Document ONNX prerequisites and add [onnx] extras_require

torch>=2.6 changed its ONNX exporter to require onnxscript, so users
hitting the export scripts on a fresh install were getting
ModuleNotFoundError. This adds:

- setup.py: extras_require={'onnx': [onnx, onnxruntime, onnxscript, onnx-simplifier]}
  so users can pip install nnunetv2_distillation[onnx].
- docs/Distillation.md: a Prerequisites callout before the ONNX export
  section listing both the extras install and the manual pip command.

Author: 77even

distillation/setup.py

@@ -7,6 +7,16 @@
         "torch>=1.6.0",
         "nnunetv2>=2.5,<3.0",
     ],
+    extras_require={
+        # ONNX export deps. torch>=2.6 needs onnxscript for the new exporter;
+        # onnx-simplifier is only used when the user passes --simplify / -sim.
+        "onnx": [
+            "onnx",
+            "onnxruntime",
+            "onnxscript",
+            "onnx-simplifier",
+        ],
+    },
     entry_points={
         'console_scripts': [
             'nnUNetv2_distillation_train=fast_nnunet_distillation_train:main',

docs/Distillation.md

@@ -222,7 +222,23 @@ Parameter description:
 
 ### 3. 📤 Export ONNX Model
 
-Export the trained student model to ONNX format for fast inference:
+Export the trained student model to ONNX format for fast inference.
+
+> **Prerequisites:** ONNX export depends on a few extra packages that are
+> not part of the core install. The easiest way is the `[onnx]` extra:
+>
+> ```bash
+> pip install -e '.[onnx]'   # quotes needed for zsh
+> ```
+>
+> Or install them manually:
+>
+> ```bash
+> pip install onnx onnxruntime onnxscript onnx-simplifier
+> ```
+>
+> `onnxscript` is required by PyTorch 2.6+ for the new ONNX exporter,
+> and `onnx-simplifier` is only needed when you pass `--simplify` / `-sim`.
 
 #### Standard Distillation Model Export