CRZbulabula opened a new pull request, #19153:
URL: https://github.com/apache/iotdb/pull/19153

   ### Summary
   
   Upgrade AINode to Torch 2.14.1, setuptools 84.2.1, and PyInstaller 6.22.3. 
Add regression tests for the Torch input-validation fixes covering 
CVE-2025-2998, CVE-2025-2999, CVE-2025-3000, and CVE-2025-3001, plus 
TorchScript model serialization and inference.
   
   Linux selects the official public CUDA 12.6 Torch wheel to retain Volta 
support; macOS and Windows resolve Torch 2.14.1 from PyPI. The 1C1D1A GPU 
workflow checks the `sm_70` architecture, runs a CUDA tensor operation and the 
Python regression suites, and uploads integration-test reports.
   
   ### Source and adaptation
   
   Adapted from TimechoDB MR !7427, source commit 
`19bc1bd7da4e91b1deb2aa328004bc5171a2da28`, based on enterprise master 
`026fbe63494fbb771164dfad877a4f8c3b92517d`. The related enterprise commits add 
Nexus-only dependency installation and authentication for enterprise CI. Those 
changes have no Apache build counterpart; this PR uses the public PyPI and 
PyTorch indexes instead. The security tests are moved from `test/` to the 
Apache AINode module's existing `tests/` directory.
   
   ### Validation
   
   - New dependency security tests: 5 passed with Torch 2.14.1.
   - Sundial flow-loss tests: 4 passed.
   - Poetry metadata check, Black, isort, workflow YAML parsing, and `git diff 
--check` passed.
   - Running the empty PackedSequence regression against the previous Torch 
2.9.1 ended in a process crash (exit 139), reproducing the failure this upgrade 
addresses.
   - GPU execution and packaged AINode cluster integration tests require the 
project's GPU runner and were not run locally.
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to