PaddlePaddle / PaddleDetection

Object Detection toolkit based on PaddlePaddle. It supports object detection, instance segmentation, multiple object tracking and real-time multi-person keypoint detection.
Apache License 2.0
12.81k stars 2.89k forks source link

ssdlite_mobilenet_v3_large_320_coco.yml加载预训练模型时部分权重shape不一致导致部分权重加载不了 #8785

Closed 535205856 closed 6 months ago

535205856 commented 10 months ago

问题确认 Search before asking

Bug组件 Bug Component

Training

Bug描述 Describe the Bug

ssdlite_mobilenet_v3_large_320_coco.yml加载预训练模型时部分权重shape不一致导致部分权重加载不了

W0108 16:48:14.314376 311413 device_context.cc:447] Please NOTE: device: 0, GPU Compute Capability: 7.0, Driver API Version: 12.2, Runtime API Version: 10.1 W0108 16:48:14.322656 311413 device_context.cc:465] device: 0, cuDNN Version: 7.6. [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [160] in pretrained weight backbone.conv14.linear_conv.bn._mean is unmatched with the shape [80] in model backbone.conv14.linear_conv.bn._mean. And the weight backbone.conv14.linear_conv.bn._mean will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [160] in pretrained weight backbone.conv14.linear_conv.bn._variance is unmatched with the shape [80] in model backbone.conv14.linear_conv.bn._variance. And the weight backbone.conv14.linear_conv.bn._variance will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [160] in pretrained weight backbone.conv14.linear_conv.bn.bias is unmatched with the shape [80] in model backbone.conv14.linear_conv.bn.bias. And the weight backbone.conv14.linear_conv.bn.bias will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [160] in pretrained weight backbone.conv14.linear_conv.bn.weight is unmatched with the shape [80] in model backbone.conv14.linear_conv.bn.weight. And the weight backbone.conv14.linear_conv.bn.weight will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [160, 672, 1, 1] in pretrained weight backbone.conv14.linear_conv.conv.weight is unmatched with the shape [80, 672, 1, 1] in model backbone.conv14.linear_conv.conv.weight. And the weight backbone.conv14.linear_conv.conv.weight will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [960] in pretrained weight backbone.conv15.bottleneck_conv.bn._mean is unmatched with the shape [480] in model backbone.conv15.bottleneck_conv.bn._mean. And the weight backbone.conv15.bottleneck_conv.bn._mean will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [960] in pretrained weight backbone.conv15.bottleneck_conv.bn._variance is unmatched with the shape [480] in model backbone.conv15.bottleneck_conv.bn._variance. And the weight backbone.conv15.bottleneck_conv.bn._variance will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [960] in pretrained weight backbone.conv15.bottleneck_conv.bn.bias is unmatched with the shape [480] in model backbone.conv15.bottleneck_conv.bn.bias. And the weight backbone.conv15.bottleneck_conv.bn.bias will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [960] in pretrained weight backbone.conv15.bottleneck_conv.bn.weight is unmatched with the shape [480] in model backbone.conv15.bottleneck_conv.bn.weight. And the weight backbone.conv15.bottleneck_conv.bn.weight will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [960, 1, 5, 5] in pretrained weight backbone.conv15.bottleneck_conv.conv.weight is unmatched with the shape [480, 1, 5, 5] in model backbone.conv15.bottleneck_conv.conv.weight. And the weight backbone.conv15.bottleneck_conv.conv.weight will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [960] in pretrained weight backbone.conv15.expand_conv.bn._mean is unmatched with the shape [480] in model backbone.conv15.expand_conv.bn._mean. And the weight backbone.conv15.expand_conv.bn._mean will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [960] in pretrained weight backbone.conv15.expand_conv.bn._variance is unmatched with the shape [480] in model backbone.conv15.expand_conv.bn._variance. And the weight backbone.conv15.expand_conv.bn._variance will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [960] in pretrained weight backbone.conv15.expand_conv.bn.bias is unmatched with the shape [480] in model backbone.conv15.expand_conv.bn.bias. And the weight backbone.conv15.expand_conv.bn.bias will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [960] in pretrained weight backbone.conv15.expand_conv.bn.weight is unmatched with the shape [480] in model backbone.conv15.expand_conv.bn.weight. And the weight backbone.conv15.expand_conv.bn.weight will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [960, 160, 1, 1] in pretrained weight backbone.conv15.expand_conv.conv.weight is unmatched with the shape [480, 80, 1, 1] in model backbone.conv15.expand_conv.conv.weight. And the weight backbone.conv15.expand_conv.conv.weight will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [160] in pretrained weight backbone.conv15.linear_conv.bn._mean is unmatched with the shape [80] in model backbone.conv15.linear_conv.bn._mean. And the weight backbone.conv15.linear_conv.bn._mean will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [160] in pretrained weight backbone.conv15.linear_conv.bn._variance is unmatched with the shape [80] in model backbone.conv15.linear_conv.bn._variance. And the weight backbone.conv15.linear_conv.bn._variance will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [160] in pretrained weight backbone.conv15.linear_conv.bn.bias is unmatched with the shape [80] in model backbone.conv15.linear_conv.bn.bias. And the weight backbone.conv15.linear_conv.bn.bias will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [160] in pretrained weight backbone.conv15.linear_conv.bn.weight is unmatched with the shape [80] in model backbone.conv15.linear_conv.bn.weight. And the weight backbone.conv15.linear_conv.bn.weight will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [160, 960, 1, 1] in pretrained weight backbone.conv15.linear_conv.conv.weight is unmatched with the shape [80, 480, 1, 1] in model backbone.conv15.linear_conv.conv.weight. And the weight backbone.conv15.linear_conv.conv.weight will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [240] in pretrained weight backbone.conv15.mid_se.conv1.bias is unmatched with the shape [120] in model backbone.conv15.mid_se.conv1.bias. And the weight backbone.conv15.mid_se.conv1.bias will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [240, 960, 1, 1] in pretrained weight backbone.conv15.mid_se.conv1.weight is unmatched with the shape [120, 480, 1, 1] in model backbone.conv15.mid_se.conv1.weight. And the weight backbone.conv15.mid_se.conv1.weight will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [960] in pretrained weight backbone.conv15.mid_se.conv2.bias is unmatched with the shape [480] in model backbone.conv15.mid_se.conv2.bias. And the weight backbone.conv15.mid_se.conv2.bias will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [960, 240, 1, 1] in pretrained weight backbone.conv15.mid_se.conv2.weight is unmatched with the shape [480, 120, 1, 1] in model backbone.conv15.mid_se.conv2.weight. And the weight backbone.conv15.mid_se.conv2.weight will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [960] in pretrained weight backbone.conv16.bottleneck_conv.bn._mean is unmatched with the shape [480] in model backbone.conv16.bottleneck_conv.bn._mean. And the weight backbone.conv16.bottleneck_conv.bn._mean will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [960] in pretrained weight backbone.conv16.bottleneck_conv.bn._variance is unmatched with the shape [480] in model backbone.conv16.bottleneck_conv.bn._variance. And the weight backbone.conv16.bottleneck_conv.bn._variance will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [960] in pretrained weight backbone.conv16.bottleneck_conv.bn.bias is unmatched with the shape [480] in model backbone.conv16.bottleneck_conv.bn.bias. And the weight backbone.conv16.bottleneck_conv.bn.bias will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [960] in pretrained weight backbone.conv16.bottleneck_conv.bn.weight is unmatched with the shape [480] in model backbone.conv16.bottleneck_conv.bn.weight. And the weight backbone.conv16.bottleneck_conv.bn.weight will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [960, 1, 5, 5] in pretrained weight backbone.conv16.bottleneck_conv.conv.weight is unmatched with the shape [480, 1, 5, 5] in model backbone.conv16.bottleneck_conv.conv.weight. And the weight backbone.conv16.bottleneck_conv.conv.weight will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [960] in pretrained weight backbone.conv16.expand_conv.bn._mean is unmatched with the shape [480] in model backbone.conv16.expand_conv.bn._mean. And the weight backbone.conv16.expand_conv.bn._mean will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [960] in pretrained weight backbone.conv16.expand_conv.bn._variance is unmatched with the shape [480] in model backbone.conv16.expand_conv.bn._variance. And the weight backbone.conv16.expand_conv.bn._variance will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [960] in pretrained weight backbone.conv16.expand_conv.bn.bias is unmatched with the shape [480] in model backbone.conv16.expand_conv.bn.bias. And the weight backbone.conv16.expand_conv.bn.bias will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [960] in pretrained weight backbone.conv16.expand_conv.bn.weight is unmatched with the shape [480] in model backbone.conv16.expand_conv.bn.weight. And the weight backbone.conv16.expand_conv.bn.weight will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [960, 160, 1, 1] in pretrained weight backbone.conv16.expand_conv.conv.weight is unmatched with the shape [480, 80, 1, 1] in model backbone.conv16.expand_conv.conv.weight. And the weight backbone.conv16.expand_conv.conv.weight will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [160] in pretrained weight backbone.conv16.linear_conv.bn._mean is unmatched with the shape [80] in model backbone.conv16.linear_conv.bn._mean. And the weight backbone.conv16.linear_conv.bn._mean will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [160] in pretrained weight backbone.conv16.linear_conv.bn._variance is unmatched with the shape [80] in model backbone.conv16.linear_conv.bn._variance. And the weight backbone.conv16.linear_conv.bn._variance will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [160] in pretrained weight backbone.conv16.linear_conv.bn.bias is unmatched with the shape [80] in model backbone.conv16.linear_conv.bn.bias. And the weight backbone.conv16.linear_conv.bn.bias will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [160] in pretrained weight backbone.conv16.linear_conv.bn.weight is unmatched with the shape [80] in model backbone.conv16.linear_conv.bn.weight. And the weight backbone.conv16.linear_conv.bn.weight will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [160, 960, 1, 1] in pretrained weight backbone.conv16.linear_conv.conv.weight is unmatched with the shape [80, 480, 1, 1] in model backbone.conv16.linear_conv.conv.weight. And the weight backbone.conv16.linear_conv.conv.weight will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [240] in pretrained weight backbone.conv16.mid_se.conv1.bias is unmatched with the shape [120] in model backbone.conv16.mid_se.conv1.bias. And the weight backbone.conv16.mid_se.conv1.bias will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [240, 960, 1, 1] in pretrained weight backbone.conv16.mid_se.conv1.weight is unmatched with the shape [120, 480, 1, 1] in model backbone.conv16.mid_se.conv1.weight. And the weight backbone.conv16.mid_se.conv1.weight will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [960] in pretrained weight backbone.conv16.mid_se.conv2.bias is unmatched with the shape [480] in model backbone.conv16.mid_se.conv2.bias. And the weight backbone.conv16.mid_se.conv2.bias will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: The shape [960, 240, 1, 1] in pretrained weight backbone.conv16.mid_se.conv2.weight is unmatched with the shape [480, 120, 1, 1] in model backbone.conv16.mid_se.conv2.weight. And the weight backbone.conv16.mid_se.conv2.weight will not be loaded [01/08 16:48:21] ppdet.utils.checkpoint INFO: Finish loading model weights: /nuwa/.cache/paddle/weights/MobileNetV3_large_x1_0_ssld_pretrained.pdparams

复现环境 Environment

Bug描述确认 Bug description confirmation

是否愿意提交PR? Are you willing to submit a PR?

lyuwenyu commented 10 months ago

什么版本的paddle和paddledet