Metrics

Definition of metrics used during training.

---

source

Metric

 Metric (name:str, device)

Base metric class.

---

source

Mean

 Mean (name:str, device)

Mean metric, used for loss.

---

source

Accuracy

 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