SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

remove check for main process for trackers initialization (#18706)

Browse Source
This commit is contained in:
Atharva Ingle
2022-08-22 20:46:27 +05:30
committed by GitHub
parent 0f257a8774
commit d90a36d192
9 changed files with 45 additions and 63 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
examples/pytorch/image-classification/run_image_classification_no_trainer.py
View File

@@ -414,14 +414,12 @@ def main():
checkpointing_steps = None checkpointing_steps = None
# We need to initialize the trackers we use, and also store our configuration. # We need to initialize the trackers we use, and also store our configuration.
# We initialize the trackers only on main process because `accelerator.log` # The trackers initializes automatically on the main process.
# only logs on main process and we don't want empty logs/runs on other processes.
if args.with_tracking: if args.with_tracking:
if accelerator.is_main_process: experiment_config = vars(args)
experiment_config = vars(args) # TensorBoard cannot log Enums, need the raw value
# TensorBoard cannot log Enums, need the raw value experiment_config["lr_scheduler_type"] = experiment_config["lr_scheduler_type"].value
experiment_config["lr_scheduler_type"] = experiment_config["lr_scheduler_type"].value accelerator.init_trackers("image_classification_no_trainer", experiment_config)
accelerator.init_trackers("image_classification_no_trainer", experiment_config)
# Get the metric function # Get the metric function
metric = evaluate.load("accuracy") metric = evaluate.load("accuracy")

12
examples/pytorch/language-modeling/run_clm_no_trainer.py
View File

@@ -516,14 +516,12 @@ def main():
checkpointing_steps = None checkpointing_steps = None
# We need to initialize the trackers we use, and also store our configuration. # We need to initialize the trackers we use, and also store our configuration.
# We initialize the trackers only on main process because `accelerator.log` # The trackers initializes automatically on the main process.
# only logs on main process and we don't want empty logs/runs on other processes.
if args.with_tracking: if args.with_tracking:
if accelerator.is_main_process: experiment_config = vars(args)
experiment_config = vars(args) # TensorBoard cannot log Enums, need the raw value
# TensorBoard cannot log Enums, need the raw value experiment_config["lr_scheduler_type"] = experiment_config["lr_scheduler_type"].value
experiment_config["lr_scheduler_type"] = experiment_config["lr_scheduler_type"].value accelerator.init_trackers("clm_no_trainer", experiment_config)
accelerator.init_trackers("clm_no_trainer", experiment_config)
# Train! # Train!
total_batch_size = args.per_device_train_batch_size * accelerator.num_processes * args.gradient_accumulation_steps total_batch_size = args.per_device_train_batch_size * accelerator.num_processes * args.gradient_accumulation_steps

12
examples/pytorch/language-modeling/run_mlm_no_trainer.py
View File

@@ -560,14 +560,12 @@ def main():
checkpointing_steps = None checkpointing_steps = None
# We need to initialize the trackers we use, and also store our configuration. # We need to initialize the trackers we use, and also store our configuration.
# We initialize the trackers only on main process because `accelerator.log` # The trackers initializes automatically on the main process.
# only logs on main process and we don't want empty logs/runs on other processes.
if args.with_tracking: if args.with_tracking:
if accelerator.is_main_process: experiment_config = vars(args)
experiment_config = vars(args) # TensorBoard cannot log Enums, need the raw value
# TensorBoard cannot log Enums, need the raw value experiment_config["lr_scheduler_type"] = experiment_config["lr_scheduler_type"].value
experiment_config["lr_scheduler_type"] = experiment_config["lr_scheduler_type"].value accelerator.init_trackers("mlm_no_trainer", experiment_config)
accelerator.init_trackers("mlm_no_trainer", experiment_config)
# Train! # Train!
total_batch_size = args.per_device_train_batch_size * accelerator.num_processes * args.gradient_accumulation_steps total_batch_size = args.per_device_train_batch_size * accelerator.num_processes * args.gradient_accumulation_steps

12
examples/pytorch/multiple-choice/run_swag_no_trainer.py
View File

@@ -513,14 +513,12 @@ def main():
checkpointing_steps = None checkpointing_steps = None
# We need to initialize the trackers we use, and also store our configuration. # We need to initialize the trackers we use, and also store our configuration.
# We initialize the trackers only on main process because `accelerator.log` # The trackers initializes automatically on the main process.
# only logs on main process and we don't want empty logs/runs on other processes.
if args.with_tracking: if args.with_tracking:
if accelerator.is_main_process: experiment_config = vars(args)
experiment_config = vars(args) # TensorBoard cannot log Enums, need the raw value
# TensorBoard cannot log Enums, need the raw value experiment_config["lr_scheduler_type"] = experiment_config["lr_scheduler_type"].value
experiment_config["lr_scheduler_type"] = experiment_config["lr_scheduler_type"].value accelerator.init_trackers("swag_no_trainer", experiment_config)
accelerator.init_trackers("swag_no_trainer", experiment_config)
# Metrics # Metrics
metric = evaluate.load("accuracy") metric = evaluate.load("accuracy")

12
examples/pytorch/question-answering/run_qa_no_trainer.py
View File

@@ -787,14 +787,12 @@ def main():
checkpointing_steps = None checkpointing_steps = None
# We need to initialize the trackers we use, and also store our configuration. # We need to initialize the trackers we use, and also store our configuration.
# We initialize the trackers only on main process because `accelerator.log` # The trackers initializes automatically on the main process.
# only logs on main process and we don't want empty logs/runs on other processes.
if args.with_tracking: if args.with_tracking:
if accelerator.is_main_process: experiment_config = vars(args)
experiment_config = vars(args) # TensorBoard cannot log Enums, need the raw value
# TensorBoard cannot log Enums, need the raw value experiment_config["lr_scheduler_type"] = experiment_config["lr_scheduler_type"].value
experiment_config["lr_scheduler_type"] = experiment_config["lr_scheduler_type"].value accelerator.init_trackers("qa_no_trainer", experiment_config)
accelerator.init_trackers("qa_no_trainer", experiment_config)
# Train! # Train!
total_batch_size = args.per_device_train_batch_size * accelerator.num_processes * args.gradient_accumulation_steps total_batch_size = args.per_device_train_batch_size * accelerator.num_processes * args.gradient_accumulation_steps

12
examples/pytorch/semantic-segmentation/run_semantic_segmentation_no_trainer.py
View File

@@ -512,14 +512,12 @@ def main():
metric = evaluate.load("mean_iou") metric = evaluate.load("mean_iou")
# We need to initialize the trackers we use, and also store our configuration. # We need to initialize the trackers we use, and also store our configuration.
# We initialize the trackers only on main process because `accelerator.log` # The trackers initializes automatically on the main process.
# only logs on main process and we don't want empty logs/runs on other processes.
if args.with_tracking: if args.with_tracking:
if accelerator.is_main_process: experiment_config = vars(args)
experiment_config = vars(args) # TensorBoard cannot log Enums, need the raw value
# TensorBoard cannot log Enums, need the raw value experiment_config["lr_scheduler_type"] = experiment_config["lr_scheduler_type"].value
experiment_config["lr_scheduler_type"] = experiment_config["lr_scheduler_type"].value accelerator.init_trackers("semantic_segmentation_no_trainer", experiment_config)
accelerator.init_trackers("semantic_segmentation_no_trainer", experiment_config)
# Train! # Train!
total_batch_size = args.per_device_train_batch_size * accelerator.num_processes * args.gradient_accumulation_steps total_batch_size = args.per_device_train_batch_size * accelerator.num_processes * args.gradient_accumulation_steps

12
examples/pytorch/summarization/run_summarization_no_trainer.py
View File

@@ -581,14 +581,12 @@ def main():
checkpointing_steps = None checkpointing_steps = None
# We need to initialize the trackers we use, and also store our configuration. # We need to initialize the trackers we use, and also store our configuration.
# We initialize the trackers only on main process because `accelerator.log` # The trackers initializes automatically on the main process.
# only logs on main process and we don't want empty logs/runs on other processes.
if args.with_tracking: if args.with_tracking:
if accelerator.is_main_process: experiment_config = vars(args)
experiment_config = vars(args) # TensorBoard cannot log Enums, need the raw value
# TensorBoard cannot log Enums, need the raw value experiment_config["lr_scheduler_type"] = experiment_config["lr_scheduler_type"].value
experiment_config["lr_scheduler_type"] = experiment_config["lr_scheduler_type"].value accelerator.init_trackers("summarization_no_trainer", experiment_config)
accelerator.init_trackers("summarization_no_trainer", experiment_config)
# Metric # Metric
metric = evaluate.load("rouge") metric = evaluate.load("rouge")

12
examples/pytorch/text-classification/run_glue_no_trainer.py
View File

@@ -459,14 +459,12 @@ def main():
checkpointing_steps = None checkpointing_steps = None
# We need to initialize the trackers we use, and also store our configuration. # We need to initialize the trackers we use, and also store our configuration.
# We initialize the trackers only on main process because `accelerator.log` # The trackers initializes automatically on the main process.
# only logs on main process and we don't want empty logs/runs on other processes.
if args.with_tracking: if args.with_tracking:
if accelerator.is_main_process: experiment_config = vars(args)
experiment_config = vars(args) # TensorBoard cannot log Enums, need the raw value
# TensorBoard cannot log Enums, need the raw value experiment_config["lr_scheduler_type"] = experiment_config["lr_scheduler_type"].value
experiment_config["lr_scheduler_type"] = experiment_config["lr_scheduler_type"].value accelerator.init_trackers("glue_no_trainer", experiment_config)
accelerator.init_trackers("glue_no_trainer", experiment_config)
# Get the metric function # Get the metric function
if args.task_name is not None: if args.task_name is not None:

12
examples/pytorch/token-classification/run_ner_no_trainer.py
View File

@@ -574,14 +574,12 @@ def main():
checkpointing_steps = None checkpointing_steps = None
# We need to initialize the trackers we use, and also store our configuration. # We need to initialize the trackers we use, and also store our configuration.
# We initialize the trackers only on main process because `accelerator.log` # The trackers initializes automatically on the main process.
# only logs on main process and we don't want empty logs/runs on other processes.
if args.with_tracking: if args.with_tracking:
if accelerator.is_main_process: experiment_config = vars(args)
experiment_config = vars(args) # TensorBoard cannot log Enums, need the raw value
# TensorBoard cannot log Enums, need the raw value experiment_config["lr_scheduler_type"] = experiment_config["lr_scheduler_type"].value
experiment_config["lr_scheduler_type"] = experiment_config["lr_scheduler_type"].value accelerator.init_trackers("ner_no_trainer", experiment_config)
accelerator.init_trackers("ner_no_trainer", experiment_config)
# Metrics # Metrics
metric = evaluate.load("seqeval") metric = evaluate.load("seqeval")