| python run_speech_recognition_ctc.py \ | |
| --dataset_name="common_voice" \ | |
| --model_name_or_path="cahya/wav2vec2-base-turkish-artificial-cv" \ | |
| --dataset_config_name="tr" \ | |
| --output_dir="./output" \ | |
| --overwrite_output_dir \ | |
| --max_steps="100" \ | |
| --per_device_train_batch_size="2" \ | |
| --learning_rate="3e-6" \ | |
| --save_total_limit="1" \ | |
| --evaluation_strategy="steps" \ | |
| --text_column_name="sentence" \ | |
| --length_column_name="input_length" \ | |
| --save_steps="5" \ | |
| --layerdrop="0.0" \ | |
| --freeze_feature_encoder \ | |
| --gradient_checkpointing \ | |
| --fp16=false \ | |
| --group_by_length \ | |
| --push_to_hub \ | |
| --use_auth_token \ | |
| --do_train --do_eval | |