NbAiLab
/

whisper-tiny-nob

Automatic Speech Recognition

Generated from Trainer

Inference Endpoints

Model card Files Files and versions Metrics Training metrics Community

pere commited on Dec 14, 2022

Commit

9e8fbe1

•

1 Parent(s): 3d86552

first

Files changed (1) hide show

run.sh +3 -3

run.sh CHANGED Viewed

@@ -7,9 +7,9 @@ python run_speech_recognition_seq2seq_streaming.py \
 	--model_index_name="Whisper Tiny Norwegian Bokmål" \
 	--max_steps="100000" \
 	--output_dir="./" \
-	--per_device_train_batch_size="64" \
-	--per_device_eval_batch_size="32" \
-	--gradient_accumulation_step="2" \
 	--logging_steps="50" \
 	--learning_rate="1e-5" \
 	--lr_scheduler_type="constant_with_warmup" \

 	--model_index_name="Whisper Tiny Norwegian Bokmål" \
 	--max_steps="100000" \
 	--output_dir="./" \
+	--per_device_train_batch_size="128" \
+	--per_device_eval_batch_size="64" \
+	--gradient_accumulation_step="1" \
 	--logging_steps="50" \
 	--learning_rate="1e-5" \
 	--lr_scheduler_type="constant_with_warmup" \