[Bugfix][MRV2] Match fast-prefill padding to active LoRA batches (#56456)

Signed-off-by: waizuichougou <2082431897@qq.com>
Signed-off-by: waizuichougou <105572299+waizuichougou@users.noreply.github.com>
This commit is contained in:
waizuichougou
2026-09-18 22:13:23 +00:00
committed by GitHub
parent 1dc2d854c1
commit a1bf8ac12d
4 changed files with 69 additions and 3 deletions
+3 -1
View File
@@ -149,7 +149,9 @@ class MMEncoderModelRunner(GPUModelRunner):
num_tokens=batch_req_state.num_tokens,
num_reqs=None,
)
self.prepare_inputs(scheduler_output, batch_req_state, batch_desc)
self.prepare_inputs(
scheduler_output, batch_req_state, batch_desc, num_active_loras=0
)
scheduled_encoder_inputs = scheduler_output.scheduled_encoder_inputs
if self.lora_config is not None: