The training wait a long time after load model.

[daiwc]

[2019-09-23 11:06:50,663-rk0-train.py#311] model prepare done
[2019-09-23 11:06:50,664-rk0-train.py#174] model
.backbone (AlexNet)
.layer1 (Sequential)
.0 (Conv2d)
- weight
- bias
.1 (BatchNorm2d)
- weight
- bias
.2 (MaxPool2d)
.3 (ReLU)
.layer2 (Sequential)
.0 (Conv2d)
- weight
- bias
.1 (BatchNorm2d)
- weight
- bias
.2 (MaxPool2d)
.3 (ReLU)
.layer3 (Sequential)
.0 (Conv2d)
- weight
- bias
.1 (BatchNorm2d)
- weight
- bias
.2 (ReLU)
.layer4 (Sequential)
.0 (Conv2d)
- weight
- bias
.1 (BatchNorm2d)
- weight
- bias
.2 (ReLU)
.layer5 (Sequential)
.0 (Conv2d)
- weight
- bias
.1 (BatchNorm2d)
- weight
- bias
.rpn_head (DepthwiseRPN)
.cls (DepthwiseXCorr)
.conv_kernel (Sequential)
.0 (Conv2d)
- weight
.1 (BatchNorm2d)
- weight
- bias
.2 (ReLU)
.conv_search (Sequential)
.0 (Conv2d)
- weight
.1 (BatchNorm2d)
- weight
- bias
.2 (ReLU)
.head (Sequential)
.0 (Conv2d)
- weight
.1 (BatchNorm2d)
- weight
- bias
.2 (ReLU)
.3 (Conv2d)
- weight
- bias
.loc (DepthwiseXCorr)
.conv_kernel (Sequential)
.0 (Conv2d)
- weight
.1 (BatchNorm2d)
- weight
- bias
.2 (ReLU)
.conv_search (Sequential)
.0 (Conv2d)
- weight
.1 (BatchNorm2d)
- weight
- bias
.2 (ReLU)
.head (Sequential)
.0 (Conv2d)
- weight
.1 (BatchNorm2d)
- weight
- bias
.2 (ReLU)
.3 (Conv2d)
- weight
- bias

the training don't go on after load model, even though i wait many hours.
the command that i used is CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.launch --nproc_per_node=2 --master_port=2333 ../../tools/train.py --cfg config.yaml

[chun-chun123]

I have met the same question.Have you managed it?

[ZhiyuanChen]

What is your batch size?

[ZhiyuanChen]

Since no further question is asked, I'm closing it for now.
Please reopen it or create a new issue if you have further question.