Update no_trainer examples to use new logger (huggingface#17044)

* Propagate and fix imports

Author: muellerzr

examples/pytorch/README.md

@@ -167,10 +167,10 @@ python xla_spawn.py --num_cores 8 \
 
 Most PyTorch example scripts have a version using the [🤗 Accelerate](https://github.com/huggingface/accelerate) library
 that exposes the training loop so it's easy for you to customize or tweak them to your needs. They all require you to
-install `accelerate` with
+install `accelerate` with the latest development version
 
 ```bash
-pip install accelerate
+pip install git+https://github.com/huggingface/accelerate
 ```
 
 Then you can easily launch any of the scripts by running

examples/pytorch/image-classification/run_image_classification_no_trainer.py

@@ -37,6 +37,7 @@
 
 import transformers
 from accelerate import Accelerator
+from accelerate.logging import get_logger
 from accelerate.utils import set_seed
 from huggingface_hub import Repository
 from transformers import (
@@ -50,7 +51,7 @@
 from transformers.utils.versions import require_version
 
 
-logger = logging.getLogger(__name__)
+logger = get_logger(__name__)
 
 require_version("datasets>=2.0.0", "To fix: pip install -r examples/pytorch/image-classification/requirements.txt")
 
@@ -188,11 +189,7 @@ def main():
         datefmt="%m/%d/%Y %H:%M:%S",
         level=logging.INFO,
     )
-    logger.info(accelerator.state)
-
-    # Setup logging, we only want one process per machine to log things on the screen.
-    # accelerator.is_local_main_process is only True for one process per machine.
-    logger.setLevel(logging.INFO if accelerator.is_local_main_process else logging.ERROR)
+    logger.info(accelerator.state, main_process_only=False)
     if accelerator.is_local_main_process:
         datasets.utils.logging.set_verbosity_warning()
         transformers.utils.logging.set_verbosity_info()

examples/pytorch/language-modeling/run_clm_no_trainer.py

@@ -39,6 +39,7 @@
 
 import transformers
 from accelerate import Accelerator, DistributedType
+from accelerate.logging import get_logger
 from accelerate.utils import set_seed
 from huggingface_hub import Repository
 from transformers import (
@@ -56,7 +57,7 @@
 from transformers.utils.versions import require_version
 
 
-logger = logging.getLogger(__name__)
+logger = get_logger(__name__)
 
 require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/language-modeling/requirements.txt")
 
@@ -234,11 +235,7 @@ def main():
         datefmt="%m/%d/%Y %H:%M:%S",
         level=logging.INFO,
     )
-    logger.info(accelerator.state)
-
-    # Setup logging, we only want one process per machine to log things on the screen.
-    # accelerator.is_local_main_process is only True for one process per machine.
-    logger.setLevel(logging.INFO if accelerator.is_local_main_process else logging.ERROR)
+    logger.info(accelerator.state, main_process_only=False)
     if accelerator.is_local_main_process:
         datasets.utils.logging.set_verbosity_warning()
         transformers.utils.logging.set_verbosity_info()

examples/pytorch/language-modeling/run_mlm_no_trainer.py

@@ -39,6 +39,7 @@
 
 import transformers
 from accelerate import Accelerator, DistributedType
+from accelerate.logging import get_logger
 from accelerate.utils import set_seed
 from huggingface_hub import Repository
 from transformers import (
@@ -56,7 +57,7 @@
 from transformers.utils.versions import require_version
 
 
-logger = logging.getLogger(__name__)
+logger = get_logger(__name__)
 require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/language-modeling/requirements.txt")
 MODEL_CONFIG_CLASSES = list(MODEL_MAPPING.keys())
 MODEL_TYPES = tuple(conf.model_type for conf in MODEL_CONFIG_CLASSES)
@@ -245,11 +246,7 @@ def main():
         datefmt="%m/%d/%Y %H:%M:%S",
         level=logging.INFO,
     )
-    logger.info(accelerator.state)
-
-    # Setup logging, we only want one process per machine to log things on the screen.
-    # accelerator.is_local_main_process is only True for one process per machine.
-    logger.setLevel(logging.INFO if accelerator.is_local_main_process else logging.ERROR)
+    logger.info(accelerator.state, main_process_only=False)
     if accelerator.is_local_main_process:
         datasets.utils.logging.set_verbosity_warning()
         transformers.utils.logging.set_verbosity_info()

examples/pytorch/multiple-choice/run_swag_no_trainer.py

@@ -37,6 +37,7 @@
 
 import transformers
 from accelerate import Accelerator
+from accelerate.logging import get_logger
 from accelerate.utils import set_seed
 from huggingface_hub import Repository
 from transformers import (
@@ -54,7 +55,7 @@
 from transformers.utils import PaddingStrategy, get_full_repo_name
 
 
-logger = logging.getLogger(__name__)
+logger = get_logger(__name__)
 # You should update this to your particular problem to have better documentation of `model_type`
 MODEL_CONFIG_CLASSES = list(MODEL_MAPPING.keys())
 MODEL_TYPES = tuple(conf.model_type for conf in MODEL_CONFIG_CLASSES)
@@ -272,11 +273,7 @@ def main():
         datefmt="%m/%d/%Y %H:%M:%S",
         level=logging.INFO,
     )
-    logger.info(accelerator.state)
-
-    # Setup logging, we only want one process per machine to log things on the screen.
-    # accelerator.is_local_main_process is only True for one process per machine.
-    logger.setLevel(logging.INFO if accelerator.is_local_main_process else logging.ERROR)
+    logger.info(accelerator.state, main_process_only=False)
     if accelerator.is_local_main_process:
         datasets.utils.logging.set_verbosity_warning()
         transformers.utils.logging.set_verbosity_info()

examples/pytorch/question-answering/run_qa_beam_search_no_trainer.py

@@ -35,6 +35,7 @@
 
 import transformers
 from accelerate import Accelerator
+from accelerate.logging import get_logger
 from accelerate.utils import set_seed
 from huggingface_hub import Repository
 from transformers import (
@@ -58,7 +59,7 @@
 
 require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/question-answering/requirements.txt")
 
-logger = logging.getLogger(__name__)
+logger = get_logger(__name__)
 
 
 def save_prefixed_metrics(results, output_dir, file_name: str = "all_results.json", metric_key_prefix: str = "eval"):
@@ -289,11 +290,7 @@ def main():
         datefmt="%m/%d/%Y %H:%M:%S",
         level=logging.INFO,
     )
-    logger.info(accelerator.state)
-
-    # Setup logging, we only want one process per machine to log things on the screen.
-    # accelerator.is_local_main_process is only True for one process per machine.
-    logger.setLevel(logging.INFO if accelerator.is_local_main_process else logging.ERROR)
+    logger.info(accelerator.state, main_process_only=False)
     if accelerator.is_local_main_process:
         datasets.utils.logging.set_verbosity_warning()
         transformers.utils.logging.set_verbosity_info()

examples/pytorch/question-answering/run_qa_no_trainer.py

@@ -35,6 +35,7 @@
 
 import transformers
 from accelerate import Accelerator
+from accelerate.logging import get_logger
 from accelerate.utils import set_seed
 from huggingface_hub import Repository
 from transformers import (
@@ -60,7 +61,7 @@
 
 require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/question-answering/requirements.txt")
 
-logger = logging.getLogger(__name__)
+logger = get_logger(__name__)
 # You should update this to your particular problem to have better documentation of `model_type`
 MODEL_CONFIG_CLASSES = list(MODEL_MAPPING.keys())
 MODEL_TYPES = tuple(conf.model_type for conf in MODEL_CONFIG_CLASSES)
@@ -318,11 +319,7 @@ def main():
         datefmt="%m/%d/%Y %H:%M:%S",
         level=logging.INFO,
     )
-    logger.info(accelerator.state)
-
-    # Setup logging, we only want one process per machine to log things on the screen.
-    # accelerator.is_local_main_process is only True for one process per machine.
-    logger.setLevel(logging.INFO if accelerator.is_local_main_process else logging.ERROR)
+    logger.info(accelerator.state, main_process_only=False)
     if accelerator.is_local_main_process:
         datasets.utils.logging.set_verbosity_warning()
         transformers.utils.logging.set_verbosity_info()

examples/pytorch/semantic-segmentation/run_semantic_segmentation_no_trainer.py

@@ -16,7 +16,6 @@
 
 import argparse
 import json
-import logging
 import math
 import os
 import random
@@ -34,6 +33,7 @@
 
 import transformers
 from accelerate import Accelerator
+from accelerate.logging import get_logger
 from accelerate.utils import set_seed
 from huggingface_hub import Repository, hf_hub_download
 from transformers import (
@@ -48,7 +48,7 @@
 from transformers.utils.versions import require_version
 
 
-logger = logging.getLogger(__name__)
+logger = get_logger(__name__)
 
 require_version("datasets>=2.0.0", "To fix: pip install -r examples/pytorch/semantic-segmentation/requirements.txt")
 
@@ -308,11 +308,7 @@ def main():
     # Initialize the accelerator. We will let the accelerator handle device placement for us in this example.
     # If we're using tracking, we also need to initialize it here and it will pick up all supported trackers in the environment
     accelerator = Accelerator(log_with="all", logging_dir=args.output_dir) if args.with_tracking else Accelerator()
-    logger.info(accelerator.state)
-
-    # Setup logging, we only want one process per machine to log things on the screen.
-    # accelerator.is_local_main_process is only True for one process per machine.
-    logger.setLevel(logging.INFO if accelerator.is_local_main_process else logging.ERROR)
+    logger.info(accelerator.state, main_process_only=False)
     if accelerator.is_local_main_process:
         datasets.utils.logging.set_verbosity_warning()
         transformers.utils.logging.set_verbosity_info()

examples/pytorch/speech-pretraining/run_wav2vec2_pretraining_no_trainer.py

@@ -16,7 +16,6 @@
 """ Pre-Training a 🤗 Wav2Vec2 model on unlabeled audio data """
 
 import argparse
-import logging
 import math
 import os
 from dataclasses import dataclass
@@ -31,6 +30,7 @@
 
 import transformers
 from accelerate import Accelerator
+from accelerate.logging import get_logger
 from huggingface_hub import Repository
 from transformers import (
     AdamW,
@@ -46,7 +46,7 @@
 from transformers.utils import get_full_repo_name
 
 
-logger = logging.getLogger(__name__)
+logger = get_logger(__name__)
 
 
 def parse_args():
@@ -362,11 +362,7 @@ def main():
 
     # Initialize the accelerator. We will let the accelerator handle device placement for us in this example.
     accelerator = Accelerator()
-    logger.info(accelerator.state)
-
-    # Setup logging, we only want one process per machine to log things on the screen.
-    # accelerator.is_local_main_process is only True for one process per machine.
-    logger.setLevel(logging.INFO if accelerator.is_local_main_process else logging.ERROR)
+    logger.info(accelerator.state, main_process_only=False)
     if accelerator.is_local_main_process:
         datasets.utils.logging.set_verbosity_warning()
         transformers.utils.logging.set_verbosity_info()

examples/pytorch/summarization/run_summarization_no_trainer.py

@@ -36,6 +36,7 @@
 
 import transformers
 from accelerate import Accelerator
+from accelerate.logging import get_logger
 from accelerate.utils import set_seed
 from filelock import FileLock
 from huggingface_hub import Repository
@@ -54,7 +55,7 @@
 from transformers.utils.versions import require_version
 
 
-logger = logging.getLogger(__name__)
+logger = get_logger(__name__)
 require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/summarization/requirements.txt")
 
 # You should update this to your particular problem to have better documentation of `model_type`
@@ -322,11 +323,7 @@ def main():
         datefmt="%m/%d/%Y %H:%M:%S",
         level=logging.INFO,
     )
-    logger.info(accelerator.state)
-
-    # Setup logging, we only want one process per machine to log things on the screen.
-    # accelerator.is_local_main_process is only True for one process per machine.
-    logger.setLevel(logging.INFO if accelerator.is_local_main_process else logging.ERROR)
+    logger.info(accelerator.state, main_process_only=False)
     if accelerator.is_local_main_process:
         datasets.utils.logging.set_verbosity_warning()
         transformers.utils.logging.set_verbosity_info()