run_eval_mm.sh

# Multimodal
CUDA_VISIBLE_DEVICES=1 python eval.py --output_dir=./save_mm/0/eval      --model_name_or_path=save_mm/0/checkpoint-4480     --train_manifest_path=dataset/mm_train_metadata.csv     --valid_manifest_path=dataset/mm_valid_metadata.csv     --test_manifest_path=dataset/mm_test_metadata.csv     --num_workers=8 --preprocessing_num_workers=8 --use_video     --audio_column_name=audio_path --text_column_name=text_path  --video_column_name=lip_image_path     --per_device_train_batch_size=16 --per_device_eval_batch_size=16     --dataloader_num_workers=32 --dataloader_pin_memory     --seed=0 --num_train_epochs=20 --learning_rate=5e-5     --logging_strategy=steps --logging_steps=10 --report_to=tensorboard     --evaluation_strategy=epoch --eval_steps=1 --eval_accumulation_steps=100     --save_steps=1 --save_strategy=epoch --save_total_limit=1     --metric_for_best_model=mer --greater_is_better=False --load_best_model_at_end=True


CUDA_VISIBLE_DEVICES=1 python eval.py --output_dir=./save_mm/1/eval      --model_name_or_path=save_mm/1/checkpoint-4480     --train_manifest_path=dataset/mm_train_metadata.csv     --valid_manifest_path=dataset/mm_valid_metadata.csv     --test_manifest_path=dataset/mm_test_metadata.csv     --num_workers=8 --preprocessing_num_workers=8 --use_video     --audio_column_name=audio_path --text_column_name=text_path  --video_column_name=lip_image_path     --per_device_train_batch_size=16 --per_device_eval_batch_size=16     --dataloader_num_workers=32 --dataloader_pin_memory     --seed=0 --num_train_epochs=20 --learning_rate=5e-5     --logging_strategy=steps --logging_steps=10 --report_to=tensorboard     --evaluation_strategy=epoch --eval_steps=1 --eval_accumulation_steps=100     --save_steps=1 --save_strategy=epoch --save_total_limit=1     --metric_for_best_model=mer --greater_is_better=False --load_best_model_at_end=True


CUDA_VISIBLE_DEVICES=1 python eval.py --output_dir=./save_mm/2/eval      --model_name_or_path=save_mm/2/checkpoint-2912     --train_manifest_path=dataset/mm_train_metadata.csv     --valid_manifest_path=dataset/mm_valid_metadata.csv     --test_manifest_path=dataset/mm_test_metadata.csv     --num_workers=8 --preprocessing_num_workers=8 --use_video     --audio_column_name=audio_path --text_column_name=text_path  --video_column_name=lip_image_path     --per_device_train_batch_size=16 --per_device_eval_batch_size=16     --dataloader_num_workers=32 --dataloader_pin_memory     --seed=0 --num_train_epochs=20 --learning_rate=5e-5     --logging_strategy=steps --logging_steps=10 --report_to=tensorboard     --evaluation_strategy=epoch --eval_steps=1 --eval_accumulation_steps=100     --save_steps=1 --save_strategy=epoch --save_total_limit=1     --metric_for_best_model=mer --greater_is_better=False --load_best_model_at_end=True


CUDA_VISIBLE_DEVICES=1 python eval.py --output_dir=./save_mm/3/eval      --model_name_or_path=save_mm/3/checkpoint-4256     --train_manifest_path=dataset/mm_train_metadata.csv     --valid_manifest_path=dataset/mm_valid_metadata.csv     --test_manifest_path=dataset/mm_test_metadata.csv     --num_workers=8 --preprocessing_num_workers=8 --use_video     --audio_column_name=audio_path --text_column_name=text_path  --video_column_name=lip_image_path     --per_device_train_batch_size=16 --per_device_eval_batch_size=16     --dataloader_num_workers=32 --dataloader_pin_memory     --seed=0 --num_train_epochs=20 --learning_rate=5e-5     --logging_strategy=steps --logging_steps=10 --report_to=tensorboard     --evaluation_strategy=epoch --eval_steps=1 --eval_accumulation_steps=100     --save_steps=1 --save_strategy=epoch --save_total_limit=1     --metric_for_best_model=mer --greater_is_better=False --load_best_model_at_end=True


CUDA_VISIBLE_DEVICES=1 python eval.py --output_dir=./save_mm/4/eval      --model_name_or_path=save_mm/4/checkpoint-3584     --train_manifest_path=dataset/mm_train_metadata.csv     --valid_manifest_path=dataset/mm_valid_metadata.csv     --test_manifest_path=dataset/mm_test_metadata.csv     --num_workers=8 --preprocessing_num_workers=8 --use_video     --audio_column_name=audio_path --text_column_name=text_path  --video_column_name=lip_image_path     --per_device_train_batch_size=16 --per_device_eval_batch_size=16     --dataloader_num_workers=32 --dataloader_pin_memory     --seed=0 --num_train_epochs=20 --learning_rate=5e-5     --logging_strategy=steps --logging_steps=10 --report_to=tensorboard     --evaluation_strategy=epoch --eval_steps=1 --eval_accumulation_steps=100     --save_steps=1 --save_strategy=epoch --save_total_limit=1     --metric_for_best_model=mer --greater_is_better=False --load_best_model_at_end=True