Apply transforms twice?

[luxunxiansheng]

Hi, there,

Suppose I have 100 image files and use the CacheDataset with a cache rate 0.5 and set it to the train laoder
When training , to get batchs like this:
for batch_data in tranin_Loader:
....

I notice that the line above is very slow. It seems to apply_transforms again besides constructing the cachedataset.

I missed something?

Thanks in advacne
Bin

[KumoLiu]

Hi @luxunxiansheng,

Not sure what you mean, you can use CacheDataset like this:

train_ds = CacheDataset(
    data=train_files, transform=train_transforms,
    cache_rate=0.5, num_workers=4)
train_loader = DataLoader(train_ds, batch_size=2, shuffle=True, num_workers=4)

for batch_data in train_loader:
.....

You can also refer to this tutorial

Hope it can help you, thanks!