1torch.monitor 2============= 3 4.. warning:: 5 6 This module is a prototype release, and its interfaces and functionality may 7 change without warning in future PyTorch releases. 8 9``torch.monitor`` provides an interface for logging events and counters from 10PyTorch. 11 12The stat interfaces are designed to be used for tracking high level metrics that 13are periodically logged out to be used for monitoring system performance. Since 14the stats aggregate with a specific window size you can log to them from 15critical loops with minimal performance impact. 16 17For more infrequent events or values such as loss, accuracy, usage tracking the 18event interface can be directly used. 19 20Event handlers can be registered to handle the events and pass them to an 21external event sink. 22 23API Reference 24------------- 25 26.. automodule:: torch.monitor 27 28.. autoclass:: torch.monitor.Aggregation 29 :members: 30 31.. autoclass:: torch.monitor.Stat 32 :members: 33 :special-members: __init__ 34 35.. autoclass:: torch.monitor.data_value_t 36 :members: 37 38.. autoclass:: torch.monitor.Event 39 :members: 40 :special-members: __init__ 41 42.. autoclass:: torch.monitor.EventHandlerHandle 43 :members: 44 45.. autofunction:: torch.monitor.log_event 46 47.. autofunction:: torch.monitor.register_event_handler 48 49.. autofunction:: torch.monitor.unregister_event_handler 50 51.. autoclass:: torch.monitor.TensorboardEventHandler 52 :members: 53 :special-members: __init__ 54