add cifar download bash

dividemix/README.md

@@ -1,5 +1,19 @@
-# DivideMix With Winning Tickets
-We add some commands for either CLK or SAME by refering to the original code from https://github.com/LiJunnan1992/DivideMix.
+# Semi-Supervised Learning based Approaches
+Overall code structure is made from original DivideMix repo.
+We refer to https://github.com/LiJunnan1992/DivideMix.
+
+## Dataset
+### CIFAR10, CIFAR100
+To download CIFAR dataset, just follow
+
+```
+bash scripts/download_cifar10.sh
+
+bash scripts/download_cifar100.sh
+```
+
+### Clothing1M
+
 
 ## Training
 To run scripts or python file, you have to download cifar-10 or cifar-100 datasets, firstly.
@@ -17,7 +31,7 @@ Run the followings:
 
     bash scripts/refinement_dynamic_cifar10
 
-### arguments
+<!-- ### arguments
 Default arguments settings are set for cifar10 experiments. Usage and other arguments are same with DivideMix original code. Options below are added only for our paper.
 
 ```
@@ -29,6 +43,6 @@ usage : python Train_cifar.py [--distill] [--distill_mode] [--refinement]
         --refinement : if not, the clean probability of all clean subset is set to one. In our paper, all experiments always use this option to make same condition with original paper.
 ```
 
-
+ -->
 <b>License</b>\
 This project is licensed under the terms of the MIT license.

dividemix/Train_cifar.py

@@ -422,11 +422,8 @@ def extract_cleanidx(model, loader, mode='fine-kmeans', p_threshold=0.6):
 if args.dataset=='cifar10':
     warm_up = 20
 elif args.dataset=='cifar100':
-<<<<<<< Updated upstream
     warm_up = 50
-=======
-    warm_up = 45
->>>>>>> Stashed changes
+
 
 loader = dataloader.cifar_dataloader(args.dataset,r=args.r,noise_mode=args.noise_mode,batch_size=args.batch_size,num_workers=5,\
     root_dir=args.data_path,log=stats_log,noise_file='%s/%.1f_%s.json'%(args.data_path,args.r,args.noise_mode))

dividemix/scripts/download_cifar10.sh

@@ -0,0 +1,6 @@
+#!/bin/bash
+
+wget https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz
+tar -zxvf cifar-10-python.tar.gz
+mv cifar-10-batches-py/ cifar-10/
+rm cifar-10-python.tar.gz

dividemix/scripts/download_cifar100.sh

@@ -0,0 +1,6 @@
+#!/bin/bash
+
+wget https://www.cs.toronto.edu/~kriz/cifar-100-python.tar.gz
+tar -zxvf cifar-100-python.tar.gz
+mv cifar-100-python/ cifar-100/
+rm cifar-100-python.tar.gz