Remove hard-dependencies of Speculative decode to CUDA workers (#10587)

Signed-off-by: Chendi Xue <chendi.xue@intel.com>
This commit is contained in:
Chendi.Xue
2024-11-26 17:57:11 -08:00
committed by GitHub
parent 2f0a0a17a4
commit 0a71900bc9
19 changed files with 219 additions and 77 deletions
+7 -1
View File
@@ -86,4 +86,10 @@ class CpuPlatform(Platform):
parallel_config.distributed_executor_backend)
parallel_config.distributed_executor_backend = "mp"
if parallel_config.worker_cls == "auto":
parallel_config.worker_cls = "vllm.worker.cpu_worker.CPUWorker"
if vllm_config.speculative_config:
parallel_config.worker_cls = \
"vllm.spec_decode.spec_decode_worker.create_spec_worker"
parallel_config.sd_worker_cls = \
"vllm.worker.cpu_worker.CPUWorker"
else:
parallel_config.worker_cls = "vllm.worker.cpu_worker.CPUWorker"