Open jacob1017 opened 4 years ago
@jonathanasdf thank you reply. This is a feature request. We have implemented this feature based on tf1.5 branch and planing to contribute this feature. Horovod support large scale training and make large audio dataset used efficient.
Thank you for opening this issue. Is this a feature request or are you planning to contribute this feature? This feature is not currently on our horizon.