Hi PyTorch Team!
I am Sundaram Gupta, a student from India. I have developed a library called torchquery built on top of PyTorch (available on PyPI).
I wanted to share my logic for Neural Healing and Vectorized Querying. Instead of using loops, I am using hardware-accelerated masking to find and fix Inf and NaN numbers in billion-scale tensors.
Hi! Can you share a link to your library? I didn’t find it on google.
"Hi Valérian! Thanks for your patience; my account was recently under review by the forum staff.
The project is quite new, so Google is still indexing it. You can find the library here:
I am currently architecting v.3 to include proactive Governance features like tq.check() for real-time tensor health monitoring. I’d love to hear your thoughts on the vectorized engine logic!"