minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

[Fix] Snapshot req.prefix_indices when the prefix cache is disabled (#34644)

Browse Source
This commit is contained in:
Liangsheng Yin
2026-08-12 20:50:46 -07:00
committed by GitHub
parent a318e16956
commit 26627e999d
2 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/srt/mem_cache/swa_radix_cache.py
+1 -1
View File
@@ -519,7 +519,7 @@ class SWARadixCache(KVCacheEventMixin, BasePrefixCache):
] ]
# `req.prefix_indices` will be used in `PrefillAdder::add_chunked_req` later # `req.prefix_indices` will be used in `PrefillAdder::add_chunked_req` later
req.prefix_indices = kv_indices req.prefix_indices = kv_indices.to(dtype=torch.int64, copy=True)
return return
token_ids = req.get_fill_ids() token_ids = req.get_fill_ids()
python/sglang/srt/mem_cache/unified_radix_cache.py
+1 -1
View File
@@ -749,7 +749,7 @@ class UnifiedRadixCache(BasePrefixCache):
kv_indices = self.req_to_token_pool.req_to_token[ kv_indices = self.req_to_token_pool.req_to_token[
req.req_pool_idx, : len(token_ids) req.req_pool_idx, : len(token_ids)
] ]
req.prefix_indices = kv_indices req.prefix_indices = kv_indices.to(dtype=torch.int64, copy=True)
return return
kv_indices_orig = self.req_to_token_pool.req_to_token[ kv_indices_orig = self.req_to_token_pool.req_to_token[