search

NVIDIA / Deep-Learning-Accelerator-SW

NVIDIA DLA-SW, the recipes and tools for running deep learning workloads on NVIDIA DLA cores for inference applications.
Other
181 stars 15 forks source link

About “Mul” #25

Open xu19971109 opened 5 months ago

xu19971109 commented 5 months ago

Hello, I see the description of Mul operators in the Readme says that DLA native supports. But when I use it, only by adjusting the Mul operator to fp16 can I get the correct result. Is there any problem with my config? image image

nvoliver commented 5 months ago

@xu19971109 Can you share a minimal repro case (ideally an ONNX model + your validation pipeline)? Thanks!