tanp loss (kl loss)

fast_adapt.py

@@ -8,7 +8,7 @@ def cl_loss(c):
     alpha = config['alpha']
     beta = config['beta']
     d = config['d']
-    a = torch.div(1, torch.add(1, torch.exp(torch.mul(-1, torch.mul(alpha, torch.sub(torch.mul(d, c), beta))))))
+    a = torch.div(1, torch.add(1, torch.exp(torch.mul(-1, torch.mul(alpha, torch.sub(torch.mul(d, c.squeeze()), beta))))))
     # a = 1 / (1 + torch.exp((-1) * alpha * (d * c - beta)))
     b = torch.mul(a, torch.mul(torch.sub(1, a), torch.sub(1, torch.mul(2, a))))
     # b = 1 * a * (1 - a) * (1 - 2 * a)

learnToLearn.py

@@ -260,7 +260,7 @@ if __name__ == '__main__':
             kmeans_model = KMeans(n_clusters=config['cluster_k'], init="k-means++").fit(user_embeddings)
             tr.cluster_module.array.data = torch.Tensor(kmeans_model.cluster_centers_).cuda()
 
-        if iteration > (0):
+        if iteration > 0:
             # indexes = data_batching(indexes, C_distribs, batch_size, training_set_size, config['cluster_k'])
             # random.shuffle(indexes)
             C_distribs = []