examples/speech-recognition/run.sh

# CUDA_VISIBLE_DEVICES="0" python run_speech_recognition_ctc.py \
#         --dataset_name="timit_asr" \
#         --model_name_or_path="facebook/wav2vec2-base" \
#         --overwrite_output_dir \
#         --output_dir="./wav2vec2-base-timit-fine-tuned" \
#         --train_split_name="train" \
#         --num_train_epochs="20" \
#         --per_device_train_batch_size="32" \
#         --per_device_eval_batch_size="1" \
#         --weight_decay="0.005" \
#         --learning_rate="1e-4" \
#         --warmup_steps="1000" \
#         --evaluation_strategy="steps" \
#         --text_column_name="text" \
#         --save_steps="400" \
#         --eval_steps="100" \
#         --logging_steps="10" \
#         --layerdrop="0.0" \
#         --save_total_limit="3" \
#         --freeze_feature_extractor \
#         --chars_to_ignore , ? . ! - \; \: \" “ % ‘ ” � \
#         --fp16 \
#         --group_by_length \
#         --push_to_hub \
#         --do_train \
#         --do_eval 


# 
CUDA_VISIBLE_DEVICES=0 python run_speech_recognition_ctc.py \
	--dataset_name="mozilla-foundation/common_voice_13_0" \
	--model_name_or_path="facebook/wav2vec2-large-xlsr-53" \
	--dataset_config_name="tr" \
	--output_dir="./wav2vec2-common_voice-tr-demo" \
	--preprocessing_num_workers="1" \
	--overwrite_output_dir \
	--num_train_epochs="15" \
	--per_device_train_batch_size="16" \
	--gradient_accumulation_steps="2" \
	--learning_rate="3e-4" \
	--warmup_steps="500" \
	--evaluation_strategy="steps" \
	--text_column_name="sentence" \
	--save_steps="400" \
	--eval_steps="100" \
	--logging_steps="1" \
	--layerdrop="0.0" \
	--save_total_limit="3" \
	--freeze_feature_encoder \
	--gradient_checkpointing \
	--chars_to_ignore , ? . ! - \; \: \" “ % ‘ ” � \
	--fp16 \
	--group_by_length \
	--push_to_hub \
	--do_train \
    --do_eval \
    --trust-remote-code \
    --apply-trp --trp-depths 1 --trp-p 0.1 --trp-lambdas 0.4 0.2 0.1


# CUDA_VISIBLE_DEVICES="0" python run_speech_recognition_ctc.py \
#     --dataset_name="multilingual_librispeech" \
#     --model_name_or_path="facebook/wav2vec2-large-xlsr-53" \
#     --train_split_name="train.9h" \
#     --dataset_config_name="german" \
#     --output_dir="./wav2vec2-xlsr-53-300m-mls-german-ft" \
#     --overwrite_output_dir \
#     --num_train_epochs="100" \
#     --per_device_train_batch_size="32" \
#     --gradient_accumulation_steps="1" \
#     --learning_rate="1e-4" \
#     --activation_dropout="0.1" \
#     --warmup_steps="1000" \
#     --evaluation_strategy="steps" \
#     --text_column_name="text" \
#     --save_steps="1000" \
#     --eval_steps="500" \
#     --logging_steps="10" \
#     --layerdrop="0.1" \
#     --hidden_dropout="0.0" \
#     --save_total_limit="1" \
#     --mask_time_prob="0.75" \
#     --mask_feature_prob="0.25" \
#     --mask_feature_length="64" \
#     --freeze_feature_extractor \
#     --fp16 \
#     --group_by_length \
#     --do_eval \
#     --gradient_checkpointing \
#     --use_auth_token \
#     --do_train --do_eval \
#     --push_to_hub \