Recommended Generation Parameter Settings
Hi,
I am currently experimenting with your Qwen3‑4B model and would like to optimize the text generation quality. Could you please clarify:
Are there any recommended generation parameter values you suggest for this model, or should we use the default values provided with Qwen3‑4B? (--temp 0.7 --top-p 0.8 --top-k 20 --min-p 0.0)
Specifically, I am interested in guidance for the main decoding parameters:
Temperature
Top-p / Nucleus Sampling
Top-k
Min-p / Minimum Probability
Repetition Penalty
Presence Penalty
Additionally, if there are any other generation settings or best practices you recommend for achieving coherent and creative outputs with Qwen3‑4B, I would greatly appreciate your guidance.
Thank you for your time and support!
Factory Qwen-3 4b non-thinking are recommended. Temp 0.6-0.8, top p 0.85, top-k 20, min p 0.0, repeat pen 1.05-1.1 - I haven't tweaked any other knobs in my testing. 🫡
qwen image is notoriously hard to prompt. does this model put the process on easy mode?