Fix: Removal of transformers logger and addition of python native logger by Abhishek-TAMU · Pull Request #270 · foundation-model-stack/fms-hf-tuning

Abhishek-TAMU · 2024-07-30T03:02:19Z

Description of the change

Aim was to move away from using logger modules from transformers and use python native logger. Now, log level is either set by CLI argument of --log_level OR environment variable TRANSFORMER_VERBOSITY OR LOG_LEVEL. The setting of log level applies to both python native logger and log level of transformers logging.

Related issue number

Closes #243
Closes #195
Closes #194
Closes #215

#994

How to verify the PR

Set environment variable LOG_LEVEL OR TRANSFORMER_VERBOSITY as desired log level and check if logs are correct according to passed value in env variable, by running tuning job for a small model like Maykeye/TinyLLama-v0.

Command:

python tuning/sft_trainer.py  \                                             
--model_name_or_path Maykeye/TinyLLama-v0  \
--training_data_path tests/data/twitter_complaints_small.json  \
--output_dir outputs/full-tuning  \
--num_train_epochs 5  \
--per_device_train_batch_size 2  \
--gradient_accumulation_steps 1  \
--learning_rate 1e-5  \
--response_template "\n### Label:"  \
--dataset_text_field "output" \
--use_flash_attn false \
--torch_dtype "float32"

Pass log_level via cli argument, and check if logs are correct according to passed LOG_LEVEL, by running tuning job for a small model like Maykeye/TinyLLama-v0.

Command:

python tuning/sft_trainer.py  \                                             
--model_name_or_path Maykeye/TinyLLama-v0  \
--training_data_path tests/data/twitter_complaints_small.json  \
--output_dir outputs/full-tuning  \
--num_train_epochs 5  \
--per_device_train_batch_size 2  \
--gradient_accumulation_steps 1  \
--learning_rate 1e-5  \
--response_template "\n### Label:"  \
--dataset_text_field "output" \
--use_flash_attn false \
--torch_dtype "float32" \
--log_level "info"

Was the PR tested

I have added >=1 unit test(s) for every new method I have added.
I have ensured all unit tests pass

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

…thon logger Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

…del-stack#258) * initial code for deleting lm_head Signed-off-by: Anh-Uong <anh.uong@ibm.com> * fix logic for copying checkpoint Signed-off-by: Anh-Uong <anh.uong@ibm.com> * fix check that embed_tokens and lm_head weights are the same Signed-off-by: Anh-Uong <anh.uong@ibm.com> * fix warning assertion Signed-off-by: Anh-Uong <anh.uong@ibm.com> * fix lm_head check, remove test Signed-off-by: Anh-Uong <anh.uong@ibm.com> * small fixes from code review Signed-off-by: Anh-Uong <anh.uong@ibm.com> * fmt Signed-off-by: Anh-Uong <anh.uong@ibm.com> --------- Signed-off-by: Anh-Uong <anh.uong@ibm.com> Co-authored-by: Anh-Uong <anh.uong@ibm.com> Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

anhuong

Thanks Abhishek for the logging update, I think this will help a lot. A few questions on how the logging works and the desired logic

anhuong · 2024-08-01T22:23:15Z

build/accelerate_launch.py

            )

+        # Configure log_level of python native logger.
+        LOGLEVEL = None


nit: I had used an uppercase variable for this since it was pulling from the environment vars. But uppercase variables are usually used for constants, and since this isn't one, we can change it to a lowercase log_level

Thank you for the PR review. This is done!

anhuong · 2024-08-01T22:28:17Z

build/accelerate_launch.py

+        if "log_level" in job_config and job_config["log_level"]:
+            LOGLEVEL = job_config["log_level"].upper()
+            logging.basicConfig(level=LOGLEVEL)
+        else:
+            LOGLEVEL = os.environ.get("LOG_LEVEL", "WARNING").upper()
+            logging.basicConfig(level=LOGLEVEL)


small refactor you can do:

log_level = job_config.get("log_level") # this will be set to either the value found or None if not log_level: # if log level not set by job_config aka by JSON, set it via env var or set default log_level = os.environ.get("LOG_LEVEL", "WARNING") log_level = log_level.upper() logging.basicConfig(level=log_level) os.environ["LOG_LEVEL"] = log_level # nice to move this up so all the parts related to log level are together

This is done! Thank you for the input!

anhuong · 2024-08-02T01:01:53Z

tuning/config/configs.py

+        default="passive",
+        metadata={
+            "help": "The log level to adopt during training. \
+            Possible values are 'debug', 'info', 'warning', 'error' and 'critical'"


Add note on what "passive" value means

plus a ‘passive’ level which doesn’t set anything and keeps the current log level for the Transformers library (which will be "warning" by default)

Added note in metadata!

anhuong · 2024-08-02T01:11:58Z

tuning/sft_trainer.py

+        # Assign value of either env var LOG_LEVEL or warning
+        LOGLEVEL = os.environ.get("LOG_LEVEL", "WARNING").upper()
+        logging.basicConfig(level=LOGLEVEL)
+        train_logger = logging.getLogger()


Since you call this on line 497, you don't need to call it within this block as well

This is done!

anhuong · 2024-08-02T01:28:10Z

tuning/sft_trainer.py

-        logger.debug(
+
+        # Function to set log level for python native logger and transformers training logger
+        training_args, _ = set_log_level(training_args)


Why are you not using the logging.getLogger() that you returned? Since you are using the root logger, running

logging.warning("Warning") # above is the same as below root_logger = logging.getLogger() root_logger.warning("Warning")

Running the above is the same. So if you are not using the getLogger, you shouldn't return it/use it.

But it looks like it is best practice to use getLogger from python docs and python logging howto.
So then instead of running logging.warning() you'd run logger.warning(). In which in the train() you would also call logging.getLogger(). It looks like its best practice to create module level loggers. So you could do logging.getLogger("sft_trainer_main") and logging.getLogger("sft_trainer_train")` and then the log messages should show like:

# named logger if you do getLogger("sft_trainer_train") WARNING:sft_trainer_train:My warning # root logger if you do getLogger() WARNING:root:My warning

In returned logger on line number 497, just to get logger in unit test cases functions. returning logger won't make sense in same file as we can set log level in same py file with returning logger.

But ya that's a good idea to use logging.getLogger. I have made the changes.

anhuong · 2024-08-02T01:39:25Z

tuning/trackers/filelogging_tracker.py

-        self.logger = logging.get_logger("file_logging_tracker")
+        # Configure log level
+        LOGLEVEL = os.environ.get("LOG_LEVEL", "WARNING").upper()
+        logging.basicConfig(level=LOGLEVEL)


logging.basicConfig() sets the log level for the root logger. But you can set individual loggers level by running self.logger.setLevel(level)

There are a few of these to fix, not going to comment in each file

I don't think you even have to set log level for every file, you only have to set it once in the sft_trainer.py after parsing user's input right?

In each file, you just need to get a logger by doing:

import logging logger = logging.getLogger(__name__)

and every time we log something, we just use that logger:

logger.info("Log this info")

@aluu317 Yes I think it makes sense. But I am not sure if all the other files other than sfttrainer.py are used standalone where they need to set their own log level. Maybe @anhuong can comment better.

I agree that having each file set their own logging level helps to ensure that the logging level is set if a file is called outside of sft_trainer.py. I do agree that using logging.getLogger(__name__) is best practice and it is unlikely these files will be called outside of sft_trainer.py so I think going with Angel's suggestion works

anhuong · 2024-08-02T01:43:49Z

tuning/sft_trainer.py

+
+    # Check if env var TRANSFORMERS_VERBOSITY is not set.
+    # Else if env var is already set then, log level of transformers is automatically set.
+    if os.environ.get("TRANSFORMERS_VERBOSITY") is None:


nit: can do

Suggested change

if os.environ.get("TRANSFORMERS_VERBOSITY") is None:

if not os.environ.get("TRANSFORMERS_VERBOSITY"):

Although do we even need the code block below? What use case is it covering? Do we have to check that TRANFORMERS_VERBOSITY at all?

is this trying to do

if os.getenv("TRANFORMERS_VERBOSITY"): parsed_training_args.log_level = os.getenv("TRANFORMERS_VERBOSITY").lower()

I don't think we are required to check "TRANFORMERS_VERBOSITY". Hence I haven't added this code in final commit.

anhuong · 2024-08-02T01:58:50Z

tuning/sft_trainer.py

+    if parsed_training_args.log_level != "passive":
+        logging.basicConfig(level=parsed_training_args.log_level.upper())
+    else:
+        # Assign value of either env var LOG_LEVEL or warning
+        LOGLEVEL = os.environ.get("LOG_LEVEL", "WARNING").upper()
+        logging.basicConfig(level=LOGLEVEL)
+        train_logger = logging.getLogger()


Could this work to attain the logic we're looking for?
Goal for logic:

TRANSFORMERS_VERBOSITY ---> should only set transformers logging, this is different than setting --log_level

--log_level --> sets both python and transformers logging

For consistency with the flag, LOG_LEVEL also sets python and transformers logging

if parsed_training_args.log_level != "passive": logging.basicConfig(level=parsed_training_args.log_level.upper()) elif os.environ.get("LOG_LEVEL"): # Assign value of either env var LOG_LEVEL or warning LOGLEVEL = os.environ.get("LOG_LEVEL") logging.basicConfig(level=LOGLEVEL) # Set log_level in TrainingArguments parsed_training_args.log_level = LOGLEVEL.lower() train_logger = logging.getLogger()

I didn't add this line parsed_training_args.log_level = LOGLEVEL.lower() in the same if-else because if env variable 'TRANSFORMERS_VERBOSITY' is set, then this would get overwritten. Hence wrote this line in separate log below checking if not os.environ.get("TRANSFORMERS_VERBOSITY"):

Also I think in your code above in else logic we can add logging.basicConfig(level="WARNING") as default.
Thanks for the optimized code idea. Taking above points in consideration too I have optimized the code to work according to our logic. Feel free to comment on it.

log_level = None if parsed_training_args.log_level != "passive": log_level = parsed_training_args.log_level elif os.environ.get("LOG_LEVEL"): log_level = os.environ.get("LOG_LEVEL") parsed_training_args.log_level = ( log_level.lower() if not os.environ.get("TRANSFORMERS_VERBOSITY") else os.environ.get("TRANSFORMERS_VERBOSITY") ) else: log_level = "WARNING" logging.basicConfig(level=log_level.upper()) train_logger = logging.getLogger()

Let's chat about this logic to clarify -- update: done!

aluu317 · 2024-08-02T16:51:36Z

tuning/sft_trainer.py

+        logging.basicConfig(level=parsed_training_args.log_level.upper())
+    else:
+        # Assign value of either env var LOG_LEVEL or warning
+        LOGLEVEL = os.environ.get("LOG_LEVEL", "WARNING").upper()


Suggested change

LOGLEVEL = os.environ.get("LOG_LEVEL", "WARNING").upper()

LOGLEVEL = os.environ.get("LOG_LEVEL", logging.WARNING).upper()

Here if LOG_LEVEL is not set, it will default to logging.WARNING, which is an integer constant from the logging module (default is 30). The use of .upper() here will cause an error because .upper() is a string method and cannot be called on an integer.

aluu317 · 2024-08-02T17:01:38Z

tuning/trackers/filelogging_tracker.py

-        self.logger = logging.get_logger("file_logging_tracker")
+        # Configure log level
+        LOGLEVEL = os.environ.get("LOG_LEVEL", "WARNING").upper()
+        logging.basicConfig(level=LOGLEVEL)


It'd be helpful also if in each file, you could do:

logger = logging.getLogger(__name__)

This will specify in the logs which module the log came from, example:

INFO:__main__:Started INFO:mylib:Doing something INFO:__main__:Finished

Yea that's a good idea. I will do that.

aluu317 · 2024-08-02T17:04:16Z

tuning/trackers/filelogging_tracker.py

-        self.logger = logging.get_logger("file_logging_tracker")
+        # Configure log level
+        LOGLEVEL = os.environ.get("LOG_LEVEL", "WARNING").upper()
+        logging.basicConfig(level=LOGLEVEL)


I don't think you even have to set log level for every file, you only have to set it once in the sft_trainer.py after parsing user's input right?

In each file, you just need to get a logger by doing:

import logging logger = logging.getLogger(__name__)

and every time we log something, we just use that logger:

logger.info("Log this info")

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

anhuong · 2024-08-06T20:56:11Z

tests/test_sft_trainer.py

+    assert training_args.log_level == "info"
+
+
+@mock.patch.dict(os.environ, {}, clear=True)


Should include TRANSFORMERS_VERBOSITY to set the env and so you don't have to set it in the test and worry about removing it

Suggested change

@mock.patch.dict(os.environ, {}, clear=True)

@mock.patch.dict(os.environ, {"TRANSFORMERS_VERBOSITY": "info"}, clear=True)

anhuong · 2024-08-06T20:58:23Z

tests/test_sft_trainer.py

 # Local
 from tuning import sft_trainer
 from tuning.config import configs, peft_config
+from tuning.utils.logging import set_log_level


Since you moved the function that you are testing, you should also move the tests. They should be in /tests/utils/test_logging.py

anhuong · 2024-08-06T20:58:59Z

tests/test_sft_trainer.py

+    assert training_args.log_level == "error"
+
+
+@mock.patch.dict(os.environ, {}, clear=True)


similarly, want to include LOG_LEVEL here

Suggested change

@mock.patch.dict(os.environ, {}, clear=True)

@mock.patch.dict(os.environ, {"LOG_LEVEL": "info"}, clear=True)

anhuong · 2024-08-06T20:59:17Z

tests/test_sft_trainer.py

+    assert training_args.log_level == "passive"
+
+
+@mock.patch.dict(os.environ, {}, clear=True)


Want to include LOG_LEVEL here to setup env

Suggested change

@mock.patch.dict(os.environ, {}, clear=True)

@mock.patch.dict(os.environ, {"LOG_LEVEL": "info"}, clear=True)

anhuong · 2024-08-06T21:02:54Z

tuning/trainercontroller/controllermetrics/history_based_metrics.py

 from tuning.trainercontroller.controllermetrics.metricshandler import MetricHandler

-logger = logging.get_logger(__name__)
+# Configure log level


remove comment, no longer relevant

@anhuong Thanks for the quick review. I have made all changes mentioned above and pushed it again. Thank you!
As discussed, I can also wait for Dushyant's comment!

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

dushyantbehl · 2024-08-07T08:39:52Z

tests/utils/test_logging.py

+from tuning.config import configs
+from tuning.utils.logging import set_log_level
+
+TRAIN_ARGS = configs.TrainingArguments(


Since this is same as the TRAIN_ARGS in sft_trainer, you can import this directly from the main test file instead of redeclaration, please see -

fms-hf-tuning/tests/trackers/test_aim_tracker.py

Line 29 in d224be6

from tests.test_sft_trainer import (

Thank you for this input. This is done!

dushyantbehl · 2024-08-07T08:43:53Z

tuning/trackers/aimstack_tracker.py

        super().__init__(name="aim", tracker_config=tracker_config)
-        self.logger = logging.get_logger("aimstack_tracker")
+        # Get logger with root log level
+        self.logger = logging.getLogger()


@Abhishek-TAMU any particular reason to remove module name from the logger I had kept it to segregate logs from various subcomponents.

Same comment for file_logging_tracker

Explained in the comment above..no need for futher explanation.

Cool. Thank you!

I still dont quite understand this. You're setting log_level at runtime to restrict the amount of logs that a user wants to see (debug vs info etc.). But if you don't do logging.get_logger("aimstack_tracker") and so on in each file, all the logs will say they come from root, not module specific, helpful but not as helpful if the module is also included.
Maybe it'd be clearer if you could share an example run of what the logs look like if user sets debug or error for example and we can see what the logs look like.

dushyantbehl · 2024-08-07T08:46:22Z

tuning/utils/logging.py

@@ -0,0 +1,46 @@
+# Standard


Needs a code ownership declaration similar to this -

fms-hf-tuning/tuning/sft_trainer.py

Line 1 in d224be6

# Copyright The FMS HF Tuning Authors

This is done!

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

Abhishek-TAMU · 2024-08-07T14:56:42Z

@anhuong Thank you again for all the multiple review on this PR. I have pushed the final changes. Feel free to have look whenever possible. Thanks!

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

anhuong

Awesome work Abhishek! This will help a lot, many thanks!

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

aluu317

@Abhishek-TAMU Some comments inline. Thank you!

aluu317 · 2024-08-08T21:38:24Z

build/accelerate_launch.py

            or 'SFT_TRAINER_CONFIG_JSON_ENV_VAR'."
            )

+        # Configure log_level of python native logger.


Let's also add a comment here that CLI arg takes precedence over env var. And if neither is set, we use default "WARNING"

Thank you! This is done

aluu317 · 2024-08-08T21:40:23Z

tests/utils/test_logging.py

+def test_set_log_level_for_logger_with_env_var_and_cli():
+    """
+    Ensure that the correct log level is being set for python native logger and
+    transformers logger when env var LOG_LEVEL is used and CLI flag is passed


Can you add a quick comment that CLI arg takes precedence and we ignore env var LOG_LEVEL in this case?

This is done!

aluu317 · 2024-08-08T22:33:43Z

tuning/utils/logging.py

+def set_log_level(train_args, logger_name=None):
+    """Set log level of python native logger and TF logger via argument from CLI or env variable.
+
+    Args:


Could you add an arg for logger_name here?

This is done!

aluu317 · 2024-08-08T23:14:10Z

tuning/trackers/aimstack_tracker.py

        super().__init__(name="aim", tracker_config=tracker_config)
-        self.logger = logging.get_logger("aimstack_tracker")
+        # Get logger with root log level
+        self.logger = logging.getLogger()


I still dont quite understand this. You're setting log_level at runtime to restrict the amount of logs that a user wants to see (debug vs info etc.). But if you don't do logging.get_logger("aimstack_tracker") and so on in each file, all the logs will say they come from root, not module specific, helpful but not as helpful if the module is also included.
Maybe it'd be clearer if you could share an example run of what the logs look like if user sets debug or error for example and we can see what the logs look like.

Abhishek-TAMU · 2024-08-08T23:28:54Z

@aluu317 The reason behind removing logging.get_logger("aimstack_tracker") or logging.get_logger(__name__) was that it doesn't instantiate logger with log level equivalent to root log level and creates a new logger with default log level, whereas logging.getLogger() gets the logger which is equivalent to root log level.

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

Abhishek-TAMU · 2024-08-09T00:18:11Z

We have used just logging.getLogger() so that logger automatically gets imported with root log level in all files but that means logs will look like this, with all logs mentioning root:

Hence, I have added an argument now in python native root logger, which logs the module name for each logging statement (Can add function name of the logging statement too), without us having to manually define logger name with module. Hence now the same above logs look like:

Thank you for the input!

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

aluu317 · 2024-08-09T15:13:35Z

Thank you @Abhishek-TAMU. This works!

Abhishek-TAMU requested review from Ssukriti, alex-jw-brooks and anhuong as code owners July 30, 2024 03:02

Abhishek-TAMU and others added 3 commits July 30, 2024 11:21

Removal of transformers logger and addition of python logger

2b4ce8d

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

FMT and lint check: Removal of transformers logger and addition of py…

76fffc5

…thon logger Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

Abhishek-TAMU force-pushed the fix_logging branch from 68a91c5 to 773f0d1 Compare July 30, 2024 15:22

Abhishek-TAMU and others added 14 commits July 30, 2024 11:22

Merge branch 'main' into fix_logging

2e36147

Fix: Addition of env var TRANSFORMERS_VERBOSITY check

8846bc2

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

FMT Fix: Addition of env var TRANSFORMERS_VERBOSITY check

4ed3878

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

Adding logging support to accelerate launch

efb8363

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

FMT_FIX: Adding logging support to accelerate launch

2ecfaf7

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

Merge branch 'main' into fix_logging

f2e8afb

Merge branch 'main' into fix_logging

c44b1dc

Merge branch 'foundation-model-stack:main' into fix_logging

abd8abc

Logging changes and unit tests added

5d08efb

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

Merge branch 'foundation-model-stack:main' into fix_logging

a4cc9c2

Solved conflict with main

7fffda7

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

FMT:Fix Solved conflict with main

ba8a972

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

enabling tests for prompt tuning

f1159f9

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

Merge branch 'main' into main

756d097

anhuong requested changes Aug 2, 2024

View reviewed changes

aluu317 requested changes Aug 2, 2024

View reviewed changes

Abhishek-TAMU added 7 commits August 5, 2024 15:30

Merge remote-tracking branch 'upstream/main'

bf30ed2

Merge branch 'main' of github.com:Abhishek-TAMU/fms-hf-tuning

cb846ca

merge with main

da7acc6

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

PR changes for changing logger

8af5792

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

Merge remote-tracking branch 'upstream/main'

697056c

Merge branch 'main' into fix_logging

768d93a

Unit Tests changes

ba489b5

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

Abhishek-TAMU added 4 commits August 6, 2024 16:34

Merge remote-tracking branch 'upstream/main'

f3a984a

Merge branch 'main' into fix_logging

024b12e

PR changes

cfeb709

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

FIX:FMT

bf36b36

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

anhuong requested changes Aug 6, 2024

View reviewed changes

Abhishek-TAMU added 2 commits August 6, 2024 17:39

PR Changes

fe4b6d5

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

PR Changes

c544f47

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

dushyantbehl reviewed Aug 7, 2024

View reviewed changes

Abhishek-TAMU added 3 commits August 7, 2024 10:31

Merge remote-tracking branch 'upstream/main'

f0fcfdb

Merge branch 'main' into fix_logging

e65cb2d

PR Changes

4841119

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

Abhishek-TAMU added 2 commits August 8, 2024 11:18

Merge remote-tracking branch 'upstream/main'

fe8bb05

Metrics file epoch indexing from 0

5ecf4dd

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

anhuong previously approved these changes Aug 8, 2024

View reviewed changes

Abhishek-TAMU added 3 commits August 8, 2024 17:03

Revert last commit

89124ac

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

Merge remote-tracking branch 'upstream/main'

bddad06

Merge branch 'main' into fix_logging

0777460

aluu317 requested changes Aug 8, 2024

View reviewed changes

PR Changes

3068f51

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

Abhishek-TAMU dismissed anhuong’s stale review via 3068f51 August 8, 2024 23:39

PR Changes

0866bce

Signed-off-by: Abhishek <maurya.abhishek@ibm.com>

aluu317 approved these changes Aug 9, 2024

View reviewed changes

Abhishek-TAMU merged commit ee25de4 into foundation-model-stack:main Aug 9, 2024

Abhishek-TAMU mentioned this pull request Aug 9, 2024

bug: info (and lower) level logs are not printed because logging level is not being set #243

Closed

	if os.environ.get("TRANSFORMERS_VERBOSITY") is None:
	if not os.environ.get("TRANSFORMERS_VERBOSITY"):

	LOGLEVEL = os.environ.get("LOG_LEVEL", "WARNING").upper()
	LOGLEVEL = os.environ.get("LOG_LEVEL", logging.WARNING).upper()

		assert training_args.log_level == "info"


		@mock.patch.dict(os.environ, {}, clear=True)

	@mock.patch.dict(os.environ, {}, clear=True)
	@mock.patch.dict(os.environ, {"TRANSFORMERS_VERBOSITY": "info"}, clear=True)

		assert training_args.log_level == "error"


		@mock.patch.dict(os.environ, {}, clear=True)

	@mock.patch.dict(os.environ, {}, clear=True)
	@mock.patch.dict(os.environ, {"LOG_LEVEL": "info"}, clear=True)

		assert training_args.log_level == "passive"


		@mock.patch.dict(os.environ, {}, clear=True)

Conversation

Abhishek-TAMU commented Jul 30, 2024 • edited by kmehant Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description of the change

Related issue number

How to verify the PR

Was the PR tested

Uh oh!

anhuong left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Abhishek-TAMU Aug 2, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Abhishek-TAMU Aug 2, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Abhishek-TAMU Aug 2, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Abhishek-TAMU Aug 2, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

anhuong Aug 5, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Abhishek-TAMU commented Jul 30, 2024 •

edited by kmehant

Loading

Abhishek-TAMU Aug 2, 2024 •

edited

Loading

Abhishek-TAMU Aug 2, 2024 •

edited

Loading

Abhishek-TAMU Aug 2, 2024 •

edited

Loading

Abhishek-TAMU Aug 2, 2024 •

edited

Loading

anhuong Aug 5, 2024 •

edited

Loading

Abhishek-TAMU commented Aug 7, 2024 •

edited

Loading