[Model BadCase]: how to disable thinking?

[jacksonjack]

Basic Information - Models Used

MiniMax2.1

Information about environment and deployment

CUDA_VISIBLE_DEVICES=2,3,4,5
vllm serve /media/1/models/MiniMax-M2.1
--served-model-name "self-pp-minimax-2-1"
--port 5011
--host 0.0.0.0
--max-num-seqs 32
--tensor-parallel-size 4
--gpu-memory-utilization 0.95
--max-model-len 196608
--tool-call-parser minimax_m2
--reasoning-parser minimax_m2
--enable-auto-tool-choice
--trust-remote-code >logs/minimax_0107.log 2>&1 &

Description

Steps to reproduce

The bug can be reproduced with the following steps:

Expected behavior

Error logs

Paste the related screenshots here

[LaoZhang-best]

I'm using Sglang for inference and share this same need. I'd love to see a configuration option to enable or disable the 'thought' (reasoning) output.

[jiangwu300]

+1