| ./run_wav2vec2_pretrain_flax.py \ | |
| --output_dir="./output50" \ | |
| --speech_file_column="path" \ | |
| --num_train_epochs="50" \ | |
| --per_device_train_batch_size="32" \ | |
| --per_device_eval_batch_size="32" \ | |
| --validation_split_percentage="5" \ | |
| --learning_rate="5e-4" \ | |
| --weight_decay="0.01" \ | |
| --warmup_steps="1000" \ | |
| --model_name_or_path="./" \ | |
| --dataset_name="common_voice" \ | |
| --dataset_config_name="dv" \ | |
| --preprocessing_num_workers="96" \ | |
| --max_duration_in_seconds="10.0" \ | |
| --adam_beta1="0.9" \ | |
| --adam_beta2="0.98" \ | |
| --dtype="bfloat16" \ | |
| --pad_to_multiple_of="16384" |