WizardLM VS openai-cookbook

This is interesting work, and a good contribution, but there is no need to mislead people.

[1] https://github.com/nlpxucan/WizardLM

If you pair this with the latest WizardCoder models, which have a fairly better performance than the standard Salesforce Codegen2 and Codegen2.5, you have a pretty solid alternative to GitHub Copilot that runs completely locally.

(We will update the demo links in our github.)

In terms of speed, we're talking about 140t/s for 7B models, and 40t/s for 33B models on a 3090/4090 now.[1] (1 token ~= 0.75 word) It's quite zippy. llama.cpp performs close on Nvidia GPUs now (but they don't have a handy chart) and you can get decent performance on 13B models on M1/M2 Macs.

You can take a look at a list of evals here: https://llm-tracker.info/books/evals/page/list-of-evals - for general usage, I think home-rolled evals like llm-jeopardy [2] and local-llm-comparison [3] by hobbyists are more useful than most of the benchmark rankings.

That being said, personally I mostly use GPT-4 for code assistance to that's what I'm most interested in, and the latest code assistants are scoring quite well: https://github.com/abacaj/code-eval - a recent replit-3b fine tune the human-eval results for open models (as a point of reference, GPT-3.5 gets 60.4 on pass@1 and 68.9 on pass@10 [4]) - I've only just started playing around with it since replit model tooling is not as good as llamas (doc here: https://llm-tracker.info/books/howto-guides/page/replit-mode...).

I'm interested in potentially applying reflexion or some of the other techniques that have been tried to even further increase coding abilities. (InterCode in particular has caught my eye https://intercode-benchmark.github.io/)

[1] https://github.com/turboderp/exllama#results-so-far

[2] https://github.com/aigoopy/llm-jeopardy

[3] https://github.com/Troyanovsky/Local-LLM-comparison/tree/mai...

[4] https://github.com/nlpxucan/WizardLM/tree/main/WizardCoder

You talking about this? https://github.com/nlpxucan/WizardLM

The smallest model that is close to competent at code is WizardCoder 15B.. https://github.com/nlpxucan/WizardLM/

WizardCoder: Empowering Code Large Language Models with Evol-Instruct (https://github.com/nlpxucan/WizardLM/tree/main/WizardCoder)

A pretrained LLM is a closed-book system: It can only access information that it was trained on. With domain fine-tuning, the system manifests additional material. An early prototype of this technique was shown in this OpenAi cookbook: For the target domain, text was embedded using an API, and then when using the LLM, embeddings were retrieved using semantic similarity search to formulate an answer. Although this approach evolved to retrieval-augmented generation, its still a technique to adapt a Gen2 (2020) or Gen3 (2022) LLM into a question-answering system.

https://github.com/openai/openai-cookbook/blob/main/examples...

It sounds like you are trying to tweak embeddings for similarity search. Rather than fine-tune the model's layers, you may want to try training a linear transformation the existing model's output embedding. Openai has a cookbook on how to do that. You will need some data though - but I think you can try it with ~20 pieces of synthetically generated data.

tutorial from OpenAI https://github.com/openai/openai-cookbook/blob/main/examples/Question_answering_using_embeddings.ipynb

OpenAI Cookbook

OpenAI community repo with lots of examples: https://github.com/openai/openai-cookbook