Tengine icon indicating copy to clipboard operation
Tengine copied to clipboard

Tengine is a lite, high performance, modular inference engine for embedded device

Results 158 Tengine issues
Sort by recently updated
recently updated
newest added

Hi, can this be compiled and used on a Windows pc? Thanks!

运行环境:Windows 编译器:minGW64 用自己的UNet分割模型进行图像分割,create_graph时候报错,信息如下: > tengine-lite library version: 1.5-dev > Tengine FATAL: Infer node(id: 40, op: Pad) shape failed. >Tengine: Infer shape of graph failed(-1). >Prerun multithread graph failed. 模型文件是Pytorch1.x训练得到的onnx文件,www.convertmodel.com转成tmfile的 对模型的部分进行了可视化:...

RK3399 Ubuntu 20.04 Latest Tengine source code, ACL v20.05 After successfully compile, run example below: sudo ./tm_mobilenet_ssd_acl -m ~/models/Tengine/mobilenet_ssd/mobilenet_ssd.tmfile -i ../../pose.jpg -r 20 -t 1 会产生core文件,崩溃的栈信息如下。没有acl的命令行可以正常工作。 #0 0x0000000000000000 in ??...

[HELP] Kernel xx backend xx has been registered! 查了很久不知道是哪不匹配,有人遇到类似的问题吗? 步骤: 按照#### 2.5.5.6 准备 VIM3/VIM3L 最新版本的编译(更新于2022.04.14)替换`galcore.ko`和`*.so` 参数: TIM-VX: V1.1.37 3rdp sdk: aarch64_A311D_6.4.9.tgz Galcore version 6.4.6.2 aml-npu/focal,now 6.4.8.7.1.1.1-1 arm64 [installed] ``` $...

我们自研的 NPU 芯片,如何借助该框架进行支持

请问一下,如果我将onnx格式转tengine格式后部署到EAIDK上,后续推理部分源码的编写有语言的限制吗?C++和Python是否都可以使用?

after compiled and installed,run as ``` ./cl_build/install/bin/tm_yolov3_opencl -m models/yolov3.tmfile -i images/ssd_dog.jpg -r 1 -t 1 ``` with gdb debug info: ``` tengine-lite library version: 1.5-dev Thread 1 "tm_yolov3_openc" received signal...

examples : 整个文件夹中未找到这个头文件,是改名了吗

尝试对三维tensor做transpose操作,run_graph()报错Bus error。错误信息具体如图所示,tensor维度为三维却调用了transpose4d。扩充tensor维度到4维后再进行transpose操作,报错Segmentation error,tengine调用transpose5d。 ![1654066554(1)](https://user-images.githubusercontent.com/82324996/171346526-6b65ddd1-4f81-47ac-9564-ff89fe1cd6af.jpg) ![1654066597(1)](https://user-images.githubusercontent.com/82324996/171346538-6ffeee4b-2b59-41df-94c7-174aaed55eee.jpg) ![1654067031(1)](https://user-images.githubusercontent.com/82324996/171346873-20167f62-d213-49aa-bbcd-f639901d9c94.jpg)

硬件平台:RV1126 模型:官方下载的yolov5模型yolov5s_uint8.tmfile 执行命令:./tm_yolov5s_timvx -m yolov5s_uint8.tmfile -i 111.jpg -r 10 官方的dog.jpg和部分图像能识别正确,但部分图像识别结果为无数个检测框。如下: ![图片](https://user-images.githubusercontent.com/20926030/160765813-6faf6f7a-c740-40fa-9c84-86c0ed62450d.png) ![图片](https://user-images.githubusercontent.com/20926030/160765893-8da094e2-c091-49c9-9dd0-3c84d5690078.png) 请问这是为什么?还望解答!谢谢!