| python run_speech_recognition_ctc.py \ | |
| --dataset_name="mozilla-foundation/common_voice_7_0" \ | |
| --model_name_or_path="facebook/wav2vec2-large-xlsr-53" \ | |
| --dataset_config_name="ur" \ | |
| --output_dir="./" \ | |
| --overwrite_output_dir \ | |
| --max_steps="3000" \ | |
| --gradient_accumulation_steps="2" \ | |
| --lr_scheduler_type="linear" \ | |
| --per_device_train_batch_size="64" \ | |
| --learning_rate="4e-4" \ | |
| --save_total_limit="5" \ | |
| --evaluation_strategy="steps" \ | |
| --text_column_name="sentence" \ | |
| --length_column_name="input_length" \ | |
| --chars_to_ignore ,\?\.\!\-\;\:\"\โ\%\โ\โ\๏ฟฝ\ู\ู\ูู,ู\ | |
| --save_steps="200" \ | |
| --num_train_epochs="1000.0" \ | |
| --freeze_feature_encoder \ | |
| --gradient_checkpointing \ | |
| --fp16 \ | |
| --group_by_length \ | |
| --push_to_hub \ | |
| --use_auth_token \ | |
| --do_train --do_eval |