Inject default thinking_token_budget and presence_penalty for vLLM, fixing the gap where --override-generation-config doesn't propagate these fields. Prevents Qwen3 thinking-mode infinite loops.
-
Updated
Apr 26, 2026 - Shell
#
thinking-budget
Here are 3 public repositories matching this topic...
vLLM v0.21 + Gemma 4 DFlash + real thinking_token_budget enforcement on Blackwell (sm_120 / sm_121a)
gemma reasoning blackwell vllm speculative-decoding nvfp4 dgx-spark cuda-graphs dflash gemma-4 thinking-budget
-
Updated
May 26, 2026 - Python
vLLM v0.21 + Qwen 3.6 DFlash + real thinking_token_budget enforcement on Blackwell (sm_120 / sm_121a)
reasoning blackwell fp8 vllm qwen speculative-decoding qwen3 dgx-spark cuda-graphs dflash thinking-budget
-
Updated
May 26, 2026 - Python
Improve this page
Add a description, image, and links to the thinking-budget topic page so that developers can more easily learn about it.
Add this topic to your repo
To associate your repository with the thinking-budget topic, visit your repo's landing page and select "manage topics."