[CI] Fix test_lora_with_spec_decode on V2 model runner (#43314)

Signed-off-by: haosdent <haosdent@gmail.com>
Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
This commit is contained in:
haosdent
2026-05-22 14:24:18 +08:00
committed by GitHub
co-authored by Wentao Ye
parent fa1ff88b31
commit 5ea76fa89a
2 changed files with 51 additions and 34 deletions
+13 -1
View File
@@ -1091,13 +1091,25 @@ class GPUModelRunner(LoRAModelRunnerMixin):
# FIXME: Replace this with LoRA warmup:
# https://github.com/vllm-project/vllm/pull/35536
assert hasattr(self, "lora_manager")
self.lora_manager._adapter_manager.set_adapter_mapping(
adapter_manager = self.lora_manager._adapter_manager
adapter_manager.set_adapter_mapping(
LoRAMapping(
index_mapping=(0,) * input_batch.num_tokens_after_padding,
prompt_mapping=(0,) * input_batch.num_reqs,
is_prefill=True,
)
)
seen_wrappers: set[int] = set()
for punica_wrapper in adapter_manager.punica_wrapper_mapping.values():
if id(punica_wrapper) in seen_wrappers:
continue
seen_wrappers.add(id(punica_wrapper))
for kernel_meta in (
punica_wrapper.token_mapping_meta, # type: ignore[attr-defined]
punica_wrapper.prompt_mapping_meta, # type: ignore[attr-defined]
):
kernel_meta.no_lora_flag_cpu[0] = False
kernel_meta.num_active_loras_cpu[0] = 1
attn_metadata = None
slot_mappings_by_layer = None