python run_speech_recognition_seq2seq_streaming.py \ --model_name_or_path="marinone94/whisper-medium-nordic" \ --dataset_train_name="mozilla-foundation/common_voice_11_0,babelbox/babelbox_voice,google/fleurs" \ --dataset_train_config_name="sv-SE,nst,sv_se" \ --language="swedish" \ --train_split_name="train+validation,train,train+validation+test" \ --dataset_eval_name="mozilla-foundation/common_voice_11_0" \ --dataset_eval_config_name="sv-SE" \ --eval_split_name="test" \ --model_index_name="Whisper Medium Swedish" \ --max_steps="5000" \ --output_dir="./" \ --per_device_train_batch_size="32" \ --per_device_eval_batch_size="16" \ --logging_steps="25" \ --learning_rate="1e-5" \ --warmup_steps="250" \ --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,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 \ --fp16 \ --predict_with_generate \ --do_normalize_eval \ --streaming \ --use_auth_token