修改batch问题

[n1-k0]

你好，我在修改Ocean.yaml的时候发现，将batch设为32或64能正常运行，如果继续调大（如128或192）则会出现如下错误

`Traceback (most recent call last): File "./tracking/train_ocean.py", line 259, in main() File "./tracking/train_ocean.py", line 250, in main model, writer_dict = ocean_train(train_loader, model, optimizer, epoch + 1, curLR, config, writer_dict, logger, device=device) File "/home/dasima/pythonwork/TracKit/tracking/../lib/core/function.py", line 38, in ocean_train cls_loss_ori, cls_loss_align, reg_loss = model(template, search, label_cls, reg_target=reg_label, reg_weight=reg_weight) File "/home/dasima/anaconda3/envs/TracKit/lib/python3.7/site-packages/torch/nn/modules/module.py", line 493, in call result = self.forward(*input, kwargs) File "/home/dasima/anaconda3/envs/TracKit/lib/python3.7/site-packages/torch/nn/parallel/data_parallel.py", line 150, in forward return self.module(*inputs[0], *kwargs[0]) File "/home/dasima/anaconda3/envs/TracKit/lib/python3.7/site-packages/torch/nn/modules/module.py", line 493, in call result = self.forward(input, kwargs) File "/home/dasima/pythonwork/TracKit/tracking/../lib/models/ocean.py", line 325, in forward cls_align = self.align_head(reg_feature, offsets) File "/home/dasima/anaconda3/envs/TracKit/lib/python3.7/site-packages/torch/nn/modules/module.py", line 493, in call result = self.forward(*input, kwargs) File "/home/dasima/pythonwork/TracKit/tracking/../lib/models/connect.py", line 219, in forward x = self.relu(self.rpn_conv(x, offset)) File "/home/dasima/anaconda3/envs/TracKit/lib/python3.7/site-packages/torch/nn/modules/module.py", line 493, in call result = self.forward(*input, *kwargs) File "/home/dasima/pythonwork/TracKit/tracking/../lib/models/connect.py", line 203, in forward x = self.conv(x, offset) File "/home/dasima/anaconda3/envs/TracKit/lib/python3.7/site-packages/torch/nn/modules/module.py", line 493, in call result = self.forward(input, kwargs) File "/home/dasima/pythonwork/TracKit/tracking/../lib/models/dcn/deform_conv.py", line 237, in forward self.dilation, self.groups, self.deformable_groups) File "/home/dasima/pythonwork/TracKit/tracking/../lib/models/dcn/deform_conv.py", line 56, in forward cur_im2col_step) RuntimeError: shape '[1, 1, 2304]' is invalid for input of size 92160000

目测是deform_conv_cuda.cpp的问题，但对这一块我不知道怎么debug，请问如何解决这个问题。

[penghouwen]

pls refer to our most recent work https://github.com/researchmm/Stark and https://github.com/researchmm/LightTrack

[Soarkey]

+1，请问后来解决了吗