christopher commited on
Commit
23020da
Β·
1 Parent(s): 2bbff8e

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +3 -3
app.py CHANGED
@@ -12,8 +12,8 @@ unnormalized_input_embeddings = np.load("unnormalized_bert_input_embeddings.npy"
12
  index_L2 = IndexFlatL2(unnormalized_input_embeddings.shape[-1])
13
  index_L2.add(unnormalized_input_embeddings)
14
 
15
- index_IP_normalized = IndexFlatIP(normalized_input_embeddings.shape[-1])
16
- index_IP_normalized.add(normalized_input_embeddings)
17
 
18
 
19
  vocab = {v:k for k,v in tokenizer.vocab.items()}
@@ -27,7 +27,7 @@ def get_first_subword(word):
27
 
28
  def search(token_to_lookup, num_neighbors):
29
  i = get_first_subword(token_to_lookup)
30
- _ , I = index_IP_normalized.search(normalized_input_embeddings[i:i+1], num_neighbors)
31
  hits = lookup_table.take(I[0])
32
  results = hits.values[1:]
33
  results = [r for r in results if not "[unused" in r]
 
12
  index_L2 = IndexFlatL2(unnormalized_input_embeddings.shape[-1])
13
  index_L2.add(unnormalized_input_embeddings)
14
 
15
+ index_IP = IndexFlatIP(normalized_input_embeddings.shape[-1])
16
+ index_IP.add(normalized_input_embeddings)
17
 
18
 
19
  vocab = {v:k for k,v in tokenizer.vocab.items()}
 
27
 
28
  def search(token_to_lookup, num_neighbors):
29
  i = get_first_subword(token_to_lookup)
30
+ _ , I = index_IP.search(normalized_input_embeddings[i:i+1], num_neighbors)
31
  hits = lookup_table.take(I[0])
32
  results = hits.values[1:]
33
  results = [r for r in results if not "[unused" in r]