Making small updates to docs and readme (facebookresearch#158)

Summary: Pull Request resolved: facebookresearch#158 Reviewed By: bottler Differential Revision: D26125924 Pulled By: prigoyal fbshipit-source-id: 8477a56c94e45a0bab0c69f2c50ba1d1fffe6d34
johnnynunez · Jan 28, 2021 · 4e50c49 · 4e50c49
1 parent cbb06ff
commit 4e50c49
Show file tree

Hide file tree

Showing 5 changed files with 13 additions and 13 deletions.
diff --git a/INSTALL.md b/INSTALL.md
@@ -33,7 +33,7 @@ This assumes you have conda 10.2.
 conda create -n vissl python=3.8
 conda activate vissl
 conda install -c pytorch pytorch=1.7.1 torchvision cudatoolkit=10.2
-conda install -c vissl apex vissl
+conda install -c vissl -c iopath -c conda-forge -c pytorch -c defaults apex vissl
 ```
 
 The package also contains code for the fairscale and ClassyVision libraries. Ensure you do not have them installed separately.

diff --git a/MODEL_ZOO.md b/MODEL_ZOO.md
@@ -174,9 +174,6 @@ To reproduce the numbers below, the experiment configuration is provided in json
 
 To reproduce the numbers below, the experiment configuration is provided in json format for each model [here](https://github.com/facebookresearch/vissl/tree/master/configs/config/model_zoo/benchmark_in1k_linear_deepclusterv2_swav.json).
 
-In the original SwAV work, authors evaluate several model checkpoints from a pre-training and evaluate the result of the best checkpoint. In the evaluations in VISSL, we used the last checkpoint.
-Further, there is also some standard deviation in linear results if we run the same eval several times. The evals reported below are for 1 run.
-
 | Method | Model | PreTrain dataset | ImageNet top-1 acc. | URL |
 | ------ | ----- | ---------------- | ------------------- | --- |
 | [DeepClusterV2](https://arxiv.org/abs/2006.09882)  |  [RN50 - 400 epochs - 2x224](https://github.com/facebookresearch/swav#model-zoo)       |  ImageNet-1K  | 70.01 | [model](https://dl.fbaipublicfiles.com/vissl/model_zoo/deepclusterv2_400ep_2x224_pretrain.pth.tar)
@@ -187,8 +184,7 @@ Further, there is also some standard deviation in linear results if we run the s
 
 To reproduce the numbers below, the experiment configuration is provided in json format for each model [here](https://github.com/facebookresearch/vissl/tree/master/configs/config/model_zoo/benchmark_in1k_linear_deepclusterv2_swav.json).
 
-In the original SwAV work, authors evaluate several model checkpoints from a pre-training and evaluate the result of the best checkpoint. In the evaluations in VISSL, we used the last checkpoint.
-Further, there is also some standard deviation in linear results if we run the same eval several times. The evals reported below are for 1 run.
+There is some standard deviation in linear results if we run the same eval several times and pre-train a SwAV model several times. The evals reported below are for 1 run.
 
 | Method | Model | PreTrain dataset | ImageNet top-1 acc. | URL |
 | ------ | ----- | ---------------- | ------------------- | --- |

diff --git a/README.md b/README.md
@@ -3,7 +3,7 @@
 [![CircleCI](https://circleci.com/gh/facebookresearch/vissl.svg?style=svg&circle-token=f15ded7b718589ad3f150355e1c37f8e74516019)](https://circleci.com/gh/facebookresearch/vissl)[![PRs Welcome](https://img.shields.io/badge/PRs-welcome-brightgreen.svg)](https://github.com/facebookresearch/vissl/blob/master/.github/CONTRIBUTING.md)
 
 ## Introduction
-VISSL is a computer **VI**sion library for state-of-the-art **S**elf-**S**upervised **L**earning research with [PyTorch](https://pytorch.org). VISSL aims to accelerate research cycle in self-supervised learning: from designing a new self-supervised task to evaluating the learned representations. Within Facebook AI, VISSL has been used to power research projects such as [SwAV](https://arxiv.org/abs/1906.02739). Key features include:
+VISSL is a computer **VI**sion library for state-of-the-art **S**elf-**S**upervised **L**earning research with [PyTorch](https://pytorch.org). VISSL aims to accelerate research cycle in self-supervised learning: from designing a new self-supervised task to evaluating the learned representations. Key features include:
 
 - **Reproducible implementation of SOTA in Self-Supervision**: All existing SOTA in Self-Supervision are implemented - [SwAV](https://arxiv.org/abs/2006.09882), [SimCLR](https://arxiv.org/abs/2002.05709), [MoCo(v2)](https://arxiv.org/abs/1911.05722), [PIRL](https://arxiv.org/abs/1912.01991), [NPID](https://arxiv.org/abs/1912.01991), [NPID++](https://arxiv.org/abs/1912.01991), [DeepClusterV2](https://arxiv.org/abs/2006.09882), [ClusterFit](https://openaccess.thecvf.com/content_CVPR_2020/papers/Yan_ClusterFit_Improving_Generalization_of_Visual_Representations_CVPR_2020_paper.pdf), [RotNet](https://arxiv.org/abs/1803.07728.), [Jigsaw](https://arxiv.org/abs/1603.09246). Also supports supervised trainings.
 
@@ -23,10 +23,12 @@ See [`INSTALL.md`](https://github.com/facebookresearch/vissl/blob/master/INSTALL
 
 ## Getting Started
 
-Follow the [installation](https://github.com/facebookresearch/vissl/blob/master/INSTALL.md) instructions to install vissl.
+Install VISSL by following the [installation instructions](https://github.com/facebookresearch/vissl/blob/master/INSTALL.md).
 After installation, please see [Getting Started with VISSL](https://github.com/facebookresearch/vissl/blob/master/GETTING_STARTED.md) and the [Colab Notebook](https://colab.research.google.com/drive/1iigQmKL_DUuBLT6BqjrGXlW9ZIqKIFmt?usp=sharing) to learn about basic usage.
 
-Learn more about VISSL at our [documentation](https://vissl.readthedocs.org). And see the [projects/](projects/) for some projects built on top of VISSL.
+## Documentation
+
+Learn more about VISSL at our [documentation](https://vissl.readthedocs.io). And see the [projects/](projects/) for some projects built on top of VISSL.
 
 ## Tutorials
 
@@ -60,7 +62,9 @@ If you find VISSL useful in your research or wish to refer to the baseline resul
 
 ```BibTeX
 @misc{goyal2021vissl,
-  author =       {Priya Goyal and Benjamin Lefaudeux and Mannat Singh and Jeremy Reizenstein and Vinicius Reis and Min Xu and and Mathilde Caron and Piotr Bojanowski and Armand Joulin and Ishan Misra},
+  author =       {Priya Goyal and Benjamin Lefaudeux and Mannat Singh and Jeremy Reizenstein and Vinicius Reis and 
+                  Min Xu and and Matthew Leavitt and Mathilde Caron and Piotr Bojanowski and Armand Joulin and 
+                  Ishan Misra},
   title =        {VISSL},
   howpublished = {\url{https://github.com/facebookresearch/vissl}},
   year =         {2021}
@@ -71,4 +75,4 @@ If you find VISSL useful in your research or wish to refer to the baseline resul
 
 Below we share, in reverse chronological order, the updates and new releases in VISSL. All VISSL releases are available [here](https://github.com/facebookresearch/vissl/releases).
 
-**[January 2021]**: VISSL v0.1 released with [blog post announcement]().
+**[January 2021]**: VISSL v0.1 released.
diff --git a/docs/source/installation.rst b/docs/source/installation.rst
@@ -28,7 +28,7 @@ This assumes you have conda 10.2.
     conda create -n vissl python=3.8
     conda activate vissl
     conda install -c pytorch pytorch=1.7.1 torchvision cudatoolkit=10.2
-    conda install -c vissl apex vissl
+    conda install -c vissl -c iopath -c conda-forge -c pytorch -c defaults apex vissl
 
 For other versions of PyTorch, Python, CUDA, please modify the above instructions with the
 desired version. VISSL provides Apex packages for all combinations of pytorch, python and compatible cuda.

diff --git a/docs/source/what_is_vissl.rst b/docs/source/what_is_vissl.rst
@@ -4,7 +4,7 @@ What is VISSL?
 .. image:: _static/img/vissl-logo.png
 
 
-VISSL is a computer VIsion library for state-of-the-art Self-Supervised Learning research with `PyTorch <https://pytorch.org>`_. VISSL aims to accelerate research cycle in self-supervised learning: from designing a new self-supervised task to evaluating the learned representations. Within Facebook AI, VISSL has been used to power research projects such as `SwAV <https://arxiv.org/abs/1906.02739>`_. Key features include:
+VISSL is a computer VIsion library for state-of-the-art Self-Supervised Learning research with `PyTorch <https://pytorch.org>`_. VISSL aims to accelerate research cycle in self-supervised learning: from designing a new self-supervised task to evaluating the learned representations. Key features include:
 
 - :code:`Reproducible implementation of SOTA in Self-Supervision`: All existing SOTA in Self-Supervision are implemented - `SwAV <https://arxiv.org/abs/2006.09882>`_, `SimCLR <https://arxiv.org/abs/2002.05709>`_, `MoCo(v2) <https://arxiv.org/abs/1911.05722>`_, `PIRL <https://arxiv.org/abs/1912.01991>`_, `NPID <https://arxiv.org/abs/1912.01991>`_, `NPID++ <https://arxiv.org/abs/1912.01991>`_, `DeepClusterV2 <https://arxiv.org/abs/2006.09882>`_, `ClusterFit <https://openaccess.thecvf.com/content_CVPR_2020/papers/Yan_ClusterFit_Improving_Generalization_of_Visual_Representations_CVPR_2020_paper.pdf>`_, `RotNet <https://arxiv.org/abs/1803.07728>`_, `Jigsaw <https://arxiv.org/abs/1603.09246>`_. Also supports supervised trainings.