search

IIC-SIG-MLsys / HDDT

Distrubuted DNN Training on Heterogeneous GPUs
0 stars 5 forks source link

Improve the simple_inference application #9

Open derekwin opened 4 weeks ago

derekwin commented 4 weeks ago
  1. Improve the existing simple_inference application with a pipeline scheduling solution.
  2. Define the tensor to be transmitted using MPI_Datatype.
  3. The current version does not support GDR, so the GPU memory data needs to be copied to the system memory for transmission.
derekwin commented 2 weeks ago

https://github.com/IIC-SIG-MLsys/HDDT/pull/16