Metrics

Definition of metrics used during training.

---

source

Metric

def Metric(
    name:str, device
):

Base metric class.

---

source

Mean

def Mean(
    name:str, device
):

Mean metric, used for loss.

---

source

Accuracy

def Accuracy(
    name:str, device
):

Accuracy metric.

Example usage:

a = Accuracy("mean", "cpu")
print(a, a.empty)

a.update_state(torch.Tensor([3,2,2,1]), torch.Tensor([1,2,2,1]))
print(a, a.empty)

a.update_state(torch.Tensor([1,2,2,3]), torch.Tensor([1,2,2,3]))
print(a, a.empty)

a.reset_state()
print(a, a.empty)

mean=nan True
mean=0.75 False
mean=0.875 False
mean=nan True