diff --git a/python/sglang/srt/disaggregation/decode.py b/python/sglang/srt/disaggregation/decode.py index 813a5e9cf4..a43704cd47 100644 --- a/python/sglang/srt/disaggregation/decode.py +++ b/python/sglang/srt/disaggregation/decode.py @@ -97,7 +97,9 @@ def __init__( self.tp_size = tp_size self.bootstrap_port = bootstrap_port - self.num_reserved_decode_tokens = 512 + self.num_reserved_decode_tokens = int( + os.environ.get("SGLANG_NUM_RESERVED_DECODE_TOKENS", "512") + ) # Queue for requests pending pre-allocation self.queue: List[DecodeRequest] = []