| python run_speech_recognition_seq2seq_streaming.py \ |
| --model_name_or_path="openai/whisper-base" \ |
| --dataset_name="google/fleurs" \ |
| --dataset_config_name="ps_af" \ |
| --language="pashto" \ |
| --train_split_name="train" \ |
| --eval_split_name="test" \ |
| --model_index_name="Whisper Tiny Pashto" \ |
| --max_steps="1300" \ |
| --output_dir="./" \ |
| --per_device_train_batch_size="32" \ |
| --per_device_eval_batch_size="16" \ |
| --gradient_accumulation_steps="2" \ |
| --logging_steps="10" \ |
| --learning_rate="1e-7" \ |
| --warmup_steps="100" \ |
| --evaluation_strategy="steps" \ |
| --eval_steps="100" \ |
| --save_strategy="steps" \ |
| --save_steps="100" \ |
| --generation_max_length="225" \ |
| --length_column_name="input_length" \ |
| --max_duration_in_seconds="30" \ |
| --text_column_name="transcription" \ |
| --freeze_feature_encoder="False" \ |
| --load_best_model_at_end \ |
| --report_to="tensorboard" \ |
| --gradient_checkpointing \ |
| --fp16 \ |
| --overwrite_output_dir="False" \ |
| --do_train \ |
| --do_eval \ |
| --predict_with_generate \ |
| --do_normalize_eval \ |
| --use_auth_token \ |
| --streaming="False" \ |
| --push_to_hub |
|
|
