Update app.py
Browse files
app.py
CHANGED
|
@@ -36,7 +36,6 @@ def predict(test_query):
|
|
| 36 |
# Add [CLS] at the front
|
| 37 |
temp_token.append('[CLS]')
|
| 38 |
token_list = tokenizer.tokenize(test_query)
|
| 39 |
-
token_list
|
| 40 |
for m,token in enumerate(token_list):
|
| 41 |
temp_token.append(token)
|
| 42 |
# Trim the token to fit the length requirement
|
|
@@ -48,13 +47,10 @@ def predict(test_query):
|
|
| 48 |
input_ids = pad_sequences([tokenizer.convert_tokens_to_ids(txt) for txt in tokenized_texts],
|
| 49 |
maxlen=max_len, dtype="long", truncating="post", padding="post")
|
| 50 |
attention_masks = [[int(i>0) for i in ii] for ii in input_ids]
|
| 51 |
-
attention_masks[0];
|
| 52 |
segment_ids = [[0] * len(input_id) for input_id in input_ids]
|
| 53 |
-
segment_ids[0];
|
| 54 |
input_ids = torch.tensor(input_ids)
|
| 55 |
attention_masks = torch.tensor(attention_masks)
|
| 56 |
segment_ids = torch.tensor(segment_ids)
|
| 57 |
-
import torch
|
| 58 |
|
| 59 |
# Assuming you have defined your model and input_ids somewhere before this
|
| 60 |
device = 'cuda' if torch.cuda.is_available() else 'cpu'
|
|
@@ -70,11 +66,9 @@ def predict(test_query):
|
|
| 70 |
# Make logits into numpy type predict result
|
| 71 |
# The predict result contain each token's all tags predict result
|
| 72 |
predict_results = logits.detach().cpu().numpy()
|
| 73 |
-
predict_results.shape
|
| 74 |
|
| 75 |
from scipy.special import softmax
|
| 76 |
result_arrays_soft = softmax(predict_results[0])
|
| 77 |
-
result_arrays_soft[0]
|
| 78 |
|
| 79 |
result_array = result_arrays_soft
|
| 80 |
result_list = np.argmax(result_array,axis=-1)
|
|
|
|
| 36 |
# Add [CLS] at the front
|
| 37 |
temp_token.append('[CLS]')
|
| 38 |
token_list = tokenizer.tokenize(test_query)
|
|
|
|
| 39 |
for m,token in enumerate(token_list):
|
| 40 |
temp_token.append(token)
|
| 41 |
# Trim the token to fit the length requirement
|
|
|
|
| 47 |
input_ids = pad_sequences([tokenizer.convert_tokens_to_ids(txt) for txt in tokenized_texts],
|
| 48 |
maxlen=max_len, dtype="long", truncating="post", padding="post")
|
| 49 |
attention_masks = [[int(i>0) for i in ii] for ii in input_ids]
|
|
|
|
| 50 |
segment_ids = [[0] * len(input_id) for input_id in input_ids]
|
|
|
|
| 51 |
input_ids = torch.tensor(input_ids)
|
| 52 |
attention_masks = torch.tensor(attention_masks)
|
| 53 |
segment_ids = torch.tensor(segment_ids)
|
|
|
|
| 54 |
|
| 55 |
# Assuming you have defined your model and input_ids somewhere before this
|
| 56 |
device = 'cuda' if torch.cuda.is_available() else 'cpu'
|
|
|
|
| 66 |
# Make logits into numpy type predict result
|
| 67 |
# The predict result contain each token's all tags predict result
|
| 68 |
predict_results = logits.detach().cpu().numpy()
|
|
|
|
| 69 |
|
| 70 |
from scipy.special import softmax
|
| 71 |
result_arrays_soft = softmax(predict_results[0])
|
|
|
|
| 72 |
|
| 73 |
result_array = result_arrays_soft
|
| 74 |
result_list = np.argmax(result_array,axis=-1)
|