async algorithm conflict with fused optimizer, can only support flattened inplac...
todo opened this issue · comments
bagua/bagua/torch_api/algorithms/async_model_average.py
Lines 84 to 89 in 8707188
Bagua Speeds up PyTorch
todo opened this issue · comments
bagua/bagua/torch_api/algorithms/async_model_average.py
Lines 84 to 89 in 8707188
TODO
comment in 8707188. It's been assigned to @wangraying because they committed the code.