Traceback (most recent call last):
File "main.py", line 73, in
main(args)
File "main.py", line 59, in main
model = get_model(num_classes, args)
File "/media/omnisky/data/wr/M3TR-master/models/init.py", line 12, in get_model
model = model_dict[args.model_name](vit, res101, num_classes)
File "/media/omnisky/data/wr/M3TR-master/models/M3TR.py", line 55, in init
self.sem_embedding = self.get_word_embedding(self.num_classes).detach()
File "/media/omnisky/data/wr/M3TR-master/models/M3TR.py", line 81, in get_word_embedding
loaded = torch.load(embedding_path)
File "/home/omnisky/anaconda2/envs/faster/lib/python3.7/site-packages/torch/serialization.py", line 595, in load
return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args)
File "/home/omnisky/anaconda2/envs/faster/lib/python3.7/site-packages/torch/serialization.py", line 766, in _legacy_load
if magic_number != MAGIC_NUMBER:
ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

您好，我使用配好的环境能够成功复现您初始代码，但是将M3TR.py中#embedding_path = './Bert1/voc_embeddings.pt' 和 #embedding_path = './Bert1/coco_embeddings.pt'分别改为embedding_path = './Glove/voc_glove_word2vec.pkl'及embedding_path = './Glove/coco_glove_word2vec.pkl'之后，报上面的错误，请问可以指导一下如何修改吗？

Hello, I think I am having weird results like mAP=1.0 after the first epoch and wanted to be sure if the model works so well on my custom dataset. But I don't know how to analyse the outputs coming from the model. Apart from outputs1, outputs2, out3 variables, only outputs is returned as model final predictions; and all the values are negative. Since its sigmoid I was thinking to print the labels having > 0 prediction, but it doesn't seem to work because everything is negative. How should we use the
outputs = self.on_forward(inputs, targets, is_train=is_train)
line in trainer.py to print or plot the predicted labels for the given image?

I add an example screenshot to show the output vector I get for batch size = 4 (while mAP is always 1.0)