Text Generation
Transformers
Safetensors
llama
Generated from Trainer
axolotl
conversational
Inference Endpoints
text-generation-inference

rope theta?

#1
by bdambrosio - opened

You said 1000000.0 on model card, but config says "rope_theta": 5000000.0,
Should I leave as is? (config max pos is 8192)

Cognitive Computations org

Yeah it should be just fine

bdambrosio changed discussion status to closed

Sign up or log in to comment