Skip to content

[BugFix] Fix tbo nan #6439

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

Open
RichardWooSJTU wants to merge 1 commit into
base: develop
Choose a base branch
from
Open

[BugFix] Fix tbo nan #6439

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
4 changes: 0 additions & 4 deletions fastdeploy/model_executor/layers/moe/fused_moe_deepgemm_backend.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -354,7 +354,6 @@ def apply_ep_prefill(
ffn_out,
m_indices,
)
del permute_input

# swiglu
ffn_out = paddle.incubate.nn.functional.swiglu(ffn_out, None)
Expand All @@ -373,7 +372,6 @@ def apply_ep_prefill(
)
ffn_in_x_scale_tensor = ffn_in_x_scale_tensor.T[: ffn_in_x.shape[0]]

del ffn_out
ffn_out = paddle.empty(
(token_all_num, getattr(layer, self.added_weight_attrs[1]).shape[1]),
dtype=paddle.bfloat16,
Expand All @@ -385,7 +383,6 @@ def apply_ep_prefill(
ffn_out,
m_indices,
)
del ffn_in_x

# prmt back per rank
tmp_ffn_out = fastdeploy.model_executor.ops.gpu.ep_moe_expert_combine(
Expand All @@ -397,7 +394,6 @@ def apply_ep_prefill(
False, # norm_topk_prob
1.0,
)
del ffn_out
else:
tmp_ffn_out = paddle.empty([0, hidden_size], paddle.bfloat16)

Expand Down
Loading