llama-cpp-python

Ah, I didn’t realize they’d upped the memory bandwidth to DDR5-6000 (vs 4800), thanks for the correction! The memory bandwidth does not double, I believe. See this random issue for a graph that has single/dual socket measurements, there is essentially no difference: https://github.com/abetlen/llama-cpp-python/issues/1098 Perhaps this is incorrect now, but I also know with 2x 4090s you don’t get higher tokens per second than 1x 4090 with llama.cpp, just more memory capacity. (All if this only applies to llama.cpp, I have no experience with other software and how memory bandwidth may scale across sockets)

Can you use https://github.com/abetlen/llama-cpp-python or you need something ollama provide ? speaking of embeddings, you saw https://jina.ai/news/jina-embeddings-v3-a-frontier-multiling... ?

There's a Python binding for llama.cpp which is actively maintained and has worked well for me: https://github.com/abetlen/llama-cpp-python

I'll share the core bit that took a while to figure out the right format, my main script is a hot mess using embeddings with SentenceTransformer, so I won't share that yet. E.g: last night I did a PR for llama-cpp-python that shows how Phi might be used with JSON only for the author to write almost exactly the same code at pretty much the same time. https://github.com/abetlen/llama-cpp-python/pull/1184 But you can see how that might work. Here is the core parser code: https://gist.github.com/lukestanley/eb1037478b1129a5ca0560ee...

You'll want to try llama.cpp [1]. The set of models that it can support is expanding [2]. Folks have also written services [3] that wrap around it. [1] https://github.com/ggerganov/llama.cpp [2] https://huggingface.co/TheBloke [3] https://github.com/abetlen/llama-cpp-python

If the model could be made to work with llama.cpp, then https://github.com/abetlen/llama-cpp-python might be more compact. llama.cpp only supports a limited list of model types though.

https://github.com/abetlen/llama-cpp-python has a web server mode that replicates openai's API iirc and the readme shows it has docker builds already.

LocalAI https://localai.io/ and LMStudio https://lmstudio.ai/ both have fairly complete OpenAI compatibility layers. llama-cpp-python has a FastAPI server as well: https://github.com/abetlen/llama-cpp-python/blob/main/llama_... (as of this moment it hasn't merged GGUF update yet though)

I have been researching methods and projects built around training and running LLMs locally. I'm interested in what others have been using on this front (including straight up Pytorch/Transformers). Here's what I've gathered so far: Engines/APIs: - vllm: Inference and serving engine for LLMs (none quantized models only?) [1] - ollama: Go project to run, create and share LLMs [2] - llama.cpp: Inference of LLaMA models in C/C++ w/UI (including quantized models) [3] - llama-cpp-python: run OpenAI compatible API bindings for llama.cpp [4] - llm-engine: engine for fine-tuning and serving LLMs [5] - Lamini: hosted? closed source? solution for training LLMs [6] - GPT4All: free, locally run chatbot [7] - SkyPilot: framework for running LLMs, AI, and batch jobs [8] - HuggingFace Transformers: APIs and tools to download and train

I see you’re using gpt4all; do you have a supported way to change the model being used for local inference? A number of apps that are designed for OpenAI’s completion/chat APIs can simply point to the endpoints served by llama-cpp-python [0], and function in (largely) the same way, while using the various models and quants supported by llama.cpp. That would allow folks to run larger models on the hardware of their choice (including Apple Silicon with Metal acceleration or NVIDIA GPUs) or using other proxies like openrouter.io. I enjoy openrouter.io myself because it supports Anthropic’s 100k models. [0]: https://github.com/abetlen/llama-cpp-python

Thanks for sharing! We've been recommending something similar with these llama-cpp-python bindings https://github.com/abetlen/llama-cpp-python#web-server, but I'll have to check that out