chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 11 Packages Projects Releases Wiki Activity

[PP] Add pp support for Qwen3-VL (#12333)

Browse Source 
Signed-off-by: Xuchun Shang <xuchun.shang@gmail.com>
Signed-off-by: Kun(llfl) <i@imux.top>
Signed-off-by: Kun(llfl) <llfl@linux.alibaba.com>
Co-authored-by: kun-llfl <i@imux.top>
Co-authored-by: Kun(llfl) <llfl@linux.alibaba.com>
This commit is contained in:
Xuchun Shang
2025-12-17 16:03:58 +08:00
committed by GitHub
co-authored by kun-llfl Kun
parent cdce516331
commit 45a959d3e9
5 changed files with 119 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/srt/managers/schedule_policy.py
+4
View File
@@ -455,6 +455,10 @@ class PrefillAdder:
def add_chunked_req(self, req: Req):
_rem_tokens = min(self.rem_chunk_tokens, int(self.rem_total_tokens))
# The chunked_req must be added to the list; otherwise, it will cause a memory leak.
# Therefore, in certain cases where _rem_tokens <= 0, it should be replaced with rem_chunk_tokens.
if _rem_tokens <= 0:
_rem_tokens = self.rem_chunk_tokens
truncated = req.extend_input_len > _rem_tokens
req.extend_input_len = min(req.extend_input_len, _rem_tokens)
req.fill_ids = req.fill_ids[: len(req.prefix_indices) + req.extend_input_len]