wassname/vllm
1
0
Fork 0
mirror of https://github.com/wassname/vllm.git synced 2026-08-10 12:40:39 +08:00
Code Issues Packages Projects Releases Wiki Activity

[ROCm][AMD] Disable auto enabling chunked prefill on ROCm (#11146)

Browse Source 
Signed-off-by: Gregory Shtrasberg <Gregory.Shtrasberg@amd.com>
This commit is contained in:
Gregory Shtrasberg
2024-12-13 05:31:26 +00:00
committed by GitHub
parent 3989a79824
commit 00c1bde5d8
1 changed files with 2 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
vllm/engine/arg_utils.py
+2 -1
View File
@@ -1074,7 +1074,8 @@ class EngineArgs:
if (is_gpu and not use_sliding_window and not use_spec_decode
and not self.enable_lora
and not self.enable_prompt_adapter
and model_config.runner_type != "pooling"):
and model_config.runner_type != "pooling"
and not current_platform.is_rocm()):
self.enable_chunked_prefill = True
logger.warning(
"Chunked prefill is enabled by default for models with "