Upwork is hiring a Llama 2 Finetuning and Walkthrough Project

Llama 2 Finetuning and Walkthrough Project

Upwork  ·  US  ·  $42k/yr - $62k/yr
over 1 year ago

Looking for experienced Machine Learning Developer for ongoing (project/consulting).

This project is aimed at finetuning an uncensored version of Llama 2 on custom data that will be provided. This assignment is a mix of project work and ML tutoring to help understand AI better. We are interested in opensource models with minimum dependency on third-party tools, preserving privacy with little to no call outs/api, data leaks.

Requirements:

-Must have deep understanding of current Opensource LLM's especially Meta's Llama 2.

-Troubleshooting common finetuning errors, runtime, loading errors, output errors, hallucinations.

-Hardware requirements for local storage and inference needs

-Ability to speak to pros and cons of current capabilities

-Ability to speak to experimental approaches to finetuning

-Understanding of HuggingFace, CLI's like Oobabooga vs Gradio Interface , Langchain

Project Phases/Scope:

-Setup for local use primarily and cloud (Colab)

-Preprocess data from targeted webscraping, existing files with file types that are PDFs, CSV, txt.

-Fine Tune the Llama 2 model 7b and 13b on custom data (Helping Draft Step by Step Guide)

-Test/Evaluate Results

Project Outcome:

The ultimate goal is to achieve a finely-tuned 7b/13b Llama 2 model on custom data that can be effectively employed for chat and instructional purposes.

Project Time: Ongoing

Job is closed

This job is already closed and no longer accepting applicants, sorry.