Fine-Tuning Open-Weight Models: A Hands-On Deep Learning Primer
Fine-Tuning Open-Weight Models: A Hands-On Deep Learning Primer
Most developers can prompt. Few know how to fine-tune.
In this free live workshop with Ravin Kumar (Deepmind, ex-SpaceX, PyMC contributor) and Hugo Bowne-Anderson (Vanishing Gradients), you’ll learn how to fine-tune small open-weight models,and in the process, pick up the deep learning essentials every builder should know.
Fine-tuning isn’t just about making a model sound different. It’s a window into how neural networks actually work: parameters, embeddings, loss curves, and evaluation.
In this hands-on session, you’ll:
🔀 Learn when to system prompt vs. fine-tune
🤔 Reason when to use a large frontier model or small model for your task
📉 Interpret train time metrics loss curves like a practitioner
🧠 Understand parameter sizes, embeddings, and training basics
🛠️ Fine-tune a 270M open-weight model in Colab
📊 Evaluate whether your fine-tune is really working
Workshop Flow
We’ll build step by step:
⚙️ Phase 0: Setup — run everything in Colab (free GPU/TPU provided).
📝 Phase 1: Prompting vs. Fine-Tuning — when each approach makes sense.
🏗️ Phase 2: Fine-tuning a 270M model — hands-on with real code.
📈 Phase 3: Deep Learning Insights — how to read training curves, think about architecture, and debug.
🧪 Phase 4: Evaluation — from vibe checks to measurable performance.
What You’ll Learn
Why (and when) you’d use a small model instead of a large API model.
How to fine-tune an open-weight model in practice.
Core deep learning knowledge: embeddings, parameter sizes, loss functions, training curves.
How to evaluate and validate fine-tunes for your use case.
Requirements
💻 Runs in Colab out of the box (free GPU/TPU).
⚡ Optional: run locally with your own Apple or NVIDIA GPU (CPU possible but very slow).
📦 No prior deep learning experience required: just curiosity.
This is the next step in our workshop series:
1️⃣ We started with prompting and local LLM apps.
2️⃣ Now, we go deeper with fine-tuning and deep learning foundations.
3️⃣ Coming soon: agents and beyond.
Bring your laptop. This is fully hands-on.
If you can’t make it, register and we’ll share the recording after.