Skip to content

Add configurable engine log level #8067

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
sfc-gh-truwase wants to merge 2 commits into
base: master
Choose a base branch
from
+14 −1
Open

Add configurable engine log level #8067

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
566214f
Add configurable engine log level
sfc-gh-truwase Jun 15, 2026
1d5b82a
Merge branch 'master' into engine-log-level
tjruwase Jun 16, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 6 additions & 0 deletions deepspeed/runtime/config.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -637,6 +637,10 @@ def get_dataloader_drop_last(param_dict):
return get_scalar_param(param_dict, DATALOADER_DROP_LAST, DATALOADER_DROP_LAST_DEFAULT)


def get_log_level(param_dict):
return get_scalar_param(param_dict, LOG_LEVEL, LOG_LEVEL_DEFAULT)


'''Write deepspeed config files by modifying basic templates.
Can be used for quickly changing parameters via command line parameters.'''

Expand Down Expand Up @@ -868,6 +872,8 @@ def _initialize_params(self, param_dict):

self.dataloader_drop_last = get_dataloader_drop_last(param_dict)

self.log_level = get_log_level(param_dict)

self.nebula_config = DeepSpeedNebulaConfig(param_dict)
self.datastates_config = DeepSpeedDataStatesConfig(param_dict)
self.checkpoint_config = get_checkpoint_config(param_dict)
Expand Down
3 changes: 3 additions & 0 deletions deepspeed/runtime/constants.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -501,3 +501,6 @@ class ValidationMode:
#########################################
USE_DATA_BEFORE_EXPERT_PARALLEL = "use_data_before_expert_parallelism"
USE_DATA_BEFORE_EXPERT_PARALLEL_DEFAULT = False

LOG_LEVEL = "log_level"
LOG_LEVEL_DEFAULT = "WARNING"
6 changes: 5 additions & 1 deletion deepspeed/runtime/engine.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -116,7 +116,7 @@
from ..git_version_info import version

from deepspeed.profiling.flops_profiler.profiler import FlopsProfiler
from deepspeed.utils.logging import print_json_dist, print_configuration
from deepspeed.utils.logging import print_json_dist, print_configuration, set_log_level_from_string

from deepspeed.accelerator import get_accelerator

Expand Down Expand Up @@ -277,6 +277,7 @@ def __init__(self,
self._do_args_sanity_check(args)
self._configure_with_arguments(args, mpu)
self._do_sanity_check()
set_log_level_from_string(self.log_level())

@chatgpt-codex-connector chatgpt-codex-connector Bot Jun 15, 2026

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Respect launcher log-level overrides

In launcher runs, deepspeed/launcher/runner.py:574 forwards --log_level and deepspeed/launcher/launch.py:151 applies it before the user script starts, but configs that omit the new key now reach this line with LOG_LEVEL_DEFAULT = "WARNING". That silently downgrades explicit --log_level debug/info and also weakens --quiet/error back to warnings during engine construction unless users duplicate the setting in ds_config.json, so the existing CLI verbosity control no longer works for ordinary launcher invocations. Only override the logger when the config key is present, or default to the current logger level instead.

Useful? React with 👍 / 👎.

self._configure_expert_parallel(model)
if self.autotp_size() > 1:
self._configure_tensor_parallel(model, self.tensor_parallel_config())
Expand Down Expand Up @@ -1228,6 +1229,9 @@ def use_node_local_storage(self):
def load_universal_checkpoint(self):
return self._config.load_universal_checkpoint

def log_level(self):
return self._config.log_level

@property
def communication_data_type(self):
res = self._config.communication_data_type
Expand Down
Loading