| python run_flax_speech_recognition_ctc.py \ | |
| --model_name_or_path="esb/wav2vec2-ctc-pretrained" \ | |
| --tokenizer_name="wav2vec2-ctc-voxpopuli-tokenizer" \ | |
| --dataset_name="esb/datasets" \ | |
| --dataset_config_name="voxpopuli" \ | |
| --output_dir="./" \ | |
| --wandb_project="wav2vec2-ctc" \ | |
| --wandb_name="wav2vec2-ctc-voxpopuli" \ | |
| --max_steps="50000" \ | |
| --save_steps="10000" \ | |
| --eval_steps="10000" \ | |
| --learning_rate="3e-4" \ | |
| --logging_steps="25" \ | |
| --warmup_steps="5000" \ | |
| --preprocessing_num_workers="1" \ | |
| --per_device_eval_batch_size="1" \ | |
| --do_train \ | |
| --do_eval \ | |
| --do_predict \ | |
| --overwrite_output_dir \ | |
| --gradient_checkpointing \ | |
| --freeze_feature_encoder \ | |
| --push_to_hub \ | |
| --use_auth_token |