Skip to content

MaxText/Tunix: Expose enable_continuous_profiling CLI flag in MaxText and Tunix profiler to allow saving >2GB profile as xplane.riegeli #5058

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
copybara-service wants to merge 1 commit into
base: main
Choose a base branch
from
Open

MaxText/Tunix: Expose enable_continuous_profiling CLI flag in MaxText and Tunix profiler to allow saving >2GB profile as xplane.riegeli #5058

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
1 change: 1 addition & 0 deletions src/maxtext/configs/base.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -913,6 +913,7 @@ skip_first_n_steps_for_profiler: 1
# Profile for a small number of steps to avoid a large profile file size.
profiler_steps: 5
hide_profiler_step_metric: false
enable_continuous_profiling: false
profile_cleanly: true # If set to true, adds a block_until_ready on train state which aligns the profile for each step.
profile_periodically_period: -1 # If set to a positive integer, profile every profile_periodically_period steps.
# This is useful to debug scenarios where performance is changing.
Expand Down
1 change: 1 addition & 0 deletions src/maxtext/configs/types.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2203,6 +2203,7 @@ class Profiling(BaseModel):
profile_cleanly: bool = Field(True, description="Add block_until_ready to align profile for each step.")
profile_periodically_period: int = Field(-1, description="If positive, profile every N steps.")
hide_profiler_step_metric: bool = Field(False, description="Whether to enable profiler step metric.")
enable_continuous_profiling: bool = Field(False, description="Enable continuous profiling in tunix profiler. Once enabled, it will support saving profile > 2GB.")
enable_jax_profiler: bool = Field(False, description="Enable the JAX live profiler.")
jax_profiler_port: int = Field(9999, description="Port for the JAX profiler.")
enable_tpu_profiling_options: bool = Field(False, description="Enable TPU advanced profiling options.")
Expand Down
2 changes: 2 additions & 0 deletions src/maxtext/trainers/post_train/rl/train_rl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -474,7 +474,9 @@ def create_rl_components( # pylint: disable=too-many-positional-arguments
log_dir=trainer_config.tensorboard_dir,
skip_first_n_steps=trainer_config.skip_first_n_steps_for_profiler,
profiler_steps=trainer_config.profiler_steps,
# Skip setting tracer levels.
set_profile_options=False,
enable_continuous_profiling=trainer_config.enable_continuous_profiling,
)

# Parse vllm_additional_config
Expand Down
Loading