the checkpoints have been reported but the return value is None when call ray.trainget_checkpoint() . I am workning on a single machine with 2.34
how are you seeing that the checkpoints have been reported? can you share the log output?
Each time when I call ray.train.get_checkpoint(), the Ray has been restart and the orginal seesion seems gone. Is this the cause ?
And you’ve submitted a Training job already? Can you share a repro script where this happens as well as the CLI commands you are running that is showing this behavior?