|
#!/usr/bin/env bash |
|
./run_wav2vec2_pretrain_flax.py \ |
|
--output_dir="./wav2vec2-spanish" \ |
|
--num_train_epochs="5" \ |
|
--per_device_train_batch_size="16" \ |
|
--per_device_eval_batch_size="16" \ |
|
--learning_rate="5e-4" \ |
|
--weight_decay="0.01" \ |
|
--warmup_steps="1000" \ |
|
--model_name_or_path="./" \ |
|
--dataset_name="common_voice" \ |
|
--dataset_config_name="es" \ |
|
--preprocessing_num_workers="32" \ |
|
--max_duration_in_seconds="10.0" \ |
|
--adam_beta1="0.9" \ |
|
--adam_beta2="0.98" \ |
|
--pad_to_multiple_of="16384" \ |
|
--validation_split_percentage="5" \ |
|
--speech_file_column="path" \ |
|
--cache_dir="./data_cache" \ |
|
--push_to_hub |
|
|
