A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 5, 09:45 UTC
[Bugfix] Fix cpu-offload-gb assertion with non-default block sizes (#36461)
Signed-off-by: AjAnubolu <anuboluajay@gmail.com> Signed-off-by: Michael Goin <mgoin64@gmail.com> Co-authored-by: Michael Goin <mgoin64@gmail.com>
Ajay Anubolu committed
Apr 8, 2026 at 19:42 UTC
2f41d6c0631c75799d40199b1787831f999d2836
1 file changed
-5
vllm/v1/worker/gpu_model_runner.py
-5
@@ -6517,11 +6517,6 @@ class GPUModelRunner(
6517
block_sizes != self._init_block_sizes
6518
or kernel_block_sizes != self._init_kernel_block_sizes
6519
):
6520
- assert self.offload_config.uva.cpu_offload_gb == 0, (
6521
- "Cannot re-initialize the input batch when CPU weight "
6522
- "offloading is enabled. See https://github.com/vllm-project/vllm/pull/18298 " # noqa: E501
6523
- "for more details."
6524
- )
6520
self._init_block_sizes = block_sizes
6521
self._init_kernel_block_sizes = kernel_block_sizes
6522
self.input_batch = InputBatch(