Hi, how would machine learning projects A.I. model training be capable of utilizing BOINC?

Does BOINC want to integrate or is a third party cluster management software used.

For instance, currently...

1. Use a utility with BOINC to manage the cluster, map and grade the network.
2. Setup a webpage to have model training jobs submitted to a task scheduler utility.
3. Alter the code to add the current cluster info, for instance tf.cluster on tensor and check for tf.distribute.
4. Execute the script on the master node utilizing the cluster directly.

Check all nodes on the same version of tensor and python and working.

This largely bypasses BOINC except for the cluster management and recruitment.

Does BOINC need to integrate the above or be upgraded in some way, or is the above the current situation?

Thanks in advance.

See my reply in link below to your question posted today.
https://boinc.berkeley.edu/forum_thread.php?id=15199

I've emailed them.
From what I can gather. I do not think they trained the models in a distributed way, and I suspect BOINC needs some additional development to distribute model training.