zmtttt

Results 26 comments of zmtttt

采用微调代码训练后得到的pytorch_model.bin达到了62g，请问有什么办法拆分成多个文件并且满足推理代码的调用格式？

同问，UU有答案了吗，想把model拆分多个

Regarding Onnx Conversion of SimpleBEV model

Have you found an answer to the problem?thanks

YOLOv9 build error

root@b37d8b2aacd1:/workspace/tensorrtx/yolov9/build# sudo ./yolov9 -s ../yolov9-c.wts yolov9-c.engine c [04/22/2024-13:15:57] [W] [TRT] The implicit batch dimension mode has been deprecated. Please create the network with NetworkDefinitionCreationFlag::kEXPLICIT_BATCH flag whenever possible. Loading weights: ../yolov9-c.wts...

YOLOv9 build error

@wang-xinyu ，谢谢，不过const static char* gCalibTablePath = "/home/zhaomt/com/tensorrtx/yolov9/calib/coco_calib"，，还是同样的错误

Bug in Yolov8 INT8 Quantization

您好，请问您解决了吗 @ChangjunDAI

s += f"{n} {names[int(c)]}{'s' * (n > 1)}, " # add to string KeyError: 595

我感觉应该是yaml的问题，我打印onnx和engine的c和det时，感觉有很大的问题

QAT does not work, this is true QAT?

have you solved it？ thanks

Could not find any implementation for node /model.1/conv/Conv + PWN(PWN(/model.1/act/Sigmoid), /model.1/act/Mul).

l also find that it is related to the method of calibration; the int8 engine can be created by using minimaxcalibratiob

[Usage] Multi-GPU training hangs: Watchdog caught collective operation timeout

have you sloved? I met the same problem

OnnxExporterError: Unsupported: ONNX export of operator GridSample with 5D volumetric input.

hi！ hi！ I export opset16 -onnx，and use onnx_graphsurgeon to directly modify the opset to 20，then use trtexec --onnx xx—engine, meeting the same problem：Error Code 3: API Usage Error (Parameter check...