search

rockchip-linux / rknn-toolkit

BSD 3-Clause "New" or "Revised" License
808 stars 173 forks source link

PyTorch的QAT模型转换失败 #396

Open yoyokitartora opened 1 year ago

yoyokitartora commented 1 year ago

PyTorch 版本：1.10.0 rknn-toolkit 版本：v1.7.1和v1.7.5

使用PyTorch做QAT训练，QAT模型trace之后做转换，报错信息是： E Traceback (most recent call last): E File "rknn/base/RKNNlib/RK_nn.py", line 167, in rknn.base.RKNNlib.RK_nn.RKnn.load_pytorch E File "rknn/base/RKNNlib/app/importer/import_pytorch.py", line 131, in rknn.base.RKNNlib.app.importer.import_pytorch.ImportPytorch.run E File "rknn/base/RKNNlib/converter/convert_pytorch_new.py", line 5245, in rknn.base.RKNNlib.converter.convert_pytorch_new.convert_pytorch.load E File "rknn/base/RKNNlib/converter/convert_pytorch_new.py", line 5353, in rknn.base.RKNNlib.converter.convert_pytorch_new.convert_pytorch.parse_nets E File "rknn/base/RKNNlib/converter/pytorch_util.py", line 543, in rknn.base.RKNNlib.converter.pytorch_util.get_weight_quant_params E AssertionError E Please feedback the detailed log file to the RKNN Toolkit development team. E You can also check github issues: https://github.com/rockchip-linux/rknn-toolkit/issues

尝试使用v1.7.1和v1.7.5的rknn-toolkit，报错信息是一样的。

请问这可能是什么问题呢，要怎么解决？

yu9s commented 8 months ago

PyTorch 版本：1.10.0 rknn-toolkit 版本：v1.7.1和v1.7.5

使用PyTorch做QAT训练，QAT模型trace之后做转换，报错信息是： E Traceback (most recent call last): E File "rknn/base/RKNNlib/RK_nn.py", line 167, in rknn.base.RKNNlib.RK_nn.RKnn.load_pytorch E File "rknn/base/RKNNlib/app/importer/import_pytorch.py", line 131, in rknn.base.RKNNlib.app.importer.import_pytorch.ImportPytorch.run E File "rknn/base/RKNNlib/converter/convert_pytorch_new.py", line 5245, in rknn.base.RKNNlib.converter.convert_pytorch_new.convert_pytorch.load E File "rknn/base/RKNNlib/converter/convert_pytorch_new.py", line 5353, in rknn.base.RKNNlib.converter.convert_pytorch_new.convert_pytorch.parse_nets E File "rknn/base/RKNNlib/converter/pytorch_util.py", line 543, in rknn.base.RKNNlib.converter.pytorch_util.get_weight_quant_params E AssertionError E Please feedback the detailed log file to the RKNN Toolkit development team. E You can also check github issues: https://github.com/rockchip-linux/rknn-toolkit/issues

尝试使用v1.7.1和v1.7.5的rknn-toolkit，报错信息是一样的。

请问这可能是什么问题呢，要怎么解决？

请问qat之后，精度损失有降低吗？