Repository navigation
Fix FSQ GPU code packing precision bug #52
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
Open
zzuzxy0430
wants to merge
1
commit into
xingchensong:main
Choose a base branch
from
zzuzxy0430:fix/fsq-gpu-code-packing
base: main
Could not load branches
Branch not found: {{ refName }}
Loading
Could not load tags
Nothing to show
Loading
Are you sure you want to change the base?
Some commits from the old base branch may be removed from the timeline,
and old review comments may become outdated.
Open
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 +1,46 @@ | ||
| import torch | ||
| from torch import nn | ||
|
|
||
| from s3tokenizer.model_v2 import FSQCodebook | ||
|
|
||
|
|
||
| class FakeProjectDown(nn.Module): | ||
|
|
||
| def __init__(self, output: torch.Tensor): | ||
| super().__init__() | ||
| self.register_buffer("output", output) | ||
|
|
||
| def forward(self, x: torch.Tensor) -> torch.Tensor: | ||
| batch, steps = x.shape[:2] | ||
| assert self.output.shape[:2] == (batch, steps) | ||
| return self.output.to(x.device) | ||
|
|
||
|
|
||
| def test_fsq_code_packing_returns_exact_token_ids(): | ||
| codebook = FSQCodebook(dim=8, level=3) | ||
| digits = torch.tensor( | ||
| [[[2, 0, 0, 0, 0, 2, 0, 0], [1, 0, 0, 0, 0, 2, 0, 0]]], | ||
| dtype=torch.int64, | ||
| ) | ||
| pre_tanh = torch.where( | ||
| digits == 2, | ||
| torch.full_like(digits, 20), | ||
| torch.where(digits == 1, torch.zeros_like(digits), torch.full_like(digits, -20)), | ||
| ).to(torch.float32) | ||
| codebook.project_down = FakeProjectDown(pre_tanh) | ||
|
|
||
| dummy_hidden = torch.zeros(1, 2, 8) | ||
| packed = codebook.encode(dummy_hidden) | ||
|
|
||
| assert torch.equal(packed, torch.tensor([[488, 487]], dtype=torch.int32)) | ||
|
|
||
|
|
||
| def test_cuda_float_pow_can_be_non_integer(): | ||
| if not torch.cuda.is_available(): | ||
| return | ||
|
|
||
| powers = torch.pow(3, torch.arange(8, device="cuda", dtype=torch.float32)).cpu() | ||
|
|
||
| # On some CUDA stacks, 3**5 becomes 242.99998 instead of 243.0. | ||
| # This test documents the motivation for avoiding float pow in code packing. | ||
| assert powers[5].item() <= 243.0 | ||
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.
Make the fake match
FSQCodebook.project_down()'s actual shape contract.FSQCodebook.encode()flattens(batch, steps, dim)to((batch * steps), dim)inpreprocess()before it callsproject_down(). Withdummy_hidden.shape == (1, 2, 8),x.shapehere is(2, 8), so this assertion fails immediately and the regression test never reaches the packing logic.🧪 Proposed fix
class FakeProjectDown(nn.Module): @@ def forward(self, x: torch.Tensor) -> torch.Tensor: - batch, steps = x.shape[:2] - assert self.output.shape[:2] == (batch, steps) - return self.output.to(x.device) + flat_output = self.output.reshape(-1, self.output.shape[-1]) + assert x.shape == flat_output.shape + return flat_output.to(x.device)📝 Committable suggestion
🤖 Prompt for AI Agents