Meet Shah

Indexing and shape incoherent with module conversion

4

**Minimum Reproducible Example** ``` import torch import torch2trt USE_TRT = True class Model(torch.nn.Module): def __init__(self): super().__init__() self.conv = torch.nn.Conv2d(32, 4, kernel_size=1) def forward(self, inp): out = self.conv(inp) return out def...

Upgrade to PyTorch 1.0

2

Planned updates * Upgrade to pytorch-1.0 * Allow support for models to have custom loss functions. * Pretrained models on S3 for common models-dataset pairs * Compatibility matrix for dataset...

Add support for torch / torch-like dataloader

1

All models do not fit in memory, thus masking a threaded dataloader makes sense. Either using torch.data.dataloader or using simple generators.

Initialization of FixedPoint<31,31> gives error.

2

The error is in the the constructor not being able to be overloaded. Steps for reproduction : Add FixedPoint in the examples.cpp and compile.

Currently, it gives incorrect values for some operations. Using pxfail for the same in py.test

Use scipy.misc.imread instead of PIL Image

1

WRT https://github.com/warmspringwinds/tf-image-segmentation/blob/master/tf_image_segmentation/utils/tf_records.py#L37 I think you should reconsider using scipy.misc.imread in place of PIL Image, because for VOC datasets, the PIL loads a grayscale image for the segmentation annotation, while a...