yeandy opened a new issue, #22895: URL: https://github.com/apache/beam/issues/22895
### What happened? When using multiple processes (when we don’t use `no_use_multiple_sdk_containers`) in DataFlow jobs, this error has been observed `Expected all tensors to be on the same device, but found at least two devices, cpu and cuda:0!` The current solution is to add the `experiment=no_use_multiple_sdk_containers` flag. We should look into how much benefit we get with multiple processes, as opposed to maxing out GPU utilization in one worker. ### Issue Priority Priority: 2 ### Issue Component Component: sdk-py-core -- 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]
