Improve weight loading and code style (#3174)

2025-01-27 03:00:41 -08:00
parent 351a72d40b
commit 53cef81587
11 changed files with 171 additions and 65 deletions
--- a/python/sglang/srt/managers/schedule_batch.py
+++ b/python/sglang/srt/managers/schedule_batch.py
@@ -247,6 +247,7 @@ class Req:
        # Each decode stage's output ids
        self.output_ids = []
        # fill_ids = origin_input_ids + output_ids. Updated if chunked.
+        self.fill_ids = None
        self.session_id = session_id
        self.input_embeds = input_embeds