| python run_speech_recognition_seq2seq_streaming.py \ | |
| --model_name_or_path="openai/whisper-large-v2" \ | |
| --dataset_name="vumichien/preprocessed_jsut_jsss_css10_common_voice_11" \ | |
| --dataset_config_name="ja" \ | |
| --language="japanese" \ | |
| --train_split_name="train" \ | |
| --eval_split_name="test" \ | |
| --model_index_name="Whisper Mix Japanese Large V2" \ | |
| --max_steps="10000" \ | |
| --output_dir="./" \ | |
| --per_device_train_batch_size="8" \ | |
| --per_device_eval_batch_size="8" \ | |
| --gradient_accumulation_steps=2 \ | |
| --logging_steps="500" \ | |
| --learning_rate="1e-5" \ | |
| --warmup_steps="500" \ | |
| --evaluation_strategy="steps" \ | |
| --eval_steps="1000" \ | |
| --save_strategy="steps" \ | |
| --save_steps="1000" \ | |
| --generation_max_length="225" \ | |
| --length_column_name="input_length" \ | |
| --max_duration_in_seconds="30" \ | |
| --text_column_name="sentence" \ | |
| --freeze_feature_encoder="False" \ | |
| --report_to="tensorboard" \ | |
| --metric_for_best_model="wer" \ | |
| --greater_is_better="False" \ | |
| --load_best_model_at_end \ | |
| --gradient_checkpointing \ | |
| --fp16 \ | |
| --overwrite_output_dir \ | |
| --do_train \ | |
| --do_eval \ | |
| --predict_with_generate \ | |
| --do_normalize_eval \ | |
| --use_auth_token \ | |
| --push_to_hub | |