youngkingdom/vllm
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

revert kv connector fix

Browse Source 
Signed-off-by: Sage Moore <sage@neuralmagic.com>
This commit is contained in:
Sage Moore
2025-08-12 22:54:47 -04:00
parent 9e16220e4e
commit 6d76bd034a
1 changed files with 4 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
vllm/v1/worker/gpu_model_runner.py
View File

@ -2031,6 +2031,7 @@ class GPUModelRunner(LoRAModelRunnerMixin, KVConnectorModelRunnerMixin):
# Run the model.
# Use persistent buffers for CUDA graphs.
<<<<<<< HEAD
# when DBO is enabled, `num_tokens_after_padding`
# represents the per-ubatch DP token count.
dp_tokens_for_forward = num_tokens_after_padding
@ -2044,6 +2045,9 @@ class GPUModelRunner(LoRAModelRunnerMixin, KVConnectorModelRunnerMixin):
num_tokens_across_dp=dp_tokens_for_forward,
skip_cuda_graphs=skip_cuda_graphs):
self.maybe_setup_kv_connector(scheduler_output)
=======
self.maybe_setup_kv_connector(scheduler_output)
>>>>>>> db77e4a92 (revert kv connector fix)
model_output = self._run_model(
attn_metadata=attn_metadata,
num_scheduled_tokens=num_input_tokens,