Created
February 16, 2019 12:05
-
-
Save GINK03/9ac2151159c6cc0887e5be36cd341dc2 to your computer and use it in GitHub Desktop.
pytorch_matrix_factorize
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
class MF(nn.Module): | |
def __init__(self, input_items, input_users): | |
super(MF, self).__init__() | |
print('item size', input_items) | |
self.l_b1 = nn.Embedding(num_embeddings=input_items, embedding_dim=768) | |
self.l_b2 = nn.Linear( | |
in_features=768, out_features=512, bias=True) | |
self.l_a1 = nn.Embedding(num_embeddings=input_users, embedding_dim=768) | |
self.l_a2 = nn.Linear( | |
in_features=768, out_features=512, bias=True) | |
self.l_l1 = nn.Linear( | |
in_features=512, out_features=1, bias=True) | |
def encode_item(self, x): | |
x = self.l_b1(x) | |
x = F.relu(self.l_b2(x)) | |
return x | |
def encode_user(self, x): | |
x = self.l_a1(x) | |
x = F.relu(self.l_a2(x)) | |
return x | |
def forward(self, inputs): | |
item_vec, user_vec = inputs | |
item_vec = self.encode_item(item_vec) | |
user_vec = self.encode_user(user_vec) | |
return F.relu(self.l_l1(user_vec * item_vec)) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment