Skip to content

optimize the performance of FlashBert Path for HPU #575

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.

Sign up for GitHub

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

Merged
merged 1 commit into from
Apr 16, 2025
Merged

optimize the performance of FlashBert Path for HPU #575

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 7 additions & 5 deletions backends/python/server/text_embeddings_server/models/flash_bert.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -323,19 +323,21 @@ def batch_type(self) -> Union[FlashBatch, PaddedBatch]:
def embed(self, batch: Union[FlashBatch, PaddedBatch]) -> List[Embedding]:
if isinstance(batch, PaddedBatch):
input_lens = batch.attention_mask.cumsum(-1)[:, -1].to(torch.int32)
max_input_lens = input_lens.max().item()
max_input_lens = 0 # This value will not be used
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
max_input_lens = 0 # This value will not be used

NIT

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hi, sorry , there may be misunderstanding. Here I commented "This value will not be used" means this variable can be any value, but we need to keep it here, as we need to pass it to L352

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@Narsil , can you help double check?

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I guess there are cases where the forward of the model does need a right value for this right? Otherwise why not removing it there?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Well, this is a common file shared by CPU/XPU andd HPU devices. On CPU/XPU, we do need this variable with exact meaning, while on HPU, we do not have real varlen_attention API, so we pass attn_mask to replace its functionality. Here we just need to set a random value for max_input_lens. This line cannot be deleted, as we need to pass it to L352.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Got it 👍

cu_seqlens = torch.cat(
(input_lens.new_tensor([0]), input_lens.cumsum(-1).int())
)
mask = batch.attention_mask.bool()
batch_size = input_lens.size(0)
bsz, tgt_len = mask.size()
min_val = torch.finfo(self.dtype).min
attn_mask = torch.full(
[batch_size, 1, 1, mask.shape[-1]],
fill_value=torch.finfo(self.dtype).min,
[bsz, 1, tgt_len, tgt_len],
fill_value=min_val,
device=self.device,
dtype=self.dtype,
)
attn_mask.masked_fill_(mask[:, None, None, :], 0)
expanded_mask = mask[:, None, None, :].expand(bsz, 1, tgt_len, tgt_len)
attn_mask = attn_mask.masked_fill(expanded_mask, 0.0)
elif isinstance(batch, FlashBatch):
cu_seqlens = batch.cu_seqlens
mask = None
Expand Down
Loading