Why do we need tkDNN? #13

New Issue

2020-05-18T10:43:04+02:00

stiv-yakovenko commented

2020-05-18 10:43:04 +02:00

(Migrated from github.com)

Sorry, for stupid question, but it is not reflected in README. If I have Tensorrt, then tkDNN is just a config reader for yolo? Why does it use CUDNN directly, does it give any benefit over TRT? Does it give any perfomance benefit, compared to TRT? It would be great to give some insight for the users in the README file.

ceccocats commented

2020-05-18 11:00:11 +02:00

(Migrated from github.com)

It is not a config reader for yolo.
tkDNN is a CUDNN and tensorrt wrapper. When I started the project there wasn't any easy method to do cudnn or tensorrt inference in plain c++.
tkDNN is small and easy to understand and if a new network come out with some new layer we can with low effort implement it.
CUDNN is always slower respect to TensorRT, but for completeness and easy debug both are implemented.

It is not a config reader for yolo. tkDNN is a CUDNN and tensorrt wrapper. When I started the project there wasn't any easy method to do cudnn or tensorrt inference in plain c++. tkDNN is small and easy to understand and if a new network come out with some new layer we can with low effort implement it. CUDNN is always slower respect to TensorRT, but for completeness and easy debug both are implemented.

👍 1

This repo is archived. You cannot comment on issues.

1 Participants

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: mmr/tkDNN#13