fix: max_seq_len 检查改为仅 prompt 超限发 STOP，max_tokens 超出部分 clamp

2026-05-10 20:17:47 +08:00 · 2026-05-10 20:17:47 +08:00 · a58fab8d6e
parent a3c8296135
commit a58fab8d6e
1 changed files with 3 additions and 1 deletions
--- a/astrai/inference/scheduler.py
+++ b/astrai/inference/scheduler.py
@ -147,11 +147,13 @@ class InferenceScheduler:
        if len(prompt_ids) > self.max_prompt_len:
            prompt_ids = prompt_ids[-self.max_prompt_len :]

-        if len(prompt_ids) + max_tokens > self.max_seq_len:
+        if len(prompt_ids) >= self.max_seq_len:
            if stream_callback:
                stream_callback(STOP)
            return task_id

+        max_tokens = min(max_tokens, self.max_seq_len - len(prompt_ids))
+
        task = Task(
            task_id=task_id,
            prompt_ids=prompt_ids,