|
python3 run_speech_recognition_seq2seq_streaming.py \ |
|
--model_name_or_path="openai/whisper-large-v2" \ |
|
--dataset_name="thomas0104/nan_tw_soap_opera" \ |
|
--language="chinese" \ |
|
--dataset_config_name="nan-tw" \ |
|
--train_split_name="train+validation" \ |
|
--eval_split_name="test" \ |
|
--max_steps="1000" \ |
|
--output_dir="./" \ |
|
--per_device_train_batch_size="2" \ |
|
--per_device_eval_batch_size="2" \ |
|
--logging_steps="1000" \ |
|
--learning_rate="1e-5" \ |
|
--warmup_steps="100" \ |
|
--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="transcription" \ |
|
--freeze_feature_encoder="False" \ |
|
--report_to="tensorboard" \ |
|
--gradient_checkpointing \ |
|
--fp16 \ |
|
--overwrite_output_dir \ |
|
--do_train \ |
|
--do_eval \ |
|
--predict_with_generate \ |
|
--do_normalize_eval \ |
|
--use_auth_token \ |
|
--optim="adamw_bnb_8bit" |
|
|
|
|
