Function TRITONSERVER_ServerOptionsSetTensorFlowGpuMemoryFraction

• Defined in File tritonserver.h

Function Documentation

TRITONSERVER_Error *TRITONSERVER_ServerOptionsSetTensorFlowGpuMemoryFraction(TRITONSERVER_ServerOptions *options, float fraction)

Set the fraction of GPU memory dedicated to TensorFlow models on each GPU visible to the inference server.

Zero (0) indicates that no memory will be dedicated to TensorFlow and that it will instead allocate memory as needed.

Return

a TRITONSERVER_Error indicating success or failure.

Parameters

• options: The server options object.

• fraction: The fraction of the GPU memory dedicated to TensorFlow.