feat(FRNet): apply softmax & fix dataset path

[amadeuszsz]

Summary

Apply softmax and deploy FRNet model with probabilities output as we do in PTv3.

Additionally fix the dataset path for test in deployment script.

[amadeuszsz]

@codex review

[Copilot]

Pull request overview

This PR updates the FRNet deployment pipeline to output per-class probabilities (via softmax) across Torch/ONNX/TensorRT backends, and adjusts the T4 dataset GT mask path handling to work with paths stored relative to the workspace root.

Changes:

• Wrap the PyTorch model with an ExportModel that returns softmax(seg_logit) and use it for ONNX export and Torch inference.
• Rename deployment output from seg_logit to pred_probs across deploy configs and TensorRT runtime wrapper.
• Fix T4 dataset semantic mask path handling by no longer blindly joining dataset_dir and instead resolving via existence checks.

Reviewed changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
projects/FRNet/deploy/trt_model.py	Switch TRT expected output tensor name to pred_probs and document probability output.
projects/FRNet/deploy/torch_model.py	Add ExportModel wrapper to expose probability output and update Torch inference.
projects/FRNet/deploy/postprocessing.py	Update docs to reflect probability-based postprocessing.
projects/FRNet/deploy/onnx_model.py	Update docs to reflect probability output from ONNX Runtime.
projects/FRNet/deploy/main.py	Export ONNX from torch_model.export_model instead of the raw model.
projects/FRNet/deploy/dataset.py	Adjust T4 mask path handling to avoid incorrect dataset_dir joins.
projects/FRNet/configs/deploy/t4dataset/frnet_tensorrt_dynamic.py	Rename ONNX/TRT output to pred_probs and update dynamic axes/config entries.
projects/FRNet/configs/deploy/nuscenes/frnet_tensorrt_dynamic.py	Rename ONNX/TRT output to pred_probs and update dynamic axes/config entries.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: dbe14f5adb

ℹ️ About Codex in GitHub

Your team has set up Codex to 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 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[KSeangTan]

LGTM overall, please resolve the comments and merge afterwards