Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
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
2 changes: 1 addition & 1 deletion vllm_ascend/spec_decode/mtp_proposer.py
Original file line number Diff line number Diff line change
Expand Up @@ -288,7 +288,7 @@ def dummy_run(self,
positions = self.positions[:num_tokens]
previous_hidden_states = self.hidden_states[:num_tokens]
for i in range(self.num_speculative_tokens):
if i > 0 and not in_graph_capturing and aclgraph_runtime_mode == CUDAGraphMode.FULL:
if i > 0 and in_graph_capturing and aclgraph_runtime_mode == CUDAGraphMode.FULL:
aclgraph_runtime_mode = CUDAGraphMode.NONE
with set_ascend_forward_context(
attn_metadata,
Expand Down
4 changes: 3 additions & 1 deletion vllm_ascend/worker/model_runner_v1.py
Original file line number Diff line number Diff line change
Expand Up @@ -2154,6 +2154,8 @@ def dummy_drafter_compute_logits(hidden_states):
dummy_compute_logits(hidden_states)

if self.drafter:
# `in_graph_capturing` indicates whether the main model is in graph capturing.
# The value is only used in `mtp_proposer.py` currently and defaults to False.
self.drafter.dummy_run(
num_tokens=num_tokens_padded,
with_prefill=with_prefill,
Expand All @@ -2162,7 +2164,7 @@ def dummy_drafter_compute_logits(hidden_states):
aclgraph_runtime_mode=aclgraph_runtime_mode,
batch_descriptor=batch_descriptor,
dummy_compute_logits=dummy_drafter_compute_logits,
in_graph_capturing=not force_attention)
in_graph_capturing=force_attention)
if is_profile and self.dynamic_eplb:
self.model.clear_all_moe_loads()
if not is_profile and self.dynamic_eplb:
Expand Down
Loading