

Free GPU Workshop: Fine-Tune Your Own Text-to-SQL Model on an Nvidia H100 GPU
Fine-tune an open-source model to turn plain-English questions into SQL, on your own private GPU lab. No setup and no cloud account. Bring a laptop and a browser; we bring the GPU.
Every participant gets a free, private JupyterLab running on an NVIDIA H100 GPU slice (about 10 GB of GPU memory). It’s ready when you log in and removed after the session.
What you will do
▪ Test a small open model (Qwen2.5-Coder 1.5B) on 50 SQL questions and record its score
▪ Fine-tune it with LoRA on real question-to-SQL examples
▪ Run the same 50 questions again and compare your before and after scores
▪ Download your trained model weights (LoRA adapter) and the notebooks to keep
What you get
▪ A free H100 GPU lab for the session, set up by labagent, our open-source lab tool
▪ Working notebooks you can rerun on your own hardware or cloud
▪ Your own trained model, plus your before/after score on the same questions
Agenda
▪ Your GPU lab (5 min): log in and check your GPU
▪ Baseline (10 min): score the base model on 50 questions
▪ Fine-tune (20 min): train with LoRA and watch the loss
▪ Measure (10 min): rerun the 50 questions and compare
▪ Take it home and Q&A (15 min): download your adapter, see how the labs were built
Who it’s for
Developers, data engineers, analysts and students who know basic Python and SQL. No machine learning background needed.
Details
▪ Wednesday, December 16, 2026, 5:00 PM ET, 1 hour
▪ Online. The join link and your lab login are sent to registered attendees
▪ Free. Seats are limited to the number of GPU labs, so please register only if you plan to attend
Hosted by Chandan Kumar (TorontoAI, beCloudReady), with labs set up by labagent: becloudready/labagent