GitLab

Co-authored-by: Nicolas Hug <nicolashug@fb.com>

Summary: We would like to track the internal usage of TorchVision components. We're tracking [datasets usage](https://fburl.com/daiquery/dqpmemn3

) now. This diff expand the tracking to all models.

Reviewed By: fmassa

Differential Revision: D31441632

fbshipit-source-id: e26072e582ac9f832c2056307ebf0eccf2ed6c9c
Co-authored-by: Kai Zhang <kaizh@fb.com>

* Added Dropout parameter of Models

* Added argument description for dropout in MobileNet v2 and v3
Updated quantization/googlenet.py as per the changes in constructor in googlenet

* Moved new dropout parameter n the end

* Updated googlenet.py
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* add ufmt as code formatter

* cleanup

* quote ufmt requirement

* split imports into more groups

* regenerate circleci config

* fix CI

* clarify local testing utils section

* use ufmt pre-commit hook

* split relative imports into local category

* Revert "split relative imports into local category"

This reverts commit f2e224cde2008c56c9347c1f69746d39065cdd51.

* pin black and usort dependencies

* fix local test utils detection

* fix ufmt rev

* add reference utils to local category

* fix usort config

* remove custom categories sorting

* Run pre-commit without fixing flake8

* got a double import in merge
Co-authored-by: Nicolas Hug <nicolashug@fb.com>

using nn.init.trunc_normal_ instead of scipy.stats.truncnorm
Co-authored-by: Vincent Moens <vmoens@fb.com>

* [doc] add minimum input size for alexnet builder

* [doc] add minimum input size for vgg builder

* [doc] add minimum input size for squeezenet builder

* [doc] add minimum input size for densenet builder

* [doc] add minimum input size for inception_v3 builder

* [doc] add minimum input size for googlenet builder
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* Replacing all torch.jit.annotations with typing

* Replacing remaining typing

* style: Added annotation typing for googlenet

* fix: Removed duplicate typing

* refactor: Moved factory function after class definition to fix typing

* fix: Fixed annotation typing

* refactor: Removed un-necessary import

* fix: Fixed typing

* refactor: Moved back up helper function and quote typed it

* partially enable mypy for .models

* fix existing errors

* ignore error instead of using Union

* remove sys.version_info == 2

* remove sys.version_info < 3

* remove from __future__ imports

* fix googlenet no aux logits

* small fix
Co-authored-by: eellison <eellison@fb.com>

* quantizable googlenet

* Minor improvements

* Rename basic_conv2d with conv_block plus additional fixes

* More renamings and fixes

* Bugfix

* Fix missing import for mypy

* Add pretrained weights

* make googlnet scriptable

* Remove typing import in favor of torch.jit.annotations

* add inceptionnet

* flake fixes

* fix asssert true

* add import division for torchscript

* fix script compilation

* fix flake, py2 division error

* fix py2 division error

`self.fc1(x)` converts the shape of `x` into "N x 1024", and `self.fc2(x)` converts into "N x num_classes".

By adding `print(x.shape)` under each comment line, the console displays as follows (batch_size is 1):

```text
torch.Size([1, 2048])
torch.Size([1, 1024])
torch.Size([1, 1024])
torch.Size([1, 1024])
torch.Size([1, 1000])
```

Fix the old flatten method which use the size(0) to caculate the batch size, the old method will intruduce Gather opertion in the onnx output, which will faild parsed by tensorRT 5.0 (#1134)

I grepped the repo for Ouputs and these were the only occurences

* added progress flag to model getters

* flake8

* bug fix

* backward commpability

* add aux_logits support to inception

it is related to pytorch/pytorch#18668

* instantiate InceptionAux only when requested

it is related to pytorch/pytorch#18668

* revert googlenet

* support and aux_logits in pretrained models

* return namedtuple when aux_logit is True

* Match Tensorflows implementation of GoogLeNet

* just disable the branch when pretrained is true

* don't use legacy code

* Added dimensions in the comments

The update provides the dimensions of the processed data following the style of inceptionV3 implementation.

* Changed docs and comments

Updated doc with the argument `transform_input`. Modified comments to match inceptionV3 style.

* Add GoogLeNet (Inception v1)

* Fix missing padding

* Add missing ReLu to aux classifier

* Add Batch normalized version of GoogLeNet

* Use ceil_mode instead of padding and initialize weights using "xavier"

* Match BVLC GoogLeNet zero initialization of classifier

* Small cleanup

* use adaptive avg pool

* adjust network to match TensorFlow

* Update url of pre-trained model and add classification results on ImageNet

* Bugfix that improves performance by 1 point