text-generation-webui/docs/llama.cpp-models.md

558 B

Using llama.cpp in the web UI

Pre-converted models

Simply place the model in the models folder, making sure that its name contains ggml somewhere and ends in .bin.

Convert LLaMA yourself

Follow the instructions in the llama.cpp README to generate the ggml-model-q4_0.bin file: https://github.com/ggerganov/llama.cpp#usage

Performance

This was the performance of llama-7b int4 on my i5-12400F:

Output generated in 33.07 seconds (6.05 tokens/s, 200 tokens, context 17)

You can change the number of threads with --threads N.