PannenetsF
commented
2 years ago I've produce a minimal code snippets
import torch
from torchvision.models import resnet18
from mqbench.prepare_by_platform import BackendType, prepare_by_platform
class model(torch.nn.Module):
def __init__(self) -> None:
super().__init__()
self.conv1 = torch.nn.Conv2d(3,3,3)
self.conv2 = torch.nn.Conv2d(3,3,3)
self.conv3 = torch.nn.Conv2d(3,3,3)
def forward(self, x):
data = x['img']
x.update({'conv1': self.conv1(data)})
x.update({'conv2': self.conv2(data)})
x.update({'conv3': self.conv3(data)})
return x
test_model = model()
test_model = prepare_by_platform(test_model, BackendType.Tengine_u8)
print(test_model)
test_model({'img': torch.rand(1,3,224,224)})And I fixed it by https://github.com/PannenetsF/MQBench/tree/tu8
RedHandLM
使用UP框架基于最新mqbench对yolox进行QAT训练，选择backbend=tengine_u8 时报错:AttributeError: 'dict' object has no attribute 'detach'
以下是使用的QAT配置文件：
以下是报错信息：
辛苦帮忙看下是什么问题？是mqbench还没有支持tengine么