HuggingFace Spaces by Midjourney This concludes the two-part blog entry on turning HuggingFace into a deep learning playground (and we have not even talked about all the LMs they host!). Mostly it will be about Spaces, but we are starting with models. Models We train tons of models as we experiment with hyperparameters or redesign … Continue reading HuggingFace. Models. Spaces. Part 2 →

https://youtu.be/fpVfTyPyPgE I have recently posted a video briefly explaining how we use GNNs (Graph Neural Nets) at Fetch to identify duplicate receipts. I have mentioned HuggingFace in passing there, but in fact I use it quite heavily to make my machine learning life simpler. The Problem: Suitcases Without Handles AI workloads are data heavy, involving … Continue reading HuggingFace. The Perfect Lab. Part 1 →

No training implementation is complete until it allows training on a cluster where each machine has multiple GPUs. Multi-node/Multi-GPU Training with PyTorch Lightning SageMaker does a great job enabling this in Script Mode, and all we have to do is write code that supports SageMaker SMDDP implementation of the distributed training DDP protocol. PyTorch Lighting … Continue reading Amazon SageMaker: Distributed Training →

At Fetch we reward you for taking pictures of store and restaurant receipts. Our app needs to read and understand crumpled, dark, smudged, warped, skewed, creased, you get the "picture" images, taken in cars, in your lap, on the way out, while walking the dog, taking out the trash, doing your nails, etc.., etc. Not … Continue reading Amazon SageMaker: What Tutorials Don’t Teach →

So I was bored, didn't know what to do, started solving problems on codewars (shut up!). There was this little problem there, where they are asking you to compute $latex \frac{\sum_{i=1}^n i!}{n!}&s=2$, for potentially large n's with $latex 10^{-6}$ precision. It's easy to see that this needs to be refactored (no need to actually compute … Continue reading F# Humor (kinda) →