Upload model
Browse files- modeling_gzipembed.py +13 -0
modeling_gzipembed.py
CHANGED
@@ -39,6 +39,19 @@ class GZIPEmbeddingModel(PreTrainedModel):
|
|
39 |
x = x.to(self.reduction_head.dtype).to(self.reduction_head.device)
|
40 |
return self.reduction_head(x)
|
41 |
return x if not return_tensor else torch.tensor(x)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
42 |
|
43 |
def normalize(self, x):
|
44 |
x = ''.join([char for char in x.lower() if char in "abcdefghijklmnopqrstuvwxyz "])
|
|
|
39 |
x = x.to(self.reduction_head.dtype).to(self.reduction_head.device)
|
40 |
return self.reduction_head(x)
|
41 |
return x if not return_tensor else torch.tensor(x)
|
42 |
+
|
43 |
+
def encode(self, sentences, batch_size=32, **kwargs):
|
44 |
+
"""
|
45 |
+
Returns a list of embeddings for the given sentences.
|
46 |
+
Args:
|
47 |
+
sentences (`List[str]`): List of sentences to encode
|
48 |
+
batch_size (`int`): Batch size for the encoding
|
49 |
+
|
50 |
+
Returns:
|
51 |
+
`List[np.ndarray]` or `List[tensor]`: List of embeddings for the given sentences
|
52 |
+
"""
|
53 |
+
x = self.forward(sentences, num_procs=batch_size, return_tensor=False)
|
54 |
+
return [torch.tensor(i) for i in x]
|
55 |
|
56 |
def normalize(self, x):
|
57 |
x = ''.join([char for char in x.lower() if char in "abcdefghijklmnopqrstuvwxyz "])
|