Vizuara Finetuning Workshop

The workshop is titled “LLM Finetuning Theory, Production and Deployment”.

Instructor: Dr. Raj Dandekar, MIT PhD

Timings: 8am IST on Saturday, Sunday (6 Sep, 7 Sep, 13 Sep, 14 Sep)

Here is what we will cover:

Part 1: Foundations

(a) Fine-tune an LLM from Scratch

Part 2: Practicals

The 3 pillars of modern LLM Finetuning: HuggingFace, Unsloth, JAX

(a) Finetune a TinyStories model using HuggingFace Trainer

(b) Finetune a Llama model using Unsloth Trainer

(c) Finetune a multimodal Gemma model using JAX Trainer

Part 3: Research

For both the below projects, we will use the OpenAI Finetuning Dashboard.

(a) Build an end-to-end research project published by Anthropic titled "Subliminal Learning"

(b) Build an end-to-end research project on RAFT (Retrieval Augmented Finetuning)

See you then!