zh/integrations/tensorrt/

[giscus[bot]]

zh/integrations/tensorrt/

探索将Ultralytics YOLOv8 模型导出为TensorRT 格式的强大功能和灵活性，以提高在英伟达™（NVIDIA®）图形处理器上的性能和效率。

https://docs.ultralytics.com/zh/integrations/tensorrt/

[abcc235]

请问在那儿可以看到
results = tensorrt_model('https://ultralytics.com/images/bus.jpg')中
results的数据结构。

[glenn-jocher]

您好！关于results的数据结构，它包含了模型推理的输出，通常包括：

• 检测框 (boxes): 每个检测对象的位置和大小，通常以[x_min, y_min, x_max, y_max]的形式表示。
• 置信度 (confidences): 模型对每个检测对象的置信度。
• 类别ID (classIDs): 每个检测对象的预测类别ID。

当您运行如下代码:

results = tensorrt_model('https://ultralytics.com/images/bus.jpg')

您可以通过results.xyxy来获取检测框，通过results.conf获取置信度，通过results.cls来获取类别ID。例如，打印所有检测框及其对应的置信度和类别ID：

# 获取xyxy格式的检测框（x_min, y_min, x_max, y_max），置信度和类别ID
for box, conf, cls_id in zip(results.xyxy[0], results.conf[0], results.cls[0]):
    print('检测框:', box.cpu().numpy(), '置信度:', conf, '类别ID:', int(cls_id))

希望这能帮助到您！🙂

[abcc235]

貌似应该这么写，直接按照您写的会报错。
、、、python
for box, conf, cls_id in zip(results[0].boxes.xyxy, results[0].boxes.conf, results[0].boxes.cls):
print('检测框:', box.cpu().numpy(), '置信度:', conf, '类别ID:', int(cls_id))
、、、

[glenn-jocher]

@abcc235 您好！看起来您已经正确地修改了访问Results对象属性的方法，确实需要适当调用属性。您的代码片段中提到将results的索引直接解包到循环中，这是非常合适的做法。以下是一个简化的代码示例，确保您可以顺利读取和打印出检测结果：

results = tensorrt_model('https://ultralytics.com/images/bus.jpg')
for box, conf, cls_id in zip(results[0].boxes.xyxy, results[0].boxes.conf, results[0].boxes.cls):
    print('检测框:', box.cpu().numpy(), '置信度:', conf.item(), '类别ID:', cls_id.item())

这里使用.item()确保从Tensor中提取标量值，使输出更加清晰。如果有进一步问题或需求，欢迎随时提问！🙂

[17638759823]

(yolov8) PS E:\ultralytics> yolo export model=./weights/yolov8n.pt format=engine
WARNING TensorRT requires GPU export, automatically assigning device=0
Ultralytics YOLOv8.0.199 Python-3.8.18 torch-2.1.0+cu121 CUDA:0 (NVIDIA GeForce RTX 3050 OEM, 8192MiB)
YOLOv8n summary (fused): 168 layers, 3151904 parameters, 0 gradients, 8.7 GFLOPs

PyTorch: starting from 'weights\yolov8n.pt' with input shape (1, 3, 640, 640) BCHW and output shape(s) (1, 84, 8400) (6.2 MB)

ONNX: starting export with onnx 1.15.0 opset 17...
到这就停止了，这是什么问题

[pderrenger]

看起来您在导出YOLOv8模型为TensorRT格式时遇到了中断。通常，这个问题可能与几个不同因素相关，包括TensorRT的安装问题、GPU兼容性，或者可能是特定版本的PyTorch或ONNX与TensorRT之间的兼容性问题。

首先，请确保您已正确安装所有必需的TensorRT依赖项，以及NVIDIA驱动和CUDA版本与您的TensorRT版本兼容。此外，建议检查您的PyTorch版本是否支持您正在尝试导出的ONNX版本。

此外，试着使用以下导出命令，并确保device被设定为可用的GPU：

yolo export model=./weights/yolov8n.pt format=engine device=0

如果问题仍然存在，请尝试更新到最新版本的PyTorch和TensorRT，或考虑在Ultralytics的GitHub问题跟踪器中查找类似的已知问题或提交新的问题。确保提供您的操作系统、PyTorch、CUDA和TensorRT版本的详细信息，以便社区更好地帮助您解决问题。🤖

有关更多细节和最佳实践，强烈建议阅读我们的TensorRT导出指南，以获取从YOLOv8模型到TensorRT模型的转换步骤、安装要求和示例代码。希望这能帮到您！

[Ares-cz]

when I use official demo to export into TensorRT, it gave me a feedback on the following error:

AttributeError: 'tensorrt_bindings.tensorrt.IBuilderConfig' object has no attribute 'max_workspace_size'
[06/30/2024-14:20:22] [TRT] [I] [MemUsageChange] Init builder kernel library: CPU +2514, GPU +288, now: CPU 14191, GPU 1967 (MiB)
TensorRT: export failure ❌ 2.1s: 'tensorrt_bindings.tensorrt.IBuilderConfig' object has no attribute 'max_workspace_size'

the system is Windows 11, and the CUDA version is 12.1, the TensorRT version is 10.1

[glenn-jocher]

@Ares-cz hello,

Thank you for reaching out and providing details about the issue you're encountering. The error you're seeing, AttributeError: 'tensorrt_bindings.tensorrt.IBuilderConfig' object has no attribute 'max_workspace_size', suggests there might be a compatibility issue with the versions of TensorRT or CUDA you are using.

Here are a few steps to help resolve this issue:

1. Update Packages: Ensure you are using the latest versions of all relevant packages, including ultralytics, tensorrt, and CUDA. You can update ultralytics using:

   pip install --upgrade ultralytics

2. Reproducible Example: If the issue persists, it would be very helpful if you could provide a minimum reproducible example. This will allow us to better understand the context and provide a more accurate solution. You can find guidelines on creating a reproducible example here: Minimum Reproducible Example.

3. Export Code: Ensure your export code follows the recommended format. Here is an example:

   from ultralytics import YOLO
   
   # Load the YOLOv8 model
   model = YOLO("yolov8n.pt")
   
   # Export the model to TensorRT format
   model.export(format="engine")  # creates 'yolov8n.engine'

4. Check TensorRT Documentation: For more detailed instructions and best practices, please refer to the TensorRT Export Guide.

If you continue to experience issues, please share the specific versions of ultralytics, tensorrt, and CUDA you are using, along with any additional error messages. This information will help us diagnose the problem more effectively.

Thank you for your cooperation, and we look forward to helping you resolve this issue! 😊