JackEasson
commented
2 years ago 验证环节的后处理部分，主要在进行obj_num_list, scores_tensor, coordinates_tensor = model(images, mode1='det_only', mode2='eval')时涉及一些张量从gpu转到cpu的操作，在多GPU情况下会遇到问题，你可以去掉验证部分最后单独验证，或者在验证环节使用单GPU验证。
请问代码需要哪些修改才能使模型使用多卡训练呢？ 我在train.py文件中直接加入了
main(parser.parse_args())
File "train.py", line 512, in main
model = train(args, model, device) # Train decoder
File "train.py", line 317, in train
obj_num_list, scores_tensor, coordinates_tensor = model(images, mode1='det_only', mode2='eval')
File "/home/guest/fhc/anaconda3/envs/slpnet/lib/python3.7/site-packages/torch/nn/modules/module.py", line 532, in call
result = self.forward(*input, *kwargs)
File "/home/guest/fhc/anaconda3/envs/slpnet/lib/python3.7/site-packages/torch/nn/parallel/data_parallel.py", line 153, in forward
return self.gather(outputs, self.output_device)
File "/home/guest/fhc/anaconda3/envs/slpnet/lib/python3.7/site-packages/torch/nn/parallel/data_parallel.py", line 165, in gather
return gather(outputs, output_device, dim=self.dim)
File "/home/guest/fhc/anaconda3/envs/slpnet/lib/python3.7/site-packages/torch/nn/parallel/scatter_gather.py", line 68, in gather
res = gather_map(outputs)
File "/home/guest/fhc/anaconda3/envs/slpnet/lib/python3.7/site-packages/torch/nn/parallel/scatter_gather.py", line 63, in gather_map
return type(out)(map(gather_map, zip(outputs)))
File "/home/guest/fhc/anaconda3/envs/slpnet/lib/python3.7/site-packages/torch/nn/parallel/scatter_gather.py", line 63, in gather_map
return type(out)(map(gather_map, zip(outputs)))
File "/home/guest/fhc/anaconda3/envs/slpnet/lib/python3.7/site-packages/torch/nn/parallel/scatter_gather.py", line 63, in gather_map
return type(out)(map(gather_map, zip(outputs)))
TypeError: zip argument #1 must support iteration
device = "cuda:0" if args.cuda: model = model.cuda(device) model = torch.nn.DataParallel(model, device_ids=[0,1,2])但是在模型验证时报错 Traceback (most recent call last): File "train.py", line 538, in不使用多GPU训练的话，感觉这个模型要训练很长时间