设备:香橙派AIPro
系统:Ubuntu 22.04
CANN版本:7.0.0
模型:Yolov8预训练模型
Python版本:自带的3.9.2
由于ATC是转换成fp16的,没有办法转换为int8模型发挥出Ascend310B4的全部算力,搜索论坛内容得知,需要AMCT进行转换,但是在安装AMCT之后进行转换时,提示错误 The given version [9] is not supported, only version 1 to 8 is supported in this build. 经搜索发现,该错误是由于onnx版本太高,于是我降了onnxruntime的版本到1.8.0,但是onnx的版本无法降为官方推荐的1.9.0,显示build失败,ERROR: Could not build wheels for onnx, which is required to install pyproject.toml-based projects
搜索了下pip源,发现python3.9.2(cp39)没有适配aarch64 linux的包,所以产生了上述错误
后续想要从onnx的源代码build一个版本,也遇到各种报错,故发帖求助有编译好的onnx1.9.0适配香橙派的包么,或者有什么其他解决办法
PS:换了官方同样推荐的onnx==1.10.0的版本,同样报错The given version [9] is not supported, only version 1 to 8 is supported in this build. Segmentation fault (core dumped)
PS2:如果使用atc直接转换int8之后的onnx模型,则会报错
The model has [8] [--domain_version] fields, but only one is allowed.
TraceBack (most recent call last):
Model parse to graph failed, graph name:yolov8n_quant_int8.[FUNC:ModelParseToGraph][FILE:onnx_parser.cc][LINE:925]
ATC model parse ret fail.[FUNC:ParseGraph][FILE:omg.cc][LINE:780]
设备:香橙派AIPro
系统:Ubuntu 22.04
CANN版本:7.0.0
模型:Yolov8预训练模型
Python版本:自带的3.9.2
由于ATC是转换成fp16的,没有办法转换为int8模型发挥出Ascend310B4的全部算力,搜索论坛内容得知,需要AMCT进行转换,但是在安装AMCT之后进行转换时,提示错误 The given version [9] is not supported, only version 1 to 8 is supported in this build. 经搜索发现,该错误是由于onnx版本太高,于是我降了onnxruntime的版本到1.8.0,但是onnx的版本无法降为官方推荐的1.9.0,显示build失败,ERROR: Could not build wheels for onnx, which is required to install pyproject.toml-based projects
搜索了下pip源,发现python3.9.2(cp39)没有适配aarch64 linux的包,所以产生了上述错误
后续想要从onnx的源代码build一个版本,也遇到各种报错,故发帖求助有编译好的onnx1.9.0适配香橙派的包么,或者有什么其他解决办法
PS:换了官方同样推荐的onnx==1.10.0的版本,同样报错The given version [9] is not supported, only version 1 to 8 is supported in this build. Segmentation fault (core dumped)
PS2:如果使用atc直接转换int8之后的onnx模型,则会报错
The model has [8] [--domain_version] fields, but only one is allowed.
TraceBack (most recent call last):
Model parse to graph failed, graph name:yolov8n_quant_int8.[FUNC:ModelParseToGraph][FILE:onnx_parser.cc][LINE:925]
ATC model parse ret fail.[FUNC:ParseGraph][FILE:omg.cc][LINE:780]