format
Browse files- handler.py +7 -4
handler.py
CHANGED
@@ -10,12 +10,13 @@ class EndpointHandler:
|
|
10 |
Initialize the model
|
11 |
"""
|
12 |
self.sign_ids = np.load(os.path.join(path, "sign_ids.npy"))
|
13 |
-
self.sign_embeddings = np.load(
|
|
|
|
|
14 |
|
15 |
hf_model_path = "openai/clip-vit-large-patch14"
|
16 |
self.model = CLIPModel.from_pretrained(hf_model_path)
|
17 |
self.tokenizer = CLIPTokenizer.from_pretrained(hf_model_path)
|
18 |
-
|
19 |
|
20 |
def __call__(self, data: Dict[str, Any]) -> List[float]:
|
21 |
"""
|
@@ -25,7 +26,9 @@ class EndpointHandler:
|
|
25 |
Return:
|
26 |
A :obj:`list` | `dict`: will be serialized and returned
|
27 |
"""
|
28 |
-
token_inputs = self.tokenizer(
|
|
|
|
|
29 |
query_embed = self.model.get_text_features(**token_inputs)
|
30 |
np_query_embed = query_embed.detach().cpu().numpy()[0]
|
31 |
np_query_embed /= np.linalg.norm(np_query_embed)
|
@@ -37,7 +40,7 @@ class EndpointHandler:
|
|
37 |
cos_similarites = w * (self.sign_embeddings @ np_query_embed)
|
38 |
count_above_threshold = np.sum(cos_similarites > threshold)
|
39 |
sign_id_arg_rankings = np.argsort(cos_similarites)[::-1]
|
40 |
-
|
41 |
threshold_id_arg_rankings = sign_id_arg_rankings[:count_above_threshold]
|
42 |
|
43 |
result_sign_ids = self.sign_ids[threshold_id_arg_rankings]
|
|
|
10 |
Initialize the model
|
11 |
"""
|
12 |
self.sign_ids = np.load(os.path.join(path, "sign_ids.npy"))
|
13 |
+
self.sign_embeddings = np.load(
|
14 |
+
os.path.join(path, "vanilla_large-patch14_image_embeddings_normalized.npy")
|
15 |
+
)
|
16 |
|
17 |
hf_model_path = "openai/clip-vit-large-patch14"
|
18 |
self.model = CLIPModel.from_pretrained(hf_model_path)
|
19 |
self.tokenizer = CLIPTokenizer.from_pretrained(hf_model_path)
|
|
|
20 |
|
21 |
def __call__(self, data: Dict[str, Any]) -> List[float]:
|
22 |
"""
|
|
|
26 |
Return:
|
27 |
A :obj:`list` | `dict`: will be serialized and returned
|
28 |
"""
|
29 |
+
token_inputs = self.tokenizer(
|
30 |
+
[data["inputs"]], padding=True, return_tensors="pt"
|
31 |
+
)
|
32 |
query_embed = self.model.get_text_features(**token_inputs)
|
33 |
np_query_embed = query_embed.detach().cpu().numpy()[0]
|
34 |
np_query_embed /= np.linalg.norm(np_query_embed)
|
|
|
40 |
cos_similarites = w * (self.sign_embeddings @ np_query_embed)
|
41 |
count_above_threshold = np.sum(cos_similarites > threshold)
|
42 |
sign_id_arg_rankings = np.argsort(cos_similarites)[::-1]
|
43 |
+
|
44 |
threshold_id_arg_rankings = sign_id_arg_rankings[:count_above_threshold]
|
45 |
|
46 |
result_sign_ids = self.sign_ids[threshold_id_arg_rankings]
|