Safetensors
qwen3

Reasoning content

#2
by RawthiL - opened

I'm deploying this model using vLLM, following your provided options:

  • "--enable-auto-tool-choice"
  • "--tool-call-parser hermes"
  • "--enforce-eager"
  • "--disable-cascade-attn"

However I'm seeing that some responses include <think> tokens (I'm prompting using /v1/chat/completions) . Does this model follow any specific reasoning template or are these only artifacts that cannot be remediated? (like olmo, qwen3, etc)

The model follows the qwen3 reasoning template

Sign up or log in to comment