Francisco Massa

If `affine` is set to false, the `weight` and `bias` [are `nil`](https://github.com/soumith/cudnn.torch/blob/master/SpatialBatchNormalization.lua#L24-L27), leading to an error in [createIODescriptors](https://github.com/soumith/cudnn.torch/blob/master/SpatialBatchNormalization.lua#L46). Maybe it would be better to assert in the constructor itself that...

Investigate if lr_scheduler from segmentation can use PyTorch's schedulers

8

Back when it was initially implemented in 2019, the LR scheduler in the segmentation reference scripts couldn't be implemented with native PyTorch schedulers, so we had to resort to `LambdaLR`...

Refactoring all the code to simplify things

5

This PR substitutes https://github.com/fmassa/object-detection.torch/pull/5. - Unified Fast-RCNN with the other frameworks (RCNN and SPP). - Tested Fast-RCNN and RCNN, both works as in the original implementations. Should test SPP again...

Fast rcnn

1

Implements fast-rcnn detection framework. This PR includes - a simple and efficient lua implementation for `ROIPooling` based on `SpatialAdaptiveMaxPooling` - train and test code - an image-centric `BatchProvider` Some code...

Improving README with more details

5

Adds more information about how the memory sharing is performed. cc @nicholas-leonard

This will help improve the memory optimizations in `training` mode, as for the moment the backward graph in `optimizeMemory` is very simplistic and has several redundant/unwanted links.

Should fix https://github.com/fmassa/optimize-net/issues/6 There is a lot of magic going on in here: - temporarily overwrite torch `__index` and `select` to keep track of the modules in which those functions...

The current way of generating the graph in `graphgen.lua` is to rely on the `input` and `self.output` of each module. The underlying assumption is that those tensors will be the...

`image` extension can't be loaded in fresh conda env with torchvision 0.13.1

2

### 🐛 Describe the bug Hi, Quick FYI that I'm getting this warning when installing torch and torchvision from a fresh conda env ``` /private/home/fmassa/.conda/envs/fairvit2/lib/python3.9/site-packages/torchvision/io/image.py:13: UserWarning: Failed to load image...

Consider removing `memory_efficient_fusion` from xformers

5

# 🐛 Bug According to @Chillee in https://github.com/pytorch/pytorch/issues/86020#issuecomment-1264119302 > Oh, we deleted the cache on AOTAutograd recently, so it doesn't recompile when the shape changes. The idea here is that...