whisper-tiny-sv / test_run_nordic.sh
marinone94's picture
debug decoding
a057c82
raw
history blame
1.67 kB
python $1run_speech_recognition_seq2seq_streaming.py \
--model_name_or_path="openai/whisper-tiny" \
--dataset_train_name="mozilla-foundation/common_voice_11_0,mozilla-foundation/common_voice_11_0,mozilla-foundation/common_voice_11_0,babelbox/babelbox_voice,NbAiLab/NST,NbAiLab/NPSC,google/fleurs,google/fleurs,google/fleurs" \
--dataset_train_config_name="sv-SE,da,nn-NO,nst,no-distant,16K_mp3_nynorsk,sv_se,da_dk,nb_no" \
--language_train="sv,da,no,sv,no,no,sv,da,no" \
--train_split_name="train+validation,train+validation,train+validation,train,train+test,train+validation,train+validation,train+validation,train+validation" \
--dataset_eval_name="mozilla-foundation/common_voice_11_0" \
--dataset_eval_config_name="sv-SE" \
--language_eval="sv" \
--eval_split_name="test" \
--model_index_name="Whisper Tiny Nordic" \
--max_train_samples="64" \
--max_eval_samples="32" \
--max_steps="1" \
--output_dir="./" \
--per_device_train_batch_size="1" \
--per_device_eval_batch_size="1" \
--logging_steps="25" \
--learning_rate="1e-5" \
--warmup_steps="500" \
--evaluation_strategy="steps" \
--eval_steps="1000" \
--save_strategy="steps" \
--save_steps="1000" \
--generation_max_length="225" \
--length_column_name="input_length" \
--max_duration_in_seconds="30" \
--text_column_name="sentence,text,raw_transcription" \
--freeze_feature_encoder="False" \
--report_to="wandb" \
--metric_for_best_model="wer" \
--greater_is_better="False" \
--load_best_model_at_end \
--gradient_checkpointing \
--overwrite_output_dir \
--do_train \
--do_eval \
--fp16 \
--predict_with_generate \
--do_normalize_eval \
--streaming \
--use_auth_token