Jiahuita
commited on
Commit
•
c356db2
1
Parent(s):
13ad768
Update pipeline to resolve tranformer issue
Browse files- pipeline.py +28 -7
pipeline.py
CHANGED
@@ -2,6 +2,7 @@ from transformers import PreTrainedModel, PretrainedConfig
|
|
2 |
from tensorflow.keras.models import load_model
|
3 |
from tensorflow.keras.preprocessing.text import tokenizer_from_json
|
4 |
from tensorflow.keras.preprocessing.sequence import pad_sequences
|
|
|
5 |
import numpy as np
|
6 |
import json
|
7 |
|
@@ -17,15 +18,35 @@ class NewsClassifier(PreTrainedModel):
|
|
17 |
|
18 |
def __init__(self, config):
|
19 |
super().__init__(config)
|
20 |
-
|
21 |
-
|
|
|
|
|
|
|
22 |
tokenizer_data = json.load(f)
|
23 |
self.tokenizer = tokenizer_from_json(tokenizer_data)
|
|
|
|
|
|
|
|
|
|
|
|
|
24 |
|
25 |
-
def forward(self, inputs):
|
26 |
-
sequences = self.tokenizer.texts_to_sequences([inputs])
|
27 |
padded = pad_sequences(sequences, maxlen=self.config.max_length)
|
28 |
predictions = self.model.predict(padded)
|
29 |
-
|
30 |
-
|
31 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
from tensorflow.keras.models import load_model
|
3 |
from tensorflow.keras.preprocessing.text import tokenizer_from_json
|
4 |
from tensorflow.keras.preprocessing.sequence import pad_sequences
|
5 |
+
import os
|
6 |
import numpy as np
|
7 |
import json
|
8 |
|
|
|
18 |
|
19 |
def __init__(self, config):
|
20 |
super().__init__(config)
|
21 |
+
model_path = os.path.join(os.path.dirname(__file__), 'news_classifier.h5')
|
22 |
+
tokenizer_path = os.path.join(os.path.dirname(__file__), 'tokenizer.json')
|
23 |
+
|
24 |
+
self.model = load_model(model_path)
|
25 |
+
with open(tokenizer_path, 'r') as f:
|
26 |
tokenizer_data = json.load(f)
|
27 |
self.tokenizer = tokenizer_from_json(tokenizer_data)
|
28 |
+
|
29 |
+
def forward(self, text_input):
|
30 |
+
if isinstance(text_input, str):
|
31 |
+
sequences = self.tokenizer.texts_to_sequences([text_input])
|
32 |
+
else:
|
33 |
+
sequences = self.tokenizer.texts_to_sequences(text_input)
|
34 |
|
|
|
|
|
35 |
padded = pad_sequences(sequences, maxlen=self.config.max_length)
|
36 |
predictions = self.model.predict(padded)
|
37 |
+
|
38 |
+
results = []
|
39 |
+
for score in predictions:
|
40 |
+
label = "foxnews" if score[0] > 0.5 else "nbc"
|
41 |
+
results.append({
|
42 |
+
"label": label,
|
43 |
+
"score": float(score[0] if label == "foxnews" else 1 - score[0])
|
44 |
+
})
|
45 |
+
|
46 |
+
return results[0] if isinstance(text_input, str) else results
|
47 |
+
|
48 |
+
@classmethod
|
49 |
+
def from_pretrained(cls, model_path, **kwargs):
|
50 |
+
config = NewsClassifierConfig.from_pretrained(model_path)
|
51 |
+
model = cls(config)
|
52 |
+
return model
|