train issue

[RichardMrLu]

/faster-rcnn.pytorch/lib/model/rpn/rpn.py:66: UserWarning: Implicit dimension choice for softmax has been deprecated. Change the call to include dim=X as an argument.
rpn_cls_prob_reshape = F.softmax(rpn_cls_score_reshape)
Traceback (most recent call last):
File "trainval_net.py", line 316, in
_, cls_prob, bbox_pred, rpn_loss, rcnn_loss = fasterRCNN(im_data, im_info, gt_boxes, num_boxes)
File "/raid/pyenv/versions/2.7.11/lib/python2.7/site-packages/torch/nn/modules/module.py", line 325, in call
result = self.forward(*input, **kwargs)
File "/raid/faster-rcnn.pytorch/lib/model/faster_rcnn/faster_rcnn_cascade.py", line 51, in forward
rois, rpn_loss_cls, rpn_loss_bbox = self.RCNN_rpn(base_feat, im_info, gt_boxes, num_boxes)
File "/raid/.pyenv/versions/2.7.11/lib/python2.7/site-packages/torch/nn/modules/module.py", line 325, in call
result = self.forward(*input, **kwargs)
File "/raid/faster-rcnn.pytorch/lib/model/rpn/rpn.py", line 76, in forward
im_info, cfg_key))
File "/raid/pyenv/versions/2.7.11/lib/python2.7/site-packages/torch/nn/modules/module.py", line 325, in call
result = self.forward(*input, **kwargs)
File "/raid/faster-rcnn.pytorch/lib/model/rpn/proposal_layer.py", line 105, in forward
proposals = bbox_transform_inv(anchors, bbox_deltas, batch_size)
File "/raid/faster-rcnn.pytorch/lib/model/rpn/bbox_transform.py", line 90, in bbox_transform_inv
pred_w = np.exp(dw) * widths.unsqueeze(2)
File "/raid/.pyenv/versions/2.7.11/lib/python2.7/site-packages/torch/tensor.py", line 309, in mul
return self.mul(other)
TypeError: mul received an invalid combination of arguments - got (torch.cuda.FloatTensor), but expected one of:

• (float value)
  didn't match because some of the arguments have invalid types: (torch.cuda.FloatTensor)
• (torch.FloatTensor other)
  didn't match because some of the arguments have invalid types: (torch.cuda.FloatTensor)

how to solve this problem?thanks for your answer

[jwyang]

it seems that the data type does not match. what is your command to run the train?

[RichardMrLu]

CUDA_VISIBLE_DEVICES=1 python trainval_net.py --dataset pascal_voc --net vgg16 --cuda --bs 8
dataset is VOC2007

[jwyang]

how many GPUs do you have on your machine? Try to remove CUDA_VISIBLE_DEVICES=1, and set batch size to 1.

[RichardMrLu]

4GPUs, I used python trainval_net.py --dataset pascal_voc --net vgg16 --cuda --bs 1
but it crashed with the same error. I think in trainval_net.py, some variable are set type in torch.cuda.FloatTensor but in original py_faster's file they are calculate in other type, so type error....

[jwyang]

can you add a breakpoint before "pred_w = np.exp(dw) * widths.unsqueeze(2)"? and show the types of dw and widths? I guess dw in your running is not a number, which caused this error.

[RichardMrLu]

I find it that type(np.exp(dw)) is torch.FloatTensor type(widths.unsqueeze(2)) is torch.cuda.FloatTensor,
so train error. I changed np.exp(dw) to torch.exp(dw), It begin to train.
Thanks for your work.

[jwyang]

Great! I also change np.exp to torch.exp in the repository.

…

On Wed, Dec 27, 2017 at 01:26 RichardMrLu ***@***.***> wrote: I find it that type(np.exp(dw)) is torch.FloatTensor type(widths.unsqueeze(2)) is torch.cuda.FloatTensor, so train error. I changed np.exp(dw) to torch.exp(dw), It begin to train. Thanks for your work. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#5 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ADtr51tDRPpuEpso3J2Q12BGgiZGKBOhks5tEeMLgaJpZM4RMsHV> .