[Wav2Vec2 Speech Event] Add speech event v2 (#15083)

* up

* up

* up

* up

* up

* up

* improve

* up

* up

* Update src/transformers/trainer.py

* up

* up

* up

examples/pytorch/speech-recognition/README.md

@@ -74,6 +74,7 @@ python run_speech_recognition_ctc.py \
 	--warmup_steps="500" \
 	--evaluation_strategy="steps" \
 	--text_column_name="sentence" \
+	--length_column_name="input_length" \
 	--save_steps="400" \
 	--eval_steps="100" \
 	--layerdrop="0.0" \
@@ -108,6 +109,7 @@ python -m torch.distributed.launch \
 	--warmup_steps="500" \
 	--evaluation_strategy="steps" \
 	--text_column_name="sentence" \
+	--length_column_name="input_length" \
 	--save_steps="400" \
 	--eval_steps="100" \
 	--logging_steps="1" \

examples/pytorch/speech-recognition/requirements.txt

@@ -2,3 +2,4 @@ datasets >= 1.13.3
 torch >= 1.5
 torchaudio
 librosa
+jiwer

examples/pytorch/speech-recognition/run_speech_recognition_ctc.py

@@ -396,7 +396,10 @@ def main():
 
     if training_args.do_train:
         raw_datasets["train"] = load_dataset(
-            data_args.dataset_name, data_args.dataset_config_name, split=data_args.train_split_name
+            data_args.dataset_name,
+            data_args.dataset_config_name,
+            split=data_args.train_split_name,
+            use_auth_token=data_args.use_auth_token,
         )
 
         if data_args.audio_column_name not in raw_datasets["train"].column_names:
@@ -418,7 +421,10 @@ def main():
 
     if training_args.do_eval:
         raw_datasets["eval"] = load_dataset(
-            data_args.dataset_name, data_args.dataset_config_name, split=data_args.eval_split_name
+            data_args.dataset_name,
+            data_args.dataset_config_name,
+            split=data_args.eval_split_name,
+            use_auth_token=data_args.use_auth_token,
         )
 
         if data_args.max_eval_samples is not None: