Comment on Beating GPT-4 on HumanEval with a Fine-Tuned CodeLlama-34B

<- View Parent
L_Acacia@lemmy.one ⁨1⁩ ⁨year⁩ ago

The best way to run Llama model locally is to run using Text generation web UI, the model will most likely be quantized to 4/5bit GGML / GPTQ today, which will make it possible to run on a “normal” computer.

Phind might make it accessible on their website soon, but it doesn’t seem to be the case yet.

source
Sort:hotnewtop