Skip to content

[Bugfix] Fix the issue when the FI Attention operator is used in graph mode with eagle enabled #6932

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

Closed
chenxi-hh wants to merge 4 commits into from
Closed

[Bugfix] Fix the issue when the FI Attention operator is used in graph mode with eagle enabled #6932

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
f58d040
[Bugfix] Fix the issue when the FI Attention operator is used in grap…
chenxi-hh Mar 2, 2026
63305fd
Merge branch 'vllm-project:main' into main
chenxi-hh Mar 3, 2026
f42f8db
[Bugfix] Fix the issue when the FI Attention operator is used in grap…
chenxi-hh Mar 3, 2026
5be8c8f
[Bugfix] Fix the issue when the FI Attention operator is used in grap…
chenxi-hh Mar 3, 2026
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
8 changes: 4 additions & 4 deletions vllm_ascend/spec_decode/eagle_proposer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -143,7 +143,7 @@ def __init__(self, vllm_config: VllmConfig, device: torch.device, runner=None):
slot_mapping_lens = self.runner.max_num_tokens + 2 * self.pcp_size * self.runner.max_num_reqs
self.slot_mapping_group = [
torch.zeros(slot_mapping_lens, dtype=torch.int32, device=device, pin_memory=self.runner.pin_memory)
for _ in range(self.num_speculative_tokens)
for _ in range(self.num_speculative_tokens + 1)
]

self._runnable = self._run_merged_draft
Expand Down Expand Up @@ -551,7 +551,7 @@ def _propose(
# The first step of speculative.
for layer_name in self.attn_layer_names:
per_layer_attn_metadata[layer_name] = attn_metadata
multi_steps_attn_metadata = [per_layer_attn_metadata]
multi_steps_attn_metadata = []

attn_metadata_i = per_layer_attn_metadata[self.attn_layer_names[0]]
if self.pcp_size * self.dcp_size > 1:
Expand Down Expand Up @@ -601,7 +601,7 @@ def _propose(
common_attn_metadata.block_table_tensor = common_attn_metadata.block_table_tensor[:batch_size]

# Copy the old attn_metadata and update
for draft_step in range(1, self.num_speculative_tokens):
for draft_step in range(1, self.num_speculative_tokens + 1):
common_attn_metadata, attn_metadata = self.attn_update_stack_num_spec_norm(
draft_step,
attn_metadata,
Expand All @@ -620,7 +620,7 @@ def _propose(
multi_steps_attn_metadata.append(per_layer_attn_metadata)
else:
# Copy the old attn_metadata and update
for draft_step in range(1, self.num_speculative_tokens):
for draft_step in range(1, self.num_speculative_tokens + 1):
common_attn_metadata, attn_metadata = self.attn_update_stack_num_spec_norm(
draft_step,
attn_metadata,
Expand Down
Loading