GitHub - Differentiable sign/round/floor for PyTorch
by•
PyTorch's sign, round, floor, and argmax break gradients. Existing fixes (STE, Gumbel-Softmax) require rewriting your model.
SLL-Core is zero-intrusive:
with sll.linearize():
y = torch.sign(x) # now differentiable!
loss.backward()
Only linearizes inside an ε-band around decision boundaries. Everywhere else stays exactly hard. Zero deployment overhead.
pip install sll-core
p/many-displays
p/runevr
p/producthunt
p/google-cloud-platform
p/reso
p/kodlo
p/ai-context-flow
p/htmldrag
p/dex-by-exmergo
p/jot-2-0
p/beamfy
p/refs
p/dino-kids
p/nexupira
p/duckfightclub
p/nudgeforme
p/seshka
p/talanoa-2
p/murror
p/supabase
Replies