Pass datasets trust_remote_code (#31406)

* Pass datasets trust_remote_code

* Pass trust_remote_code in more tests

* Add trust_remote_dataset_code arg to some tests

* Revert "Temporarily pin datasets upper version to fix CI"

This reverts commit b7672826.

* Pass trust_remote_code in librispeech_asr_dummy docstrings

* Revert "Pin datasets<2.20.0 for examples"

This reverts commit 833fc17a.

* Pass trust_remote_code to all examples

* Revert "Add trust_remote_dataset_code arg to some tests" to research_projects

* Pass trust_remote_code to tests

* Pass trust_remote_code to docstrings

* Fix flax examples tests requirements

* Pass trust_remote_dataset_code arg to tests

* Replace trust_remote_dataset_code with trust_remote_code in one example

* Fix duplicate trust_remote_code

* Replace args.trust_remote_dataset_code with args.trust_remote_code

* Replace trust_remote_dataset_code with trust_remote_code in parser

* Replace trust_remote_dataset_code with trust_remote_code in dataclasses

* Replace trust_remote_dataset_code with trust_remote_code arg

examples/pytorch/speech-recognition/run_speech_recognition_ctc_adapter.py

@@ -245,9 +245,9 @@ class DataTrainingArguments:
         default=False,
         metadata={
             "help": (
-                "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option "
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
-                "execute code present on the Hub on your local machine."
+                "Whether to trust the execution of code from datasets/models defined on the Hub."
+                " This option should only be set to `True` for repositories you trust and in which you have read the"
+                " code, as it will execute code present on the Hub on your local machine."
             )
         },
     )
@@ -434,6 +434,7 @@ def main():
             data_args.dataset_config_name,
             split=data_args.train_split_name,
             token=data_args.token,
+            trust_remote_code=data_args.trust_remote_code,
         )
 
         if data_args.audio_column_name not in raw_datasets["train"].column_names:
@@ -459,6 +460,7 @@ def main():
             data_args.dataset_config_name,
             split=data_args.eval_split_name,
             token=data_args.token,
+            trust_remote_code=data_args.trust_remote_code,
         )
 
         if data_args.max_eval_samples is not None:

examples/pytorch/speech-recognition/run_speech_recognition_seq2seq.py

@@ -98,9 +98,9 @@ class ModelArguments:
         default=False,
         metadata={
             "help": (
-                "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option "
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
-                "execute code present on the Hub on your local machine."
+                "Whether to trust the execution of code from datasets/models defined on the Hub."
+                " This option should only be set to `True` for repositories you trust and in which you have read the"
+                " code, as it will execute code present on the Hub on your local machine."
             )
         },
     )
@@ -347,6 +347,7 @@ def main():
             split=data_args.train_split_name,
             cache_dir=model_args.cache_dir,
             token=model_args.token,
+            trust_remote_code=model_args.trust_remote_code,
         )
 
     if training_args.do_eval:
@@ -356,6 +357,7 @@ def main():
             split=data_args.eval_split_name,
             cache_dir=model_args.cache_dir,
             token=model_args.token,
+            trust_remote_code=model_args.trust_remote_code,
         )
 
     if data_args.audio_column_name not in next(iter(raw_datasets.values())).column_names:

examples/pytorch/summarization/run_summarization.py

@@ -112,9 +112,9 @@ class ModelArguments:
         default=False,
         metadata={
             "help": (
-                "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option "
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
-                "execute code present on the Hub on your local machine."
+                "Whether to trust the execution of code from datasets/models defined on the Hub."
+                " This option should only be set to `True` for repositories you trust and in which you have read the"
+                " code, as it will execute code present on the Hub on your local machine."
             )
         },
     )
@@ -397,6 +397,7 @@ def main():
             data_args.dataset_config_name,
             cache_dir=model_args.cache_dir,
             token=model_args.token,
+            trust_remote_code=model_args.trust_remote_code,
         )
     else:
         data_files = {}

examples/pytorch/summarization/run_summarization_no_trainer.py

@@ -268,12 +268,11 @@ def parse_args():
     parser.add_argument("--hub_token", type=str, help="The token to use to push to the Model Hub.")
     parser.add_argument(
         "--trust_remote_code",
-        type=bool,
-        default=False,
+        action="store_true",
         help=(
-            "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option "
-            "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
-            "execute code present on the Hub on your local machine."
+            "Whether to trust the execution of code from datasets/models defined on the Hub."
+            " This option should only be set to `True` for repositories you trust and in which you have read the"
+            " code, as it will execute code present on the Hub on your local machine."
         ),
     )
     parser.add_argument(
@@ -398,7 +397,9 @@ def main():
     # download the dataset.
     if args.dataset_name is not None:
         # Downloading and loading a dataset from the hub.
-        raw_datasets = load_dataset(args.dataset_name, args.dataset_config_name)
+        raw_datasets = load_dataset(
+            args.dataset_name, args.dataset_config_name, trust_remote_code=args.trust_remote_code
+        )
     else:
         data_files = {}
         if args.train_file is not None:

examples/pytorch/test_accelerate_examples.py

@@ -313,6 +313,7 @@ class ExamplesTestsNoTrainer(TestCasePlus):
             {self.examples_dir}/pytorch/image-classification/run_image_classification_no_trainer.py
             --model_name_or_path google/vit-base-patch16-224-in21k
             --dataset_name hf-internal-testing/cats_vs_dogs_sample
+            --trust_remote_code
             --learning_rate 1e-4
             --per_device_train_batch_size 2
             --per_device_eval_batch_size 1

examples/pytorch/test_pytorch_examples.py

@@ -391,6 +391,7 @@ class ExamplesTests(TestCasePlus):
             --output_dir {tmp_dir}
             --model_name_or_path google/vit-base-patch16-224-in21k
             --dataset_name hf-internal-testing/cats_vs_dogs_sample
+            --trust_remote_code
             --do_train
             --do_eval
             --learning_rate 1e-4
@@ -424,6 +425,7 @@ class ExamplesTests(TestCasePlus):
             --dataset_config_name clean
             --train_split_name validation
             --eval_split_name validation
+            --trust_remote_code
             --do_train
             --do_eval
             --learning_rate 1e-4
@@ -454,6 +456,7 @@ class ExamplesTests(TestCasePlus):
             --dataset_config_name clean
             --train_split_name validation
             --eval_split_name validation
+            --trust_remote_code
             --do_train
             --do_eval
             --learning_rate 1e-4
@@ -486,6 +489,7 @@ class ExamplesTests(TestCasePlus):
             --dataset_config_name clean
             --train_split_name validation
             --eval_split_name validation
+            --trust_remote_code
             --do_train
             --do_eval
             --learning_rate 1e-4
@@ -513,6 +517,7 @@ class ExamplesTests(TestCasePlus):
             --output_dir {tmp_dir}
             --model_name_or_path hf-internal-testing/tiny-random-wav2vec2
             --dataset_name anton-l/superb_demo
+            --trust_remote_code
             --dataset_config_name ks
             --train_split_name test
             --eval_split_name test
@@ -547,6 +552,7 @@ class ExamplesTests(TestCasePlus):
             --dataset_name hf-internal-testing/librispeech_asr_dummy
             --dataset_config_names clean
             --dataset_split_names validation
+            --trust_remote_code
             --learning_rate 1e-4
             --per_device_train_batch_size 4
             --per_device_eval_batch_size 4
@@ -567,6 +573,7 @@ class ExamplesTests(TestCasePlus):
             run_mae.py
             --output_dir {tmp_dir}
             --dataset_name hf-internal-testing/cats_vs_dogs_sample
+            --trust_remote_code
             --do_train
             --do_eval
             --learning_rate 1e-4

examples/pytorch/text-classification/run_classification.py

@@ -240,9 +240,9 @@ class ModelArguments:
         default=False,
         metadata={
             "help": (
-                "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option "
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
-                "execute code present on the Hub on your local machine."
+                "Whether to trust the execution of code from datasets/models defined on the Hub."
+                " This option should only be set to `True` for repositories you trust and in which you have read the"
+                " code, as it will execute code present on the Hub on your local machine."
             )
         },
     )
@@ -338,6 +338,7 @@ def main():
             data_args.dataset_config_name,
             cache_dir=model_args.cache_dir,
             token=model_args.token,
+            trust_remote_code=model_args.trust_remote_code,
         )
         # Try print some info about the dataset
         logger.info(f"Dataset loaded: {raw_datasets}")

examples/pytorch/text-classification/run_glue.py

@@ -201,9 +201,9 @@ class ModelArguments:
         default=False,
         metadata={
             "help": (
-                "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option "
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
-                "execute code present on the Hub on your local machine."
+                "Whether to trust the execution of code from datasets/models defined on the Hub."
+                " This option should only be set to `True` for repositories you trust and in which you have read the"
+                " code, as it will execute code present on the Hub on your local machine."
             )
         },
     )
@@ -300,6 +300,7 @@ def main():
             data_args.dataset_config_name,
             cache_dir=model_args.cache_dir,
             token=model_args.token,
+            trust_remote_code=model_args.trust_remote_code,
         )
     else:
         # Loading a dataset from your local files.

examples/pytorch/token-classification/run_ner.py

@@ -92,9 +92,9 @@ class ModelArguments:
         default=False,
         metadata={
             "help": (
-                "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option "
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
-                "execute code present on the Hub on your local machine."
+                "Whether to trust the execution of code from datasets/models defined on the Hub."
+                " This option should only be set to `True` for repositories you trust and in which you have read the"
+                " code, as it will execute code present on the Hub on your local machine."
             )
         },
     )
@@ -290,6 +290,7 @@ def main():
             data_args.dataset_config_name,
             cache_dir=model_args.cache_dir,
             token=model_args.token,
+            trust_remote_code=model_args.trust_remote_code,
         )
     else:
         data_files = {}

examples/pytorch/token-classification/run_ner_no_trainer.py

@@ -212,12 +212,11 @@ def parse_args():
     parser.add_argument("--hub_token", type=str, help="The token to use to push to the Model Hub.")
     parser.add_argument(
         "--trust_remote_code",
-        type=bool,
-        default=False,
+        action="store_true",
         help=(
-            "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option "
-            "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
-            "execute code present on the Hub on your local machine."
+            "Whether to trust the execution of code from datasets/models defined on the Hub."
+            " This option should only be set to `True` for repositories you trust and in which you have read the"
+            " code, as it will execute code present on the Hub on your local machine."
         ),
     )
     parser.add_argument(
@@ -333,7 +332,9 @@ def main():
     # download the dataset.
     if args.dataset_name is not None:
         # Downloading and loading a dataset from the hub.
-        raw_datasets = load_dataset(args.dataset_name, args.dataset_config_name)
+        raw_datasets = load_dataset(
+            args.dataset_name, args.dataset_config_name, trust_remote_code=args.trust_remote_code
+        )
     else:
         data_files = {}
         if args.train_file is not None:

examples/pytorch/translation/run_translation.py

@@ -102,9 +102,9 @@ class ModelArguments:
         default=False,
         metadata={
             "help": (
-                "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option "
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
-                "execute code present on the Hub on your local machine."
+                "Whether to trust the execution of code from datasets/models defined on the Hub."
+                " This option should only be set to `True` for repositories you trust and in which you have read the"
+                " code, as it will execute code present on the Hub on your local machine."
             )
         },
     )
@@ -346,6 +346,7 @@ def main():
             data_args.dataset_config_name,
             cache_dir=model_args.cache_dir,
             token=model_args.token,
+            trust_remote_code=model_args.trust_remote_code,
         )
     else:
         data_files = {}

examples/pytorch/translation/run_translation_no_trainer.py

@@ -76,7 +76,6 @@ def parse_args():
         default=None,
         help="The name of the dataset to use (via the datasets library).",
     )
-
     parser.add_argument(
         "--predict_with_generate",
         type=bool,
@@ -259,12 +258,11 @@ def parse_args():
     parser.add_argument("--hub_token", type=str, help="The token to use to push to the Model Hub.")
     parser.add_argument(
         "--trust_remote_code",
-        type=bool,
-        default=False,
+        action="store_true",
         help=(
-            "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option "
-            "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
-            "execute code present on the Hub on your local machine."
+            "Whether to trust the execution of code from datasets/models defined on the Hub."
+            " This option should only be set to `True` for repositories you trust and in which you have read the"
+            " code, as it will execute code present on the Hub on your local machine."
         ),
     )
     parser.add_argument(
@@ -378,7 +376,9 @@ def main():
     # download the dataset.
     if args.dataset_name is not None:
         # Downloading and loading a dataset from the hub.
-        raw_datasets = load_dataset(args.dataset_name, args.dataset_config_name)
+        raw_datasets = load_dataset(
+            args.dataset_name, args.dataset_config_name, trust_remote_code=args.trust_remote_code
+        )
     else:
         data_files = {}
         if args.train_file is not None:

examples/tensorflow/_tests_requirements.txt

@@ -14,7 +14,7 @@ streamlit
 elasticsearch
 nltk
 pandas
-datasets >= 1.13.3,<2.20.0 # Temporary upper version
+datasets >= 1.13.3
 fire
 pytest<8.0.1
 conllu

examples/tensorflow/contrastive-image-text/run_clip.py

@@ -105,9 +105,9 @@ class ModelArguments:
         default=False,
         metadata={
             "help": (
-                "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option "
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
-                "execute code present on the Hub on your local machine."
+                "Whether to trust the execution of code from datasets/models defined on the Hub."
+                " This option should only be set to `True` for repositories you trust and in which you have read the"
+                " code, as it will execute code present on the Hub on your local machine."
             )
         },
     )
@@ -326,6 +326,7 @@ def main():
             keep_in_memory=False,
             data_dir=data_args.data_dir,
             token=model_args.token,
+            trust_remote_code=model_args.trust_remote_code,
         )
     else:
         data_files = {}

examples/tensorflow/image-classification/run_image_classification.py

@@ -171,9 +171,9 @@ class ModelArguments:
         default=False,
         metadata={
             "help": (
-                "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option "
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
-                "execute code present on the Hub on your local machine."
+                "Whether to trust the execution of code from datasets/models defined on the Hub."
+                " This option should only be set to `True` for repositories you trust and in which you have read the"
+                " code, as it will execute code present on the Hub on your local machine."
             )
         },
     )
@@ -284,6 +284,7 @@ def main():
             cache_dir=model_args.cache_dir,
             task="image-classification",
             token=model_args.token,
+            trust_remote_code=model_args.trust_remote_code,
         )
     else:
         data_files = {}

examples/tensorflow/language-modeling-tpu/prepare_tfrecord_shards.py

@@ -42,6 +42,15 @@ def parse_args():
     parser.add_argument(
         "--dataset_config", type=str, default="wikitext-103-raw-v1", help="Configuration name of the dataset."
     )
+    parser.add_argument(
+        "--trust_remote_code",
+        action="store_true",
+        help=(
+            "Whether to trust the execution of code from datasets/models defined on the Hub."
+            " This option should only be set to `True` for repositories you trust and in which you have read the"
+            " code, as it will execute code present on the Hub on your local machine."
+        ),
+    )
     parser.add_argument(
         "--tokenizer_name_or_path",
         type=str,
@@ -105,7 +114,9 @@ def get_serialized_examples(tokenized_data):
 
 
 def main(args):
-    dataset = datasets.load_dataset(args.dataset_name, args.dataset_config, split=args.split)
+    dataset = datasets.load_dataset(
+        args.dataset_name, args.dataset_config, split=args.split, trust_remote_code=args.trust_remote_code
+    )
 
     if args.limit is not None:
         max_samples = min(len(dataset), args.limit)

examples/tensorflow/language-modeling-tpu/train_unigram.py

@@ -41,6 +41,15 @@ def parse_args():
     parser.add_argument(
         "--dataset_config", type=str, default="wikitext-103-raw-v1", help="Configuration name of the dataset."
     )
+    parser.add_argument(
+        "--trust_remote_code",
+        action="store_true",
+        help=(
+            "Whether to trust the execution of code from datasets/models defined on the Hub."
+            " This option should only be set to `True` for repositories you trust and in which you have read the"
+            " code, as it will execute code present on the Hub on your local machine."
+        ),
+    )
     parser.add_argument(
         "--batch_size",
         type=int,
@@ -69,7 +78,9 @@ def parse_args():
 
 
 def main(args):
-    dataset = datasets.load_dataset(args.dataset_name, args.dataset_config, split="train")
+    dataset = datasets.load_dataset(
+        args.dataset_name, args.dataset_config, split="train", trust_remote_code=args.trust_remote_code
+    )
 
     if args.limit is not None:
         max_train_samples = min(len(dataset), args.limit)

examples/tensorflow/language-modeling/run_clm.py

@@ -125,9 +125,9 @@ class ModelArguments:
         default=False,
         metadata={
             "help": (
-                "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option "
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
-                "execute code present on the Hub on your local machine."
+                "Whether to trust the execution of code from datasets/models defined on the Hub."
+                " This option should only be set to `True` for repositories you trust and in which you have read the"
+                " code, as it will execute code present on the Hub on your local machine."
             )
         },
     )
@@ -298,6 +298,7 @@ def main():
             data_args.dataset_config_name,
             cache_dir=model_args.cache_dir,
             token=model_args.token,
+            trust_remote_code=model_args.trust_remote_code,
         )
         if "validation" not in raw_datasets.keys():
             raw_datasets["validation"] = load_dataset(
@@ -306,6 +307,7 @@ def main():
                 split=f"train[:{data_args.validation_split_percentage}%]",
                 cache_dir=model_args.cache_dir,
                 token=model_args.token,
+                trust_remote_code=model_args.trust_remote_code,
             )
             raw_datasets["train"] = load_dataset(
                 data_args.dataset_name,
@@ -313,6 +315,7 @@ def main():
                 split=f"train[{data_args.validation_split_percentage}%:]",
                 cache_dir=model_args.cache_dir,
                 token=model_args.token,
+                trust_remote_code=model_args.trust_remote_code,
             )
     else:
         data_files = {}

examples/tensorflow/language-modeling/run_mlm.py

@@ -123,9 +123,9 @@ class ModelArguments:
         default=False,
         metadata={
             "help": (
-                "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option "
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
-                "execute code present on the Hub on your local machine."
+                "Whether to trust the execution of code from datasets/models defined on the Hub."
+                " This option should only be set to `True` for repositories you trust and in which you have read the"
+                " code, as it will execute code present on the Hub on your local machine."
             )
         },
     )
@@ -307,6 +307,7 @@ def main():
             data_args.dataset_name,
             data_args.dataset_config_name,
             token=model_args.token,
+            trust_remote_code=model_args.trust_remote_code,
         )
         if "validation" not in raw_datasets.keys():
             raw_datasets["validation"] = load_dataset(
@@ -314,12 +315,14 @@ def main():
                 data_args.dataset_config_name,
                 split=f"train[:{data_args.validation_split_percentage}%]",
                 token=model_args.token,
+                trust_remote_code=model_args.trust_remote_code,
             )
             raw_datasets["train"] = load_dataset(
                 data_args.dataset_name,
                 data_args.dataset_config_name,
                 split=f"train[{data_args.validation_split_percentage}%:]",
                 token=model_args.token,
+                trust_remote_code=model_args.trust_remote_code,
             )
     else:
         data_files = {}

examples/tensorflow/question-answering/run_qa.py

@@ -104,9 +104,9 @@ class ModelArguments:
         default=False,
         metadata={
             "help": (
-                "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option "
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
-                "execute code present on the Hub on your local machine."
+                "Whether to trust the execution of code from datasets/models defined on the Hub."
+                " This option should only be set to `True` for repositories you trust and in which you have read the"
+                " code, as it will execute code present on the Hub on your local machine."
             )
         },
     )
@@ -329,6 +329,7 @@ def main():
             data_args.dataset_config_name,
             cache_dir=model_args.cache_dir,
             token=model_args.token,
+            trust_remote_code=model_args.trust_remote_code,
         )
     else:
         data_files = {}