wrong reasoning parser
#1
by mratsim - opened
Main model mentions Qwen3 as a reasoning parser
Thank you for your attention! In fact, according to our tests, when deploying the model with vllm and setting --reasoning-parser qwen3, the parsing of reasoning malfunctions. Conversely, setting --reasoning-parser deepseek_r1 results in normal parsing.
Luukus changed pull request status to closed
Additionally, in the vllm deployment instructions in the main model's README, the reasoning parser has already been changed to deepseek_r1.