def calculate_multiple_f1_scores(preds, labels):
if len(labels.shape) == 1:
return calculate_multiclass_f1_score(preds, labels)
else:
return [
compute_f1_score(preds[:, i], labels[:, i], average="weighted")
for i in range(labels.shape[-1])
]
def flat_f1(settings, preds, labels):
pred_flat = np.argmax(preds, axis=1).flatten()
labels_flat = labels.flatten()
labels = list(range(0, settings.get_num_classifier_labels()))
f1 = compute_f1_score(labels_flat,
pred_flat,
labels,
average='weighted',
zero_division=0)
return f1
def calculate_multiclass_f1_score(preds, labels):
preds = torch.argmax(preds, dim=1).detach().numpy()
labels = labels.numpy()
f1score = compute_f1_score(labels, preds, average="weighted")
return f1score
def calculate_multiclass_f1_score(preds, labels):
f1score = compute_f1_score(labels, preds, average="weighted")
return f1score
def calculate_f1_score(logits, labels):
preds = np.argmax(logits, axis=1)
f1score = compute_f1_score(labels, preds, average="weighted")
return f1score
