量化导入自己pth模型时候出现 KeyError: 'state_dict'

[maziyi234]

(pytorch1) maziyi@kpyf:~/python/yolox5/micronet/micronet/compression/quantization/wqaq/iao$ python main.py --resume "/home/maziyi/python/yolox5/yolox_best.pth" --q_type 0 --q_level 0 --bn_fuse --qaft --lr 0.00001
==> Options: Namespace(a_bits=8, batch_size=32, bn_fuse=True, bn_fuse_calib=False, cpu=False, data='../../../../data', end_epochs=300, eval_batch_size=32, gpu_id='', lr='0.00001', model_type=1, num_workers=2, percentile=0.999999, pretrained_model=False, prune_qaft='', prune_quant='', ptq=False, ptq_batch=200, ptq_control=False, q_level=0, q_type=0, qaft=True, refine='', resume='/home/maziyi/python/yolox5/yolox_best.pth', start_epochs=1, w_bits=8, wd=1e-05, weight_observer=0)
==> Preparing data..
Files already downloaded and verified
Files already downloaded and verified
Reume model
ori_model
Net(
(model): Sequential(
(0): ConvBNReLU(
(conv): Conv2d(3, 256, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))
(bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
(1): ConvBNReLU(
(conv): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1), groups=2)
(bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
(2): ConvBNReLU(
(conv): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1), groups=2)
(bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
(3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
(4): ConvBNReLU(
(conv): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=16)
(bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
(5): ConvBNReLU(
(conv): Conv2d(512, 512, kernel_size=(1, 1), stride=(1, 1), groups=4)
(bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
(6): ConvBNReLU(
(conv): Conv2d(512, 512, kernel_size=(1, 1), stride=(1, 1), groups=4)
(bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
(7): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
(8): ConvBNReLU(
(conv): Conv2d(512, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32)
(bn): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
(9): ConvBNReLU(
(conv): Conv2d(1024, 1024, kernel_size=(1, 1), stride=(1, 1), groups=8)
(bn): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
(10): ConvBNReLU(
(conv): Conv2d(1024, 10, kernel_size=(1, 1), stride=(1, 1))
(bn): BatchNorm2d(10, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
(11): AvgPool2d(kernel_size=8, stride=1, padding=0)
)
)

quant_model
Net(
(model): Sequential(
(0): ConvBNReLU(
(conv): QuantBNFuseConv2d(
3, 256, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2)
(activation_quantizer): SymmetricQuantizer(
(observer): MovingAverageMinMaxObserver()
)
(weight_quantizer): SymmetricQuantizer(
(observer): MinMaxObserver()
)
)
(bn): Identity()
(relu): ReLU(inplace=True)
)
(1): ConvBNReLU(
(conv): QuantBNFuseConv2d(
256, 256, kernel_size=(1, 1), stride=(1, 1), groups=2
(activation_quantizer): SymmetricQuantizer(
(observer): MovingAverageMinMaxObserver()
)
(weight_quantizer): SymmetricQuantizer(
(observer): MinMaxObserver()
)
)
(bn): Identity()
(relu): ReLU(inplace=True)
)
(2): ConvBNReLU(
(conv): QuantBNFuseConv2d(
256, 256, kernel_size=(1, 1), stride=(1, 1), groups=2
(activation_quantizer): SymmetricQuantizer(
(observer): MovingAverageMinMaxObserver()
)
(weight_quantizer): SymmetricQuantizer(
(observer): MinMaxObserver()
)
)
(bn): Identity()
(relu): ReLU(inplace=True)
)
(3): QuantMaxPool2d(
kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False
(activation_quantizer): SymmetricQuantizer(
(observer): MovingAverageMinMaxObserver()
)
)
(4): ConvBNReLU(
(conv): QuantBNFuseConv2d(
256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=16
(activation_quantizer): SymmetricQuantizer(
(observer): MovingAverageMinMaxObserver()
)
(weight_quantizer): SymmetricQuantizer(
(observer): MinMaxObserver()
)
)
(bn): Identity()
(relu): ReLU(inplace=True)
)
(5): ConvBNReLU(
(conv): QuantBNFuseConv2d(
512, 512, kernel_size=(1, 1), stride=(1, 1), groups=4
(activation_quantizer): SymmetricQuantizer(
(observer): MovingAverageMinMaxObserver()
)
(weight_quantizer): SymmetricQuantizer(
(observer): MinMaxObserver()
)
)
(bn): Identity()
(relu): ReLU(inplace=True)
)
(6): ConvBNReLU(
(conv): QuantBNFuseConv2d(
512, 512, kernel_size=(1, 1), stride=(1, 1), groups=4
(activation_quantizer): SymmetricQuantizer(
(observer): MovingAverageMinMaxObserver()
)
(weight_quantizer): SymmetricQuantizer(
(observer): MinMaxObserver()
)
)
(bn): Identity()
(relu): ReLU(inplace=True)
)
(7): QuantMaxPool2d(
kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False
(activation_quantizer): SymmetricQuantizer(
(observer): MovingAverageMinMaxObserver()
)
)
(8): ConvBNReLU(
(conv): QuantBNFuseConv2d(
512, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32
(activation_quantizer): SymmetricQuantizer(
(observer): MovingAverageMinMaxObserver()
)
(weight_quantizer): SymmetricQuantizer(
(observer): MinMaxObserver()
)
)
(bn): Identity()
(relu): ReLU(inplace=True)
)
(9): ConvBNReLU(
(conv): QuantBNFuseConv2d(
1024, 1024, kernel_size=(1, 1), stride=(1, 1), groups=8
(activation_quantizer): SymmetricQuantizer(
(observer): MovingAverageMinMaxObserver()
)
(weight_quantizer): SymmetricQuantizer(
(observer): MinMaxObserver()
)
)
(bn): Identity()
(relu): ReLU(inplace=True)
)
(10): ConvBNReLU(
(conv): QuantBNFuseConv2d(
1024, 10, kernel_size=(1, 1), stride=(1, 1)
(activation_quantizer): SymmetricQuantizer(
(observer): MovingAverageMinMaxObserver()
)
(weight_quantizer): SymmetricQuantizer(
(observer): MinMaxObserver()
)
)
(bn): Identity()
(relu): ReLU(inplace=True)
)
(11): QuantAvgPool2d(
kernel_size=8, stride=1, padding=0
(activation_quantizer): SymmetricQuantizer(
(observer): MovingAverageMinMaxObserver()
)
)
)
)
Traceback (most recent call last):
File "main.py", line 479, in
model.load_state_dict(checkpoint["state_dict"])
KeyError: 'state_dict'
求大佬解答

[250zhanghu]

加载出问题了，先定义自己的模型