Problems iterating through loader when using sampler

Laura_Montalvo · February 17, 2021, 12:21pm

**Hi there! **

I have a custom dataset in which (even though I have more or less the same number of samples for each class) it missclassified some of the data for some specific classes. So what I’m trying to do is oversample by using “sampler” in the dataloader and WeightedRandomSampler as follows:

from sklearn.utils.class_weight import compute_class_weight

def load_data(root_path, batch_size):
transform1 = transforms.Compose(
[transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
])
transform2 = transforms.ToTensor();
data = datasets.ImageFolder(root=root_path, transform= transform1)

class_weights = compute_class_weight('balanced', np.unique(data.targets), data.targets)
class_weights = []
for root, subdir, files in os.walk(root_path):
    if len(files) > 0:
        class_weights.append(1/len(files))
sample_weights = [0]*len(data)

for idx, (data,label) in enumerate(data):
    class_weight = class_weights[label]
    sample_weights[idx] = class_weight
sampler = WeightedRandomSampler(sample_weights, num_samples=len(sample_weights), replacement=True)    
data_loader = torch.utils.data.DataLoader(data, batch_size=batch_size, sampler=sampler)#, drop_last=False, num_workers=4)
return data_loader

trainloader = load_data(root_path_train, batch_size)
testloader = load_data(root_path_test, batch_size)
validloader = load_data(root_path_valid, batch_size)

for x, y in trainloader:
print(y)
break

However, when running the code in the for x, y in trainloader: (where x = data and y = targets) I get an error:

/home/laura/anaconda3/lib/python3.7/site-packages/sklearn/utils/validation.py:70: FutureWarning: targets as keyword args. From version 0.25 passing these as positional arguments will result in an error
FutureWarning)
Traceback (most recent call last):

File “/home/laura/Documentos/DatasetAppLGG3v6/Sin título0.py”, line 104, in
for x,y in trainloader:

ValueError: too many values to unpack (expected 2)

or when i write instead: for i, (x,y) in trainloader: I get:

Traceback (most recent call last):

File “/home/laura/Documentos/DatasetAppLGG3v6/Sin título0.py”, line 104, in
for x,y in trainloader:

File “/home/laura/anaconda3/lib/python3.7/site-packages/torch/utils/data/dataloader.py”, line 345, in next
data = self._next_data()

File “/home/laura/anaconda3/lib/python3.7/site-packages/torch/utils/data/dataloader.py”, line 385, in _next_data
data = self._dataset_fetcher.fetch(index) # may raise StopIteration

File “/home/laura/anaconda3/lib/python3.7/site-packages/torch/utils/data/_utils/fetch.py”, line 44, in fetch
data = [self.dataset[idx] for idx in possibly_batched_index]

File “/home/laura/anaconda3/lib/python3.7/site-packages/torch/utils/data/_utils/fetch.py”, line 44, in
data = [self.dataset[idx] for idx in possibly_batched_index]

IndexError: index 2312 is out of bounds for dimension 0 with size 3

Can anybody help me please???

Dwight_Foster · February 17, 2021, 1:12pm

In the for loop do you say

for i, (x,y) in trainloader: 
or 
for i (x,y) in enumerate(trainloader):

the second one should work.

Laura_Montalvo · February 17, 2021, 2:38pm

Hi! Thanks, but with both options I still get:

Traceback (most recent call last):
File “/home/laura/Documentos/DatasetAppLGG3v6/Sin título0.py”, line 104, in
for x,y in trainloader:
File “/home/laura/anaconda3/lib/python3.7/site-packages/torch/utils/data/dataloader.py”, line 345, in next
data = self._next_data()
File “/home/laura/anaconda3/lib/python3.7/site-packages/torch/utils/data/dataloader.py”, line 385, in _next_data
data = self._dataset_fetcher.fetch(index) # may raise StopIteration
File “/home/laura/anaconda3/lib/python3.7/site-packages/torch/utils/data/_utils/fetch.py”, line 44, in fetch
data = [self.dataset[idx] for idx in possibly_batched_index]
File “/home/laura/anaconda3/lib/python3.7/site-packages/torch/utils/data/_utils/fetch.py”, line 44, in
data = [self.dataset[idx] for idx in possibly_batched_index]
IndexError: index 2312 is out of bounds for dimension 0 with size 3

Dwight_Foster · February 17, 2021, 8:09pm

Ok I think I see the problem in your for loop that defines the data here:

for idx, (data,label) in enumerate(data):

you define one of the outputs as data and your input is data, so when you put data as the dataset for the dataloader it is just using the last output of the for loop which doesn’t work. You just need to change the name of your outputs like (inputs, labels) or something.

Laura_Montalvo · February 18, 2021, 8:15am

That was it!! Thanks so much! Could I ask you something else just in case you know the problem?

When loading validation set into the loader I created above (the same code of the main question of this topic) I get this warning and I don’t know if it may affect something:

/home/laura/anaconda3/lib/python3.7/site-packages/sklearn/utils/validation.py:70: FutureWarning: Pass classes=[ 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
24], y=[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 17, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24] as keyword args. From version 0.25 passing these as positional arguments will result in an error
FutureWarning)

Thanks so much!!!

Dwight_Foster · February 18, 2021, 1:31pm

Sorry I have no idea what this future warning is saying. It probably won’t affect your training but you might want to start a new forum post about it because it might be important in the future.

ptrblck · February 19, 2021, 7:24am

The warning is raised by scikit-learn and asks you to use the keyword arg for the passed classes instead of a positional argument.
I don’t know which method you are calling, but try to use it via: sklearn.some_method(classes=classes).

Laura_Montalvo · February 19, 2021, 8:26am

Hi there! And thank you for answering. I was using “compute_class_weight” from scikit-learn like this:
class_weights = compute_class_weight(‘balanced’, np.unique(data.targets), data.targets)
then I added as you told me (classes = classes) as follows:

NUM_POSITIONS = 25
EPOCH = 17 #config.epochs
optimizer = 2 #config.optimizer
input_channels = 3

from sklearn.utils.class_weight import compute_class_weight

def load_data(root_path, batch_size):
transform1 = transforms.Compose( [transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
])
transform2 = transforms.ToTensor();
data = datasets.ImageFolder(root=root_path, transform= transform1)

class_weights = compute_class_weight('balanced', np.unique(data.targets), data.targets,classes=np.arange(NUM_POSITIONS))
class_weights = []

However I’m still getting the warning… Thanks!

ptrblck · February 19, 2021, 8:41am

Which sklearn version are you using, as I’m getting an error running your code with random target data:

targets = np.random.randint(0, 100, (100,))
class_weights = compute_class_weight('balanced', np.unique(targets), targets,classes=np.arange(NUM_POSITIONS))
> TypeError: compute_class_weight() got multiple values for argument 'classes'

This work for me without a warning:

targets = np.random.randint(0, 100, (100,))
class_weights = compute_class_weight('balanced', classes=np.unique(targets), y=targets)

Laura_Montalvo · February 19, 2021, 8:52am

I’m using sklearn.version 0.23.2.

targets = np.random.randint(0, 100, (100,))
class_weights = compute_class_weight('balanced', classes=np.unique(targets), y=targets)

This way works for me either but I still get the warning…

ptrblck · February 19, 2021, 8:59am

Are you sure this line of code raises the warning or are you using it somewhere else in your script as well (without the changes)?
I’m using a newer version without any warning:

>>> sklearn.__version__
'0.24.1'
>>> from sklearn.utils.class_weight import compute_class_weight
>>> import numpy as np
>>> targets = np.random.randint(0, 100, (100,))
>>> class_weights = compute_class_weight('balanced', classes=np.unique(targets), y=targets)
>>>

and even downgrading to 0.23.2 gives no warnings:

>>> sklearn.__version__
'0.23.2'
>>> from sklearn.utils.class_weight import compute_class_weight
>>> import numpy as np
>>> targets = np.random.randint(0, 100, (100,))
>>> class_weights = compute_class_weight('balanced', classes=np.unique(targets), y=targets)
>>>

If I don’t use the classes=... argument, I get your warning.

Laura_Montalvo · February 19, 2021, 9:08am

Thanks so much! I fixed it, I forgot introducing ‘y=’.

raharth · February 25, 2021, 9:42am

Just a question out of curiosity: Why did you change the sample probability instead of weighting the loss? I would have expected it to be more efficient in terms of runtime?

Laura_Montalvo · February 25, 2021, 10:01am

Actually I’m trying both, I just wanted to see the effect of both: changing sample probability and weighting the loss. That is because I do have a balanced dataset, but for some of the data that has less information, the network classifies worse (like for example it classifies classes 1,2,3,6 and 7 with accuracy 98% but classes 4 and 5 are classified with an accuracy of 20-30%).

raharth · February 25, 2021, 10:23am

Ah ok, got it! So did you find a difference between resampling and weighting in terms of accuracy and computational efficiency?

I usually use loss weighting, but once I had an extremely unbalanced data set, so that most batches would have not seen any samples from one class - that’s where I used the sampling, but I would be really interested in what the overall differences are! Maybe you could share your findings

Laura_Montalvo · February 25, 2021, 10:27am

Using sampling with my dataset didn’t enhanced the results, they were more or less the same and the error in the miss-classified classes was the same. However, giving higher weight to the loss of the classes that were miss-classified achieved better results, however they were hardly noticeable. So I think none of this techniques are the solution to my problem of miss-classified classes because my dataset is balanced

raharth · February 25, 2021, 10:42am

Interesting! I would have expected it to have impact - at least by theory it should have…

Thanks for sharing!