使用自定义损失函数编译Keras模型时发生TypeError

问题描述 投票:0回答:1

我使用细分模型库使用Jaccard用自定义损失函数BinaryCrossEntropy训练了一个模型,当我尝试加载和编译模型以开始预测时,出现此错误:model = compile(optimizer='adam', loss=bce_jaccard_loss, metrics=['iou_score']) TypeError: compile() missing required argument 'source' (pos 1)

我该如何解决?

import keras
import segmentation_models as sm
from segmentation_models.losses import bce_jaccard_loss
from segmentation_models.metrics import iou_score

model = keras.models.load_model("ResNet_34_Aug.h5", compile = False)
model = compile(optimizer='adam', loss=bce_jaccard_loss, metrics=['iou_score'])
python keras image-segmentation loss-function pre-trained-model
1个回答
0
投票
整个问题在代码的最后一行,缺少的必需参数'source'(位置1)是模型本身,所以我将其固定为:

model.compile(optimizer='adam', loss=bce_jaccard_loss, metrics=['iou_score'])

© www.soinside.com 2019 - 2024. All rights reserved.