python run_speech_recognition_ctc.py \ --dataset_name="common_voice" \ --model_name_or_path="facebook/wav2vec2-large-xlsr-53" \ --dataset_config_name="es" \ --output_dir="./" \ --overwrite_output_dir \ --num_train_epochs="15" \ --per_device_train_batch_size="16" \ --gradient_accumulation_steps="2" \ --learning_rate="3e-4" \ --warmup_steps="500" \ --evaluation_strategy="steps" \ --text_column_name="sentence" \ --length_column_name="input_length" \ --save_steps="400" \ --eval_steps="100" \ --layerdrop="0.0" \ --save_total_limit="3" \ --freeze_feature_encoder \ --gradient_checkpointing \ --chars_to_ignore , ? . ! - \; \: \" “ % ‘ ” � \ --fp16 \ --group_by_length \ --push_to_hub \ --do_train --do_eval