New NVIDIA models (nvidia/Llama3-70B-SteerLM-Chat, nvidia/Llama3-70B-PPO-Chat, nvidia/Llama3-70B-DPO-Chat )

#98
by AuriAetherwiing - opened

+1 this, its really rare for a model to be this open with its code, dataset and weights to be open source fully with a paper too

Unfortunately, they are all in nvidias proprietary nemo framework format, and as such not supported by llama.cpp. With "open" you mean "vendor lock-in", right? :)

If somebody converts this to transformers format, or support is added in llama.cpp, I am fair game to quantize these, of course :)

mradermacher changed discussion status to closed

Sign up or log in to comment