From 4a05bdfa869c80fdcac2d1b8fb48656f743a1fac Mon Sep 17 00:00:00 2001 From: Lianmin Zheng Date: Sun, 9 Mar 2025 18:53:33 -0700 Subject: [PATCH] Revert "Check eagle server args" (#4242) --- python/sglang/srt/server_args.py | 8 ++------ 1 file changed, 2 insertions(+), 6 deletions(-) diff --git a/python/sglang/srt/server_args.py b/python/sglang/srt/server_args.py index 480a415e8..4e6fbdd49 100644 --- a/python/sglang/srt/server_args.py +++ b/python/sglang/srt/server_args.py @@ -284,13 +284,9 @@ class ServerArgs: "Overlap scheduler are disabled because of using " "eagle speculative decoding." ) - # The token generated from the verify step is counted in speculative_num_draft_tokens. + # The token generated from the verify step is counted. # If sepculative_num_steps >= speculative_num_draft_tokens, the additional tokens will definitely be discarded. - assert self.speculative_num_steps < self.speculative_num_draft_tokens - assert ( - self.speculative_num_draft_tokens - 1 - <= self.speculative_num_steps * self.speculative_eagle_topk - ) + # assert self.speculative_num_steps < self.speculative_num_draft_tokens # GGUF if (