Add examples telemetry (#17552)

* Add examples telemetry

* Alternative approach

* Add to all other examples

* Add to templates as well

* Put framework separately

* Same for TensorFlow

Author: sgugger

examples/flax/image-captioning/run_image_captioning_flax.py

@@ -52,7 +52,7 @@
     HfArgumentParser,
     is_tensorboard_available,
 )
-from transformers.utils import get_full_repo_name, is_offline_mode
+from transformers.utils import get_full_repo_name, is_offline_mode, send_example_telemetry
 
 
 logger = logging.getLogger(__name__)
@@ -388,6 +388,10 @@ def main():
     else:
         model_args, data_args, training_args = parser.parse_args_into_dataclasses()
 
+    # Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
+    # information sent is the one passed as arguments along with your Python/PyTorch versions.
+    send_example_telemetry("run_image_captioning", model_args, data_args, framework="flax")
+
     if (
         os.path.exists(training_args.output_dir)
         and os.listdir(training_args.output_dir)

examples/flax/language-modeling/run_clm_flax.py

@@ -58,7 +58,7 @@
     set_seed,
 )
 from transformers.testing_utils import CaptureLogger
-from transformers.utils import get_full_repo_name
+from transformers.utils import get_full_repo_name, send_example_telemetry
 
 
 logger = logging.getLogger(__name__)
@@ -328,6 +328,10 @@ def main():
     else:
         model_args, data_args, training_args = parser.parse_args_into_dataclasses()
 
+    # Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
+    # information sent is the one passed as arguments along with your Python/PyTorch versions.
+    send_example_telemetry("run_clm", model_args, data_args, framework="flax")
+
     if (
         os.path.exists(training_args.output_dir)
         and os.listdir(training_args.output_dir)

examples/flax/language-modeling/run_mlm_flax.py

@@ -58,7 +58,7 @@
     is_tensorboard_available,
     set_seed,
 )
-from transformers.utils import get_full_repo_name
+from transformers.utils import get_full_repo_name, send_example_telemetry
 
 
 MODEL_CONFIG_CLASSES = list(FLAX_MODEL_FOR_MASKED_LM_MAPPING.keys())
@@ -365,6 +365,10 @@ def main():
     else:
         model_args, data_args, training_args = parser.parse_args_into_dataclasses()
 
+    # Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
+    # information sent is the one passed as arguments along with your Python/PyTorch versions.
+    send_example_telemetry("run_mlm", model_args, data_args, framework="flax")
+
     if (
         os.path.exists(training_args.output_dir)
         and os.listdir(training_args.output_dir)

examples/flax/language-modeling/run_t5_mlm_flax.py

@@ -57,7 +57,7 @@
     set_seed,
 )
 from transformers.models.t5.modeling_flax_t5 import shift_tokens_right
-from transformers.utils import get_full_repo_name
+from transformers.utils import get_full_repo_name, send_example_telemetry
 
 
 MODEL_CONFIG_CLASSES = list(FLAX_MODEL_FOR_MASKED_LM_MAPPING.keys())
@@ -498,6 +498,10 @@ def main():
     else:
         model_args, data_args, training_args = parser.parse_args_into_dataclasses()
 
+    # Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
+    # information sent is the one passed as arguments along with your Python/PyTorch versions.
+    send_example_telemetry("run_t5_mlm", model_args, data_args, framework="flax")
+
     if (
         os.path.exists(training_args.output_dir)
         and os.listdir(training_args.output_dir)

examples/flax/question-answering/run_qa.py

@@ -53,7 +53,7 @@
     PreTrainedTokenizerFast,
     is_tensorboard_available,
 )
-from transformers.utils import check_min_version, get_full_repo_name
+from transformers.utils import check_min_version, get_full_repo_name, send_example_telemetry
 from utils_qa import postprocess_qa_predictions
 
 
@@ -424,6 +424,10 @@ def main():
         model_args, data_args, training_args = parser.parse_json_file(json_file=os.path.abspath(sys.argv[1]))
     else:
         model_args, data_args, training_args = parser.parse_args_into_dataclasses()
+
+    # Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
+    # information sent is the one passed as arguments along with your Python/PyTorch versions.
+    send_example_telemetry("run_qa", model_args, data_args, framework="flax")
     # endregion
 
     # region Logging

examples/flax/summarization/run_summarization_flax.py

@@ -54,7 +54,7 @@
     HfArgumentParser,
     is_tensorboard_available,
 )
-from transformers.utils import get_full_repo_name, is_offline_mode
+from transformers.utils import get_full_repo_name, is_offline_mode, send_example_telemetry
 
 
 logger = logging.getLogger(__name__)
@@ -399,6 +399,10 @@ def main():
     else:
         model_args, data_args, training_args = parser.parse_args_into_dataclasses()
 
+    # Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
+    # information sent is the one passed as arguments along with your Python/PyTorch versions.
+    send_example_telemetry("run_summarization", model_args, data_args, framework="flax")
+
     if (
         os.path.exists(training_args.output_dir)
         and os.listdir(training_args.output_dir)

examples/flax/text-classification/run_flax_glue.py

@@ -48,7 +48,7 @@
     TrainingArguments,
     is_tensorboard_available,
 )
-from transformers.utils import check_min_version, get_full_repo_name
+from transformers.utils import check_min_version, get_full_repo_name, send_example_telemetry
 
 
 logger = logging.getLogger(__name__)
@@ -308,6 +308,10 @@ def main():
     else:
         model_args, data_args, training_args = parser.parse_args_into_dataclasses()
 
+    # Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
+    # information sent is the one passed as arguments along with your Python/PyTorch versions.
+    send_example_telemetry("run_glue", model_args, data_args, framework="flax")
+
     # Make one log on every process with the configuration for debugging.
     logging.basicConfig(
         format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",

examples/flax/token-classification/run_flax_ner.py

@@ -47,7 +47,7 @@
     HfArgumentParser,
     is_tensorboard_available,
 )
-from transformers.utils import check_min_version, get_full_repo_name
+from transformers.utils import check_min_version, get_full_repo_name, send_example_telemetry
 from transformers.utils.versions import require_version
 
 
@@ -366,6 +366,10 @@ def main():
     else:
         model_args, data_args, training_args = parser.parse_args_into_dataclasses()
 
+    # Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
+    # information sent is the one passed as arguments along with your Python/PyTorch versions.
+    send_example_telemetry("run_ner", model_args, data_args, framework="flax")
+
     # Make one log on every process with the configuration for debugging.
     logging.basicConfig(
         format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",

examples/flax/vision/run_image_classification.py

@@ -53,7 +53,7 @@
     is_tensorboard_available,
     set_seed,
 )
-from transformers.utils import get_full_repo_name
+from transformers.utils import get_full_repo_name, send_example_telemetry
 
 
 logger = logging.getLogger(__name__)
@@ -256,6 +256,10 @@ def main():
     else:
         model_args, data_args, training_args = parser.parse_args_into_dataclasses()
 
+    # Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
+    # information sent is the one passed as arguments along with your Python/PyTorch versions.
+    send_example_telemetry("run_image_classification", model_args, data_args, framework="flax")
+
     if (
         os.path.exists(training_args.output_dir)
         and os.listdir(training_args.output_dir)

examples/pytorch/audio-classification/run_audio_classification.py

@@ -37,7 +37,7 @@
     set_seed,
 )
 from transformers.trainer_utils import get_last_checkpoint
-from transformers.utils import check_min_version
+from transformers.utils import check_min_version, send_example_telemetry
 from transformers.utils.versions import require_version
 
 
@@ -197,6 +197,10 @@ def main():
     else:
         model_args, data_args, training_args = parser.parse_args_into_dataclasses()
 
+    # Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
+    # information sent is the one passed as arguments along with your Python/PyTorch versions.
+    send_example_telemetry("run_audio_classification", model_args, data_args)
+
     # Setup logging
     logging.basicConfig(
         format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",

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

@@ -47,7 +47,7 @@
     set_seed,
 )
 from transformers.trainer_utils import get_last_checkpoint
-from transformers.utils import check_min_version
+from transformers.utils import check_min_version, send_example_telemetry
 from transformers.utils.versions import require_version
 
 
@@ -233,6 +233,10 @@ def main():
     else:
         model_args, data_args, training_args = parser.parse_args_into_dataclasses()
 
+    # Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
+    # information sent is the one passed as arguments along with your Python/PyTorch versions.
+    send_example_telemetry("run_clip", model_args, data_args)
+
     # 2. Setup logging
     logging.basicConfig(
         format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",

examples/pytorch/image-classification/run_image_classification.py

@@ -45,7 +45,7 @@
     TrainingArguments,
 )
 from transformers.trainer_utils import get_last_checkpoint
-from transformers.utils import check_min_version
+from transformers.utils import check_min_version, send_example_telemetry
 from transformers.utils.versions import require_version
 
 
@@ -175,6 +175,10 @@ def main():
     else:
         model_args, data_args, training_args = parser.parse_args_into_dataclasses()
 
+    # Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
+    # information sent is the one passed as arguments along with your Python/PyTorch versions.
+    send_example_telemetry("run_image_classification", model_args, data_args)
+
     # Setup logging
     logging.basicConfig(
         format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",

examples/pytorch/image-classification/run_image_classification_no_trainer.py

@@ -47,7 +47,7 @@
     SchedulerType,
     get_scheduler,
 )
-from transformers.utils import get_full_repo_name
+from transformers.utils import get_full_repo_name, send_example_telemetry
 from transformers.utils.versions import require_version
 
 
@@ -201,6 +201,10 @@ def parse_args():
 def main():
     args = parse_args()
 
+    # Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
+    # information sent is the one passed as arguments along with your Python/PyTorch versions.
+    send_example_telemetry("run_image_classification_no_trainer", args)
+
     # 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 by default pick up all supported trackers
     # in the environment

examples/pytorch/image-pretraining/run_mae.py

@@ -34,7 +34,7 @@
     ViTMAEForPreTraining,
 )
 from transformers.trainer_utils import get_last_checkpoint
-from transformers.utils import check_min_version
+from transformers.utils import check_min_version, send_example_telemetry
 from transformers.utils.versions import require_version
 
 
@@ -175,6 +175,10 @@ def main():
     else:
         model_args, data_args, training_args = parser.parse_args_into_dataclasses()
 
+    # Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
+    # information sent is the one passed as arguments along with your Python/PyTorch versions.
+    send_example_telemetry("run_mae", model_args, data_args)
+
     # Setup logging
     logging.basicConfig(
         format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",

examples/pytorch/image-pretraining/run_mim.py

@@ -37,7 +37,7 @@
     TrainingArguments,
 )
 from transformers.trainer_utils import get_last_checkpoint
-from transformers.utils import check_min_version
+from transformers.utils import check_min_version, send_example_telemetry
 from transformers.utils.versions import require_version
 
 
@@ -239,6 +239,10 @@ def main():
     else:
         model_args, data_args, training_args = parser.parse_args_into_dataclasses()
 
+    # Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
+    # information sent is the one passed as arguments along with your Python/PyTorch versions.
+    send_example_telemetry("run_mim", model_args, data_args)
+
     # Setup logging
     logging.basicConfig(
         format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",

examples/pytorch/language-modeling/run_clm.py

@@ -48,7 +48,7 @@
 )
 from transformers.testing_utils import CaptureLogger
 from transformers.trainer_utils import get_last_checkpoint
-from transformers.utils import check_min_version
+from transformers.utils import check_min_version, send_example_telemetry
 from transformers.utils.versions import require_version
 
 
@@ -214,6 +214,10 @@ def main():
     else:
         model_args, data_args, training_args = parser.parse_args_into_dataclasses()
 
+    # Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
+    # information sent is the one passed as arguments along with your Python/PyTorch versions.
+    send_example_telemetry("run_clm", model_args, data_args)
+
     # Setup logging
     logging.basicConfig(
         format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",

examples/pytorch/language-modeling/run_clm_no_trainer.py

@@ -52,7 +52,7 @@
     default_data_collator,
     get_scheduler,
 )
-from transformers.utils import get_full_repo_name
+from transformers.utils import get_full_repo_name, send_example_telemetry
 from transformers.utils.versions import require_version
 
 
@@ -239,6 +239,10 @@ def parse_args():
 def main():
     args = parse_args()
 
+    # Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
+    # information sent is the one passed as arguments along with your Python/PyTorch versions.
+    send_example_telemetry("run_clm_no_trainer", args)
+
     # 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 by default pick up all supported trackers
     # in the environment

examples/pytorch/language-modeling/run_mlm.py

@@ -47,7 +47,7 @@
     set_seed,
 )
 from transformers.trainer_utils import get_last_checkpoint
-from transformers.utils import check_min_version
+from transformers.utils import check_min_version, send_example_telemetry
 from transformers.utils.versions import require_version
 
 
@@ -224,6 +224,10 @@ def main():
     else:
         model_args, data_args, training_args = parser.parse_args_into_dataclasses()
 
+    # Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
+    # information sent is the one passed as arguments along with your Python/PyTorch versions.
+    send_example_telemetry("run_mlm", model_args, data_args)
+
     # Setup logging
     logging.basicConfig(
         format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",

examples/pytorch/language-modeling/run_mlm_no_trainer.py

@@ -52,7 +52,7 @@
     SchedulerType,
     get_scheduler,
 )
-from transformers.utils import get_full_repo_name
+from transformers.utils import get_full_repo_name, send_example_telemetry
 from transformers.utils.versions import require_version
 
 
@@ -248,6 +248,10 @@ def parse_args():
 def main():
     args = parse_args()
 
+    # Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
+    # information sent is the one passed as arguments along with your Python/PyTorch versions.
+    send_example_telemetry("run_mlm_no_trainer", args)
+
     # 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 by default pick up all supported trackers
     # in the environment

examples/pytorch/language-modeling/run_plm.py

@@ -42,7 +42,7 @@
     set_seed,
 )
 from transformers.trainer_utils import get_last_checkpoint
-from transformers.utils import check_min_version
+from transformers.utils import check_min_version, send_example_telemetry
 from transformers.utils.versions import require_version
 
 
@@ -220,6 +220,10 @@ def main():
     else:
         model_args, data_args, training_args = parser.parse_args_into_dataclasses()
 
+    # Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
+    # information sent is the one passed as arguments along with your Python/PyTorch versions.
+    send_example_telemetry("run_plm", model_args, data_args)
+
     # Setup logging
     logging.basicConfig(
         format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",