第一遍用cuda10.2不成功，改用cuda11.7遇到了这样的问题 about etsauto HOT 5 CLOSED

@lihuang218 FileNotFoundError: Could not find: cublas64_10.dll. Is it on your PATH? 找一下cuda安装目录下有没有cublas64_10.dll这个文件，或者有没有cublas64_xx.dll，如果有改成cublas64_10.dll试试。

from etsauto.

有一个cublas64_11.dll，修改后依旧报错

PS D:\ETSAuto\ETSAuto-1.1> python ./tools/export.py -o ./engines/yolov6s_bdd_60.onnx -e ./engines/yolov6s_bdd_60.engine --end2end
Traceback (most recent call last):
File "./tools/export.py", line 7, in
import tensorrt as trt
File "E:\python\lib\site-packages\tensorrt_init_.py", line 129, in
ctypes.CDLL(find_lib(lib))
File "E:\python\lib\ctypes_init_.py", line 373, in init
self._handle = _dlopen(self._name, mode)
FileNotFoundError: Could not find module 'C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.7\bin\cublas64_10.dll' (or one of its dependencies). Try using the full path with constructor syntax.
PS D:\ETSAuto\ETSAuto-1.1>

from etsauto.

@lihuang218 C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.7\bin在这个目录下吗

from etsauto.

在我C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.7\bin在这个目录下。

将cuda10.2中的cublas64_10.dll复制到C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.7\bin中，可以成功构建 YOLOV6 的 TensorRT 文件。

但是在构建 环境感知 的 TensorRT 文件时报错如下：
[04/14/2023-20:25:03] [TRT] [V] --------------- Timing Runner: Conv_9 + Add_10 (CudnnConvolution)
[04/14/2023-20:25:03] [TRT] [V] CudnnConvolution has no valid tactics for this config, skipping
[04/14/2023-20:25:03] [TRT] [V] --------------- Timing Runner: Conv_9 + Add_10 (CublasConvolution)
[04/14/2023-20:25:03] [TRT] [V] CublasConvolution has no valid tactics for this config, skipping
[04/14/2023-20:25:03] [TRT] [V] --------------- Timing Runner: Conv_9 + Add_10 (CaskConvolution)
[04/14/2023-20:25:03] [TRT] [V] CaskConvolution has no valid tactics for this config, skipping
[04/14/2023-20:25:03] [TRT] [V] Deleting timing cache: 435 entries, served 806 hits since creation.
[04/14/2023-20:25:03] [TRT] [E] 10: [optimizer.cpp::nvinfer1::builder::cgraph::LeafCNode::computeCosts::3626] Error Code 10: Internal Error (Could not find any implementation for node Conv_9 + Add_10.)
Traceback (most recent call last):
File "./tools/onnx2trt.py", line 55, in
build_engine(ONNX_SIM_MODEL_PATH, TENSORRT_ENGINE_PATH_PY)
File "./tools/onnx2trt.py", line 49, in build_engine
f.write(engine.serialize())
AttributeError: 'NoneType' object has no attribute 'serialize'

from etsauto.

@lihuang218 用trtexec转试试

from etsauto.