-
I followed this guide https://github.com/NVIDIA-AI-IOT/yolo_deepstream/tree/main/yolov7_qat
I tried to do QAT with multiple GPUs with `torch.nn.DataParallel`, but I got an error
```
Traceback (mo…
-
The official recommendation for the post-training quantization with accuracy control of OpenVINO models includes two methods: NNCF and Pot. I am wondering which one is better. It appears that the offi…
-
### 🐛 Describe the bug
I tried to benchmark the code from the post-training quantization [tutorial]() and noticed significant performance degradation for quantized models vs. non-quantized ones.
H…
-
### Search before asking
- [X] I have searched the YOLOv8 [issues](https://github.com/ultralytics/ultralytics/issues) and [discussions](https://github.com/ultralytics/ultralytics/discussions) and fou…
-
**Describe the bug**
otx optimize command returns errors for action detection task
**Steps to Reproduce**
1. otx build --task action_detection --train-data-roots tests/assets/cvat_dataset…
-
I have a calibration file obtained with PTQ. It have several lines like this:
> TRT-8401-EntropyCalibration2
image: 3cd7f65a
batch_norm_0.tmp_2: 3e22ff2b
sigmoid_0.tmp_0: 3c010a14
p2o.Mul.1: 3e…
-
[27.07.2023 12:17:50] jasiu-rozowyczlowiek
-
I test the code on the examples of text-image and I found that load_quantized_model will use the original model to replace the quantized model. Can you tell me the version of the packages for this …
-
I use the version of otx1.4.1, running pot optimize, an error is reported (the error log is below), how to solve it, I see that the template is marked in trainingextensions-1.4.1/sources/training_exte…
-
### 一、BackGround 📚
飞桨新IR（PIR）功能建设已经基本完成，当前CI流水线上静态图依然是以飞桨旧IR运行，我们想将CI默认运行的IR切换至PIR，从而能顺利支持未来飞桨基于PIR下的代码提交与验证。但是当前依然有很多单测在PIR模式下的运行会存在问题，修复这些问题成为实现默认切换PIR的必要条件。
### 二、Task Introduction📚
对于存在问题的单测，我们已统一…