This pull request introduces a temporary workaround for Issue #53, where the system defaults to using GPU ID 0 when multiple GPUs are available. This interim solution is intended to ensure stability and consistent behavior in multi-GPU environments until we develop and deploy a comprehensive testing suite for distributed inference.
Key Changes:
Enforce GPU ID 0 selection in multi-GPU setups.
Log a message indicating that one GPU (ID 0) is being used.
Future Work:
Develop an extensive test suite for distributed inference.
Implement a robust mechanism for GPU selection and utilization.
This approach allows us to maintain operational integrity while laying the groundwork for more sophisticated GPU handling in the future.
This pull request introduces a temporary workaround for Issue #53, where the system defaults to using GPU ID 0 when multiple GPUs are available. This interim solution is intended to ensure stability and consistent behavior in multi-GPU environments until we develop and deploy a comprehensive testing suite for distributed inference.
Key Changes:
Future Work:
This approach allows us to maintain operational integrity while laying the groundwork for more sophisticated GPU handling in the future.