Remove automatic ONNX export from training

[Copilot]

Training automatically exported models to ONNX format regardless of whether they would go to production, wasting compute time.

Changes

• Removed export_to_onnx() call and supporting code from Train.run() in src/hyrax/verbs/train.py
• Removed import of export_to_onnx from train verb
• The function remains available in model_exporters.py for explicit manual use

Impact

Training completes faster. Models are only exported to ONNX when users explicitly call export_to_onnx() instead of automatically after every training run.

Original prompt

---

This section details on the original issue you should resolve

<issue_title>Remove export_to_onnx from train</issue_title>
<issue_description>In the interest of avoiding surprises, let's remove the automatic ONNX-ification of models immediately after training.

Often the trained model is not the one that will go to production, and thus we're wasting compute time ONNX-ifying them prepmaturely. </issue_description>

Comments on the Issue (you are @copilot in this section)

• Fixes Remove export_to_onnx from train #519

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 53.55%. Comparing base (25b020c) to head (d334335).
⚠️ Report is 63 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #520      +/-   ##
==========================================
- Coverage   54.05%   53.55%   -0.51%     
==========================================
  Files          48       48              
  Lines        4767     4760       -7     
==========================================
- Hits         2577     2549      -28     
- Misses       2190     2211      +21     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[drewoldag]

This looks correct to me. The goal of the PR is to stop attempting to automatically produce an ONNX model at the end of train. This does just that.

[Copilot]

Pull Request Overview

This pull request removes the automatic ONNX model export functionality from the training verb. After training completes, the model will no longer be automatically exported to ONNX format.

• Removed ONNX export import statement
• Removed post-training ONNX export code block

Comments suppressed due to low confidence (1)

src/hyrax/verbs/train.py:1

• The removal of ONNX export functionality from the training verb leaves the [onnx] configuration section in hyrax_default_config.toml (lines 160-164) unused. This section includes the opset_version parameter that is no longer referenced anywhere in the codebase. Consider either removing this configuration section or documenting that ONNX export has been moved/deprecated.

import logging

[gitosaurus]

LGTM

[aritraghsh09]

🚢 it