CALL gds.graph.project not get the default value while training gds.beta.graphSage.train

The following is the code for my graph
CALL gds.graph.project(
'scamTraining',
{
Transaction: {
label: 'Transaction',
properties: {
num_input: {
property:'num_input',
defaultValue: -100
},
num_output: {
property: 'num_output',
defaultValue: -100
},
total_received: {
property:'total_received',
defaultValue: -100000.0
},
total_sent: {
property:'total_sent',
defaultValue: -100000.0
},
fee: {
property:'fee',
defaultValue: -10
},
min_input: {
property:'min_input',
defaultValue: -100000.0
},
max_input: {
property:'max_input',
defaultValue: -100000.0
},
avg_input: {
property:'avg_input',
defaultValue: -100000.0
}
}
}
}, {
SPENT: {
type: 'SPENT',
orientation: 'NATURAL',
properties: ['value']
}
})
Some of my transaction does not have all properties
But I assigned default value as above
When I try for the gds.beta.graphSage.train as follows
CALL gds.beta.graphSage.train(
'scamTraining',
{
modelName: 'scamTrainingModel',
featureProperties: ['num_input','num_output', 'total_received', 'total_sent', 'fee', 'min_input', 'max_input', 'avg_input'],
aggregator: 'pool',
activationFunction: 'sigmoid',
sampleSizes: [25, 50],
relationshipWeightProperty: 'value',
nodeLabels: ['Transaction'],
relationshipTypes: ['SPENT'],
embeddingDimension: 20
}
) YIELD modelInfo as info
I received the following
{
"modelName": "scamTrainingModel",
"modelType": "graphSage",
"metrics": {
"didConverge": false,
"ranEpochs": 1,
"epochLosses": [
NaN
]
}
}
It cause null embedding value
But the same one worked in neo4j 4.2.1 version the newer version 4.4.5 not working

Please let me know the cause of this issue

Thank you