Functions — github.com/KinglittleQ/torch-batch-svd

CodeHub MCPcopy Create free account

hub / github.com/KinglittleQ/torch-batch-svd / functions

Functions12 in github.com/KinglittleQ/torch-batch-svd

⨍ Functions 12 ◇ Types & classes 2

↓ 7 callersMethodbackward

(ctx, grad_u: torch.Tensor, grad_s: torch.Tensor, grad_v: torch.Tensor)

torch_batch_svd/batch_svd.py:57

↓ 2 callersFunctionbench_speed

benchs/bench_speed.py:5

↓ 2 callersFunctionunique_allocate

, class A = Status(*)(P), class D = Status(*)(T)>

torch_batch_svd/include/utils.h:25

FunctionPYBIND11_MODULE

torch_batch_svd/csrc/bindings.cpp:3

Functionbatch_svd_backward

https://j-towns.github.io/papers/svd-derivative.pdf This makes no assumption on the signs of sigma.

torch_batch_svd/csrc/torch_batch_svd.cpp:118

Functionbatch_svd_forward

solve U S V = svd(A) a.k.a. syevj, where A (b, m, n), U (b, m, m), S (b, min(m, n)), V (b, n, n) see also https://docs.nvidia.com/cuda/cusolver/index

torch_batch_svd/csrc/torch_batch_svd.cpp:14

This function returns `(U, S, V)` which is the singular value decomposition of a input real matrix or batches of real matrice

torch_batch_svd/batch_svd.py:8

Methodtest_double

tests/test.py:33

Methodtest_float

tests/test.py:12

Methodtest_half

tests/test.py:59

Methodtest_multiple_gpus

tests/test.py:78

Functionunique_cuda_ptr

torch_batch_svd/include/utils.h:34