| export HF_TOKEN=`cat /home/huggingface.token` | |
| export HF_HOME="/home/Work/common_huggingface" | |
| python run_speech_recognition_ctc.py \ | |
| --token="${HF_TOKEN}" \ | |
| --dataset_name="timit_asr" \ | |
| --dataset_path="/home/Work_/common_darpa/Timit_data/data" \ | |
| --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_encoder \ | |
| --chars_to_ignore , ? . ! - \; \: \" “ % ‘ ” � \ | |
| --fp16 \ | |
| --group_by_length \ | |
| --push_to_hub \ | |
| --do_train --do_eval \ | |