training code

run.sh

@@ -0,0 +1,34 @@
+python run_speech_recognition_ctc.py \
+	--dataset_name="mozilla-foundation/common_voice_8_0" \
+	--model_name_or_path="facebook/wav2vec2-xls-r-300m" \
+	--dataset_config_name="ab" \
+	--output_dir="./wav2vec2-large-xls-r-300m-abkhaz-cv8" \
+	--overwrite_output_dir \
+	--num_train_epochs="50" \
+	--per_device_train_batch_size="32" \
+	--per_device_eval_batch_size="16" \
+	--gradient_accumulation_steps="1" \
+	--learning_rate="7e-5" \
+	--warmup_steps="4000" \
+	--length_column_name="input_length" \
+	--evaluation_strategy="steps" \
+	--text_column_name="sentence" \
+	--chars_to_ignore , ? . ! \- \; \: \" “ % ‘ ” � — ’ … – \' \c \
+	--save_steps="4000" \
+	--eval_steps="4000" \
+	--logging_steps="1000" \
+	--layerdrop="0.0" \
+	--activation_dropout="0.1" \
+	--save_total_limit="2" \
+	--freeze_feature_encoder \
+	--feat_proj_dropout="0.0" \
+	--mask_time_prob="0.75" \
+	--mask_time_length="10" \
+	--mask_feature_prob="0.25" \
+	--mask_feature_length="64" \
+	--gradient_checkpointing \
+	--use_auth_token \
+	--fp16 \
+	--group_by_length \
+	--do_train --do_eval \
+    --push_to_hub

run_speech_recognition_ctc.py

@@ -444,9 +444,6 @@ def main():
             batch["target_text"] = re.sub(chars_to_ignore_regex, "", batch[text_column_name]).lower() + " "
         else:
             batch["target_text"] = batch[text_column_name].lower() + " "
-
-        batch["target_text"] = batch[text_column_name].lower().replace('c', '')
-
         return batch
 
     with training_args.main_process_first(desc="dataset map special characters removal"):