Views
No views yet
get_ctc_tokenizer.py to train the CTC tokenizer and then execute the following command to train the CTC system:1#!/usr/bin/env bash
2python run_flax_speech_recognition_ctc.py \
3 --model_name_or_path="esb/wav2vec2-ctc-pretrained" \
4 --tokenizer_name="wav2vec2-ctc-gigaspeech-tokenizer" \
5 --dataset_name="esb/datasets" \
6 --dataset_config_name="gigaspeech" \
7 --output_dir="./" \
8 --wandb_project="wav2vec2-ctc" \
9 --wandb_name="wav2vec2-ctc-gigaspeech" \
10 --max_steps="50000" \
11 --save_steps="10000" \
12 --eval_steps="10000" \
13 --learning_rate="3e-4" \
14 --logging_steps="25" \
15 --warmup_steps="5000" \
16 --preprocessing_num_workers="1" \
17 --do_train \
18 --do_eval \
19 --do_predict \
20 --overwrite_output_dir \
21 --gradient_checkpointing \
22 --freeze_feature_encoder \
23 --push_to_hub \
24 --use_auth_token