To reproduce this run, first call get_ctc_tokenizer.py
to train the CTC tokenizer and then execute the following command to train the CTC system:
#!/usr/bin/env bash
python run_flax_speech_recognition_ctc.py \
--model_name_or_path="esb/wav2vec2-ctc-pretrained" \
--tokenizer_name="wav2vec2-ctc-gigaspeech-tokenizer" \
--dataset_name="esb/datasets" \
--dataset_config_name="gigaspeech" \
--output_dir="./" \
--wandb_project="wav2vec2-ctc" \
--wandb_name="wav2vec2-ctc-gigaspeech" \
--max_steps="50000" \
--save_steps="10000" \
--eval_steps="10000" \
--learning_rate="3e-4" \
--logging_steps="25" \
--warmup_steps="5000" \
--preprocessing_num_workers="1" \
--do_train \
--do_eval \
--do_predict \
--overwrite_output_dir \
--gradient_checkpointing \
--freeze_feature_encoder \
--push_to_hub \
--use_auth_token
- Downloads last month
- 14
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.