NeuriCam运行报错：卷积层通道数不匹配问题求助

使用NeuriCam项目代码，已安装全部所需依赖库，用自定义数据集运行时出现如下错误：

Traceback (most recent call last):
  File "evaluate.py", line 184, in <module>
    args.output_dir, args.file_fmt, args.profile)
  File "evaluate.py", line 82, in evaluate
    output_batch = model(train_batch)
  File "/tools2/opensource/proj/test/lib64/python3.6/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
    return forward_call(*input, **kwargs)
  File "/tools2/opensource/proj/NeuriCam/model/net.py", line 27, in forward
    return self.model(model_in)
  File "/tools2/opensource/proj/test/lib64/python3.6/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
    return forward_call(*input, **kwargs)
  File "/tools2/opensource/proj/NeuriCam/model/keyvsrc/net.py", line 151, in forward
    feats = self.propagate(lr_s, feats, key_frame_int)
  File "/tools2/opensource/proj/NeuriCam/model/keyvsrc/net.py", line 95, in propagate
    feats = self.basicvsr_pp.propagate(feats, flows, module)
  File "/tools2/opensource/proj/NeuriCam/model/basicvsr_pp/basicvsr_pp.py", line 252, in propagate
    feat_prop = feat_prop + self.backbone[module_name](feat)
  File "/tools2/opensource/proj/test/lib64/python3.6/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
    return forward_call(*input, **kwargs)
  File "/tools2/opensource/proj/NeuriCam/model/basicvsr_pp/basicvsr_net.py", line 234, in forward
    return self.main(feat)
  File "/tools2/opensource/proj/test/lib64/python3.6/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
    return forward_call(*input, **kwargs)
  File "/tools2/opensource/proj/test/lib64/python3.6/site-packages/torch/nn/modules/container.py", line 141, in forward
    input = module(input)
  File "/tools2/opensource/proj/test/lib64/python3.6/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
    return forward_call(*input, **kwargs)
  File "/tools2/opensource/proj/test/lib64/python3.6/site-packages/torch/nn/modules/conv.py", line 446, in forward
    return self._conv_forward(input, self.weight, self.bias)
  File "/tools2/opensource/proj/test/lib64/python3.6/site-packages/torch/nn/modules/conv.py", line 443, in _conv_forward
    self.padding, self.dilation, self.groups)
RuntimeError: Given groups=1, weight of size [64, 128, 3, 3], expected input[1, 64, 124, 328] to have 128 channels, but got 64 channels instead

解决思路 #

• 检查自定义数据集预处理流程：确认数据预处理后的通道数是否符合模型要求，比如是否遗漏了特征拼接、通道扩展等步骤，导致输入通道数从128变成64。
• 核对模型配置参数：查看NeuriCam中basicvsr_pp模块的配置文件，确认卷积层的输入输出通道设置是否和你的自定义数据集适配，是否有需要调整的参数。
• 对比官方数据集格式：将自定义数据集的维度、预处理后的数据结构和项目提供的官方数据集做对比，定位通道数差异的根源。
• 排查数据加载代码：检查evaluate.py或数据加载脚本中，是否在处理自定义数据时错误修改了通道维度，比如误做了通道减半操作。

内容的提问来源于stack exchange，提问作者Ashish Kumar