pablouribe's picture
Training in progress, step 400
e93d482
raw
history blame
739 Bytes
python run_speech_recognition_ctc.py \
--dataset_name="common_voice" \
--model_name_or_path="facebook/wav2vec2-large-xlsr-53" \
--dataset_config_name="es" \
--output_dir="./" \
--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" \
--length_column_name="input_length" \
--save_steps="400" \
--eval_steps="100" \
--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