File size: 1,200 Bytes
5da3530 a19ef19 81c740d 5da3530 2584f56 5da3530 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 |
#!/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-xlsr-53-espeak-cv-ft" \
--tokenizer_name_or_path="facebook/wav2vec2-xlsr-53-espeak-cv-ft" \
--dataset_config_name="tr" \
--phoneme_language="tr" \
--output_dir="./wav2vec2-xls-r-phoneme-300m-tr" \
--overwrite_output_dir \
--num_train_epochs="3" \
--per_device_train_batch_size="1" \
--gradient_accumulation_steps="8" \
--learning_rate="5e-5" \
--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 , ? . ! - \; \: \" “ % ‘ ” � \
--fp16 \
--group_by_length \
--do_eval \
--gradient_checkpointing \
--use_auth_token \
--do_train --do_eval \
--push_to_hub \ |