Closed wouterzwerink closed 1 year ago
Fixes #316
Now, when running the example code in the issue and printing args:
{'fname': '/home/ec2-user/ffcv/tmp.beton', 'batch_size': 100, 'num_workers': 1, 'os_cache': 0, 'order': <OrderOption.RANDOM: 2>, 'distributed': False, 'seed': 1500678110, 'indices': None, 'pipelines': {'value': [<ffcv.fields.rgb_image.RandomResizedCropRGBImageDecoder at 0x7fc3bc9523a0>, <ffcv.transforms.flip.RandomHorizontalFlip at 0x7fc3bc952550>, <ffcv.transforms.ops.ToTensor at 0x7fc3bc952940>, <ffcv.transforms.ops.ToDevice at 0x7fc3bc9527c0>, <ffcv.transforms.ops.ToTorchImage at 0x7fc3bc952b50>, <ffcv.transforms.normalize.NormalizeImage at 0x7fc3bc952220>], 'index': [<ffcv.fields.basics.IntDecoder at 0x7fc3bc952190>, <ffcv.transforms.ops.ToTensor at 0x7fc3bc975550>, <ffcv.transforms.common.Squeeze at 0x7fc3bc951ca0>, <ffcv.transforms.ops.ToDevice at 0x7fc3bc951490>]}, 'drop_last': True, 'batches_ahead': 3, 'recompile': False}
Sorry about my previous PR being in here, let me know if you want me to properly split them
Hi @wouterzwerink ! Thanks for the PR! Do you think you could open a new PR into the v1.1.0 branch that only contains the deepcopy fix (I've already merged your other PR into that branch)
Fixes #316
Now, when running the example code in the issue and printing args:
Sorry about my previous PR being in here, let me know if you want me to properly split them