run.sh

#!/usr/bin/env bash
#python -m torch.distributed.launch \
#	--nproc_per_node 8 run_speech_recognition_ctc.py \
CUDA_VISIBLE_DEVICES="0" python run_speech_recognition_ctc.py \
	--dataset_name="common_voice" \
	--model_name_or_path="./" \
	--tokenizer_name_or_path="./" \
	--dataset_name="mozilla-foundation/common_voice_3_0" \
	--path_to_eval_file="./valSeqs_1.0_uniform_new_version.text" \
	--path_to_train_file="./trainSeqs_1.0_uniform_new_version.text" \
	--dataset_config_name="it" \
	--text_column_name="sentence" \
	--output_dir="./" \
	--overwrite_output_dir \
	--num_train_epochs="150" \
	--per_device_train_batch_size="4" \
	--per_device_eval_batch_size="4" \
	--gradient_accumulation_steps="1" \
	--learning_rate="7.5e-5" \
	--warmup_steps="1000" \
	--evaluation_strategy="steps" \
	--save_steps="500" \
	--eval_steps="2" \
	--logging_steps="100" \
	--layerdrop="0.0" \
	--activation_dropout="0.1" \
	--save_total_limit="3" \
	--freeze_feature_extractor \
	--feat_proj_dropout="0.0" \
	--mask_time_prob="0.75" \
	--mask_time_length="10" \
	--mask_feature_prob="0.25" \
	--mask_feature_length="64" \
	--gradient_checkpointing \
	--fp16 \
	--group_by_length \
	--do_train --do_eval