项目地址：https://aistudio.baidu.com/bjcpu02/user/192331/3199266/notebooks/3199266.ipynb
软件代码：
num_classes = len(train_dataset.labels)
print('num_class: ', num_classes)
# 定义PPYOLO模型
model = pdx.det.PPYOLO(num_classes=num_classes)
# 开始进行训练
model.train(
num_epochs=10, # 设置训练轮数
train_dataset=train_dataset, # 设置训练数据集
train_batch_size=8, # 设置Bs
eval_dataset=eval_dataset, # 设置评估验证数据集
learning_rate=3e-5, # 设置学习率
warmup_steps=90,
warmup_start_lr=0.0,
# 定义保村间隔轮数，即每7轮保存一次训练的模型结果
save_interval_epochs=7,
lr_decay_epochs=[42, 70], # 定义学习率衰减轮数范围
save_dir='output/PPYOLO', # 定义模型保存输出文件目录
# 定义预训练模型参数文件
use_vdl=True) # 定义使用Visual DL可视化工具
运行上述代码时，执行默认的预训练模型ResNet50_vd_ssld_pretrained.pdparams，却为什么有下面warnings:
num_class: 6
2021-12-07 16:39:09 [INFO] Loading pretrained model from output/PPYOLO/pretrain/ResNet50_vd_ssld_pretrained.pdparams
2021-12-07 16:39:10 [WARNING] backbone.res5.res5a.branch2b.conv_offset.weight is not in pretrained model
2021-12-07 16:39:10 [WARNING] backbone.res5.res5a.branch2b.conv_offset.bias is not in pretrained model
2021-12-07 16:39:10 [WARNING] backbone.res5.res5b.branch2b.conv_offset.weight is not in pretrained model
2021-12-07 16:39:10 [WARNING] backbone.res5.res5b.branch2b.conv_offset.bias is not in pretrained model
2021-12-07 16:39:10 [WARNING] backbone.res5.res5c.branch2b.conv_offset.weight is not in pretrained model
2021-12-07 16:39:10 [WARNING] backbone.res5.res5c.branch2b.conv_offset.bias is not in pretrained model
2021-12-07 16:39:10 [WARNING] neck.yolo_block.0.conv_module.conv0.conv.conv.weight is not in pretrained model
2021-12-07 16:39:10 [WARNING] neck.yolo_block.0.conv_module.conv0.conv.batch_norm.weight is not in pretrained model
2021-12-07 16:39:10 [WARNING] neck.yolo_block.0.conv_module.conv0.conv.batch_norm.bias is not in pretrained model
2021-12-07 16:39:10 [WARNING] neck.yolo_block.0.conv_module.conv0.conv.batch_norm._mean is not in pretrained model
2021-12-07 16:39:10 [WARNING] neck.yolo_block.0.conv_module.conv0.conv.batch_norm._variance is not in pretrained mode
....
2021-12-07 16:39:10 [WARNING] yolo_head.yolo_output.0.weight is not in pretrained model
2021-12-07 16:39:10 [WARNING] yolo_head.yolo_output.0.bias is not in pretrained model
2021-12-07 16:39:10 [WARNING] yolo_head.yolo_output.1.weight is not in pretrained model
2021-12-07 16:39:10 [WARNING] yolo_head.yolo_output.1.bias is not in pretrained model
2021-12-07 16:39:10 [WARNING] yolo_head.yolo_output.2.weight is not in pretrained model
2021-12-07 16:39:10 [WARNING] yolo_head.yolo_output.2.bias is not in pretrained model
这是预训练模型参数和实际模型有一些不匹配的地方，这些不匹配的参数会随机初始化，而不会使用预训练参数。
比如你根据实际任务更改了模型的输出层，你实际的输出层与预训练模型训练时候类别不一致，因此输出层参数矩阵是不一样的。在加载预训练参数时候，就没办法加载，因此会打印一个警告信息来提示你咯。
不过这种情况是正常的，如果不想看到这个警告信息，可以使用下面的代码：
谢谢！