Fix: RL base feature parity #2133
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
NanoCode012
force-pushed
the
fix/orpo_feature_parity
branch
2 times, most recently
from
0719188 to
4b8f65b
Compare
NanoCode012
force-pushed
the
fix/orpo_feature_parity
branch
from
1b15a11 to
62d04e4
Compare
bursteratom
force-pushed
the
fix/orpo_feature_parity
branch
from
65a83b7 to
93a2ecc
Compare
winglian
approved these changes
Feb 21, 2025
|
Thanks @NanoCode012 ! This should be good to go once the multi gpu tests pass too https://github.com/axolotl-ai-cloud/axolotl/actions/runs/13457763772 |
winglian
force-pushed
the
fix/orpo_feature_parity
branch
from
fc04dcf to
4321607
Compare
|
looks like the multi-gpu GRPO tests are failing https://github.com/axolotl-ai-cloud/axolotl/actions/runs/13469023440/job/37640091555 |
winglian
reviewed
Feb 22, 2025
src/axolotl/core/trainer_builder.py
Outdated
Comment on lines
234
to
252
| warmup_steps = max(int(self.cfg.warmup_ratio * total_num_steps), 0) | ||
| else: | ||
| warmup_steps = min(int(0.03 * total_num_steps), 100) | ||
| if warmup_steps == 1: | ||
| warmup_steps = 2 | ||
|
|
||
| logging_steps = ( | ||
| self.cfg.logging_steps | ||
| if self.cfg.logging_steps is not None | ||
| else max(min(int(0.005 * total_num_steps), 10), 1) |
There was a problem hiding this comment.
for many of the RL trainers, we can't rely on total_num_steps (set to None) and we pass -1 as max_steps to let the trainer figure out the total number of steps
winglian
reviewed
Feb 23, 2025
Comment on lines
+237
to
+245
| if warmup_steps == 1: | ||
| warmup_steps = 2 |
There was a problem hiding this comment.
winglian
force-pushed
the
fix/orpo_feature_parity
branch
from
4321607 to
8331312
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
RL trainer was not loading some basic configs like
logging_stepsetc. This PR consolidates the setting of these params and cleans them up.This PR also fixes a case where we did not call
.mapwithnum_proc.To discuss:
use_reentrantremove_unused_columnsAddresses a lot of points in #2105
Motivation and Context
How has this been tested?
Untested!Screenshots (if appropriate)
Types of changes
Social Handles (Optional)