Repository navigation
Conversation
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 8f3a24d86e
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
There was a problem hiding this comment.
When this feature is used through ONNX, Caffe, or MXNet conversion, the new batch branch is unreachable: ONNX writes source axis 0 as 0 (tools/onnx/onnx2ncnn.cpp:4234-4237), while Caffe and MXNet subtract one and write -1 (tools/caffe/caffe2ncnn.cpp:334-338 and tools/mxnet/mxnet2ncnn.cpp:1889-1892). With batched inputs those values are interpreted as per-sample axes, so axis-0 concat either rejects unequal batches or concatenates channels/width instead of summing Mat::n; map the source batch axis to -233 in these converters as pnnx now does.
Useful? React with 👍 / 👎.
No description provided.