[Bugfix][MRV2] Keep input_ids on later PP stages for raw-token models (#61016)

Signed-off-by: booth-algo <kevinlauofficial01@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Kevin H. Luu <khluu000@gmail.com>
This commit is contained in:
Kevin
2026-10-15 12:54:21 -07:00
committed by GitHub
co-authored by Claude Opus 5.5 Kevin H. Luu
parent d486219fc2
commit d12ea8e309
6 changed files with 51 additions and 5 deletions
+3 -1
View File
@@ -33,6 +33,7 @@ from vllm.distributed.parallel_state import (
)
from vllm.forward_context import BatchDescriptor, set_forward_context
from vllm.logger import init_logger
from vllm.model_executor.models.interfaces import requires_raw_input_tokens
from vllm.model_executor.offloader.base import get_offloader
from vllm.platforms import current_platform
from vllm.sequence import IntermediateTensors
@@ -690,7 +691,8 @@ class ModelCudaGraphManager(CudaGraphManager):
}
if not self.is_first_pp_rank:
# Update for non-first PP ranks.
model_inputs["input_ids"] = None
if not requires_raw_input_tokens(model):
model_inputs["input_ids"] = None
model_inputs["inputs_embeds"] = None
assert intermediate_tensors is not None
model_inputs["intermediate_tensors"] = intermediate_tensors[:num_tokens]
+2 -1
View File
@@ -1970,7 +1970,8 @@ class GPUModelRunner(LoRAModelRunnerMixin):
}
if not self.is_first_pp_rank:
# Update for non-first PP ranks.
model_inputs["input_ids"] = None
if not requires_raw_input_tokens(self.model):
model_inputs["input_ids"] = None
model_inputs["inputs_embeds"] = None
# Prepare the intermediate tensors.