thomasahle/tensorgrad
View on GitHub ↗ ★ 373 Python
Machine Learning with Symbolic Tensors
5 mentions on Hacker News by 1 person: 0 stories, 5 comments · first 2024-08-05 · latest 2025-11-02
20082026
-
…"x_channels", "y_channels"]: return x.unsqueeze(-1) * y.unsqueeze(-2) There's also https://github.com/thomasahle/tensorgrad which uses sympy for "axis" dimension variables: b, x, y = sp.symbols("b x y") X = tg.Variable("X", b, x) Y =…
comment by thomasahle · 2025-11-02 · in Why don't you use dependent types? -
This is neat! I've been thinking about how to best express recursion with diagrams for tensorgrad ( https://github.com/thomasahle/tensorgrad ) and this might just be it. So much computation these days are done on tensors, mostly for easy…
comment by thomasahle · 2025-02-18 · in Tensor evolution: A framework for fast tensor computations using recurrences -
…to it (and manim back then was less well-documented than it is now). Manim is on the TODO list :-) https://github.com/thomasahle/tensorgrad/issues/7 If you feel like writing some code, I think it could work very well with tensorgrad's…
comment by thomasahle · 2025-01-07 · in Einsum in Depth -
…notation. I also wrote this tensorgrad library to do symbolic calculations using tensor networks: https://github.com/thomasahle/tensorgrad it keeps track of what you are taking derivatives with respect to. But I agree it would be nice to…
comment by thomasahle · 2025-01-06 · in Einsum in Depth -
…been working on a library for doing automatic manipulation and differentiation of tensor diagrams ( https://github.com/thomasahle/tensorgrad ), and to me they are clearly a cleaner notation. For a beautiful introduction to tensor networks,…
comment by thomasahle · 2024-08-05 · in A new type of neural network is more interpretable