Open dota2mhxy opened 2 years ago
@hytseng0509
Do you follow the instruction here to prepare the dataset? data_file
should point to the location of a .json
file, it should not be a list.
yes,I downloaded five datasets and formed JSON files according to the steps,
I used params.dataset == 'multi' in my training, so I will use the four datasets except the testset,
the variable base_file is has four json file,is list ,
this cuase the error in this mehod
@hytseng0509
The current code does not support training the baseline approach with multiple datasets. If you want to you can consider manually modify the code similar to this.
Hello, when I run the code train_ baseline.py, I use multiple domains during training, but the code reports an error。
File "/home/wangchengcheng/WCC/CrossDomainFewShot/train_baseline.py", line 96, in
base_loader = base_datamgr.get_data_loader( base_file , aug=params.train_aug )
File "/home/wangchengcheng/WCC/CrossDomainFewShot/data/datamgr.py", line 57, in get_data_loader
dataset = SimpleDataset(data_file, transform)
File "/home/wangchengcheng/WCC/CrossDomainFewShot/data/dataset.py", line 15, in init
with open(data_file, 'r') as f:
TypeError: expected str, bytes or os.PathLike object, not list
The error code is in dataset py
class SimpleDataset: def init(self, data_file, transform, target_transform=identity): with open(data_file, 'r') as f: self.meta = json.load(f) self.transform = transform self.target_transform = target_transform
I saw data_file is list , so I reported an error. Maybe you have corrected this problem, but there is a problem with the code version. Thank you for your reply