spec: size the speculative buffers from the bags, not the startup record (#35024)

2026-08-17 16:16:56 -07:00
parent 3d7ec00179
commit d2bc697396
28 changed files with 216 additions and 179 deletions
@@ -88,13 +88,9 @@ patches:
  - target: sglang.srt.mem_cache.allocation_sizing.get_req_to_token_extra_context_len
    edits:
      - match: |
-          if (
-              server_args.speculative_algorithm is not None
-              and server_args.page_size > 1
-              and (server_args.speculative_eagle_topk or 1) > 1
-          ):
-              extra = max(extra, get_alloc_reserve_per_decode(server_args))
-        replacement: ""
+          extra = max(extra, get_alloc_reserve_per_decode() + page_size - 1)
+        replacement: |
+          pass
 """