Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
|
@@ -41,12 +41,15 @@ classifier = pipeline("zero-shot-classification", model="facebook/bart-large-mnl
|
|
| 41 |
|
| 42 |
|
| 43 |
#defs
|
| 44 |
-
def classify_reviews(reviews):
|
| 45 |
-
|
| 46 |
-
|
| 47 |
-
|
|
|
|
|
|
|
|
|
|
| 48 |
return probabilities
|
| 49 |
-
|
| 50 |
def top_rating(scores):
|
| 51 |
return scores.index(max(scores)) + 1
|
| 52 |
|
|
@@ -62,8 +65,13 @@ def filter_dataframe(df, review_column, filter_words):
|
|
| 62 |
# Return full DataFrame if filter_words is empty or contains only spaces
|
| 63 |
if not filter_words or all(word.isspace() for word in filter_words):
|
| 64 |
return df
|
| 65 |
-
|
| 66 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 67 |
|
| 68 |
|
| 69 |
|
|
@@ -73,16 +81,15 @@ def process_filter_words(filter_words_input):
|
|
| 73 |
|
| 74 |
|
| 75 |
# Function for classifying with the new model
|
| 76 |
-
def classify_with_new_classes(reviews, class_names):
|
| 77 |
class_scores = []
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
result = classifier(
|
| 81 |
scores_dict = dict(zip(result['labels'], result['scores']))
|
| 82 |
# Reorder scores to match the original class_names order
|
| 83 |
scores = [scores_dict[name] for name in class_names]
|
| 84 |
-
class_scores.
|
| 85 |
-
|
| 86 |
return class_scores
|
| 87 |
|
| 88 |
|
|
@@ -98,7 +105,9 @@ def main():
|
|
| 98 |
|
| 99 |
if file is not None:
|
| 100 |
try:
|
| 101 |
-
|
|
|
|
|
|
|
| 102 |
# Drop rows where all columns are NaN
|
| 103 |
df = df.dropna(how='all')
|
| 104 |
# Replace blank spaces with NaN, then drop rows where all columns are NaN again
|
|
|
|
| 41 |
|
| 42 |
|
| 43 |
#defs
|
| 44 |
+
def classify_reviews(reviews, batch_size=100):
|
| 45 |
+
probabilities = []
|
| 46 |
+
for i in range(0, len(reviews), batch_size):
|
| 47 |
+
batch_reviews = reviews[i:i+batch_size]
|
| 48 |
+
inputs = tokenizer(batch_reviews, return_tensors='pt', truncation=True, padding=True, max_length=512)
|
| 49 |
+
outputs = model(**inputs)
|
| 50 |
+
probabilities.extend(F.softmax(outputs.logits, dim=1).tolist())
|
| 51 |
return probabilities
|
| 52 |
+
|
| 53 |
def top_rating(scores):
|
| 54 |
return scores.index(max(scores)) + 1
|
| 55 |
|
|
|
|
| 65 |
# Return full DataFrame if filter_words is empty or contains only spaces
|
| 66 |
if not filter_words or all(word.isspace() for word in filter_words):
|
| 67 |
return df
|
| 68 |
+
|
| 69 |
+
# Use multiprocessing to speed up fuzz token set ratio calculation
|
| 70 |
+
from multiprocessing import Pool
|
| 71 |
+
with Pool() as pool:
|
| 72 |
+
filter_scores = pool.map(lambda x: max([fuzz.token_set_ratio(x, word) for word in filter_words]), df[review_column])
|
| 73 |
+
return df[np.array(filter_scores) > 70] # Adjust this threshold as necessary
|
| 74 |
+
|
| 75 |
|
| 76 |
|
| 77 |
|
|
|
|
| 81 |
|
| 82 |
|
| 83 |
# Function for classifying with the new model
|
| 84 |
+
def classify_with_new_classes(reviews, class_names, batch_size=100):
|
| 85 |
class_scores = []
|
| 86 |
+
for i in range(0, len(reviews), batch_size):
|
| 87 |
+
batch_reviews = reviews[i:i+batch_size]
|
| 88 |
+
result = classifier(batch_reviews, class_names)
|
| 89 |
scores_dict = dict(zip(result['labels'], result['scores']))
|
| 90 |
# Reorder scores to match the original class_names order
|
| 91 |
scores = [scores_dict[name] for name in class_names]
|
| 92 |
+
class_scores.extend(scores)
|
|
|
|
| 93 |
return class_scores
|
| 94 |
|
| 95 |
|
|
|
|
| 105 |
|
| 106 |
if file is not None:
|
| 107 |
try:
|
| 108 |
+
chunk_size = 10000 # adjust this value depending on your available memory
|
| 109 |
+
df = pd.concat(pd.read_excel(file, chunksize=chunk_size))
|
| 110 |
+
|
| 111 |
# Drop rows where all columns are NaN
|
| 112 |
df = df.dropna(how='all')
|
| 113 |
# Replace blank spaces with NaN, then drop rows where all columns are NaN again
|