|
#!/usr/bin/env bash |
|
python -m torch.distributed.launch \ |
|
--nproc_per_node 2 run_speech_recognition_ctc.py \ |
|
--dataset_name="common_voice" \ |
|
--model_name_or_path="facebook/wav2vec2-xls-r-300m" \ |
|
--tokenizer_name_or_path="./created_tokenizer_from_cv_sv" \ |
|
--dataset_config_name="sv-SE" \ |
|
--phoneme_language="sv" \ |
|
--output_dir="./wav2vec2-xls-r-300m-phoneme-sv" \ |
|
--overwrite_output_dir \ |
|
--num_train_epochs="100" \ |
|
--per_device_train_batch_size="16" \ |
|
--gradient_accumulation_steps="1" \ |
|
--learning_rate="5e-4" \ |
|
--warmup_steps="500" \ |
|
--evaluation_strategy="steps" \ |
|
--text_column_name="sentence" \ |
|
--pad_token="<pad>" \ |
|
--unk_token="<unk>" \ |
|
--save_steps="100" \ |
|
--eval_steps="100" \ |
|
--logging_steps="1" \ |
|
--layerdrop="0.0" \ |
|
--save_total_limit="3" \ |
|
--freeze_feature_extractor \ |
|
--chars_to_ignore , ? . ! - \; \: \" “ % ‘ ” � \ |
|
--mask_time_prob="0.3" \ |
|
--mask_time_length="10" \ |
|
--mask_feature_prob="0.1" \ |
|
--mask_feature_length="64" \ |
|
--fp16 \ |
|
--group_by_length \ |
|
--do_eval \ |
|
--gradient_checkpointing \ |
|
--use_auth_token \ |
|
--do_train --do_eval \ |
|
--push_to_hub \ |
|
|
