My eval tests "command-r" vs "L3-TheSpice-8b-v0.1.3-GGUF-IQ-Imatrix"
#2
by
AlekseiPravdin
- opened
Wow! 🥰
It's just myself auto-eval test 😇 and manual testing in koboldcpp and LM Studio, looked awesome! 😊
Author feedback: @cgato
I am pleased to report, after having tried a few different models, including the Cat one you posted. This one here is the king. I have played no L3 model for RP. The cat one just was too bland, and kinda clinical in a way.
Good to hear this version is delivering. I'm looking forward to cgato's new iterations.
They've been popping off at the Chaiverse leaderboard:
https://console.chaiverse.com/
I have been using the standard chatml for RP , but it seems that with the presets from Virt-io/SillyTavern-Presets it does looks constructive and good, especially v2.0 prompts chatml