sinezhan / deepalplus Goto Github PK

Dear Repository owners,

I would like to use your deepALplus to do experiments with Deep Active Learning.
However, I always get stuck at line 23 in nets.py. It takes ages to execute but should normally be milliseconds:
self.clf = self.net(dim = dim, pretrained = self.params['pretrained'], num_classes = self.params['num_class']).to(self.device)
The script fails after ~20 min with RuntimeError: cuDNN error: CUDNN_STATUS_EXECUTION_FAILED.
Do you have any recommendations? Executing the code in https://github.com/ej0cl6/deep-active-learning works fine for me.
Could this be because of cudnn and cuda versions? Are the certain versions one has to use?

I installed using conda. I used cudnn/8.0_v7.0 and cuda/11.0.2 as well as cudnn/11.7_v8.6 amd cuda/11.7.0 and had the same behavior with both.

Thank you.

Hello,

Is there any way to test this code with a custom image dataset?

Thank you

In line 298 and line 299 of data.py, train data are used for testing.

hello,@SineZHAN .
def get_BreakHis(handler, args_task):
download data from https://www.kaggle.com/datasets/ambarish/breakhis and unzip it in data/BreakHis/
data_dir = './data/BreakHis/BreaKHis_v1/BreaKHis_v1/histology_slides/breast'
data = datasets.ImageFolder(root=data_dir, transform=None).imgs
train_ratio = 0.7
test_ratio = 0.3
data_idx = list(range(len(data)))
random.shuffle(data_idx)
train_idx = data_idx[:int(len(data) * train_ratio)]
test_idx = data_idx[int(len(data) * train_ratio):]
X_tr = [np.array(Image.open(data[i][0])) for i in train_idx]
Y_tr = [data[i][1] for i in train_idx]
X_te = [np.array(Image.open(data[i][0])) for i in test_idx]
Y_te = [data[i][1] for i in test_idx]
X_tr = np.array(X_tr, dtype=object)
X_te = np.array(X_te, dtype=object)
Y_tr = torch.from_numpy(np.array(Y_tr))
Y_te = torch.from_numpy(np.array(Y_te))
return Data(X_tr, Y_tr, X_te, Y_te, handler, args_task)

A memory error occurred while running this code！

MemoryError: Unable to allocate 943. KiB for an array with shape (460, 700, 3) and data type uint8

X_tr keeps storing Data, which causes the memory to be full. How can I modify it to work with class data？

Hi is VAAL implementation incorrect in this toolkit. I see that it is very different from the author's original implementation and it also looks incomplete, unusable. Can you please let me know if it is?

Dear Repository owners,

I would like to use your deepALplus to do experiments with Deep Active Learning, but i got a comprehension problem with net_waal.py.
Indeed, at line 81, you recompute the features for labeled and unlabeled data in "with torch.no_grad" loop. Then you compute the gradient penalty and add it to the loss.
Since you used the with torch.no_grad loop, the contibution of the gradient penalty when updating the weight of your features_extractor will be null. and Since at line 64 you set the requires_grad=False for the discriminator the weigths of the discriminator will not be update.

I would like to know why you recomputed the features in the "with torch.no_grad" loop since it seems to make the gradient penalty to have no impact when updating the weight of your entire model.
Thank you.