Cadene / vqa.pytorch

Visual Question Answering in Pytorch
717 stars 178 forks source link

How to train on VQA 2? #14

Closed gaopeng-eugene closed 7 years ago

gaopeng-eugene commented 7 years ago

I trained on VQA 1 successfully. When I switch to VQA 2, I run into the error below :+1:

Traceback (most recent call last): File "train.py", line 373, in main() File "train.py", line 120, in main options['vgenome']) File "/home/pgao/vqa_real_train/vqa.pytorch/vqa/datasets/vqa.py", line 253, in factory dataset_img = coco.factory(data_split, opt_coco) File "/home/pgao/vqa_real_train/vqa.pytorch/vqa/datasets/coco.py", line 93, in factory trainset = factory('train', opt, transform) File "/home/pgao/vqa_real_train/vqa.pytorch/vqa/datasets/coco.py", line 102, in factory return FeaturesDataset(data_split, opt) File "/home/pgao/vqa_real_train/vqa.pytorch/vqa/datasets/features.py", line 20, in init 'File not found in {}, you must extract the features first with extract.py'.format(self.path_hdf5)

VQA 1 and VQA 2 have the same image feature, right?