up (#13733)

examples/pytorch/speech-recognition/README.md

@@ -81,7 +81,7 @@ of **0.35**.
 The following command shows how to fine-tune [XLSR-Wav2Vec2](https://huggingface.co/transformers/master/model_doc/xlsr_wav2vec2.html) on [Common Voice](https://huggingface.co/datasets/common_voice) using 8 GPUs in half-precision.
 
 ```bash
-python -m torch.distributed.launch \
+OMP_NUM_THREADS=1 python -m torch.distributed.launch \
 	--nproc_per_node 8 run_speech_recognition_ctc.py \
 	--dataset_name="common_voice" \
 	--model_name_or_path="facebook/wav2vec2-large-xlsr-53" \

examples/pytorch/speech-recognition/run_speech_recognition_ctc.py

@@ -103,12 +103,6 @@ class ModelArguments:
             "vectors will be masked along the time axis. This is only relevant if ``apply_spec_augment is True``."
         },
     )
-    gradient_checkpointing: Optional[bool] = field(
-        default=False,
-        metadata={
-            "help": "If True, use gradient checkpointing to save memory at the expense of slower backward pass."
-        },
-    )
     layerdrop: Optional[float] = field(default=0.0, metadata={"help": "The LayerDrop probability."})
     ctc_loss_reduction: Optional[str] = field(
         default="mean", metadata={"help": "The way the ctc loss should be reduced. Should be one of 'mean' or 'sum'."}
@@ -438,7 +432,7 @@ def main():
             "hidden_dropout": model_args.hidden_dropout,
             "final_dropout": model_args.final_dropout,
             "mask_time_prob": model_args.mask_time_prob,
-            "gradient_checkpointing": model_args.gradient_checkpointing,
+            "gradient_checkpointing": training_args.gradient_checkpointing,
             "layerdrop": model_args.layerdrop,
             "ctc_loss_reduction": model_args.ctc_loss_reduction,
             "pad_token_id": processor.tokenizer.pad_token_id,