ResNeSt plus (zhanghang1989#256)

.github/workflows/pythonpublish.yml → .github/workflows/pypi.yml

@@ -4,9 +4,8 @@
 name: Upload Python Package
 
 on:
-  push:
-    branches:
-      - master
+  schedule:
+    - cron: "0 12 * * *"
 
 jobs:
   deploy:

.github/workflows/build_sphix_master.yml → .github/workflows/sphix_build_master.yml

@@ -47,5 +47,5 @@ jobs:
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
         with:
-          build_dir: "docs/build/html/*"
+          build_dir: docs/build/html/
           target_branch: gh-pages

.github/workflows/sphix_build_pr.yml

@@ -9,16 +9,21 @@ on:
 
 jobs:
   docs:
-    runs-on: ubuntu-latest
+    runs-on: self-hosted
     steps:
       - uses: actions/checkout@v2
       - uses: seanmiddleditch/gha-setup-ninja@master
-        
-      - name: Set up Python
-        uses: actions/setup-python@v1
+
+      - name: Set PR Number
+        uses: actions/github-script@0.3.0
         with:
-          python-version: 3.7
-
+          github-token: ${{github.token}}
+          script: |
+            const core = require('@actions/core')
+            const prNumber = context.payload.number;
+            core.exportVariable('PULL_NUMBER', prNumber);
+            core.exportVariable("PATH", "/home/ubuntu/anaconda3/bin:/usr/local/bin:/usr/bin/:/bin:$PATH")
+        
       - name: Install dependencies
         run: |
           python -m pip install --upgrade pip
@@ -39,35 +44,11 @@ jobs:
           cd docs/
           make html
           touch build/html/.nojekyll
-          
-      - name: Set PR Number
-        uses: actions/github-script@0.3.0
-        with:
-          github-token: ${{github.token}}
-          script: |
-            const core = require('@actions/core')
-            const prNumber = context.payload.number;
-            core.exportVariable('PULL_NUMBER', prNumber);
-        
-
-      # https://github.com/marketplace/actions/github-pages
-      - name: Deploy
-        if: success()
-        uses: jakejarvis/s3-sync-action@master
-        with:
-          args: --acl public-read --follow-symlinks --delete
-        env:
-          AWS_S3_BUCKET: ${{ secrets.AWS_S3_BUCKET }}
-          AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
-          AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
-          AWS_REGION: ${{ secrets.AWS_REGION }}
-          DEST_DIR: "${{ secrets.DEST_DIR }}/${PULL_NUMBER}"
-          SOURCE_DIR: 'docs/build/html/'
-
+          aws s3 sync build/html/ s3://hangzh/encoding/docs/${{ env.PULL_NUMBER }}/ --acl public-read --follow-symlinks --delete
           
       - name: Comment
         if: success()
         uses: thollander/actions-comment-pull-request@master
         with:
-          message: "The docs are uploaded and can be previewed at http://${{ secrets.AWS_S3_BUCKET }}.s3.amazonaws.com/${{ secrets.DEST_DIR }}/${{ env.PULL_NUMBER }}/index.html"
+          message: "The docs are uploaded and can be previewed at http://hangzh.s3.amazonaws.com/encoding/docs/${{ env.PULL_NUMBER }}/index.html"
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

.gitignore

@@ -8,3 +8,6 @@ docs/src/
 docs/html/
 encoding/_ext/
 encoding.egg-info/
+*.o
+*.so
+*.ninja*

README.md

@@ -1,7 +1,8 @@
 [![PyPI](https://img.shields.io/pypi/v/torch-encoding.svg)](https://pypi.python.org/pypi/torch-encoding)
-[![PyPI Pre-release](https://img.shields.io/badge/pypi--prerelease-v1.1.0-ff69b4.svg)](https://pypi.org/project/torch-encoding/#history)
+[![PyPI Pre-release](https://img.shields.io/badge/pypi--prerelease-v1.2.0-ff69b4.svg)](https://pypi.org/project/torch-encoding/#history)
 [![Upload Python Package](https://github.com/zhanghang1989/PyTorch-Encoding/workflows/Upload%20Python%20Package/badge.svg)](https://github.com/zhanghang1989/PyTorch-Encoding/actions)
 [![Downloads](http://pepy.tech/badge/torch-encoding)](http://pepy.tech/project/torch-encoding)
+[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
 [![Build Docs](https://github.com/zhanghang1989/PyTorch-Encoding/workflows/Build%20Docs/badge.svg)](https://github.com/zhanghang1989/PyTorch-Encoding/actions)
 # PyTorch-Encoding
 
@@ -11,10 +12,23 @@ created by [Hang Zhang](http://hangzh.com/)
 
 - Please visit the [**Docs**](http://hangzh.com/PyTorch-Encoding/) for detail instructions of installation and usage. 
 
-- Please visit the [link](http://hangzh.com/PyTorch-Encoding/experiments/segmentation.html) to examples of semantic segmentation.
+- Please visit the [link](http://hangzh.com/PyTorch-Encoding/model_zoo/imagenet.html) to image classification models.
+
+- Please visit the [link](http://hangzh.com/PyTorch-Encoding/model_zoo/segmentation.html) to semantic segmentation models.
 
 ## Citations
 
+**ResNeSt: Split-Attention Networks** [[arXiv]]()  
+  [Hang Zhang](http://hangzh.com/), Chongruo Wu, Zhongyue Zhang, Yi Zhu, Zhi Zhang, Haibin Lin, Yue Sun, Tong He, Jonas Muller, R. Manmatha, Mu Li and Alex Smola
+```
+@article{zhang2020resnest,
+title={ResNeSt: Split-Attention Networks},
+author={Zhang, Hang and Wu, Chongruo and Zhang, Zhongyue and Zhu, Yi and Zhang, Zhi and Lin, Haibin and Sun, Yue and He, Tong and Muller, Jonas and Manmatha, R. and Li, Mu and Smola, Alexander},
+journal={arXiv preprint},
+year={2020}
+}
+```
+
 **Context Encoding for Semantic Segmentation** [[arXiv]](https://arxiv.org/pdf/1803.08904.pdf)  
  [Hang Zhang](http://hangzh.com/), [Kristin Dana](http://eceweb1.rutgers.edu/vision/dana.html), [Jianping Shi](http://shijianping.me/), [Zhongyue Zhang](http://zhongyuezhang.com/), [Xiaogang Wang](http://www.ee.cuhk.edu.hk/~xgwang/), [Ambrish Tyagi](https://scholar.google.com/citations?user=GaSWCoUAAAAJ&hl=en), [Amit Agrawal](http://www.amitkagrawal.com/)
 ```

docs/source/index.rst

@@ -13,24 +13,30 @@ An optimized PyTorch package with CUDA backend.
 .. toctree::
    :glob:
    :maxdepth: 1
-   :caption: Notes
+   :caption: Installation
 
    notes/*
 
 .. toctree::
    :glob:
    :maxdepth: 1
-   :caption: Experiment Systems
+   :caption: Model Zoo
 
-   experiments/*
+   model_zoo/*
+
+.. toctree::
+   :glob:
+   :maxdepth: 1
+   :caption: Other Tutorials
+
+   tutorials/*
 
 .. toctree::
    :maxdepth: 1
    :caption: Package Reference
 
    nn
    parallel
-   models
    utils
 
 Indices and tables

docs/source/model_zoo/imagenet.rst

@@ -0,0 +1,83 @@
+Image Classification
+====================
+
+Install Package
+---------------
+
+- Clone the GitHub repo::
+    
+    git clone https://github.com/zhanghang1989/PyTorch-Encoding
+
+- Install PyTorch Encoding (if not yet). Please follow the installation guide `Installing PyTorch Encoding <../notes/compile.html>`_.
+
+Get Pre-trained Model
+---------------------
+
+.. hint::
+    How to get pretrained model, for example ``ResNeSt50``::
+
+        model = encoding.models.get_model('ResNeSt50', pretrained=True)
+
+    After clicking ``cmd`` in the table, the command for training the model can be found below the table.
+
+.. role:: raw-html(raw)
+   :format: html
+
+
+ResNeSt
+~~~~~~~
+
+.. note::
+    The provided models were trained using MXNet Gluon, this PyTorch implementation is slightly worse than the original implementation.
+
+===============================  ==============    ==============    =========================================================================================================
+Model                            crop-size         Acc               Command                                                                                      
+===============================  ==============    ==============    =========================================================================================================
+ResNeSt-50                       224               81.03             :raw-html:`<a href="javascript:toggleblock('cmd_resnest50')" class="toggleblock">cmd</a>`
+ResNeSt-101                      256               82.83             :raw-html:`<a href="javascript:toggleblock('cmd_resnest101')" class="toggleblock">cmd</a>`
+ResNeSt-200                      320               83.84             :raw-html:`<a href="javascript:toggleblock('cmd_resnest200')" class="toggleblock">cmd</a>`
+ResNeSt-269                      416               84.54             :raw-html:`<a href="javascript:toggleblock('cmd_resnest269')" class="toggleblock">cmd</a>`
+===============================  ==============    ==============    =========================================================================================================
+
+.. raw:: html
+
+    <code xml:space="preserve" id="cmd_resnest50" style="display: none; text-align: left; white-space: pre-wrap">
+    # change the rank for worker node
+    python train_dist.py --dataset imagenet --model resnest50 --lr-scheduler cos --epochs 270 --checkname resnest50 --lr 0.025 --batch-size 64 --dist-url tcp://MASTER:NODE:IP:ADDRESS:23456 --world-size 4 --label-smoothing 0.1 --mixup 0.2 --no-bn-wd --last-gamma --warmup-epochs 5 --rand-aug --rank 0
+    </code>
+
+    <code xml:space="preserve" id="cmd_resnest101" style="display: none; text-align: left; white-space: pre-wrap">
+    # change the rank for worker node
+    python train_dist.py --dataset imagenet --model resnest101 --lr-scheduler cos --epochs 270 --checkname resnest101 --lr 0.025 --batch-size 64 --dist-url tcp://MASTER:NODE:IP:ADDRESS:23456 --world-size 4 --label-smoothing 0.1 --mixup 0.2 --no-bn-wd --last-gamma --warmup-epochs 5 --rand-aug --rank 0
+    </code>
+
+    <code xml:space="preserve" id="cmd_resnest200" style="display: none; text-align: left; white-space: pre-wrap">
+    # change the rank for worker node
+    python train_dist.py --dataset imagenet --model resnest200 --lr-scheduler cos --epochs 270 --checkname resnest200 --lr 0.0125 --batch-size 32 --dist-url tcp://MASTER:NODE:IP:ADDRESS:23456 --world-size 8 --label-smoothing 0.1 --mixup 0.2 --no-bn-wd --last-gamma --warmup-epochs 5 --rand-aug --crop-size 256 --rank 0
+    </code>
+
+    <code xml:space="preserve" id="cmd_resnest269" style="display: none; text-align: left; white-space: pre-wrap">
+    # change the rank for worker node
+    python train_dist.py --dataset imagenet --model resnest269 --lr-scheduler cos --epochs 270 --checkname resnest269 --lr 0.0125 --batch-size 32 --dist-url tcp://MASTER:NODE:IP:ADDRESS:23456 --world-size 8 --label-smoothing 0.1 --mixup 0.2 --no-bn-wd --last-gamma --warmup-epochs 5 --rand-aug --crop-size 320 --rank 0
+    </code>
+
+Test Pretrained
+~~~~~~~~~~~~~~~
+
+- Prepare the datasets by downloading the data into current folder and then runing the scripts in the ``scripts/`` folder::
+
+      python scripts/prepare_imagenet.py --data-dir ./
+  
+- The test script is in the ``experiments/recognition/`` folder. For evaluating the model (using MS),
+  for example ``ResNeSt50``::
+
+      python test.py --dataset imagenet --model-zoo ResNeSt50 --crop-size 224 --eval
+
+Train Your Own Model
+--------------------
+
+- Prepare the datasets by downloading the data into current folder and then runing the scripts in the ``scripts/`` folder::
+
+    python scripts/prepare_imagenet.py --data-dir ./
+
+- The training script is in the ``experiments/recognition/`` folder. Commands for reproducing pre-trained models can be found in the table.