Upload folder using huggingface_hub

scripts/.ipynb_checkpoints/ViViT-checkpoint.ipynb

@@ -0,0 +1,829 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "id": "18122027-63d4-45bc-a155-11d941da97b9",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "DATASET_PATH = \"/media/khmt/HDD1/TQKhang/datasets/xarac\""
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "id": "bc5b5096-93dd-4661-bd32-e1c66a3facfc",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "%load_ext autoreload\n",
+    "%autoreload 2"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "id": "59965104-6b47-4bd4-ae52-9924a91feed3",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "/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",
+      "  from .autonotebook import tqdm as notebook_tqdm\n"
+     ]
+    }
+   ],
+   "source": [
+    "from torch.optim import AdamW\n",
+    "from video_transformers import VideoModel\n",
+    "from video_transformers.backbones.transformers import TransformersBackbone\n",
+    "from video_transformers.backbones.timm import TimmBackbone\n",
+    "from video_transformers.data import VideoDataModule\n",
+    "from video_transformers.heads import LinearHead\n",
+    "from video_transformers.trainer import trainer_factory"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "id": "2f5c18af-5a5a-4e6d-91dd-446123218792",
+   "metadata": {
+    "scrolled": true
+   },
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "Some weights of the model checkpoint at facebook/timesformer-base-finetuned-k400 were not used when initializing TimesformerModel: ['classifier.weight', 'classifier.bias']\n",
+      "- 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",
+      "- 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",
+      "/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",
+      "  FutureWarning,\n"
+     ]
+    }
+   ],
+   "source": [
+    "backbone = TransformersBackbone(\"facebook/timesformer-base-finetuned-k400\", num_unfrozen_stages=1)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "id": "04eb1ed9-0497-48d0-a72e-7223939f257b",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "datamodule = VideoDataModule(\n",
+    "    train_root=\"/media/khmt/HDD1/TQKhang/datasets/xarac/train\",\n",
+    "    val_root=\"/media/khmt/HDD1/TQKhang/datasets/xarac/val\",\n",
+    "    test_root=\"/media/khmt/HDD1/TQKhang/datasets/xarac/val\",\n",
+    "    batch_size=4,\n",
+    "    num_workers=4,\n",
+    "    num_timesteps=8,\n",
+    "    preprocess_input_size=224,\n",
+    "    preprocess_clip_duration=1,\n",
+    "    preprocess_means=backbone.mean,\n",
+    "    preprocess_stds=backbone.std,\n",
+    "    preprocess_min_short_side=256,\n",
+    "    preprocess_max_short_side=320,\n",
+    "    preprocess_horizontal_flip_p=0.5,\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "id": "12b51c8e-8420-456a-96fb-a185165c990a",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "2"
+      ]
+     },
+     "execution_count": 6,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "datamodule.num_classes"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "id": "a6189aca-e28c-4147-8efc-fae604663d14",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "head = LinearHead(hidden_size=backbone.num_features, num_classes=datamodule.num_classes)\n",
+    "model = VideoModel(backbone, head)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "id": "f3ed0a1e-d90c-49b1-a213-55815d0cdf15",
+   "metadata": {
+    "scrolled": true
+   },
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "VideoModel(\n",
+       "  (backbone): TransformersBackbone(\n",
+       "    (model): TimesformerModel(\n",
+       "      (embeddings): TimesformerEmbeddings(\n",
+       "        (patch_embeddings): TimesformerPatchEmbeddings(\n",
+       "          (projection): Conv2d(3, 768, kernel_size=(16, 16), stride=(16, 16))\n",
+       "        )\n",
+       "        (pos_drop): Dropout(p=0.0, inplace=False)\n",
+       "        (time_drop): Dropout(p=0.0, inplace=False)\n",
+       "      )\n",
+       "      (encoder): TimesformerEncoder(\n",
+       "        (layer): ModuleList(\n",
+       "          (0): TimesformerLayer(\n",
+       "            (drop_path): Identity()\n",
+       "            (attention): TimeSformerAttention(\n",
+       "              (attention): TimesformerSelfAttention(\n",
+       "                (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
+       "                (attn_drop): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): TimesformerSelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (intermediate): TimesformerIntermediate(\n",
+       "              (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              (intermediate_act_fn): GELUActivation()\n",
+       "            )\n",
+       "            (output): TimesformerOutput(\n",
+       "              (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "            )\n",
+       "            (layernorm_before): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (layernorm_after): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (temporal_layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (temporal_attention): TimeSformerAttention(\n",
+       "              (attention): TimesformerSelfAttention(\n",
+       "                (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
+       "                (attn_drop): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): TimesformerSelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (temporal_dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "          )\n",
+       "          (1): TimesformerLayer(\n",
+       "            (drop_path): Identity()\n",
+       "            (attention): TimeSformerAttention(\n",
+       "              (attention): TimesformerSelfAttention(\n",
+       "                (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
+       "                (attn_drop): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): TimesformerSelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (intermediate): TimesformerIntermediate(\n",
+       "              (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              (intermediate_act_fn): GELUActivation()\n",
+       "            )\n",
+       "            (output): TimesformerOutput(\n",
+       "              (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "            )\n",
+       "            (layernorm_before): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (layernorm_after): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (temporal_layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (temporal_attention): TimeSformerAttention(\n",
+       "              (attention): TimesformerSelfAttention(\n",
+       "                (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
+       "                (attn_drop): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): TimesformerSelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (temporal_dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "          )\n",
+       "          (2): TimesformerLayer(\n",
+       "            (drop_path): Identity()\n",
+       "            (attention): TimeSformerAttention(\n",
+       "              (attention): TimesformerSelfAttention(\n",
+       "                (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
+       "                (attn_drop): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): TimesformerSelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (intermediate): TimesformerIntermediate(\n",
+       "              (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              (intermediate_act_fn): GELUActivation()\n",
+       "            )\n",
+       "            (output): TimesformerOutput(\n",
+       "              (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "            )\n",
+       "            (layernorm_before): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (layernorm_after): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (temporal_layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (temporal_attention): TimeSformerAttention(\n",
+       "              (attention): TimesformerSelfAttention(\n",
+       "                (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
+       "                (attn_drop): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): TimesformerSelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (temporal_dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "          )\n",
+       "          (3): TimesformerLayer(\n",
+       "            (drop_path): Identity()\n",
+       "            (attention): TimeSformerAttention(\n",
+       "              (attention): TimesformerSelfAttention(\n",
+       "                (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
+       "                (attn_drop): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): TimesformerSelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (intermediate): TimesformerIntermediate(\n",
+       "              (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              (intermediate_act_fn): GELUActivation()\n",
+       "            )\n",
+       "            (output): TimesformerOutput(\n",
+       "              (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "            )\n",
+       "            (layernorm_before): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (layernorm_after): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (temporal_layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (temporal_attention): TimeSformerAttention(\n",
+       "              (attention): TimesformerSelfAttention(\n",
+       "                (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
+       "                (attn_drop): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): TimesformerSelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (temporal_dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "          )\n",
+       "          (4): TimesformerLayer(\n",
+       "            (drop_path): Identity()\n",
+       "            (attention): TimeSformerAttention(\n",
+       "              (attention): TimesformerSelfAttention(\n",
+       "                (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
+       "                (attn_drop): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): TimesformerSelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (intermediate): TimesformerIntermediate(\n",
+       "              (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              (intermediate_act_fn): GELUActivation()\n",
+       "            )\n",
+       "            (output): TimesformerOutput(\n",
+       "              (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "            )\n",
+       "            (layernorm_before): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (layernorm_after): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (temporal_layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (temporal_attention): TimeSformerAttention(\n",
+       "              (attention): TimesformerSelfAttention(\n",
+       "                (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
+       "                (attn_drop): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): TimesformerSelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (temporal_dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "          )\n",
+       "          (5): TimesformerLayer(\n",
+       "            (drop_path): Identity()\n",
+       "            (attention): TimeSformerAttention(\n",
+       "              (attention): TimesformerSelfAttention(\n",
+       "                (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
+       "                (attn_drop): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): TimesformerSelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (intermediate): TimesformerIntermediate(\n",
+       "              (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              (intermediate_act_fn): GELUActivation()\n",
+       "            )\n",
+       "            (output): TimesformerOutput(\n",
+       "              (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "            )\n",
+       "            (layernorm_before): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (layernorm_after): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (temporal_layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (temporal_attention): TimeSformerAttention(\n",
+       "              (attention): TimesformerSelfAttention(\n",
+       "                (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
+       "                (attn_drop): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): TimesformerSelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (temporal_dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "          )\n",
+       "          (6): TimesformerLayer(\n",
+       "            (drop_path): Identity()\n",
+       "            (attention): TimeSformerAttention(\n",
+       "              (attention): TimesformerSelfAttention(\n",
+       "                (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
+       "                (attn_drop): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): TimesformerSelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (intermediate): TimesformerIntermediate(\n",
+       "              (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              (intermediate_act_fn): GELUActivation()\n",
+       "            )\n",
+       "            (output): TimesformerOutput(\n",
+       "              (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "            )\n",
+       "            (layernorm_before): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (layernorm_after): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (temporal_layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (temporal_attention): TimeSformerAttention(\n",
+       "              (attention): TimesformerSelfAttention(\n",
+       "                (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
+       "                (attn_drop): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): TimesformerSelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (temporal_dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "          )\n",
+       "          (7): TimesformerLayer(\n",
+       "            (drop_path): Identity()\n",
+       "            (attention): TimeSformerAttention(\n",
+       "              (attention): TimesformerSelfAttention(\n",
+       "                (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
+       "                (attn_drop): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): TimesformerSelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (intermediate): TimesformerIntermediate(\n",
+       "              (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              (intermediate_act_fn): GELUActivation()\n",
+       "            )\n",
+       "            (output): TimesformerOutput(\n",
+       "              (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "            )\n",
+       "            (layernorm_before): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (layernorm_after): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (temporal_layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (temporal_attention): TimeSformerAttention(\n",
+       "              (attention): TimesformerSelfAttention(\n",
+       "                (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
+       "                (attn_drop): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): TimesformerSelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (temporal_dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "          )\n",
+       "          (8): TimesformerLayer(\n",
+       "            (drop_path): Identity()\n",
+       "            (attention): TimeSformerAttention(\n",
+       "              (attention): TimesformerSelfAttention(\n",
+       "                (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
+       "                (attn_drop): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): TimesformerSelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (intermediate): TimesformerIntermediate(\n",
+       "              (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              (intermediate_act_fn): GELUActivation()\n",
+       "            )\n",
+       "            (output): TimesformerOutput(\n",
+       "              (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "            )\n",
+       "            (layernorm_before): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (layernorm_after): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (temporal_layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (temporal_attention): TimeSformerAttention(\n",
+       "              (attention): TimesformerSelfAttention(\n",
+       "                (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
+       "                (attn_drop): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): TimesformerSelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (temporal_dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "          )\n",
+       "          (9): TimesformerLayer(\n",
+       "            (drop_path): Identity()\n",
+       "            (attention): TimeSformerAttention(\n",
+       "              (attention): TimesformerSelfAttention(\n",
+       "                (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
+       "                (attn_drop): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): TimesformerSelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (intermediate): TimesformerIntermediate(\n",
+       "              (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              (intermediate_act_fn): GELUActivation()\n",
+       "            )\n",
+       "            (output): TimesformerOutput(\n",
+       "              (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "            )\n",
+       "            (layernorm_before): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (layernorm_after): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (temporal_layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (temporal_attention): TimeSformerAttention(\n",
+       "              (attention): TimesformerSelfAttention(\n",
+       "                (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
+       "                (attn_drop): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): TimesformerSelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (temporal_dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "          )\n",
+       "          (10): TimesformerLayer(\n",
+       "            (drop_path): Identity()\n",
+       "            (attention): TimeSformerAttention(\n",
+       "              (attention): TimesformerSelfAttention(\n",
+       "                (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
+       "                (attn_drop): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): TimesformerSelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (intermediate): TimesformerIntermediate(\n",
+       "              (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              (intermediate_act_fn): GELUActivation()\n",
+       "            )\n",
+       "            (output): TimesformerOutput(\n",
+       "              (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "            )\n",
+       "            (layernorm_before): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (layernorm_after): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (temporal_layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (temporal_attention): TimeSformerAttention(\n",
+       "              (attention): TimesformerSelfAttention(\n",
+       "                (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
+       "                (attn_drop): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): TimesformerSelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (temporal_dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "          )\n",
+       "          (11): TimesformerLayer(\n",
+       "            (drop_path): Identity()\n",
+       "            (attention): TimeSformerAttention(\n",
+       "              (attention): TimesformerSelfAttention(\n",
+       "                (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
+       "                (attn_drop): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): TimesformerSelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (intermediate): TimesformerIntermediate(\n",
+       "              (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              (intermediate_act_fn): GELUActivation()\n",
+       "            )\n",
+       "            (output): TimesformerOutput(\n",
+       "              (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "            )\n",
+       "            (layernorm_before): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (layernorm_after): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (temporal_layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "            (temporal_attention): TimeSformerAttention(\n",
+       "              (attention): TimesformerSelfAttention(\n",
+       "                (qkv): Linear(in_features=768, out_features=2304, bias=True)\n",
+       "                (attn_drop): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): TimesformerSelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (temporal_dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "          )\n",
+       "        )\n",
+       "      )\n",
+       "      (layernorm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)\n",
+       "    )\n",
+       "  )\n",
+       "  (head): LinearHead(\n",
+       "    (linear): Linear(in_features=768, out_features=2, bias=True)\n",
+       "  )\n",
+       ")"
+      ]
+     },
+     "execution_count": 8,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "model"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "id": "a9d97e70-f539-4f34-8859-f69f1033a57e",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "optimizer = AdamW(model.parameters(), lr=1e-4)\n",
+    "\n",
+    "Trainer = trainer_factory(\"single_label_classification\")\n",
+    "trainer = Trainer(datamodule, model, optimizer=optimizer, max_epochs=10, \n",
+    "                  mixed_precision=\"fp16\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 10,
+   "id": "552054ca-55bd-4692-90ae-5f36b1c4d030",
+   "metadata": {
+    "scrolled": true
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Trainable parameteres: 10045442\n",
+      "Total parameteres: 121260290\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "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",
+      "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",
+      "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",
+      "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",
+      "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",
+      "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",
+      "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",
+      "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",
+      "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",
+      "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"
+     ]
+    }
+   ],
+   "source": [
+    "trainer.fit()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 11,
+   "id": "7f378e22-d85b-4714-981e-4b68c3fdeb73",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "['khongxarac', 'xarac']"
+      ]
+     },
+     "execution_count": 11,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "datamodule.labels"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 12,
+   "id": "ef16f8e1-6ad1-446a-a817-b01d4259e60b",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "{'num_timesteps': 8,\n",
+       " 'input_size': 224,\n",
+       " 'means': [0.45, 0.45, 0.45],\n",
+       " 'stds': [0.225, 0.225, 0.225],\n",
+       " 'min_short_side': 256,\n",
+       " 'max_short_side': 320,\n",
+       " 'horizontal_flip_p': 0.5,\n",
+       " 'clip_duration': 1}"
+      ]
+     },
+     "execution_count": 12,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "datamodule.preprocessor_config"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 13,
+   "id": "8e9c61dc-8c52-4280-8a02-becadafd304b",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import os, glob\n",
+    "from pathlib import Path\n",
+    "from tqdm import tqdm\n",
+    "from sklearn.metrics import accuracy_score, f1_score, classification_report, confusion_matrix"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 14,
+   "id": "a215c096-f988-4cb8-a5bd-7d320cc0d86b",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "test_loader = datamodule.test_dataloader\n",
+    "class_names = datamodule.labels"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 15,
+   "id": "ba8fa1bc-3796-4076-ab46-ab1cbb44ce66",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "Testing: 32it [00:12,  2.47it/s]                                                                                                                                                    \n"
+     ]
+    }
+   ],
+   "source": [
+    "import torch\n",
+    "model.eval()\n",
+    "y_true, y_pred = [], []\n",
+    "with torch.no_grad():\n",
+    "    for batch in tqdm(test_loader, desc=\"Testing\"):\n",
+    "        inputs = batch[\"video\"].to(\"cuda\")\n",
+    "        labels = batch[\"label\"].to(\"cuda\")\n",
+    "        outputs = model(inputs)\n",
+    "        probabilities = torch.nn.functional.softmax(outputs, dim=1)\n",
+    "        predictions = probabilities.argmax(dim=-1)\n",
+    "        y_true.extend(labels.cpu().tolist())\n",
+    "        y_pred.extend(predictions.cpu().tolist())"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 16,
+   "id": "6fbcc06f-d230-4c46-9cd8-c2aa83aeca19",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Accuracy: 0.7692 | F1-macro: 0.7608\n",
+      "              precision    recall  f1-score   support\n",
+      "\n",
+      "  khongxarac       0.79      0.82      0.81        68\n",
+      "       xarac       0.74      0.69      0.72        49\n",
+      "\n",
+      "    accuracy                           0.77       117\n",
+      "   macro avg       0.76      0.76      0.76       117\n",
+      "weighted avg       0.77      0.77      0.77       117\n",
+      "\n",
+      "Confusion matrix:\n",
+      " [[56 12]\n",
+      " [15 34]]\n"
+     ]
+    }
+   ],
+   "source": [
+    "acc = accuracy_score(y_true, y_pred)\n",
+    "f1m = f1_score(y_true, y_pred, average=\"macro\")\n",
+    "print(f\"Accuracy: {acc:.4f} | F1-macro: {f1m:.4f}\")\n",
+    "print(classification_report(y_true, y_pred, target_names=class_names))\n",
+    "print(\"Confusion matrix:\\n\", confusion_matrix(y_true, y_pred))"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "0b78a3c0-72ef-4207-bb66-03a8a3b02782",
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "7acdbc58-f8c8-4841-84c5-78fbe837f4fb",
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "TQKhang-ViViT",
+   "language": "python",
+   "name": "tqkhang-vivit"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.7.16"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

scripts/.ipynb_checkpoints/VideoMAE-checkpoint.ipynb

@@ -0,0 +1,713 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "id": "18122027-63d4-45bc-a155-11d941da97b9",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "DATASET_PATH = \"/media/khmt/HDD1/TQKhang/datasets/xarac\""
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "id": "bc5b5096-93dd-4661-bd32-e1c66a3facfc",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "%load_ext autoreload\n",
+    "%autoreload 2"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "id": "59965104-6b47-4bd4-ae52-9924a91feed3",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "/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",
+      "  from .autonotebook import tqdm as notebook_tqdm\n"
+     ]
+    }
+   ],
+   "source": [
+    "from torch.optim import AdamW\n",
+    "from video_transformers import VideoModel\n",
+    "from video_transformers.backbones.transformers import TransformersBackbone\n",
+    "from video_transformers.backbones.timm import TimmBackbone\n",
+    "from video_transformers.data import VideoDataModule\n",
+    "from video_transformers.heads import LinearHead\n",
+    "from video_transformers.trainer import trainer_factory"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "id": "2f5c18af-5a5a-4e6d-91dd-446123218792",
+   "metadata": {
+    "scrolled": true
+   },
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "Downloading: 22.9kB [00:00, 10.8MB/s]\n",
+      "Downloading: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 346M/346M [00:29<00:00, 11.9MB/s]\n",
+      "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",
+      "- 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",
+      "- 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",
+      "Downloading: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 271/271 [00:00<00:00, 220kB/s]\n",
+      "/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",
+      "  FutureWarning,\n"
+     ]
+    }
+   ],
+   "source": [
+    "backbone = TransformersBackbone(\"MCG-NJU/videomae-base-finetuned-kinetics\", num_unfrozen_stages=1)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 12,
+   "id": "04eb1ed9-0497-48d0-a72e-7223939f257b",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "datamodule = VideoDataModule(\n",
+    "    train_root=\"/media/khmt/HDD1/TQKhang/datasets/xarac/train\",\n",
+    "    val_root=\"/media/khmt/HDD1/TQKhang/datasets/xarac/val\",\n",
+    "    test_root=\"/media/khmt/HDD1/TQKhang/datasets/xarac/val\",\n",
+    "    batch_size=4,\n",
+    "    num_workers=4,\n",
+    "    num_timesteps=16,\n",
+    "    preprocess_input_size=224,\n",
+    "    preprocess_clip_duration=1,\n",
+    "    preprocess_means=backbone.mean,\n",
+    "    preprocess_stds=backbone.std,\n",
+    "    preprocess_min_short_side=256,\n",
+    "    preprocess_max_short_side=320,\n",
+    "    preprocess_horizontal_flip_p=0.5,\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 13,
+   "id": "12b51c8e-8420-456a-96fb-a185165c990a",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "2"
+      ]
+     },
+     "execution_count": 13,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "datamodule.num_classes"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 14,
+   "id": "a6189aca-e28c-4147-8efc-fae604663d14",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "head = LinearHead(hidden_size=backbone.num_features, num_classes=datamodule.num_classes)\n",
+    "model = VideoModel(backbone, head)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 15,
+   "id": "f3ed0a1e-d90c-49b1-a213-55815d0cdf15",
+   "metadata": {
+    "scrolled": true
+   },
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "VideoModel(\n",
+       "  (backbone): TransformersBackbone(\n",
+       "    (model): VideoMAEModel(\n",
+       "      (embeddings): VideoMAEEmbeddings(\n",
+       "        (patch_embeddings): VideoMAEPatchEmbeddings(\n",
+       "          (projection): Conv3d(3, 768, kernel_size=(2, 16, 16), stride=(2, 16, 16))\n",
+       "        )\n",
+       "      )\n",
+       "      (encoder): VideoMAEEncoder(\n",
+       "        (layer): ModuleList(\n",
+       "          (0): VideoMAELayer(\n",
+       "            (attention): VideoMAEAttention(\n",
+       "              (attention): VideoMAESelfAttention(\n",
+       "                (query): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (key): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (value): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): VideoMAESelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (intermediate): VideoMAEIntermediate(\n",
+       "              (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
+       "              (intermediate_act_fn): GELUActivation()\n",
+       "            )\n",
+       "            (output): VideoMAEOutput(\n",
+       "              (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "            )\n",
+       "            (layernorm_before): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
+       "            (layernorm_after): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
+       "          )\n",
+       "          (1): VideoMAELayer(\n",
+       "            (attention): VideoMAEAttention(\n",
+       "              (attention): VideoMAESelfAttention(\n",
+       "                (query): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (key): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (value): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): VideoMAESelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (intermediate): VideoMAEIntermediate(\n",
+       "              (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
+       "              (intermediate_act_fn): GELUActivation()\n",
+       "            )\n",
+       "            (output): VideoMAEOutput(\n",
+       "              (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "            )\n",
+       "            (layernorm_before): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
+       "            (layernorm_after): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
+       "          )\n",
+       "          (2): VideoMAELayer(\n",
+       "            (attention): VideoMAEAttention(\n",
+       "              (attention): VideoMAESelfAttention(\n",
+       "                (query): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (key): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (value): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): VideoMAESelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (intermediate): VideoMAEIntermediate(\n",
+       "              (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
+       "              (intermediate_act_fn): GELUActivation()\n",
+       "            )\n",
+       "            (output): VideoMAEOutput(\n",
+       "              (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "            )\n",
+       "            (layernorm_before): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
+       "            (layernorm_after): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
+       "          )\n",
+       "          (3): VideoMAELayer(\n",
+       "            (attention): VideoMAEAttention(\n",
+       "              (attention): VideoMAESelfAttention(\n",
+       "                (query): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (key): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (value): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): VideoMAESelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (intermediate): VideoMAEIntermediate(\n",
+       "              (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
+       "              (intermediate_act_fn): GELUActivation()\n",
+       "            )\n",
+       "            (output): VideoMAEOutput(\n",
+       "              (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "            )\n",
+       "            (layernorm_before): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
+       "            (layernorm_after): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
+       "          )\n",
+       "          (4): VideoMAELayer(\n",
+       "            (attention): VideoMAEAttention(\n",
+       "              (attention): VideoMAESelfAttention(\n",
+       "                (query): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (key): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (value): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): VideoMAESelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (intermediate): VideoMAEIntermediate(\n",
+       "              (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
+       "              (intermediate_act_fn): GELUActivation()\n",
+       "            )\n",
+       "            (output): VideoMAEOutput(\n",
+       "              (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "            )\n",
+       "            (layernorm_before): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
+       "            (layernorm_after): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
+       "          )\n",
+       "          (5): VideoMAELayer(\n",
+       "            (attention): VideoMAEAttention(\n",
+       "              (attention): VideoMAESelfAttention(\n",
+       "                (query): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (key): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (value): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): VideoMAESelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (intermediate): VideoMAEIntermediate(\n",
+       "              (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
+       "              (intermediate_act_fn): GELUActivation()\n",
+       "            )\n",
+       "            (output): VideoMAEOutput(\n",
+       "              (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "            )\n",
+       "            (layernorm_before): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
+       "            (layernorm_after): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
+       "          )\n",
+       "          (6): VideoMAELayer(\n",
+       "            (attention): VideoMAEAttention(\n",
+       "              (attention): VideoMAESelfAttention(\n",
+       "                (query): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (key): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (value): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): VideoMAESelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (intermediate): VideoMAEIntermediate(\n",
+       "              (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
+       "              (intermediate_act_fn): GELUActivation()\n",
+       "            )\n",
+       "            (output): VideoMAEOutput(\n",
+       "              (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "            )\n",
+       "            (layernorm_before): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
+       "            (layernorm_after): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
+       "          )\n",
+       "          (7): VideoMAELayer(\n",
+       "            (attention): VideoMAEAttention(\n",
+       "              (attention): VideoMAESelfAttention(\n",
+       "                (query): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (key): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (value): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): VideoMAESelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (intermediate): VideoMAEIntermediate(\n",
+       "              (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
+       "              (intermediate_act_fn): GELUActivation()\n",
+       "            )\n",
+       "            (output): VideoMAEOutput(\n",
+       "              (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "            )\n",
+       "            (layernorm_before): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
+       "            (layernorm_after): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
+       "          )\n",
+       "          (8): VideoMAELayer(\n",
+       "            (attention): VideoMAEAttention(\n",
+       "              (attention): VideoMAESelfAttention(\n",
+       "                (query): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (key): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (value): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): VideoMAESelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (intermediate): VideoMAEIntermediate(\n",
+       "              (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
+       "              (intermediate_act_fn): GELUActivation()\n",
+       "            )\n",
+       "            (output): VideoMAEOutput(\n",
+       "              (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "            )\n",
+       "            (layernorm_before): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
+       "            (layernorm_after): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
+       "          )\n",
+       "          (9): VideoMAELayer(\n",
+       "            (attention): VideoMAEAttention(\n",
+       "              (attention): VideoMAESelfAttention(\n",
+       "                (query): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (key): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (value): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): VideoMAESelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (intermediate): VideoMAEIntermediate(\n",
+       "              (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
+       "              (intermediate_act_fn): GELUActivation()\n",
+       "            )\n",
+       "            (output): VideoMAEOutput(\n",
+       "              (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "            )\n",
+       "            (layernorm_before): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
+       "            (layernorm_after): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
+       "          )\n",
+       "          (10): VideoMAELayer(\n",
+       "            (attention): VideoMAEAttention(\n",
+       "              (attention): VideoMAESelfAttention(\n",
+       "                (query): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (key): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (value): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): VideoMAESelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (intermediate): VideoMAEIntermediate(\n",
+       "              (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
+       "              (intermediate_act_fn): GELUActivation()\n",
+       "            )\n",
+       "            (output): VideoMAEOutput(\n",
+       "              (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "            )\n",
+       "            (layernorm_before): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
+       "            (layernorm_after): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
+       "          )\n",
+       "          (11): VideoMAELayer(\n",
+       "            (attention): VideoMAEAttention(\n",
+       "              (attention): VideoMAESelfAttention(\n",
+       "                (query): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (key): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (value): Linear(in_features=768, out_features=768, bias=False)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "              (output): VideoMAESelfOutput(\n",
+       "                (dense): Linear(in_features=768, out_features=768, bias=True)\n",
+       "                (dropout): Dropout(p=0.0, inplace=False)\n",
+       "              )\n",
+       "            )\n",
+       "            (intermediate): VideoMAEIntermediate(\n",
+       "              (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
+       "              (intermediate_act_fn): GELUActivation()\n",
+       "            )\n",
+       "            (output): VideoMAEOutput(\n",
+       "              (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
+       "              (dropout): Dropout(p=0.0, inplace=False)\n",
+       "            )\n",
+       "            (layernorm_before): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
+       "            (layernorm_after): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
+       "          )\n",
+       "        )\n",
+       "      )\n",
+       "    )\n",
+       "  )\n",
+       "  (head): LinearHead(\n",
+       "    (linear): Linear(in_features=768, out_features=2, bias=True)\n",
+       "  )\n",
+       ")"
+      ]
+     },
+     "execution_count": 15,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "model"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 16,
+   "id": "a9d97e70-f539-4f34-8859-f69f1033a57e",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "optimizer = AdamW(model.parameters(), lr=1e-4)\n",
+    "\n",
+    "Trainer = trainer_factory(\"single_label_classification\")\n",
+    "trainer = Trainer(datamodule, model, optimizer=optimizer, max_epochs=10, \n",
+    "                  mixed_precision=\"fp16\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 17,
+   "id": "552054ca-55bd-4692-90ae-5f36b1c4d030",
+   "metadata": {
+    "scrolled": true
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Trainable parameteres: 7088642\n",
+      "Total parameteres: 86227202\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "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",
+      "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",
+      "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",
+      "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",
+      "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",
+      "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",
+      "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",
+      "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",
+      "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",
+      "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"
+     ]
+    }
+   ],
+   "source": [
+    "trainer.fit()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 18,
+   "id": "7f378e22-d85b-4714-981e-4b68c3fdeb73",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "['khongxarac', 'xarac']"
+      ]
+     },
+     "execution_count": 18,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "datamodule.labels"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 19,
+   "id": "ef16f8e1-6ad1-446a-a817-b01d4259e60b",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "{'num_timesteps': 16,\n",
+       " 'input_size': 224,\n",
+       " 'means': [0.485, 0.456, 0.406],\n",
+       " 'stds': [0.229, 0.224, 0.225],\n",
+       " 'min_short_side': 256,\n",
+       " 'max_short_side': 320,\n",
+       " 'horizontal_flip_p': 0.5,\n",
+       " 'clip_duration': 1}"
+      ]
+     },
+     "execution_count": 19,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "datamodule.preprocessor_config"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 20,
+   "id": "8e9c61dc-8c52-4280-8a02-becadafd304b",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import os, glob\n",
+    "from pathlib import Path\n",
+    "from tqdm import tqdm\n",
+    "from sklearn.metrics import accuracy_score, f1_score, classification_report, confusion_matrix"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 21,
+   "id": "a215c096-f988-4cb8-a5bd-7d320cc0d86b",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "test_loader = datamodule.test_dataloader\n",
+    "class_names = datamodule.labels"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 26,
+   "id": "ba8fa1bc-3796-4076-ab46-ab1cbb44ce66",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "Testing:   0%|                                                                                                                                               | 0/30 [00:00<?, ?it/s]Exception ignored in: <function _MultiProcessingDataLoaderIter.__del__ at 0x75ce96bc04d0>\n",
+      "Traceback (most recent call last):\n",
+      "  File \"/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/site-packages/torch/utils/data/dataloader.py\", line 1358, in __del__\n",
+      "    self._shutdown_workers()\n",
+      "  File \"/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/site-packages/torch/utils/data/dataloader.py\", line 1341, in _shutdown_workers\n",
+      "    if w.is_alive():\n",
+      "  File \"/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/multiprocessing/process.py\", line 151, in is_alive\n",
+      "    assert self._parent_pid == os.getpid(), 'can only test a child process'\n",
+      "AssertionError: can only test a child process\n",
+      "Exception ignored in: <function _MultiProcessingDataLoaderIter.__del__ at 0x75ce96bc04d0>\n",
+      "Traceback (most recent call last):\n",
+      "  File \"/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/site-packages/torch/utils/data/dataloader.py\", line 1358, in __del__\n",
+      "    self._shutdown_workers()\n",
+      "  File \"/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/site-packages/torch/utils/data/dataloader.py\", line 1341, in _shutdown_workers\n",
+      "    if w.is_alive():\n",
+      "  File \"/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/multiprocessing/process.py\", line 151, in is_alive\n",
+      "    assert self._parent_pid == os.getpid(), 'can only test a child process'\n",
+      "AssertionError: can only test a child process\n",
+      "Exception ignored in: <function _MultiProcessingDataLoaderIter.__del__ at 0x75ce96bc04d0>\n",
+      "Traceback (most recent call last):\n",
+      "  File \"/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/site-packages/torch/utils/data/dataloader.py\", line 1358, in __del__\n",
+      "    self._shutdown_workers()\n",
+      "  File \"/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/site-packages/torch/utils/data/dataloader.py\", line 1341, in _shutdown_workers\n",
+      "    if w.is_alive():\n",
+      "  File \"/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/multiprocessing/process.py\", line 151, in is_alive\n",
+      "    assert self._parent_pid == os.getpid(), 'can only test a child process'\n",
+      "AssertionError: can only test a child process\n",
+      "Exception ignored in: <function _MultiProcessingDataLoaderIter.__del__ at 0x75ce96bc04d0>\n",
+      "Traceback (most recent call last):\n",
+      "  File \"/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/site-packages/torch/utils/data/dataloader.py\", line 1358, in __del__\n",
+      "    self._shutdown_workers()\n",
+      "  File \"/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/site-packages/torch/utils/data/dataloader.py\", line 1341, in _shutdown_workers\n",
+      "    if w.is_alive():\n",
+      "  File \"/home/khmt/anaconda3/envs/tqkhang-vivit/lib/python3.7/multiprocessing/process.py\", line 151, in is_alive\n",
+      "    assert self._parent_pid == os.getpid(), 'can only test a child process'\n",
+      "AssertionError: can only test a child process\n",
+      "Testing: 32it [00:15,  2.06it/s]                                                                                                                                                    \n"
+     ]
+    }
+   ],
+   "source": [
+    "import torch\n",
+    "model.eval()\n",
+    "y_true, y_pred = [], []\n",
+    "with torch.no_grad():\n",
+    "    for batch in tqdm(test_loader, desc=\"Testing\"):\n",
+    "        inputs = batch[\"video\"].to(\"cuda\")\n",
+    "        labels = batch[\"label\"].to(\"cuda\")\n",
+    "        outputs = model(inputs)\n",
+    "        probabilities = torch.nn.functional.softmax(outputs, dim=1)\n",
+    "        predictions = probabilities.argmax(dim=-1)\n",
+    "        y_true.extend(labels.cpu().tolist())\n",
+    "        y_pred.extend(predictions.cpu().tolist())"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 27,
+   "id": "6fbcc06f-d230-4c46-9cd8-c2aa83aeca19",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Accuracy: 0.7607 | F1-macro: 0.7426\n",
+      "              precision    recall  f1-score   support\n",
+      "\n",
+      "  khongxarac       0.75      0.88      0.81        68\n",
+      "       xarac       0.78      0.59      0.67        49\n",
+      "\n",
+      "    accuracy                           0.76       117\n",
+      "   macro avg       0.77      0.74      0.74       117\n",
+      "weighted avg       0.76      0.76      0.75       117\n",
+      "\n",
+      "Confusion matrix:\n",
+      " [[60  8]\n",
+      " [20 29]]\n"
+     ]
+    }
+   ],
+   "source": [
+    "acc = accuracy_score(y_true, y_pred)\n",
+    "f1m = f1_score(y_true, y_pred, average=\"macro\")\n",
+    "print(f\"Accuracy: {acc:.4f} | F1-macro: {f1m:.4f}\")\n",
+    "print(classification_report(y_true, y_pred, target_names=class_names))\n",
+    "print(\"Confusion matrix:\\n\", confusion_matrix(y_true, y_pred))"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "0b78a3c0-72ef-4207-bb66-03a8a3b02782",
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "TQKhang-ViViT",
+   "language": "python",
+   "name": "tqkhang-vivit"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.7.16"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

scripts/.ipynb_checkpoints/run_self_influence-checkpoint.sh

@@ -0,0 +1,13 @@
+#!/usr/bin/env bash
+
+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'
+
+python -m soups.self_influence \
+    --seed 42 \
+    --checkpoint_path "${CHECKPOINTS_DIR}" \
+    --device cpu \
+    --output_file self_influence_results/_test.json \
+    --model timm/caformer_s36.sail_in22k_ft_in1k \
+    --dataset_dir data/ich-split-renamed \
+    --eval_batch_size 16 \
+    --num_workers 8

scripts/.ipynb_checkpoints/soup_caformer_b36-checkpoint.sh

@@ -0,0 +1,15 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-19_19-37-23-caformer_m36-one_cycle_lr'
+
+python -m soups.run_test_with_model_soups \
+  --seed 42 \
+  --checkpoint_path "$CHECKPOINTS_DIR" \
+  --model timm/caformer_b36.sail_in22k_ft_in1k \
+  --uniform_soup \
+  --greedy_soup \
+  --pruned_soup \
+  --pruned_soup_num_iters 64 \
+  --greedy_soup_comparison_metric f1 \
+  --dataset_dir data/ich-split-renamed \
+  --output_dir "${CHECKPOINTS_DIR}/soups_results"

scripts/.ipynb_checkpoints/soup_caformer_m36-checkpoint.sh

@@ -0,0 +1,17 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-19_19-37-23-caformer_m36-one_cycle_lr'
+
+python -m soups.run_test_with_model_soups \
+  --seed 42 \
+  --checkpoint_path '/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-19_19-37-23-caformer_m36-one_cycle_lr/f1-top8' \
+      '/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-19_19-37-23-caformer_m36-one_cycle_lr/loss-top8' \
+      '/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-19_19-37-23-caformer_m36-one_cycle_lr/accuracy-top8' \
+  --model timm/caformer_m36.sail_in22k_ft_in1k \
+  --uniform_soup \
+  --greedy_soup \
+  --pruned_soup \
+  --pruned_soup_num_iters 64 \
+  --greedy_soup_comparison_metric f1 \
+  --dataset_dir data/ich-split-renamed \
+  --output_dir "${CHECKPOINTS_DIR}/soups_results"

scripts/.ipynb_checkpoints/soup_caformer_s36-checkpoint.sh

@@ -0,0 +1,12 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-01_16-22-06-caformer_s36/f1'
+
+python -m soups.run_test_with_model_soups \
+  --seed 42 \
+  --checkpoint_path "$CHECKPOINTS_DIR" \
+  --model timm/caformer_s36.sail_in22k_ft_in1k \
+  --uniform_soup \
+  --greedy_soup \
+  --dataset_dir data/ich-split-renamed \
+  --output_dir "${CHECKPOINTS_DIR}/soups_results"

scripts/.ipynb_checkpoints/soup_coatnet_0-merged_4_15-checkpoint.sh

@@ -0,0 +1,12 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-26_20-43-44-coatnet_0_merged_4_15/loss'
+
+python -m soups.run_test_with_model_soups \
+  --seed 42 \
+  --checkpoint_path "$CHECKPOINTS_DIR" \
+  --model timm/coatnet_0_rw_224.sw_in1k \
+  --uniform_soup \
+  --greedy_soup \
+  --dataset_dir data/ich-split-renamed-merged-4-15 \
+  --output_dir "${CHECKPOINTS_DIR}/soups_results"

scripts/.ipynb_checkpoints/soup_coatnet_0_co_teaching-checkpoint.sh

@@ -0,0 +1,12 @@
+#!/usr/bin/env bash
+
+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'
+
+python -m soups.run_test_with_model_soups \
+  --seed 42 \
+  --checkpoint_path "$CHECKPOINTS_DIR" \
+  --model timm/coatnet_0_rw_224.sw_in1k \
+  --uniform_soup \
+  --greedy_soup \
+  --dataset_dir data/ich-split-renamed \
+  --output_dir "${CHECKPOINTS_DIR}/soups_results"

scripts/.ipynb_checkpoints/soup_coatnet_0_random-checkpoint.sh

@@ -0,0 +1,12 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-23_10-38-52-coatnet_0_random/accuracy'
+
+python -m soups.run_test_with_model_soups \
+  --seed 42 \
+  --checkpoint_path "$CHECKPOINTS_DIR" \
+  --model timm/coatnet_0_rw_224.sw_in1k \
+  --uniform_soup \
+  --greedy_soup \
+  --dataset_dir data/ich-split-renamed \
+  --output_dir "${CHECKPOINTS_DIR}/soups_results"

scripts/.ipynb_checkpoints/soup_coatnet_2-merged_4_15-checkpoint.sh

@@ -0,0 +1,12 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-26_20-57-11-coatnet_2_merged_4_15/loss'
+
+python -m soups.run_test_with_model_soups \
+  --seed 42 \
+  --checkpoint_path "$CHECKPOINTS_DIR" \
+  --model timm/coatnet_2_rw_224.sw_in12k_ft_in1k \
+  --uniform_soup \
+  --greedy_soup \
+  --dataset_dir data/ich-split-renamed-merged-4-15 \
+  --output_dir "${CHECKPOINTS_DIR}/soups_results"

scripts/.ipynb_checkpoints/soup_coatnet_2_random-checkpoint.sh

@@ -0,0 +1,12 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-23_10-41-55-coatnet_2_random/accuracy'
+
+python -m soups.run_test_with_model_soups \
+  --seed 42 \
+  --checkpoint_path "$CHECKPOINTS_DIR" \
+  --model timm/coatnet_2_rw_224.sw_in12k_ft_in1k \
+  --uniform_soup \
+  --greedy_soup \
+  --dataset_dir data/ich-split-renamed \
+  --output_dir "${CHECKPOINTS_DIR}/soups_results"

scripts/.ipynb_checkpoints/soup_convnext_base-checkpoint.sh

@@ -0,0 +1,15 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-10_16-27-06-convnext_base'
+
+python -m soups.run_test_with_model_soups \
+  --seed 42 \
+  --checkpoint_path "$CHECKPOINTS_DIR" \
+  --model timm/convnext_base.fb_in22k_ft_in1k \
+  --uniform_soup \
+  --greedy_soup \
+  --pruned_soup \
+  --pruned_soup_num_iters 64 \
+  --greedy_soup_comparison_metric f1 \
+  --dataset_dir data/ich-split-renamed \
+  --output_dir "${CHECKPOINTS_DIR}/soups_results"

scripts/.ipynb_checkpoints/soup_convnext_small-checkpoint.sh

@@ -0,0 +1,15 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-10_14-54-53-convnext_small'
+
+python -m soups.run_test_with_model_soups \
+  --seed 42 \
+  --checkpoint_path "$CHECKPOINTS_DIR" \
+  --model timm/convnext_small.in12k_ft_in1k \
+  --uniform_soup \
+  --greedy_soup \
+  --pruned_soup \
+  --pruned_soup_num_iters 64 \
+  --greedy_soup_comparison_metric f1 \
+  --dataset_dir data/ich-split-renamed \
+  --output_dir "${CHECKPOINTS_DIR}/soups_results"

scripts/.ipynb_checkpoints/soup_convnext_tiny-checkpoint.sh

@@ -0,0 +1,15 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-10_15-23-36-convnext_tiny'
+
+python -m soups.run_test_with_model_soups \
+  --seed 42 \
+  --checkpoint_path "$CHECKPOINTS_DIR" \
+  --model timm/convnext_tiny.in12k_ft_in1k \
+  --uniform_soup \
+  --greedy_soup \
+  --pruned_soup \
+  --pruned_soup_num_iters 64 \
+  --greedy_soup_comparison_metric f1 \
+  --dataset_dir data/ich-split-renamed \
+  --output_dir "${CHECKPOINTS_DIR}/soups_results"

scripts/.ipynb_checkpoints/soup_eva02_base-checkpoint.sh

@@ -0,0 +1,13 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-02_00-11-46-eva02_base/accuracy'
+
+python -m soups.run_test_with_model_soups \
+  --seed 42 \
+  --checkpoint_path "$CHECKPOINTS_DIR" \
+  --model timm/eva02_base_patch14_224.mim_in22k  \
+  --uniform_soup \
+  --greedy_soup \
+  --greedy_soup_comparison_metric f1 \
+  --dataset_dir data/ich-split-renamed \
+  --output_dir "${CHECKPOINTS_DIR}/soups_results"

scripts/.ipynb_checkpoints/soup_eva02_large-checkpoint.sh

@@ -0,0 +1,16 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-02_01-21-34-eva02_large'
+
+python -m soups.run_test_with_model_soups \
+  --seed 42 \
+  --checkpoint_path "$CHECKPOINTS_DIR" \
+  --model timm/eva02_large_patch14_224.mim_in22k  \
+  --uniform_soup \
+  --greedy_soup \
+  --pruned_soup \
+  --pruned_soup_num_iters 64 \
+  --greedy_soup_comparison_metric f1 \
+  --eval_batch_size 4 \
+  --dataset_dir data/ich-split-renamed \
+  --output_dir "${CHECKPOINTS_DIR}/soups_results"

scripts/.ipynb_checkpoints/soup_eva02_small-checkpoint.sh

@@ -0,0 +1,16 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-20_14-45-41-eva02_small-reproduced'
+
+python -m soups.run_test_with_model_soups \
+  --seed 42 \
+  --checkpoint_path "$CHECKPOINTS_DIR" \
+  --model timm/eva02_small_patch14_224.mim_in22k  \
+  --uniform_soup \
+  --greedy_soup \
+  --pruned_soup \
+  --pruned_soup_num_iters 64 \
+  --remove_duplicate_checkpoints \
+  --greedy_soup_comparison_metric f1 \
+  --dataset_dir data/ich-split-renamed \
+  --output_dir "${CHECKPOINTS_DIR}/soups_results"

scripts/.ipynb_checkpoints/soup_focalnet_base_srf-checkpoint.sh

@@ -0,0 +1,15 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-10_14-12-59-focalnet_base_srf/accuracy'
+
+python -m soups.run_test_with_model_soups \
+  --seed 42 \
+  --checkpoint_path "$CHECKPOINTS_DIR" \
+  --model timm/focalnet_base_srf.ms_in1k \
+  --uniform_soup \
+  --greedy_soup \
+  --pruned_soup \
+  --pruned_soup_num_iters 64 \
+  --greedy_soup_comparison_metric f1 \
+  --dataset_dir data/ich-split-renamed \
+  --output_dir "${CHECKPOINTS_DIR}/soups_results"

scripts/.ipynb_checkpoints/soup_focalnet_small_lrf-checkpoint.sh

@@ -0,0 +1,15 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-10_14-43-07-focalnet_small_lrf/accuracy'
+
+python -m soups.run_test_with_model_soups \
+  --seed 42 \
+  --checkpoint_path "$CHECKPOINTS_DIR" \
+  --model timm/focalnet_small_lrf.ms_in1k \
+  --uniform_soup \
+  --greedy_soup \
+  --pruned_soup \
+  --pruned_soup_num_iters 64 \
+  --greedy_soup_comparison_metric f1 \
+  --dataset_dir data/ich-split-renamed \
+  --output_dir "${CHECKPOINTS_DIR}/soups_results"

scripts/.ipynb_checkpoints/soup_focalnet_small_srf-checkpoint.sh

@@ -0,0 +1,15 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-02_10-13-09-focalnet_small_srf/accuracy'
+
+python -m soups.run_test_with_model_soups \
+  --seed 42 \
+  --checkpoint_path "$CHECKPOINTS_DIR" \
+  --model timm/focalnet_small_srf.ms_in1k \
+  --uniform_soup \
+  --greedy_soup \
+  --pruned_soup \
+  --pruned_soup_num_iters 64 \
+  --greedy_soup_comparison_metric f1 \
+  --dataset_dir data/ich-split-renamed \
+  --output_dir "${CHECKPOINTS_DIR}/soups_results"

scripts/.ipynb_checkpoints/soup_hgnetv2_b6_ssld_stage2-checkpoint.sh

@@ -0,0 +1,15 @@
+#!/usr/bin/env bash
+
+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'
+
+python -m soups.run_test_with_model_soups \
+  --seed 42 \
+  --checkpoint_path "$CHECKPOINTS_DIR" \
+  --model timm/hgnetv2_b6.ssld_stage2_ft_in1k  \
+  --uniform_soup \
+  --greedy_soup \
+  --pruned_soup \
+  --pruned_soup_num_iters 64 \
+  --greedy_soup_comparison_metric f1 \
+  --dataset_dir data/ich-split-renamed \
+  --output_dir "${CHECKPOINTS_DIR}/soups_results"

scripts/.ipynb_checkpoints/soup_maxvit_base-checkpoint.sh

@@ -0,0 +1,15 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-01_15-07-03-maxvit_base'
+
+python -m soups.run_test_with_model_soups \
+  --seed 42 \
+  --checkpoint_path "$CHECKPOINTS_DIR" \
+  --model timm/maxvit_base_tf_224.in1k \
+  --uniform_soup \
+  --greedy_soup \
+  --pruned_soup \
+  --pruned_soup_num_iters 64 \
+  --greedy_soup_comparison_metric f1 \
+  --dataset_dir data/ich-split-renamed \
+  --output_dir "${CHECKPOINTS_DIR}/soups_results"

scripts/.ipynb_checkpoints/soup_swin-checkpoint.sh

@@ -0,0 +1,14 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-31_11-32-23-swin'
+
+python -m soups.run_test_with_model_soups \
+  --seed 42 \
+  --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' \
+  --model timm/swin_base_patch4_window7_224.ms_in22k_ft_in1k \
+  --uniform_soup \
+  --greedy_soup \
+  --pruned_soup \
+  --pruned_soup_num_iters 64 \
+  --dataset_dir data/ich-split-renamed \
+  --output_dir "${CHECKPOINTS_DIR}/soups_results"

scripts/.ipynb_checkpoints/soup_tiny_vit_21m-checkpoint.sh

@@ -0,0 +1,15 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-01_13-56-39-tiny_vit_21m_dist'
+
+python -m soups.run_test_with_model_soups \
+  --seed 42 \
+  --checkpoint_path "$CHECKPOINTS_DIR" \
+  --model timm/tiny_vit_21m_224.dist_in22k_ft_in1k \
+  --uniform_soup \
+  --greedy_soup \
+  --pruned_soup \
+  --pruned_soup_num_iters 64 \
+  --greedy_soup_comparison_metric f1 \
+  --dataset_dir data/ich-split-renamed \
+  --output_dir "${CHECKPOINTS_DIR}/soups_results"

scripts/.ipynb_checkpoints/soup_vit_base-checkpoint.sh

@@ -0,0 +1,15 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-17_20-27-15-vit_base'
+
+python -m soups.run_test_with_model_soups \
+  --seed 42 \
+  --checkpoint_path "$CHECKPOINTS_DIR" \
+  --model timm/vit_base_patch16_224.augreg2_in21k_ft_in1k \
+  --uniform_soup \
+  --greedy_soup \
+  --pruned_soup \
+  --pruned_soup_num_iters 64 \
+  --greedy_soup_comparison_metric f1 \
+  --dataset_dir data/ich-split-renamed \
+  --output_dir "${CHECKPOINTS_DIR}/soups_results"

scripts/.ipynb_checkpoints/soup_vit_base_laion2b-checkpoint.sh

@@ -0,0 +1,15 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-17_20-07-52-vit_base_laion2b'
+
+python -m soups.run_test_with_model_soups \
+  --seed 42 \
+  --checkpoint_path "$CHECKPOINTS_DIR" \
+  --model timm/vit_base_patch16_clip_224.laion2b_ft_in1k \
+  --uniform_soup \
+  --greedy_soup \
+  --pruned_soup \
+  --pruned_soup_num_iters 64 \
+  --greedy_soup_comparison_metric f1 \
+  --dataset_dir data/ich-split-renamed \
+  --output_dir "${CHECKPOINTS_DIR}/soups_results"

scripts/.ipynb_checkpoints/test_caformer_b36-checkpoint.sh

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-14_09-18-13-caformer_b36-seed_7-one_cycle_lr'
+
+python -m soups.run_test_multiple_checkpoints \
+  --seed 42 \
+  --checkpoint_path "$CHECKPOINTS_DIR" \
+  --model timm/caformer_b36.sail_in22k_ft_in1k \
+  --dataset_dir data/ich-split-renamed \
+  --output_file "${CHECKPOINTS_DIR}/test_results.json"

scripts/.ipynb_checkpoints/test_caformer_m36-checkpoint.sh

@@ -0,0 +1,12 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-19_19-37-23-caformer_m36-one_cycle_lr'
+
+python -m soups.run_test_multiple_checkpoints \
+  --seed 42 \
+  --checkpoint_paths '/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-19_19-37-23-caformer_m36-one_cycle_lr/f1-top8' \
+      '/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-19_19-37-23-caformer_m36-one_cycle_lr/accuracy-top8' \
+      '/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-19_19-37-23-caformer_m36-one_cycle_lr/loss-top8' \
+  --model timm/caformer_m36.sail_in22k_ft_in1k \
+  --dataset_dir data/ich-split-renamed \
+  --output_file "${CHECKPOINTS_DIR}/test_results.json"

scripts/.ipynb_checkpoints/test_caformer_s36-checkpoint.sh

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='//home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-01_16-22-06-caformer_s36'
+
+python -m soups.run_test_multiple_checkpoints \
+  --seed 42 \
+  --checkpoints_dir "$CHECKPOINTS_DIR" \
+  --model timm/caformer_s36.sail_in22k_ft_in1k \
+  --dataset_dir data/ich-split-renamed \
+  --output_file "${CHECKPOINTS_DIR}/test_results.json"

scripts/.ipynb_checkpoints/test_coatnet_0-checkpoint.sh

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-31_11-30-41-coatnet_0_filtered_1000'
+
+python -m soups.run_test_multiple_checkpoints \
+  --seed 42 \
+  --checkpoints_dir "$CHECKPOINTS_DIR" \
+  --model timm/coatnet_0_rw_224.sw_in1k \
+  --dataset_dir data/ich-split-renamed \
+  --output_file "${CHECKPOINTS_DIR}/test_results.json"

scripts/.ipynb_checkpoints/test_coatnet_0-merged_4_15-checkpoint.sh

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-26_20-43-44-coatnet_0_merged_4_15'
+
+python -m soups.run_test_multiple_checkpoints \
+  --seed 42 \
+  --checkpoints_dir "$CHECKPOINTS_DIR" \
+  --model timm/coatnet_0_rw_224.sw_in1k \
+  --dataset_dir data/ich-split-renamed-merged-4-15 \
+  --output_file "${CHECKPOINTS_DIR}/test_results.json"

scripts/.ipynb_checkpoints/test_coatnet_0_co_teaching-checkpoint.sh

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+
+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'
+
+python -m soups.run_test_multiple_checkpoints \
+  --seed 42 \
+  --checkpoints_dir "$CHECKPOINTS_DIR" \
+  --model timm/coatnet_0_rw_224.sw_in1k \
+  --dataset_dir data/ich-split-renamed \
+  --output_file "${CHECKPOINTS_DIR}/test_results.json"

scripts/.ipynb_checkpoints/test_coatnet_0_random-checkpoint.sh

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-23_10-38-52-coatnet_0_random'
+
+python -m soups.run_test_multiple_checkpoints \
+  --seed 42 \
+  --checkpoints_dir "$CHECKPOINTS_DIR" \
+  --model timm/coatnet_0_rw_224.sw_in1k \
+  --dataset_dir data/ich-split-renamed \
+  --output_file "${CHECKPOINTS_DIR}/test_results.json"

scripts/.ipynb_checkpoints/test_coatnet_2-merged_4_15-checkpoint.sh

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-26_20-57-11-coatnet_2_merged_4_15'
+
+python -m soups.run_test_multiple_checkpoints \
+  --seed 42 \
+  --checkpoints_dir "$CHECKPOINTS_DIR" \
+  --model timm/coatnet_2_rw_224.sw_in12k_ft_in1k \
+  --dataset_dir data/ich-split-renamed-merged-4-15 \
+  --output_file "${CHECKPOINTS_DIR}/test_results.json"

scripts/.ipynb_checkpoints/test_coatnet_2_random-checkpoint.sh

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-08-23_10-41-55-coatnet_2_random'
+
+python -m soups.run_test_multiple_checkpoints \
+  --seed 42 \
+  --checkpoints_dir "$CHECKPOINTS_DIR" \
+  --model timm/coatnet_2_rw_224.sw_in12k_ft_in1k \
+  --dataset_dir data/ich-split-renamed \
+  --output_file "${CHECKPOINTS_DIR}/test_results.json"

scripts/.ipynb_checkpoints/test_convnext_base-checkpoint.sh

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-10_16-27-06-convnext_base'
+
+python -m soups.run_test_multiple_checkpoints \
+  --seed 42 \
+  --checkpoints_dir "$CHECKPOINTS_DIR" \
+  --model timm/convnext_base.fb_in22k_ft_in1k \
+  --dataset_dir data/ich-split-renamed \
+  --output_file "${CHECKPOINTS_DIR}/test_results.json"

scripts/.ipynb_checkpoints/test_convnext_femto-checkpoint.sh

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-10_15-31-58-convnext_femto'
+
+python -m soups.run_test_multiple_checkpoints \
+  --seed 42 \
+  --checkpoints_dir "$CHECKPOINTS_DIR" \
+  --model timm/convnext_femto.d1_in1k \
+  --dataset_dir data/ich-split-renamed \
+  --output_file "${CHECKPOINTS_DIR}/test_results.json"

scripts/.ipynb_checkpoints/test_convnext_large-checkpoint.sh

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-11_16-47-37-convnext_large6'
+
+python -m soups.run_test_multiple_checkpoints \
+  --seed 42 \
+  --checkpoints_dir "$CHECKPOINTS_DIR" \
+  --model timm/convnext_large.fb_in22k_ft_in1k \
+  --dataset_dir data/ich-split-renamed \
+  --output_file "${CHECKPOINTS_DIR}/test_results.json"

scripts/.ipynb_checkpoints/test_convnext_nano-checkpoint.sh

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-10_14-49-36-convnext_nano'
+
+python -m soups.run_test_multiple_checkpoints \
+  --seed 42 \
+  --checkpoints_dir "$CHECKPOINTS_DIR" \
+  --model timm/convnext_nano.in12k_ft_in1k \
+  --dataset_dir data/ich-split-renamed \
+  --output_file "${CHECKPOINTS_DIR}/test_results.json"

scripts/.ipynb_checkpoints/test_convnext_pico-checkpoint.sh

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-10_15-24-36-convnext_pico'
+
+python -m soups.run_test_multiple_checkpoints \
+  --seed 42 \
+  --checkpoints_dir "$CHECKPOINTS_DIR" \
+  --model timm/convnext_pico.d1_in1k \
+  --dataset_dir data/ich-split-renamed \
+  --output_file "${CHECKPOINTS_DIR}/test_results.json"

scripts/.ipynb_checkpoints/test_convnext_small-checkpoint.sh

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-10_14-54-53-convnext_small'
+
+python -m soups.run_test_multiple_checkpoints \
+  --seed 42 \
+  --checkpoints_dir "$CHECKPOINTS_DIR" \
+  --model timm/convnext_small.in12k_ft_in1k \
+  --dataset_dir data/ich-split-renamed \
+  --output_file "${CHECKPOINTS_DIR}/test_results.json"

scripts/.ipynb_checkpoints/test_convnext_tiny-checkpoint.sh

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-10_15-23-36-convnext_tiny'
+
+python -m soups.run_test_multiple_checkpoints \
+  --seed 42 \
+  --checkpoints_dir "$CHECKPOINTS_DIR" \
+  --model timm/convnext_tiny.in12k_ft_in1k \
+  --dataset_dir data/ich-split-renamed \
+  --output_file "${CHECKPOINTS_DIR}/test_results.json"

scripts/.ipynb_checkpoints/test_convnext_v2_large-checkpoint.sh

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync2/2025-09-11_17-25-36-convnext_v2_large'
+
+python -m soups.run_test_multiple_checkpoints \
+  --seed 42 \
+  --checkpoints_dir "$CHECKPOINTS_DIR" \
+  --model timm/convnextv2_large.fcmae_ft_in22k_in1k \
+  --dataset_dir data/ich-split-renamed \
+  --output_file "${CHECKPOINTS_DIR}/test_results.json"

scripts/.ipynb_checkpoints/test_convnext_xlarge-checkpoint.sh

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync2/2025-09-11_11-58-22-convnext_xlarge'
+
+python -m soups.run_test_multiple_checkpoints \
+  --seed 42 \
+  --checkpoints_dir "$CHECKPOINTS_DIR" \
+  --model timm/convnext_xlarge.fb_in22k_ft_in1k \
+  --dataset_dir data/ich-split-renamed \
+  --output_file "${CHECKPOINTS_DIR}/test_results.json"

scripts/.ipynb_checkpoints/test_convnext_xxlarge-checkpoint.sh

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync2/2025-09-11_13-53-47-convnext_xxlarge'
+
+python -m soups.run_test_multiple_checkpoints \
+  --seed 42 \
+  --checkpoints_dir "$CHECKPOINTS_DIR" \
+  --model timm/convnext_xxlarge.clip_laion2b_soup_ft_in1k \
+  --dataset_dir data/ich-split-renamed \
+  --output_file "${CHECKPOINTS_DIR}/test_results.json"

scripts/.ipynb_checkpoints/test_eva02_base-checkpoint.sh

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-02_00-11-46-eva02_base'
+
+python -m soups.run_test_multiple_checkpoints \
+  --seed 42 \
+  --checkpoints_dir "$CHECKPOINTS_DIR" \
+  --model timm/eva02_base_patch14_224.mim_in22k \
+  --dataset_dir data/ich-split-renamed \
+  --output_file "${CHECKPOINTS_DIR}/test_results.json"

scripts/.ipynb_checkpoints/test_eva02_large-checkpoint.sh

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-14_00-21-30-eva02_large-one_cycle_lr'
+
+python -m soups.run_test_multiple_checkpoints \
+  --seed 42 \
+  --checkpoints_dir "$CHECKPOINTS_DIR" \
+  --model timm/eva02_large_patch14_224.mim_in22k \
+  --dataset_dir data/ich-split-renamed \
+  --output_file "${CHECKPOINTS_DIR}/test_results.json"

scripts/.ipynb_checkpoints/test_eva02_small-checkpoint.sh

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/home/khmt/Documents/TQKhangT/mthien/soups/checkpoints-sync/2025-09-20_14-45-41-eva02_small-reproduced'
+
+python -m soups.run_test_multiple_checkpoints \
+  --seed 42 \
+  --checkpoint_paths "$CHECKPOINTS_DIR" \
+  --model timm/eva02_small_patch14_224.mim_in22k \
+  --dataset_dir data/ich-split-renamed \
+  --output_file "${CHECKPOINTS_DIR}/test_results.json"

scripts/.ipynb_checkpoints/test_focalnet_base_srf-checkpoint.sh

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/media/khmt/HDD1/TQKhang/mthien/soups/checkpoints/2025-09-10_14-12-59-focalnet_base_srf'
+
+python -m soups.run_test_multiple_checkpoints \
+  --seed 42 \
+  --checkpoint_path "$CHECKPOINTS_DIR" \
+  --model timm/focalnet_base_srf.ms_in1k \
+  --dataset_dir data/ich-split-renamed \
+  --output_file "${CHECKPOINTS_DIR}/test_results.json"

scripts/.ipynb_checkpoints/test_focalnet_small_lrf-checkpoint.sh

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+
+CHECKPOINTS_DIR='/media/khmt/HDD1/TQKhang/mthien/soups/checkpoints/2025-09-10_14-43-07-focalnet_small_lrf'
+
+python -m soups.run_test_multiple_checkpoints \
+  --seed 42 \
+  --checkpoint_path "$CHECKPOINTS_DIR" \
+  --model timm/focalnet_small_lrf.ms_in1k \
+  --dataset_dir data/ich-split-renamed \
+  --output_file "${CHECKPOINTS_DIR}/test_results.json"