update pnnx torch-2.11#6701
Conversation
|
|
There was a problem hiding this comment.
Pull request overview
Updates pnnx to be compatible with Torch 2.11 by handling bias-less nn.MultiheadAttention exports more robustly, and refreshes the pnnx CI workflow to build/test against newer Torch/TorchVision/ONNXRuntime versions.
Changes:
- Update
nn_MultiheadAttentionNCNN pass to always splitin_proj_weight, and to synthesize zero biases whenbias=False(includingout_proj.bias). - Bump pnnx workflow build dependencies to LibTorch 2.11 / TorchVision 0.26 and ONNXRuntime 1.25.1.
- Extend CI test matrix with a Torch 2.11 / Python 3.14 configuration.
Reviewed changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.
| File | Description |
|---|---|
| tools/pnnx/src/pass_ncnn/nn_MultiheadAttention.cpp | Makes MultiheadAttention weight/bias handling tolerant of bias=False by defaulting to zero bias and avoiding reads of missing/empty bias attrs. |
| .github/workflows/pnnx.yml | Updates pinned dependency versions and adds a Torch 2.11 test entry to CI. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
|
@codex review |
|
Codex Review: Didn't find any major issues. Already looking forward to the next diff. ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
No description provided.