/// <summary>
/// Runs inference on Kon model for a batch of inputs.
/// The shape of each input is the same as that for the non-batch case above.
/// </summary>
public IEnumerable <IEnumerable <string> > Infer(IEnumerable <IEnumerable <float> > dataBatch)
{
List <float> dataCombined = new List <float>();
foreach (var input in dataBatch)
{
dataCombined.AddRange(input);
}
List <Tensor> result = manager.RunModel(
modelName,
int.MaxValue,
inferInputNames,
new List <Tensor> {
new Tensor(dataCombined, new List <long> {
dataBatch.LongCount(), 3, 227, 227
})
},
inferOutputNames
);
List <string> r0 = new List <string>();
result[0].CopyTo(r0);
List <List <string> > results = new List <List <string> >();
results.Add(r0);
return(results);
}
/// <summary>
/// Runs inference on Bear model for a batch of inputs.
/// The shape of each input is the same as that for the non-batch case above.
/// </summary>
public IEnumerable <IEnumerable <string> > Infer(IEnumerable <IEnumerable <float> > dataBatch)
{
List <float> dataCombined = new List <float>();
long batchCount = 0;
foreach (var input in dataBatch)
{
dataCombined.AddRange(input);
++batchCount;
}
List <Tensor> result = manager.RunModel(
modelName,
int.MaxValue,
inferInputNames,
new List <Tensor> {
new Tensor(dataCombined, new List <long> {
batchCount, 3, 227, 227
})
},
inferOutputNames
);
int classLabelBatchNum = (int)result[0].GetShape()[0];
int classLabelBatchSize = (int)result[0].GetShape().Aggregate((a, x) => a * x) / classLabelBatchNum;
for (int batchNum = 0, offset = 0; batchNum < classLabelBatchNum; batchNum++, offset += classLabelBatchSize)
{
List <string> tmp = new List <string>();
result[0].CopyTo(tmp, offset, classLabelBatchSize);
yield return(tmp);
}
}
