I’m using python API for tensorrt. I’m reading through documentation and can’t find in which units is max_workspace. I often see “1<<30” as value which amounts to 1073741824 bytes which is 1,07 GB and sounds reasonable, so any layer can have at most 1,07 GB during inference?
Related topics
Topic | Replies | Views | Activity | |
---|---|---|---|---|
The significance of builder.max_workspace_size. | 2 | 5518 | April 15, 2019 | |
Minimize gpu usage with Tensorrt | 2 | 554 | October 18, 2021 | |
why tensorrt engine inference speed become much slower if I increase the input image size | 1 | 1066 | December 18, 2019 | |
TensorRT 6.0 Float32 engine will use MaxWorkspace | 1 | 919 | January 16, 2020 | |
How to set workspace in Tensorrt Python API when converting from onnx to engine model | 5 | 1337 | June 29, 2023 | |
How to set parameters when using TensorRT for optimizing InceptionV4 using Jetson TX2? | 2 | 418 | October 18, 2021 | |
How to calculate proper tensorrt workspace size? | 3 | 5693 | January 28, 2021 | |
TensorRT 5 builder when set max_batch_size to 8 the output shape? | 3 | 1434 | October 18, 2021 | |
AssertionError: Max workspace size for TensorRT inference should be positive, got 0 | 5 | 1152 | October 4, 2021 | |
setMaxWorkspaceSize parameter clarification | 1 | 2140 | February 25, 2021 |