INT8 conversion of EfficientDet-lite1 to tensort fails to produce any output
Open
@zerollzeng is already working on this.
Since Nov 4, 2023.
triaged
- Dominant language
- C++
- Stars
- 13.4k
- Forks
- 2.4k
- Avg merge
- 5d 3h
- Merged PRs (30d)
- 2
Description
I am experiencing a similar issue to #1464
I have a locally trained model efficientdet-lite1 and have exported to onnx and then tensorrt using the methods described in the readme.
Converting to 16bit, I get performance as expected however, with int8 I get no bboxes or junk results.
I have tried a pre-trained checkpoint and the same results, using coco images for quantisation.
This is using TensorRT 8.5.2.2.
Using the latest container nvcr.io/nvidia/tensorrt:23.10-py3 TensorRT 8.6.1.6, the pre-trained checkpoint works with int8, but my locally trained version does not.
Any suggestions?
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Assessment
This issue has not been assessed yet.