkpoints/soup_caformer_b36-checkpoint.sh ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-19_19-37-23-caformer_m36-one_cycle_lr'
4
+
5
+ python -m soups.run_test_with_model_soups \
6
+ --seed 42 \
7
+ --checkpoint_path "$CHECKPOINTS_DIR" \
8
+ --model timm/caformer_b36.sail_in22k_ft_in1k \
9
+ --uniform_soup \
10
+ --greedy_soup \
11
+ --pruned_soup \
12
+ --pruned_soup_num_iters 64 \
13
+ --greedy_soup_comparison_metric f1 \
14
+ --dataset_dir data/ich-split-renamed \
15
+ --output_dir "${CHECKPOINTS_DIR}/soups_results"
scripts/.ipynb_checkpoints/soup_caformer_m36-checkpoint.sh ADDED
@@ -0,0 +1,17 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-19_19-37-23-caformer_m36-one_cycle_lr'
4
+
5
+ python -m soups.run_test_with_model_soups \
6
+ --seed 42 \
7
+ --checkpoint_path '/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-19_19-37-23-caformer_m36-one_cycle_lr/f1-top8' \
8
+ '/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-19_19-37-23-caformer_m36-one_cycle_lr/loss-top8' \
9
+ '/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-19_19-37-23-caformer_m36-one_cycle_lr/accuracy-top8' \
10
+ --model timm/caformer_m36.sail_in22k_ft_in1k \
11
+ --uniform_soup \
12
+ --greedy_soup \
13
+ --pruned_soup \
14
+ --pruned_soup_num_iters 64 \
15
+ --greedy_soup_comparison_metric f1 \
16
+ --dataset_dir data/ich-split-renamed \
17
+ --output_dir "${CHECKPOINTS_DIR}/soups_results"
scripts/.ipynb_checkpoints/soup_caformer_s36-checkpoint.sh ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-01_16-22-06-caformer_s36/f1'
4
+
5
+ python -m soups.run_test_with_model_soups \
6
+ --seed 42 \
7
+ --checkpoint_path "$CHECKPOINTS_DIR" \
8
+ --model timm/caformer_s36.sail_in22k_ft_in1k \
9
+ --uniform_soup \
10
+ --greedy_soup \
11
+ --dataset_dir data/ich-split-renamed \
12
+ --output_dir "${CHECKPOINTS_DIR}/soups_results"
scripts/.ipynb_checkpoints/soup_coatnet_0-merged_4_15-checkpoint.sh ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-26_20-43-44-coatnet_0_merged_4_15/loss'
4
+
5
+ python -m soups.run_test_with_model_soups \
6
+ --seed 42 \
7
+ --checkpoint_path "$CHECKPOINTS_DIR" \
8
+ --model timm/coatnet_0_rw_224.sw_in1k \
9
+ --uniform_soup \
10
+ --greedy_soup \
11
+ --dataset_dir data/ich-split-renamed-merged-4-15 \
12
+ --output_dir "${CHECKPOINTS_DIR}/soups_results"
scripts/.ipynb_checkpoints/soup_coatnet_0_co_teaching-checkpoint.sh ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-23_15-36-45-coatnet_0_co_teaching_forget_0_2/model_1_loss'
4
+
5
+ python -m soups.run_test_with_model_soups \
6
+ --seed 42 \
7
+ --checkpoint_path "$CHECKPOINTS_DIR" \
8
+ --model timm/coatnet_0_rw_224.sw_in1k \
9
+ --uniform_soup \
10
+ --greedy_soup \
11
+ --dataset_dir data/ich-split-renamed \
12
+ --output_dir "${CHECKPOINTS_DIR}/soups_results"
scripts/.ipynb_checkpoints/soup_coatnet_0_random-checkpoint.sh ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-23_10-38-52-coatnet_0_random/accuracy'
4
+
5
+ python -m soups.run_test_with_model_soups \
6
+ --seed 42 \
7
+ --checkpoint_path "$CHECKPOINTS_DIR" \
8
+ --model timm/coatnet_0_rw_224.sw_in1k \
9
+ --uniform_soup \
10
+ --greedy_soup \
11
+ --dataset_dir data/ich-split-renamed \
12
+ --output_dir "${CHECKPOINTS_DIR}/soups_results"
scripts/.ipynb_checkpoints/soup_coatnet_2-merged_4_15-checkpoint.sh ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-26_20-57-11-coatnet_2_merged_4_15/loss'
4
+
5
+ python -m soups.run_test_with_model_soups \
6
+ --seed 42 \
7
+ --checkpoint_path "$CHECKPOINTS_DIR" \
8
+ --model timm/coatnet_2_rw_224.sw_in12k_ft_in1k \
9
+ --uniform_soup \
10
+ --greedy_soup \
11
+ --dataset_dir data/ich-split-renamed-merged-4-15 \
12
+ --output_dir "${CHECKPOINTS_DIR}/soups_results"
scripts/.ipynb_checkpoints/soup_coatnet_2_random-checkpoint.sh ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-23_10-41-55-coatnet_2_random/accuracy'
4
+
5
+ python -m soups.run_test_with_model_soups \
6
+ --seed 42 \
7
+ --checkpoint_path "$CHECKPOINTS_DIR" \
8
+ --model timm/coatnet_2_rw_224.sw_in12k_ft_in1k \
9
+ --uniform_soup \
10
+ --greedy_soup \
11
+ --dataset_dir data/ich-split-renamed \
12
+ --output_dir "${CHECKPOINTS_DIR}/soups_results"
scripts/.ipynb_checkpoints/soup_convnext_base-checkpoint.sh ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-10_16-27-06-convnext_base'
4
+
5
+ python -m soups.run_test_with_model_soups \
6
+ --seed 42 \
7
+ --checkpoint_path "$CHECKPOINTS_DIR" \
8
+ --model timm/convnext_base.fb_in22k_ft_in1k \
9
+ --uniform_soup \
10
+ --greedy_soup \
11
+ --pruned_soup \
12
+ --pruned_soup_num_iters 64 \
13
+ --greedy_soup_comparison_metric f1 \
14
+ --dataset_dir data/ich-split-renamed \
15
+ --output_dir "${CHECKPOINTS_DIR}/soups_results"
scripts/.ipynb_checkpoints/soup_convnext_small-checkpoint.sh ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-10_14-54-53-convnext_small'
4
+
5
+ python -m soups.run_test_with_model_soups \
6
+ --seed 42 \
7
+ --checkpoint_path "$CHECKPOINTS_DIR" \
8
+ --model timm/convnext_small.in12k_ft_in1k \
9
+ --uniform_soup \
10
+ --greedy_soup \
11
+ --pruned_soup \
12
+ --pruned_soup_num_iters 64 \
13
+ --greedy_soup_comparison_metric f1 \
14
+ --dataset_dir data/ich-split-renamed \
15
+ --output_dir "${CHECKPOINTS_DIR}/soups_results"
scripts/.ipynb_checkpoints/soup_convnext_tiny-checkpoint.sh ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-10_15-23-36-convnext_tiny'
4
+
5
+ python -m soups.run_test_with_model_soups \
6
+ --seed 42 \
7
+ --checkpoint_path "$CHECKPOINTS_DIR" \
8
+ --model timm/convnext_tiny.in12k_ft_in1k \
9
+ --uniform_soup \
10
+ --greedy_soup \
11
+ --pruned_soup \
12
+ --pruned_soup_num_iters 64 \
13
+ --greedy_soup_comparison_metric f1 \
14
+ --dataset_dir data/ich-split-renamed \
15
+ --output_dir "${CHECKPOINTS_DIR}/soups_results"
scripts/.ipynb_checkpoints/soup_eva02_base-checkpoint.sh ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-02_00-11-46-eva02_base/accuracy'
4
+
5
+ python -m soups.run_test_with_model_soups \
6
+ --seed 42 \
7
+ --checkpoint_path "$CHECKPOINTS_DIR" \
8
+ --model timm/eva02_base_patch14_224.mim_in22k \
9
+ --uniform_soup \
10
+ --greedy_soup \
11
+ --greedy_soup_comparison_metric f1 \
12
+ --dataset_dir data/ich-split-renamed \
13
+ --output_dir "${CHECKPOINTS_DIR}/soups_results"
scripts/.ipynb_checkpoints/soup_eva02_large-checkpoint.sh ADDED
@@ -0,0 +1,16 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-02_01-21-34-eva02_large'
4
+
5
+ python -m soups.run_test_with_model_soups \
6
+ --seed 42 \
7
+ --checkpoint_path "$CHECKPOINTS_DIR" \
8
+ --model timm/eva02_large_patch14_224.mim_in22k \
9
+ --uniform_soup \
10
+ --greedy_soup \
11
+ --pruned_soup \
12
+ --pruned_soup_num_iters 64 \
13
+ --greedy_soup_comparison_metric f1 \
14
+ --eval_batch_size 4 \
15
+ --dataset_dir data/ich-split-renamed \
16
+ --output_dir "${CHECKPOINTS_DIR}/soups_results"
scripts/.ipynb_checkpoints/soup_eva02_small-checkpoint.sh ADDED
@@ -0,0 +1,16 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-20_14-45-41-eva02_small-reproduced'
4
+
5
+ python -m soups.run_test_with_model_soups \
6
+ --seed 42 \
7
+ --checkpoint_path "$CHECKPOINTS_DIR" \
8
+ --model timm/eva02_small_patch14_224.mim_in22k \
9
+ --uniform_soup \
10
+ --greedy_soup \
11
+ --pruned_soup \
12
+ --pruned_soup_num_iters 64 \
13
+ --remove_duplicate_checkpoints \
14
+ --greedy_soup_comparison_metric f1 \
15
+ --dataset_dir data/ich-split-renamed \
16
+ --output_dir "${CHECKPOINTS_DIR}/soups_results"
scripts/.ipynb_checkpoints/soup_focalnet_base_srf-checkpoint.sh ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-10_14-12-59-focalnet_base_srf/accuracy'
4
+
5
+ python -m soups.run_test_with_model_soups \
6
+ --seed 42 \
7
+ --checkpoint_path "$CHECKPOINTS_DIR" \
8
+ --model timm/focalnet_base_srf.ms_in1k \
9
+ --uniform_soup \
10
+ --greedy_soup \
11
+ --pruned_soup \
12
+ --pruned_soup_num_iters 64 \
13
+ --greedy_soup_comparison_metric f1 \
14
+ --dataset_dir data/ich-split-renamed \
15
+ --output_dir "${CHECKPOINTS_DIR}/soups_results"
scripts/.ipynb_checkpoints/soup_focalnet_small_lrf-checkpoint.sh ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-10_14-43-07-focalnet_small_lrf/accuracy'
4
+
5
+ python -m soups.run_test_with_model_soups \
6
+ --seed 42 \
7
+ --checkpoint_path "$CHECKPOINTS_DIR" \
8
+ --model timm/focalnet_small_lrf.ms_in1k \
9
+ --uniform_soup \
10
+ --greedy_soup \
11
+ --pruned_soup \
12
+ --pruned_soup_num_iters 64 \
13
+ --greedy_soup_comparison_metric f1 \
14
+ --dataset_dir data/ich-split-renamed \
15
+ --output_dir "${CHECKPOINTS_DIR}/soups_results"
scripts/.ipynb_checkpoints/soup_focalnet_small_srf-checkpoint.sh ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-02_10-13-09-focalnet_small_srf/accuracy'
4
+
5
+ python -m soups.run_test_with_model_soups \
6
+ --seed 42 \
7
+ --checkpoint_path "$CHECKPOINTS_DIR" \
8
+ --model timm/focalnet_small_srf.ms_in1k \
9
+ --uniform_soup \
10
+ --greedy_soup \
11
+ --pruned_soup \
12
+ --pruned_soup_num_iters 64 \
13
+ --greedy_soup_comparison_metric f1 \
14
+ --dataset_dir data/ich-split-renamed \
15
+ --output_dir "${CHECKPOINTS_DIR}/soups_results"
scripts/.ipynb_checkpoints/soup_hgnetv2_b6_ssld_stage2-checkpoint.sh ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-19_11-57-53-hgnetv2_b6.ssld_stage2_ft_in1k-one_cycle_lr'
4
+
5
+ python -m soups.run_test_with_model_soups \
6
+ --seed 42 \
7
+ --checkpoint_path "$CHECKPOINTS_DIR" \
8
+ --model timm/hgnetv2_b6.ssld_stage2_ft_in1k \
9
+ --uniform_soup \
10
+ --greedy_soup \
11
+ --pruned_soup \
12
+ --pruned_soup_num_iters 64 \
13
+ --greedy_soup_comparison_metric f1 \
14
+ --dataset_dir data/ich-split-renamed \
15
+ --output_dir "${CHECKPOINTS_DIR}/soups_results"
scripts/.ipynb_checkpoints/soup_maxvit_base-checkpoint.sh ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-01_15-07-03-maxvit_base'
4
+
5
+ python -m soups.run_test_with_model_soups \
6
+ --seed 42 \
7
+ --checkpoint_path "$CHECKPOINTS_DIR" \
8
+ --model timm/maxvit_base_tf_224.in1k \
9
+ --uniform_soup \
10
+ --greedy_soup \
11
+ --pruned_soup \
12
+ --pruned_soup_num_iters 64 \
13
+ --greedy_soup_comparison_metric f1 \
14
+ --dataset_dir data/ich-split-renamed \
15
+ --output_dir "${CHECKPOINTS_DIR}/soups_results"
scripts/.ipynb_checkpoints/soup_swin-checkpoint.sh ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-31_11-32-23-swin'
4
+
5
+ python -m soups.run_test_with_model_soups \
6
+ --seed 42 \
7
+ --checkpoint_path '/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-31_11-32-23-swin/accuracy' '/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-31_11-32-23-swin/f1' '/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-31_11-32-23-swin/loss' \
8
+ --model timm/swin_base_patch4_window7_224.ms_in22k_ft_in1k \
9
+ --uniform_soup \
10
+ --greedy_soup \
11
+ --pruned_soup \
12
+ --pruned_soup_num_iters 64 \
13
+ --dataset_dir data/ich-split-renamed \
14
+ --output_dir "${CHECKPOINTS_DIR}/soups_results"
scripts/.ipynb_checkpoints/soup_tiny_vit_21m-checkpoint.sh ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-01_13-56-39-tiny_vit_21m_dist'
4
+
5
+ python -m soups.run_test_with_model_soups \
6
+ --seed 42 \
7
+ --checkpoint_path "$CHECKPOINTS_DIR" \
8
+ --model timm/tiny_vit_21m_224.dist_in22k_ft_in1k \
9
+ --uniform_soup \
10
+ --greedy_soup \
11
+ --pruned_soup \
12
+ --pruned_soup_num_iters 64 \
13
+ --greedy_soup_comparison_metric f1 \
14
+ --dataset_dir data/ich-split-renamed \
15
+ --output_dir "${CHECKPOINTS_DIR}/soups_results"
scripts/.ipynb_checkpoints/soup_vit_base-checkpoint.sh ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-17_20-27-15-vit_base'
4
+
5
+ python -m soups.run_test_with_model_soups \
6
+ --seed 42 \
7
+ --checkpoint_path "$CHECKPOINTS_DIR" \
8
+ --model timm/vit_base_patch16_224.augreg2_in21k_ft_in1k \
9
+ --uniform_soup \
10
+ --greedy_soup \
11
+ --pruned_soup \
12
+ --pruned_soup_num_iters 64 \
13
+ --greedy_soup_comparison_metric f1 \
14
+ --dataset_dir data/ich-split-renamed \
15
+ --output_dir "${CHECKPOINTS_DIR}/soups_results"
scripts/.ipynb_checkpoints/soup_vit_base_laion2b-checkpoint.sh ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-17_20-07-52-vit_base_laion2b'
4
+
5
+ python -m soups.run_test_with_model_soups \
6
+ --seed 42 \
7
+ --checkpoint_path "$CHECKPOINTS_DIR" \
8
+ --model timm/vit_base_patch16_clip_224.laion2b_ft_in1k \
9
+ --uniform_soup \
10
+ --greedy_soup \
11
+ --pruned_soup \
12
+ --pruned_soup_num_iters 64 \
13
+ --greedy_soup_comparison_metric f1 \
14
+ --dataset_dir data/ich-split-renamed \
15
+ --output_dir "${CHECKPOINTS_DIR}/soups_results"
scripts/.ipynb_checkpoints/test_caformer_b36-checkpoint.sh ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-14_09-18-13-caformer_b36-seed_7-one_cycle_lr'
4
+
5
+ python -m soups.run_test_multiple_checkpoints \
6
+ --seed 42 \
7
+ --checkpoint_path "$CHECKPOINTS_DIR" \
8
+ --model timm/caformer_b36.sail_in22k_ft_in1k \
9
+ --dataset_dir data/ich-split-renamed \
10
+ --output_file "${CHECKPOINTS_DIR}/test_results.json"
scripts/.ipynb_checkpoints/test_caformer_m36-checkpoint.sh ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-19_19-37-23-caformer_m36-one_cycle_lr'
4
+
5
+ python -m soups.run_test_multiple_checkpoints \
6
+ --seed 42 \
7
+ --checkpoint_paths '/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-19_19-37-23-caformer_m36-one_cycle_lr/f1-top8' \
8
+ '/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-19_19-37-23-caformer_m36-one_cycle_lr/accuracy-top8' \
9
+ '/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-19_19-37-23-caformer_m36-one_cycle_lr/loss-top8' \
10
+ --model timm/caformer_m36.sail_in22k_ft_in1k \
11
+ --dataset_dir data/ich-split-renamed \
12
+ --output_file "${CHECKPOINTS_DIR}/test_results.json"
scripts/.ipynb_checkpoints/test_caformer_s36-checkpoint.sh ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='//home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-01_16-22-06-caformer_s36'
4
+
5
+ python -m soups.run_test_multiple_checkpoints \
6
+ --seed 42 \
7
+ --checkpoints_dir "$CHECKPOINTS_DIR" \
8
+ --model timm/caformer_s36.sail_in22k_ft_in1k \
9
+ --dataset_dir data/ich-split-renamed \
10
+ --output_file "${CHECKPOINTS_DIR}/test_results.json"
scripts/.ipynb_checkpoints/test_coatnet_0-checkpoint.sh ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-31_11-30-41-coatnet_0_filtered_1000'
4
+
5
+ python -m soups.run_test_multiple_checkpoints \
6
+ --seed 42 \
7
+ --checkpoints_dir "$CHECKPOINTS_DIR" \
8
+ --model timm/coatnet_0_rw_224.sw_in1k \
9
+ --dataset_dir data/ich-split-renamed \
10
+ --output_file "${CHECKPOINTS_DIR}/test_results.json"
scripts/.ipynb_checkpoints/test_coatnet_0-merged_4_15-checkpoint.sh ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-26_20-43-44-coatnet_0_merged_4_15'
4
+
5
+ python -m soups.run_test_multiple_checkpoints \
6
+ --seed 42 \
7
+ --checkpoints_dir "$CHECKPOINTS_DIR" \
8
+ --model timm/coatnet_0_rw_224.sw_in1k \
9
+ --dataset_dir data/ich-split-renamed-merged-4-15 \
10
+ --output_file "${CHECKPOINTS_DIR}/test_results.json"
scripts/.ipynb_checkpoints/test_coatnet_0_co_teaching-checkpoint.sh ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-23_15-36-45-coatnet_0_co_teaching_forget_0_2/model_1'
4
+
5
+ python -m soups.run_test_multiple_checkpoints \
6
+ --seed 42 \
7
+ --checkpoints_dir "$CHECKPOINTS_DIR" \
8
+ --model timm/coatnet_0_rw_224.sw_in1k \
9
+ --dataset_dir data/ich-split-renamed \
10
+ --output_file "${CHECKPOINTS_DIR}/test_results.json"
scripts/.ipynb_checkpoints/test_coatnet_0_random-checkpoint.sh ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-23_10-38-52-coatnet_0_random'
4
+
5
+ python -m soups.run_test_multiple_checkpoints \
6
+ --seed 42 \
7
+ --checkpoints_dir "$CHECKPOINTS_DIR" \
8
+ --model timm/coatnet_0_rw_224.sw_in1k \
9
+ --dataset_dir data/ich-split-renamed \
10
+ --output_file "${CHECKPOINTS_DIR}/test_results.json"
scripts/.ipynb_checkpoints/test_coatnet_2-merged_4_15-checkpoint.sh ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-26_20-57-11-coatnet_2_merged_4_15'
4
+
5
+ python -m soups.run_test_multiple_checkpoints \
6
+ --seed 42 \
7
+ --checkpoints_dir "$CHECKPOINTS_DIR" \
8
+ --model timm/coatnet_2_rw_224.sw_in12k_ft_in1k \
9
+ --dataset_dir data/ich-split-renamed-merged-4-15 \
10
+ --output_file "${CHECKPOINTS_DIR}/test_results.json"
scripts/.ipynb_checkpoints/test_coatnet_2_random-checkpoint.sh ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-23_10-41-55-coatnet_2_random'
4
+
5
+ python -m soups.run_test_multiple_checkpoints \
6
+ --seed 42 \
7
+ --checkpoints_dir "$CHECKPOINTS_DIR" \
8
+ --model timm/coatnet_2_rw_224.sw_in12k_ft_in1k \
9
+ --dataset_dir data/ich-split-renamed \
10
+ --output_file "${CHECKPOINTS_DIR}/test_results.json"
scripts/.ipynb_checkpoints/test_convnext_base-checkpoint.sh ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-10_16-27-06-convnext_base'
4
+
5
+ python -m soups.run_test_multiple_checkpoints \
6
+ --seed 42 \
7
+ --checkpoints_dir "$CHECKPOINTS_DIR" \
8
+ --model timm/convnext_base.fb_in22k_ft_in1k \
9
+ --dataset_dir data/ich-split-renamed \
10
+ --output_file "${CHECKPOINTS_DIR}/test_results.json"
scripts/.ipynb_checkpoints/test_convnext_femto-checkpoint.sh ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-10_15-31-58-convnext_femto'
4
+
5
+ python -m soups.run_test_multiple_checkpoints \
6
+ --seed 42 \
7
+ --checkpoints_dir "$CHECKPOINTS_DIR" \
8
+ --model timm/convnext_femto.d1_in1k \
9
+ --dataset_dir data/ich-split-renamed \
10
+ --output_file "${CHECKPOINTS_DIR}/test_results.json"
scripts/.ipynb_checkpoints/test_convnext_large-checkpoint.sh ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-11_16-47-37-convnext_large6'
4
+
5
+ python -m soups.run_test_multiple_checkpoints \
6
+ --seed 42 \
7
+ --checkpoints_dir "$CHECKPOINTS_DIR" \
8
+ --model timm/convnext_large.fb_in22k_ft_in1k \
9
+ --dataset_dir data/ich-split-renamed \
10
+ --output_file "${CHECKPOINTS_DIR}/test_results.json"
scripts/.ipynb_checkpoints/test_convnext_nano-checkpoint.sh ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-10_14-49-36-convnext_nano'
4
+
5
+ python -m soups.run_test_multiple_checkpoints \
6
+ --seed 42 \
7
+ --checkpoints_dir "$CHECKPOINTS_DIR" \
8
+ --model timm/convnext_nano.in12k_ft_in1k \
9
+ --dataset_dir data/ich-split-renamed \
10
+ --output_file "${CHECKPOINTS_DIR}/test_results.json"
scripts/.ipynb_checkpoints/test_convnext_pico-checkpoint.sh ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-10_15-24-36-convnext_pico'
4
+
5
+ python -m soups.run_test_multiple_checkpoints \
6
+ --seed 42 \
7
+ --checkpoints_dir "$CHECKPOINTS_DIR" \
8
+ --model timm/convnext_pico.d1_in1k \
9
+ --dataset_dir data/ich-split-renamed \
10
+ --output_file "${CHECKPOINTS_DIR}/test_results.json"
scripts/.ipynb_checkpoints/test_convnext_small-checkpoint.sh ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-10_14-54-53-convnext_small'
4
+
5
+ python -m soups.run_test_multiple_checkpoints \
6
+ --seed 42 \
7
+ --checkpoints_dir "$CHECKPOINTS_DIR" \
8
+ --model timm/convnext_small.in12k_ft_in1k \
9
+ --dataset_dir data/ich-split-renamed \
10
+ --output_file "${CHECKPOINTS_DIR}/test_results.json"
scripts/.ipynb_checkpoints/test_convnext_tiny-checkpoint.sh ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-10_15-23-36-convnext_tiny'
4
+
5
+ python -m soups.run_test_multiple_checkpoints \
6
+ --seed 42 \
7
+ --checkpoints_dir "$CHECKPOINTS_DIR" \
8
+ --model timm/convnext_tiny.in12k_ft_in1k \
9
+ --dataset_dir data/ich-split-renamed \
10
+ --output_file "${CHECKPOINTS_DIR}/test_results.json"
scripts/.ipynb_checkpoints/test_convnext_v2_large-checkpoint.sh ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync2/2025-09-11_17-25-36-convnext_v2_large'
4
+
5
+ python -m soups.run_test_multiple_checkpoints \
6
+ --seed 42 \
7
+ --checkpoints_dir "$CHECKPOINTS_DIR" \
8
+ --model timm/convnextv2_large.fcmae_ft_in22k_in1k \
9
+ --dataset_dir data/ich-split-renamed \
10
+ --output_file "${CHECKPOINTS_DIR}/test_results.json"
scripts/.ipynb_checkpoints/test_convnext_xlarge-checkpoint.sh ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync2/2025-09-11_11-58-22-convnext_xlarge'
4
+
5
+ python -m soups.run_test_multiple_checkpoints \
6
+ --seed 42 \
7
+ --checkpoints_dir "$CHECKPOINTS_DIR" \
8
+ --model timm/convnext_xlarge.fb_in22k_ft_in1k \
9
+ --dataset_dir data/ich-split-renamed \
10
+ --output_file "${CHECKPOINTS_DIR}/test_results.json"
scripts/.ipynb_checkpoints/test_convnext_xxlarge-checkpoint.sh ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync2/2025-09-11_13-53-47-convnext_xxlarge'
4
+
5
+ python -m soups.run_test_multiple_checkpoints \
6
+ --seed 42 \
7
+ --checkpoints_dir "$CHECKPOINTS_DIR" \
8
+ --model timm/convnext_xxlarge.clip_laion2b_soup_ft_in1k \
9
+ --dataset_dir data/ich-split-renamed \
10
+ --output_file "${CHECKPOINTS_DIR}/test_results.json"
scripts/.ipynb_checkpoints/test_eva02_base-checkpoint.sh ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-02_00-11-46-eva02_base'
4
+
5
+ python -m soups.run_test_multiple_checkpoints \
6
+ --seed 42 \
7
+ --checkpoints_dir "$CHECKPOINTS_DIR" \
8
+ --model timm/eva02_base_patch14_224.mim_in22k \
9
+ --dataset_dir data/ich-split-renamed \
10
+ --output_file "${CHECKPOINTS_DIR}/test_results.json"
scripts/.ipynb_checkpoints/test_eva02_large-checkpoint.sh ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-14_00-21-30-eva02_large-one_cycle_lr'
4
+
5
+ python -m soups.run_test_multiple_checkpoints \
6
+ --seed 42 \
7
+ --checkpoints_dir "$CHECKPOINTS_DIR" \
8
+ --model timm/eva02_large_patch14_224.mim_in22k \
9
+ --dataset_dir data/ich-split-renamed \
10
+ --output_file "${CHECKPOINTS_DIR}/test_results.json"
scripts/.ipynb_checkpoints/test_eva02_small-checkpoint.sh ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-20_14-45-41-eva02_small-reproduced'
4
+
5
+ python -m soups.run_test_multiple_checkpoints \
6
+ --seed 42 \
7
+ --checkpoint_paths "$CHECKPOINTS_DIR" \
8
+ --model timm/eva02_small_patch14_224.mim_in22k \
9
+ --dataset_dir data/ich-split-renamed \
10
+ --output_file "${CHECKPOINTS_DIR}/test_results.json"
scripts/.ipynb_checkpoints/test_focalnet_base_srf-checkpoint.sh ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/media/khmt/HDD1/TQKhang/mthien/soups/checkpoints/2025-09-10_14-12-59-focalnet_base_srf'
4
+
5
+ python -m soups.run_test_multiple_checkpoints \
6
+ --seed 42 \
7
+ --checkpoint_path "$CHECKPOINTS_DIR" \
8
+ --model timm/focalnet_base_srf.ms_in1k \
9
+ --dataset_dir data/ich-split-renamed \
10
+ --output_file "${CHECKPOINTS_DIR}/test_results.json"
scripts/.ipynb_checkpoints/test_focalnet_small_lrf-checkpoint.sh ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+
3
+ CHECKPOINTS_DIR='/media/khmt/HDD1/TQKhang/mthien/soups/checkpoints/2025-09-10_14-43-07-focalnet_small_lrf'
4
+
5
+ python -m soups.run_test_multiple_checkpoints \
6
+ --seed 42 \
7
+ --checkpoint_path "$CHECKPOINTS_DIR" \
8
+ --model timm/focalnet_small_lrf.ms_in1k \
9
+ --dataset_dir data/ich-split-renamed \
10
+ --output_file "${CHECKPOINTS_DIR}/test_results.json"