Add **quantization_kwargs to FrozenNF4Linear and LoRALinear and DoRALinear
#1987
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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/torchtune/1987
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 1196f8a with merge base ac4f88e ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #1987 +/- ##
===========================================
+ Coverage 24.79% 64.96% +40.16%
===========================================
Files 318 318
Lines 17597 17610 +13
===========================================
+ Hits 4364 11440 +7076
+ Misses 13233 6170 -7063 ☔ View full report in Codecov by Sentry. |
joecummings
changed the title
FrozenNF4Linear and LoRALinearFrozenNF4Linear and LoRALinear and DoRALinear
pbontrager
approved these changes
Nov 12, 2024
drisspg
approved these changes
Nov 12, 2024
| """Test that passing in non-default quantization kwargs works as expected.""" | ||
| quantization_kwargs = { | ||
| "block_size": 16, | ||
| "scaler_block_size": 256, |
There was a problem hiding this comment.
Sad my spelling mistake is now surfaced lol
joecummings
added a commit
that referenced
this pull request
Nov 13, 2024
…nd ``DoRALinear`` (#1987)
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Outdated suggestions cannot be applied.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.
Context
What is the purpose of this PR? Is it to
The original error that sparked this investigation was that I was able to run Llama3.2V QLoRA on 4 GPUs, but it failed on 8 GPUs with the following error:
Weird, right? In digging further @pbontrager figured out that the default block_size and scaler_block_size used when converting the weights to NF4 were the wrong size when the model was sharded to 8 GPUs. After consulting with @drisspg, it seems relatively harmless to modify the scaler_block_size when quantizing these weights. Therefore, the fix here is to expose the quantization_kwargs for both FrozenNF4Linear and LoRALinear. A follow up PR will actually land the changes in the model builders to resolve the initial error.
Changelog
What are the changes made in this PR?
quantization_kwargstoFrozenNF4Linearquantization_kwargstoLoRALinearquantization_kwargstoDoRALinearTest plan
Please make sure to do each of the following if applicable to your PR. If you're unsure about any one of these just ask and we will happily help. We also have a contributing page for some guidance on contributing.
pre-commit install)pytest testspytest tests -m integration_testUX
If your function changed a public API, please add a dummy example of what the user experience will look like when calling it.
Here is a docstring example
and a tutorial example