tencentyouturesearch / imagecolorization-colorformer Goto Github PK

I'm trying to train the model from scratch on a custom subset of Imagenet, the training works fine on a single gpu, but when running on multiple gpus I get the following error:

Expected tensor for argument #1 'input' to have the same device as tensor for argument #2 'weight'; but device 3 does not equal 0 (while checking arguments for cudnn_batch_norm)

my configuration file looks like this:

name: train_colorformer
model_type: LABGANRGBModel
scale: 1
num_gpu: 4
manual_seed: 0
queue_size: 64

and I'm using CUDA_VISIBLE_DEVICES to specify the gpus to be used.
I tried looking for any inputs that are not moved to cuda but without success.

想问下配置文件中的'meta_info_file'是具体是要配置什么

Thank you for the source code. I am confused about the performance,
When I trained the network, the colorfulness score was only 32.51, much lower than the value of the paper.
Did you use tricks during training?
Besides, I also apply your pre-trained model, the metric is also lower than the paper.