Count of DLA loadables

How can I know the count of dla loadables in my trt engine?

Hi,

You can find this information when creating the TensorRT engine.

For example, there is one loadable for mnist.onnx model:

$ /usr/src/tensorrt/bin/trtexec --onnx=/usr/src/tensorrt/data/mnist/mnist.onnx --useDLACore=0 --allowGPUFallback --buildOnly
&&&& RUNNING TensorRT.trtexec [TensorRT v8401] # /usr/src/tensorrt/bin/trtexec --onnx=/usr/src/tensorrt/data/mnist/mnist.onnx --useDLACore=0 --allowGPUFallback --buildOnly
...
[04/24/2022-16:44:13] [I] Finish parsing network model
[04/24/2022-16:44:13] [W] [TRT] Layer 'Parameter193': Unsupported on DLA. Switching this layer's device type to GPU.
[04/24/2022-16:44:13] [W] [TRT] Layer 'Times212_reshape1': Unsupported on DLA. Switching this layer's device type to GPU.
[04/24/2022-16:44:13] [W] [TRT] Layer 'Parameter6': Unsupported on DLA. Switching this layer's device type to GPU.
[04/24/2022-16:44:13] [W] [TRT] Layer '(Unnamed Layer* 4) [Shuffle]': Unsupported on DLA. Switching this layer's device type to GPU.
[04/24/2022-16:44:13] [W] [TRT] Layer 'Parameter88': Unsupported on DLA. Switching this layer's device type to GPU.
[04/24/2022-16:44:13] [W] [TRT] Layer '(Unnamed Layer* 10) [Shuffle]': Unsupported on DLA. Switching this layer's device type to GPU.
[04/24/2022-16:44:13] [W] [TRT] Layer 'Times212_reshape0': Unsupported on DLA. Switching this layer's device type to GPU.
[04/24/2022-16:44:13] [W] [TRT] Times212: MatMul is unsupported on DLA, relax constraints or use Convolution layer instead.
[04/24/2022-16:44:13] [W] [TRT] Layer 'Times212': Unsupported on DLA. Switching this layer's device type to GPU.
[04/24/2022-16:44:13] [W] [TRT] Layer 'Parameter194': Unsupported on DLA. Switching this layer's device type to GPU.
[04/24/2022-16:44:13] [I] [TRT] ---------- Layers Running on DLA ----------
[04/24/2022-16:44:13] [I] [TRT] [DlaLayer] {ForeignNode[Convolution28...Pooling160]}
[04/24/2022-16:44:13] [I] [TRT] ---------- Layers Running on GPU ----------
[04/24/2022-16:44:13] [I] [TRT] [GpuLayer] SHUFFLE: Times212_reshape0 + reshape_before_Times212
[04/24/2022-16:44:13] [I] [TRT] [GpuLayer] CONVOLUTION: Times212
[04/24/2022-16:44:13] [I] [TRT] [GpuLayer] SHUFFLE: reshape_after_Times212
...
&&&& PASSED TensorRT.trtexec [TensorRT v8401] # /usr/src/tensorrt/bin/trtexec --onnx=/usr/src/tensorrt/data/mnist/mnist.onnx --useDLACore=0 --allowGPUFallback --buildOnly

Thanks.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.