[WIP][AWQ] Support accumulation for reduced memory usage #1435
Conversation
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
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
Co-authored-by: Kyle Sayers <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Kyle Sayers <[email protected]>
Signed-off-by: Kyle Sayers <[email protected]>
| int_w_output = self._forward_input_with_kwargs( | ||
| module=module2inspect, inputs=x, input_kwargs=self._module_kwargs | ||
| ) | ||
| int_w_output = self._run_samples(parent_layer) |
There was a problem hiding this comment.
inputs here should be x, not the dataset
There was a problem hiding this comment.
x is the dataset, right?
There was a problem hiding this comment.
my bad, misread what self._samples was. can we keep that as a different name to distinguish it from actual data?
There was a problem hiding this comment.
Yeah I'm always happy to take suggestions on names for any of the code I push 🙂
| default_factory=IntermediatesCache | ||
| ) | ||
| _sample_means: Dict[Module, float] = PrivateAttr(default_factory=dict) | ||
| _num_samples: Dict[Module, int] = PrivateAttr(default_factory=dict) |
There was a problem hiding this comment.
I know this is in GPTQ, but a field named num_samples indicates an int in virtually any context i've seen it. what about _sample_counts so it's more similar to sample_means?
There was a problem hiding this comment.
That's good too!
Signed-off-by: Kyle Sayers <[email protected]>
rahul-tuli
reviewed
May 15, 2025
| AWQ_PRECISION = torch.float32 | ||
|
|
||
|
|
||
| def accumulate_mean( |
Comment on lines
+126
to
+133
| def iter( | ||
| self, input_names: Optional[List[str]] = None | ||
| ) -> Generator[Any, None, None]: | ||
| for batch_index in self.batch_intermediates: | ||
| yield self.fetch(batch_index, input_names) | ||
|
|
||
| def __iter__(self) -> Generator[Any, None, None]: | ||
| yield from self.iter() |
There was a problem hiding this comment.
Why not just have iter?
There was a problem hiding this comment.
Saves an extra function call
for batch in cache:
...for batch in cache.iter():
...
1 task
brian-dellabetta
force-pushed
the
bdellabe/awq-quantization-mixin
branch
from
c4cd97c to
2659c22
Compare
brian-dellabetta
added a commit
that referenced
this pull request
May 15, 2025
SUMMARY: - Add QuantizationMixin to AWQModifier so we don't have redundant inputs (num_bits, symmetric, group_size) - Move AWQModifier to sequential pipelines, to avoid huge memory requirements of caching all activations at once. Regression test results are acceptable, results are all roughly the same, and within stderr, see test plan below. Resolves #1409 Resolves #1369 Related to #1383 Related to #1406 Related to #1368 Related to #1410 More improvements split into #1435 TEST PLAN: - [x] Rerun tests to validate No regression in tests, comparing against those reported in [original AWQ PR](#1177 (comment)). All gsm8k results are within stderr: | Type | gsm8k | wikitext | ------ | ------ | ----- | Old AWQ+QuantModifier Sym | .1054, .1069 | 9.1931 | New AWQ+QuantMixin Sym | .1077, .1084 | 9.1841 | Old AWQ+QuantModifier Asym | .1274, .1281 | 9.0281 | New AWQ+QuantMixin Asym | .1312, .1350 | 9.0288 --------- Signed-off-by: Brian Dellabetta <[email protected]> Co-authored-by: Kyle Sayers <[email protected]>
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.
No description provided.