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 3, 22:45 UTC
bugfix(flashinfer,dcp): remove kv_cache_layout for BatchDCPPrefillWrapper._new_tokens. (#38895)
Signed-off-by: QiuChunshuo <qiuchunshuo@huawei.com>
Qiu committed
May 11, 2026 at 16:11 UTC
b1b59720b252b422368deb99aadc888fb397db40
1 file changed
+1
-3
vllm/v1/attention/backends/flashinfer.py
+1
-3
@@ -223,9 +223,7 @@ class BatchDCPPrefillWrapper:
223
self._context = BatchPrefillWithPagedKVCacheWrapper(
224
workspace_buffer, get_kv_cache_layout()
225
)
226
- self._new_tokens = BatchPrefillWithRaggedKVCacheWrapper(
227
- workspace_buffer, get_kv_cache_layout()
228
- )
226
+ self._new_tokens = BatchPrefillWithRaggedKVCacheWrapper(workspace_buffer)
227
228
def plan(
229
self,