Tensor size mismatch in usage of captum to explain graph - Chapter 14

未关闭
  • 哪里不对？
#15 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

还没有人认领这个 Issue。

评估

难度
4/5
预计耗时
3-5 天
新手友好度
30/100
Issue 类型
缺陷
描述清晰度
需要澄清
活跃度
停滞
技术栈
jupyter-notebook, python
领域
machine-learning

调研方向

The traceback points to the Chapter 14 notebook, the GCN.forward call in cell 87, and the custom GCNConv implementation in cell 11. Re-run the notebook and inspect the tensor shapes at the failing message-passing step. Done means the model's forward pass and loss calculation complete without the tensor-size mismatch.

由索引模型根据 Issue 内容生成。

描述

I am receiving the following error:

RuntimeError: The size of tensor a (84900) must match the size of tensor b (7126) at non-singleton dimension 0

{
"name": "RuntimeError",
"message": "The size of tensor a (84900) must match the size of tensor b (7126) at non-singleton dimension 0",
"stack": "---------------------------------------------------------------------------
RuntimeError Traceback (most recent call last)
Cell In[87], line 21
19 model.train()
20 optimizer.zero_grad()
---> 21 log_logits = model(data.x, data.edge_index)
22 loss = F.nll_loss(log_logits, data.y)
23 loss.backward()

File ~/Desktop/Extras/Graph-Neural-Network/.venv/lib/python3.12/site-packages/torch/nn/modules/module.py:1553, in Module._wrapped_call_impl(self, *args, **kwargs)
1551 return self._compiled_call_impl(*args, **kwargs) # type: ignore[misc]
1552 else:
-> 1553 return self._call_impl(*args, **kwargs)

File ~/Desktop/Extras/Graph-Neural-Network/.venv/lib/python3.12/site-packages/torch/nn/modules/module.py:1562, in Module._call_impl(self, *args, **kwargs)
1557 # If we don't have any hooks, we want to skip the rest of the logic in
1558 # this function, and just call forward.
1559 if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks
1560 or _global_backward_pre_hooks or _global_backward_hooks
1561 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1562 return forward_call(*args, **kwargs)
1564 try:
1565 result = None

Cell In[87], line 8, in GCN.forward(self, x, edge_index)
7 def forward(self, x, edge_index):
----> 8 h = self.conv1(x, edge_index).relu()
9 h = F.dropout(h, p=0.5, training=self.training)
10 h = self.conv2(h, edge_index)

File ~/Desktop/Extras/Graph-Neural-Network/.venv/lib/python3.12/site-packages/torch/nn/modules/module.py:1553, in Module._wrapped_call_impl(self, *args, **kwargs)
1551 return self._compiled_call_impl(*args, **kwargs) # type: ignore[misc]
1552 else:
-> 1553 return self._call_impl(*args, **kwargs)

File ~/Desktop/Extras/Graph-Neural-Network/.venv/lib/python3.12/site-packages/torch/nn/modules/module.py:1562, in Module._call_impl(self, *args, **kwargs)
1557 # If we don't have any hooks, we want to skip the rest of the logic in
1558 # this function, and just call forward.
1559 if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks
1560 or _global_backward_pre_hooks or _global_backward_hooks
1561 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1562 return forward_call(*args, **kwargs)
1564 try:
1565 result = None

Cell In[11], line 20, in GCNConv.forward(self, x, edge_index)
17 deg_inv_sqrt[deg_inv_sqrt == float('inf')] = 0
18 norm = deg_inv_sqrt[row] * deg_inv_sqrt[col]
---> 20 out = self.propagate(edge_index, x=x, norm=norm)
22 return out

File ~/Desktop/Extras/Graph-Neural-Network/.venv/lib/python3.12/site-packages/torch_geometric/nn/conv/message_passing.py:523, in MessagePassing.propagate(self, edge_index, size, **kwargs)
521 if res is not None:
522 msg_kwargs = res[0] if isinstance(res, tuple) else res
--> 523 out = self.message(**msg_kwargs)
524 for hook in self._message_forward_hooks.values():
525 res = hook(self, (msg_kwargs, ), out)

Cell In[11], line 25, in GCNConv.message(self, x, norm)
24 def message(self, x, norm):
---> 25 return norm.view(-1, 1) * x

RuntimeError: The size of tensor a (84900) must match the size of tensor b (7126) at non-singleton dimension 0"
}

主要语言
Jupyter Notebook
星标
1.1k
派生
300
30 天内没有已合并 PR

环境准备

这个项目没有提供开发容器、Dockerfile 或贡献指南，环境需要你自己搭建：先看它的 README，通用步骤见我们的新手贡献指南。

从这里开始

  1. 先读完整个 Issue，再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库，在一个分支上完成修改。
  4. 提交 Pull Request，并在描述里引用这个 Issue 编号。

PacktPublishing/Hands-On-Graph-Neural-Networks-Using-Python 的其他 Issue

查看 PacktPublishing/Hands-On-Graph-Neural-Networks-Using-Python 的全部 Issue

相似的 Issue

更多 Machine Learning Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。