pytorch-minimize

pytorch-minimize copied to clipboard

Output the hess_inv from bfgs using the Optimizer API

1

Hi Reuben, I would like to ask if it is possible to extract the hessian_inv from BFGS using the Optimizer API (Minimizer) as is possible with the functional API?

richinex

Hi rfeinman, I want to thank you for this incredible package. I especially like the fact that I can pass in a custom call back function to minimize. I would...

richinex

Distribution on pypi

7

This is awesome! It'd be great if you could make this available on PyPI.

calvinmccarter

make `minimize` jit-able?

1

Thanks for this great package, it seems that`minimize` is not jit-able? for example ``` from torchmin import minimize import torch from functools import partial def bar(y, x): return (x +...

alibool

Multiple constraints minimization

7

First of all - thank you for the great repo! Now, I have a question: is there any support in multiple constraints minimization? It is often the case, since I...

dudifrid

Multiple GPUs

2

Can multiple gpus be utilized? Since input or model cannot be parallelized? Running trust_ncg on a model with mult-million rows input exceeds single GPU memory.

yubobao27

Memory leak when optimizing a pytorch module using hess or hessp.

1

Hello, When I ran "examples/train_mnist_Minimizer.py", the following warning was output. ``` /home/user/.pyenv/versions/py38-pytorch/lib/python3.8/site-packages/torch/autograd/__init__.py:200: UserWarning: Using backward() with create_graph=True will create a reference cycle between the parameter and its gradient which can...

tatsuhiko-inoue

Using _vmap in PyTorch to compute the Hessian-vector product (hvp) encounters a runtime error

1

Trying to use the minimize function with methods - trust-ncg - dogleg - newton-exact - trust-exact - trust-krylov But succeeds with the other methods. Presumably, the other methods aren't computing...

bfialkoff

LSMR fails with NaN output for trivial problems

2

Thanks for the nice library. I wanted to try LSMR but my first attempt to use it with a trivial problem failed with NaN output. Steps to reproduce: ```python import...

tvercaut

I've implemented Frank-Wolfe solvers for the constraints $||X||_{tr} < t$ and $X \in \mathcal{B}_n$, the Birkhoff polytope. In both cases, these are just using the vanilla Frank-Wolfe method. I've added...

calvinmccarter