capeblog

I am Thomas, Machine Learning engineer at Weights and Biases working on the AI Applied Team.

These days I build AI applications for the enterprise. Mostly supporting Weights & Biases customers trying to get started on GenAI.

Mont Granier, Savoie, France

🙃 Articles

Custom Huggingface Pipeline

transformers

huggingface

pytorch

How to create a custom pipeline for a model.

Flagging Toxic Content on Bluesky

LLMs

Bluesky

How to flag toxic content on Bluesky using an LLM toxicity scorer

How to Fine-tune an LLM Part 3: The HuggingFace Trainer

LLMs

PyTorch

HF

SFT

Exploring how to get the best out of the Hugging Face Trainer and subclasses

How to Fine-Tune an LLM Part 2: Instruction Tuning Llamas

LLMs

PyTorch

HF

SFT

Instruction tuning in pure PyTorch

Translating Weights & Biases Documentation with GPT-4

LLMs

GPT

OpenAI

How to use GPT-4 to translate Weights & Biases documentation to Japanese.

How to Fine-Tune an LLM Part 1: Preparing a Dataset for Instruction Tuning

LLMs

PyTorch

HF

SFT

How to prepare a dataset for instruction tuning.

How to Run LLMs Locally With llama-cpp and GGML

LLMs

llama-cpp

GGML

GGUF and llama-cpp are a game changer for running LLMs locally.

Diffusion on the Clouds: Short-term solar energy forecasting with Diffusion Models

Diffusion

PyTorch

Using diffusion models to forecast solar energy production.

Is the New M2Pro Mac Mini a Deep Learning Workstation?

PyTorch

Apple

MPS

Testing the new M2Pro Mac Mini as a deep learning workstation.

How To Train a Conditional Diffusion Model From Scratch

Diffusion

PyTorch

How to train and evaluate a conditional diffusion model in pure PyTorch.

Export Segmentation Masks

Pytorch

fastai

cv

How to save the outputs of your model to disk

Moving MNIST

Pytorch

fastai

cv

How to use fastai to predict the movement of numbers

Using fastai on sequences of Images

Pytorch

fastai

cv

How to use fastai to train an image sequence to image sequence job.

The Devil lives in the details

Pytorch

fastai

Resizing method matters…