Closed rchan26 closed 2 months ago
Mistral's Mixtral 8x7b model is a new open-source model which is competitive with GPT3.5: https://mistral.ai/news/mixtral-of-experts/
There are now quantised versions of this model available on Huggingface which we can experiment with and using with our implementation of llama-index-llama-cpp.
llama-index-llama-cpp
Alternatively, we can try out Mistral's Mistral 7B model (with llama-index-hf) or quantised versions of it.
llama-index-hf
Feel free to experiment with Mistral's models in a notebook setting like here or with the full pipeline.
Closing since we can easily try out different VMs using the Ollama model implemented in #179
Mistral's Mixtral 8x7b model is a new open-source model which is competitive with GPT3.5: https://mistral.ai/news/mixtral-of-experts/
There are now quantised versions of this model available on Huggingface which we can experiment with and using with our implementation of
llama-index-llama-cpp
.Alternatively, we can try out Mistral's Mistral 7B model (with
llama-index-hf
) or quantised versions of it.Feel free to experiment with Mistral's models in a notebook setting like here or with the full pipeline.