first

run.sh

@@ -1,10 +1,11 @@
 python run_speech_recognition_ctc.py \
         --dataset_name="NbAiLab/NPSC" \
         --model_name_or_path="facebook/wav2vec2-xls-r-300m" \
-        --dataset_config_name="16K_mp3" \
+        --hub_model_id="NbAiLab/wav2vec2-xlsr-300M-NPSC" \
+	--dataset_config_name="16K_mp3" \
         --output_dir="./" \
         --overwrite_output_dir \
-        --num_train_epochs="50" \
+        --num_train_epochs="75" \
         --per_device_train_batch_size="8" \
         --per_device_eval_batch_size="8" \
         --gradient_accumulation_steps="4" \

run_speech_recognition_ctc.py

@@ -393,7 +393,7 @@ def main():
 
     # Pre-processing dataset
     import re
-     def filter_inaudible(entry):
+    def filter_inaudible(entry):
         return not re.search("\d|<inaudible>", entry["text"], flags=re.IGNORECASE)
 
     def filter_nynorsk(entry):