Repository navigation
Fix DynamicLossScaler never lowering scale when hysteresis <= 0 #8993
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.
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
Closed
siddhant-rajhans
wants to merge
1
commit into
deepspeedai:master
from
siddhant-rajhans:fix/dynamic-loss-scaler-hysteresis
Closed
Changes from all commits
Commits
File filter
Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
There are no files selected for viewing
This file contains hidden or 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
This file contains hidden or 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
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,62 @@ | ||
| # Copyright (c) Microsoft Corporation. | ||
| # SPDX-License-Identifier: Apache-2.0 | ||
|
|
||
| # DeepSpeed Team | ||
|
|
||
| import pytest | ||
|
|
||
| from deepspeed.runtime.fp16.loss_scaler import DynamicLossScaler | ||
|
|
||
|
|
||
| @pytest.fixture(autouse=True) | ||
| def _patch_dist(monkeypatch): | ||
| """Stub out dist.get_rank so update_scale can run without a process group.""" | ||
| import deepspeed.comm as dist | ||
| monkeypatch.setattr(dist, "get_rank", lambda: 0) | ||
|
|
||
|
|
||
| INIT_SCALE = 2**16 | ||
| SCALE_WINDOW = 1000 | ||
| MIN_SCALE = 1 | ||
|
|
||
|
|
||
| def _make_scaler(delayed_shift, consecutive_hysteresis=False): | ||
| return DynamicLossScaler( | ||
| init_scale=INIT_SCALE, | ||
| scale_window=SCALE_WINDOW, | ||
| min_scale=MIN_SCALE, | ||
| delayed_shift=delayed_shift, | ||
| consecutive_hysteresis=consecutive_hysteresis, | ||
| ) | ||
|
|
||
|
|
||
| class TestDynamicLossScalerHysteresis: | ||
| """Ensure DynamicLossScaler lowers its scale on overflow for all hysteresis values.""" | ||
|
|
||
| @pytest.mark.parametrize("delayed_shift", [0, -1, -5]) | ||
| def test_zero_or_negative_hysteresis_lowers_scale(self, delayed_shift): | ||
| scaler = _make_scaler(delayed_shift) | ||
| scaler.update_scale(overflow=True) | ||
| assert scaler.cur_scale < INIT_SCALE | ||
|
|
||
| def test_hysteresis_one_lowers_scale_immediately(self): | ||
| scaler = _make_scaler(delayed_shift=1) | ||
| scaler.update_scale(overflow=True) | ||
| assert scaler.cur_scale == INIT_SCALE / 2 | ||
|
|
||
| def test_hysteresis_two_delays_one_step(self): | ||
| scaler = _make_scaler(delayed_shift=2) | ||
| scaler.update_scale(overflow=True) | ||
| assert scaler.cur_scale == INIT_SCALE | ||
| scaler.update_scale(overflow=True) | ||
| assert scaler.cur_scale == INIT_SCALE / 2 | ||
|
|
||
| @pytest.mark.parametrize("delayed_shift", [0, 1, 2]) | ||
| def test_repeated_overflow_reaches_min_scale(self, delayed_shift): | ||
| scaler = _make_scaler(delayed_shift) | ||
| for _ in range(200): | ||
| try: | ||
| scaler.update_scale(overflow=True) | ||
| except Exception: | ||
| break | ||
| assert scaler.cur_scale == MIN_SCALE | ||
Oops, something went wrong.
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.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
For
fp16.hysteresis <= 0, the user-visible fix is that ZeRO training can recover after overflow, but these tests invokeupdate_scale()directly and stub distributed rank. They never exercise configuration propagation, overflow detection, step skipping, or subsequent optimizer progress; the existing ZeRO overflow tests use positive hysteresis. Add aDistributedTestwithSimpleModelthroughdeepspeed.initialize()that verifies an overflowing step lowers the scale without updating weights, then a finite step updates weights for zero and negative hysteresis, and execute it on an fp16-capable device. RootAGENTS.mdexplicitly requires integration tests for training-loop contract changes and actual-device execution.AGENTS.md reference: AGENTS.md:L81-L82
Useful? React with 👍 / 👎.