ArmNN Delegate GPU execution error
jishminor opened this issue · comments
When using the ArmNN gpu optimization option, if two models are loaded using ArmNN acceleration (either gpu or cpu for the second model) the gpu model fails to execute. Only when the gpu model is the only loaded model does it behave properly. It seems that only one tflite armnn delegate is created across all the model instances. There may be an issue here with shared state.