docs: update speech recognition examples

examples/pytorch/speech-recognition/README.md

@@ -66,7 +66,7 @@ The following command shows how to fine-tune [XLSR-Wav2Vec2](https://huggingface
 
 ```bash
 python run_speech_recognition_ctc.py \
-	--dataset_name="common_voice" \
+	--dataset_name="mozilla-foundation/common_voice_17_0" \
 	--model_name_or_path="facebook/wav2vec2-large-xlsr-53" \
 	--dataset_config_name="tr" \
 	--output_dir="./wav2vec2-common_voice-tr-demo" \
@@ -102,7 +102,7 @@ The following command shows how to fine-tune [XLSR-Wav2Vec2](https://huggingface
 ```bash
 torchrun \
 	--nproc_per_node 8 run_speech_recognition_ctc.py \
-	--dataset_name="common_voice" \
+	--dataset_name="mozilla-foundation/common_voice_17_0" \
 	--model_name_or_path="facebook/wav2vec2-large-xlsr-53" \
 	--dataset_config_name="tr" \
 	--output_dir="./wav2vec2-common_voice-tr-demo-dist" \
@@ -149,7 +149,7 @@ However, the `--shuffle_buffer_size` argument controls how many examples we can
 ```bash
 **torchrun \
 	--nproc_per_node 4 run_speech_recognition_ctc_streaming.py \
-	--dataset_name="common_voice" \
+	--dataset_name="mozilla-foundation/common_voice_17_0" \
 	--model_name_or_path="facebook/wav2vec2-xls-r-300m" \
 	--tokenizer_name_or_path="anton-l/wav2vec2-tokenizer-turkish" \
 	--dataset_config_name="tr" \
@@ -314,7 +314,7 @@ below 27%.
 For an example run, you can have a look at [`patrickvonplaten/wav2vec2-common_voice-tr-mms-demo`](https://huggingface.co/patrickvonplaten/wav2vec2-common_voice-tr-mms-demo).
 
 
-If you'd like to train another adapter model with the same base model, you can simply re-use the same `--output_dir`,
+If you'd like to train another adapter model with the same base model, you can simply reuse the same `--output_dir`,
 but make sure to pass the `--output_dir` folder also to `--tokenizer_name_or_path` so that the vocabulary is not 
 overwritten but **extended**. Assuming you would like to train adapter weights on Swedish in addition to Turkish and save 
 the adapter weights in the same model repo, you can run:

examples/pytorch/speech-recognition/run_speech_recognition_ctc.py

@@ -63,7 +63,10 @@
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
 check_min_version("4.57.0.dev0")
 
-require_version("datasets>=1.18.0", "To fix: pip install -r examples/pytorch/speech-recognition/requirements.txt")
+require_version(
+    "datasets>=1.18.0",
+    "To fix: pip install -r examples/pytorch/speech-recognition/requirements.txt",
+)
 
 
 logger = logging.getLogger(__name__)
@@ -91,13 +94,16 @@ class ModelArguments:
         metadata={"help": "Where do you want to store the pretrained models downloaded from huggingface.co"},
     )
     freeze_feature_encoder: bool = field(
-        default=True, metadata={"help": "Whether to freeze the feature encoder layers of the model."}
+        default=True,
+        metadata={"help": "Whether to freeze the feature encoder layers of the model."},
     )
     attention_dropout: float = field(
-        default=0.0, metadata={"help": "The dropout ratio for the attention probabilities."}
+        default=0.0,
+        metadata={"help": "The dropout ratio for the attention probabilities."},
     )
     activation_dropout: float = field(
-        default=0.0, metadata={"help": "The dropout ratio for activations inside the fully connected layer."}
+        default=0.0,
+        metadata={"help": "The dropout ratio for activations inside the fully connected layer."},
     )
     feat_proj_dropout: float = field(default=0.0, metadata={"help": "The dropout ratio for the projected features."})
     hidden_dropout: float = field(
@@ -140,7 +146,8 @@ class ModelArguments:
     )
     layerdrop: 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'."}
+        default="mean",
+        metadata={"help": "The way the ctc loss should be reduced. Should be one of 'mean' or 'sum'."},
     )
     ctc_zero_infinity: Optional[bool] = field(
         default=False,
@@ -169,10 +176,13 @@ class DataTrainingArguments:
     """
 
     dataset_name: str = field(
-        metadata={"help": "The configuration name of the dataset to use (via the datasets library)."}
+        metadata={"help": "Path or name of the dataset (cf `load_dataset` method of the Datasets library)."}
     )
     dataset_config_name: str = field(
-        default=None, metadata={"help": "The configuration name of the dataset to use (via the datasets library)."}
+        default=None,
+        metadata={
+            "help": "The configuration name of the dataset to use (cf `load_dataset` method of the Datasets library)."
+        },
     )
     train_split_name: str = field(
         default="train+validation",
@@ -198,7 +208,8 @@ class DataTrainingArguments:
         metadata={"help": "The name of the dataset column containing the text data. Defaults to 'text'"},
     )
     overwrite_cache: bool = field(
-        default=False, metadata={"help": "Overwrite the cached preprocessed datasets or not."}
+        default=False,
+        metadata={"help": "Overwrite the cached preprocessed datasets or not."},
     )
     preprocessing_num_workers: Optional[int] = field(
         default=None,
@@ -240,7 +251,8 @@ class DataTrainingArguments:
         },
     )
     min_duration_in_seconds: float = field(
-        default=0.0, metadata={"help": "Filter audio files that are shorter than `min_duration_in_seconds` seconds"}
+        default=0.0,
+        metadata={"help": "Filter audio files that are shorter than `min_duration_in_seconds` seconds"},
     )
     preprocessing_only: bool = field(
         default=False,
@@ -383,7 +395,8 @@ def extract_all_chars(batch):
 
     # take union of all unique characters in each dataset
     vocab_set = functools.reduce(
-        lambda vocab_1, vocab_2: set(vocab_1["vocab"][0]) | set(vocab_2["vocab"][0]), vocabs.values()
+        lambda vocab_1, vocab_2: set(vocab_1["vocab"][0]) | set(vocab_2["vocab"][0]),
+        vocabs.values(),
     )
 
     vocab_dict = {v: k for k, v in enumerate(sorted(vocab_set))}
@@ -571,7 +584,7 @@ def remove_special_characters(batch):
         # it is defined by `tokenizer_class` if present in config else by `model_type`
         tokenizer_kwargs = {
             "config": config if config.tokenizer_class is not None else None,
-            "tokenizer_type": config.model_type if config.tokenizer_class is None else None,
+            "tokenizer_type": (config.model_type if config.tokenizer_class is None else None),
             "unk_token": unk_token,
             "pad_token": pad_token,
             "word_delimiter_token": word_delimiter_token,
@@ -639,7 +652,8 @@ def remove_special_characters(batch):
     dataset_sampling_rate = next(iter(raw_datasets.values())).features[data_args.audio_column_name].sampling_rate
     if dataset_sampling_rate != feature_extractor.sampling_rate:
         raw_datasets = raw_datasets.cast_column(
-            data_args.audio_column_name, datasets.features.Audio(sampling_rate=feature_extractor.sampling_rate)
+            data_args.audio_column_name,
+            datasets.features.Audio(sampling_rate=feature_extractor.sampling_rate),
         )
 
     # derive max & min input length for sample rate & max duration

examples/pytorch/speech-recognition/run_speech_recognition_ctc_adapter.py

@@ -66,7 +66,10 @@
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
 check_min_version("4.57.0.dev0")
 
-require_version("datasets>=1.18.0", "To fix: pip install -r examples/pytorch/speech-recognition/requirements.txt")
+require_version(
+    "datasets>=1.18.0",
+    "To fix: pip install -r examples/pytorch/speech-recognition/requirements.txt",
+)
 
 
 logger = logging.getLogger(__name__)
@@ -127,7 +130,8 @@ class ModelArguments:
     )
     layerdrop: 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'."}
+        default="mean",
+        metadata={"help": "The way the ctc loss should be reduced. Should be one of 'mean' or 'sum'."},
     )
     adapter_attn_dim: int = field(
         default=16,
@@ -148,9 +152,9 @@ class DataTrainingArguments:
     """
 
     dataset_name: str = field(
-        metadata={"help": "The configuration name of the dataset to use (via the datasets library)."}
+        metadata={"help": "Path or name of the dataset (cf `load_dataset` method of the Datasets library)."}
     )
-    target_language: Optional[str] = field(
+    target_language: str = field(
         metadata={
             "help": (
                 "The target language on which the adapter attention layers"
@@ -162,7 +166,10 @@ class DataTrainingArguments:
         },
     )
     dataset_config_name: str = field(
-        default=None, metadata={"help": "The configuration name of the dataset to use (via the datasets library)."}
+        default=None,
+        metadata={
+            "help": "The configuration name of the dataset to use (cf `load_dataset` method of the Datasets library)."
+        },
     )
     train_split_name: str = field(
         default="train+validation",
@@ -188,7 +195,8 @@ class DataTrainingArguments:
         metadata={"help": "The name of the dataset column containing the text data. Defaults to 'text'"},
     )
     overwrite_cache: bool = field(
-        default=False, metadata={"help": "Overwrite the cached preprocessed datasets or not."}
+        default=False,
+        metadata={"help": "Overwrite the cached preprocessed datasets or not."},
     )
     preprocessing_num_workers: Optional[int] = field(
         default=None,
@@ -230,7 +238,8 @@ class DataTrainingArguments:
         },
     )
     min_duration_in_seconds: float = field(
-        default=0.0, metadata={"help": "Filter audio files that are shorter than `min_duration_in_seconds` seconds"}
+        default=0.0,
+        metadata={"help": "Filter audio files that are shorter than `min_duration_in_seconds` seconds"},
     )
     preprocessing_only: bool = field(
         default=False,
@@ -363,7 +372,8 @@ def extract_all_chars(batch):
 
     # take union of all unique characters in each dataset
     vocab_set = functools.reduce(
-        lambda vocab_1, vocab_2: set(vocab_1["vocab"][0]) | set(vocab_2["vocab"][0]), vocabs.values()
+        lambda vocab_1, vocab_2: set(vocab_1["vocab"][0]) | set(vocab_2["vocab"][0]),
+        vocabs.values(),
     )
 
     vocab_dict = {v: k for k, v in enumerate(sorted(vocab_set))}
@@ -578,7 +588,7 @@ def remove_special_characters(batch):
         # it is defined by `tokenizer_class` if present in config else by `model_type`
         tokenizer_kwargs = {
             "config": config if config.tokenizer_class is not None else None,
-            "tokenizer_type": config.model_type if config.tokenizer_class is None else None,
+            "tokenizer_type": (config.model_type if config.tokenizer_class is None else None),
             "unk_token": unk_token,
             "pad_token": pad_token,
             "word_delimiter_token": word_delimiter_token,
@@ -650,7 +660,8 @@ def remove_special_characters(batch):
     dataset_sampling_rate = next(iter(raw_datasets.values())).features[data_args.audio_column_name].sampling_rate
     if dataset_sampling_rate != feature_extractor.sampling_rate:
         raw_datasets = raw_datasets.cast_column(
-            data_args.audio_column_name, datasets.features.Audio(sampling_rate=feature_extractor.sampling_rate)
+            data_args.audio_column_name,
+            datasets.features.Audio(sampling_rate=feature_extractor.sampling_rate),
         )
 
     # derive max & min input length for sample rate & max duration

examples/pytorch/speech-recognition/run_speech_recognition_seq2seq.py

@@ -62,7 +62,10 @@
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
 check_min_version("4.57.0.dev0")
 
-require_version("datasets>=1.18.0", "To fix: pip install -r examples/pytorch/speech-recognition/requirements.txt")
+require_version(
+    "datasets>=1.18.0",
+    "To fix: pip install -r examples/pytorch/speech-recognition/requirements.txt",
+)
 
 logger = logging.getLogger(__name__)
 
@@ -77,13 +80,16 @@ class ModelArguments:
         metadata={"help": "Path to pretrained model or model identifier from huggingface.co/models"}
     )
     config_name: Optional[str] = field(
-        default=None, metadata={"help": "Pretrained config name or path if not the same as model_name"}
+        default=None,
+        metadata={"help": "Pretrained config name or path if not the same as model_name"},
     )
     tokenizer_name: Optional[str] = field(
-        default=None, metadata={"help": "Pretrained tokenizer name or path if not the same as model_name"}
+        default=None,
+        metadata={"help": "Pretrained tokenizer name or path if not the same as model_name"},
     )
     feature_extractor_name: Optional[str] = field(
-        default=None, metadata={"help": "feature extractor name or path if not the same as model_name"}
+        default=None,
+        metadata={"help": "feature extractor name or path if not the same as model_name"},
     )
     cache_dir: Optional[str] = field(
         default=None,
@@ -117,10 +123,12 @@ class ModelArguments:
         },
     )
     freeze_feature_encoder: bool = field(
-        default=True, metadata={"help": "Whether to freeze the feature encoder layers of the model."}
+        default=True,
+        metadata={"help": "Whether to freeze the feature encoder layers of the model."},
     )
     freeze_encoder: bool = field(
-        default=False, metadata={"help": "Whether to freeze the entire encoder of the seq2seq model."}
+        default=False,
+        metadata={"help": "Whether to freeze the entire encoder of the seq2seq model."},
     )
     forced_decoder_ids: list[list[int]] = field(
         default=None,
@@ -150,13 +158,17 @@ class DataTrainingArguments:
     """
 
     dataset_name: str = field(
-        default=None, metadata={"help": "The name of the dataset to use (via the datasets library)."}
+        metadata={"help": "Path or name of the dataset (cf `load_dataset` method of the Datasets library)."}
     )
-    dataset_config_name: Optional[str] = field(
-        default=None, metadata={"help": "The configuration name of the dataset to use (via the datasets library)."}
+    dataset_config_name: str = field(
+        default=None,
+        metadata={
+            "help": "The configuration name of the dataset to use (cf `load_dataset` method of the Datasets library)."
+        },
     )
     overwrite_cache: bool = field(
-        default=False, metadata={"help": "Overwrite the cached training and evaluation sets"}
+        default=False,
+        metadata={"help": "Overwrite the cached training and evaluation sets"},
     )
     preprocessing_num_workers: Optional[int] = field(
         default=None,
@@ -198,7 +210,8 @@ class DataTrainingArguments:
         },
     )
     min_duration_in_seconds: float = field(
-        default=0.0, metadata={"help": "Filter audio files that are shorter than `min_duration_in_seconds` seconds"}
+        default=0.0,
+        metadata={"help": "Filter audio files that are shorter than `min_duration_in_seconds` seconds"},
     )
     preprocessing_only: bool = field(
         default=False,
@@ -387,7 +400,7 @@ def main():
     # Distributed training:
     # The .from_pretrained methods guarantee that only one local process can concurrently
     config = AutoConfig.from_pretrained(
-        model_args.config_name if model_args.config_name else model_args.model_name_or_path,
+        (model_args.config_name if model_args.config_name else model_args.model_name_or_path),
         cache_dir=model_args.cache_dir,
         revision=model_args.model_revision,
         token=model_args.token,
@@ -399,14 +412,14 @@ def main():
         config.update({"apply_spec_augment": model_args.apply_spec_augment})
 
     feature_extractor = AutoFeatureExtractor.from_pretrained(
-        model_args.feature_extractor_name if model_args.feature_extractor_name else model_args.model_name_or_path,
+        (model_args.feature_extractor_name if model_args.feature_extractor_name else model_args.model_name_or_path),
         cache_dir=model_args.cache_dir,
         revision=model_args.model_revision,
         token=model_args.token,
         trust_remote_code=model_args.trust_remote_code,
     )
     tokenizer = AutoTokenizer.from_pretrained(
-        model_args.tokenizer_name if model_args.tokenizer_name else model_args.model_name_or_path,
+        (model_args.tokenizer_name if model_args.tokenizer_name else model_args.model_name_or_path),
         cache_dir=model_args.cache_dir,
         use_fast=model_args.use_fast_tokenizer,
         revision=model_args.model_revision,
@@ -465,7 +478,8 @@ def main():
     dataset_sampling_rate = next(iter(raw_datasets.values())).features[data_args.audio_column_name].sampling_rate
     if dataset_sampling_rate != feature_extractor.sampling_rate:
         raw_datasets = raw_datasets.cast_column(
-            data_args.audio_column_name, datasets.features.Audio(sampling_rate=feature_extractor.sampling_rate)
+            data_args.audio_column_name,
+            datasets.features.Audio(sampling_rate=feature_extractor.sampling_rate),
         )
 
     # 7. Preprocessing the datasets.
@@ -494,7 +508,9 @@ def prepare_dataset(batch):
         # process audio
         sample = batch[audio_column_name]
         inputs = feature_extractor(
-            sample["array"], sampling_rate=sample["sampling_rate"], return_attention_mask=forward_attention_mask
+            sample["array"],
+            sampling_rate=sample["sampling_rate"],
+            return_attention_mask=forward_attention_mask,
         )
         # process audio length
         batch[model_input_name] = inputs.get(model_input_name)[0]
@@ -579,7 +595,7 @@ def compute_metrics(pred):
         eval_dataset=vectorized_datasets["eval"] if training_args.do_eval else None,
         processing_class=feature_extractor,
         data_collator=data_collator,
-        compute_metrics=compute_metrics if training_args.predict_with_generate else None,
+        compute_metrics=(compute_metrics if training_args.predict_with_generate else None),
     )
 
     # 12. Training
@@ -621,7 +637,10 @@ def compute_metrics(pred):
         trainer.save_metrics("eval", metrics)
 
     # 14. Write Training Stats
-    kwargs = {"finetuned_from": model_args.model_name_or_path, "tasks": "automatic-speech-recognition"}
+    kwargs = {
+        "finetuned_from": model_args.model_name_or_path,
+        "tasks": "automatic-speech-recognition",
+    }
     if data_args.dataset_name is not None:
         kwargs["dataset_tags"] = data_args.dataset_name
         if data_args.dataset_config_name is not None: