|
#!/bin/sh |
|
|
|
export CUDA_VISIBLE_DEVICES=1,2 |
|
python src/run_speech_recognition_ctc_bnb.py \ |
|
--dataset_name="mozilla-foundation/common_voice_7_0" \ |
|
--model_name_or_path="facebook/wav2vec2-xls-r-300m" \ |
|
--dataset_config_name="et" \ |
|
--output_dir="./" \ |
|
--overwrite_output_dir \ |
|
--max_steps="10" \ |
|
--per_device_train_batch_size="4" \ |
|
--learning_rate="3e-4" \ |
|
--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 \ |
|
--group_by_length \ |
|
--push_to_hub \ |
|
--use_auth_token \ |
|
--do_train --do_eval |
|
|
