File size: 4,331 Bytes
250d4b8 10789df 250d4b8 24bacd6 250d4b8 e55b6a5 ff77f42 e55b6a5 10789df e55b6a5 250d4b8 e55b6a5 10789df 250d4b8 10789df 250d4b8 10789df 250d4b8 10789df 250d4b8 e55b6a5 250d4b8 10789df |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 |
from transformers import pipeline
from sklearn.preprocessing import LabelEncoder
import joblib
import torch
import os
# Debugging: Print current directory and contents
print("Current working directory:", os.getcwd())
print("Contents of the directory:", os.listdir())
# Load the label encoder
label_encoder = joblib.load('/repository/label_encoder.pkl') # Use absolute path
print("Label encoder loaded successfully.")
# Load the model and tokenizer from Hugging Face
model_name = "SCANSKY/distilbertTourism-multilingual-sentiment"
sentiment_analyzer = pipeline(
'sentiment-analysis',
model=model_name,
tokenizer=model_name,
device=0 if torch.cuda.is_available() else -1 # Use GPU if available
)
def get_average_sentiment(positive_count, negative_count, neutral_count):
total = positive_count + negative_count + neutral_count
if total == 0:
return "neutral"
positive_pct = (positive_count / total) * 100
negative_pct = (negative_count / total) * 100
neutral_pct = (neutral_count / total) * 100
max_sentiment = max(positive_pct, negative_pct, neutral_pct)
if max_sentiment == positive_pct:
return "positive"
elif max_sentiment == negative_pct:
return "negative"
else:
return "neutral"
class EndpointHandler:
def __init__(self, model_dir=None):
# Model and tokenizer are loaded globally, so no need to reinitialize here
# The `model_dir` argument is required by Hugging Face's inference toolkit
pass
def preprocess(self, data):
# Extract the input text from the request
text = data.get("inputs", "")
return text
def inference(self, text):
if not text.strip():
return {"error": "Please enter some text for sentiment analysis."}
# Split text into lines
lines = [line.strip() for line in text.split('\n') if line.strip()]
if not lines:
return {"error": "Please enter valid text for sentiment analysis."}
# Analyze each line
total_confidence = 0
positive_count = 0
negative_count = 0
neutral_count = 0
line_results = [] # Store results for each line
for line in lines:
result = sentiment_analyzer(line)
predicted_label_encoded = int(result[0]['label'].split('_')[-1])
predicted_label = label_encoder.inverse_transform([predicted_label_encoded])[0]
confidence = result[0]['score'] * 100
# Store line and its sentiment result
line_results.append({
'text': line,
'sentiment': predicted_label,
'confidence': confidence
})
if predicted_label == 'positive':
positive_count += 1
elif predicted_label == 'negative':
negative_count += 1
else:
neutral_count += 1
total_confidence += confidence
# Calculate averages
avg_confidence = total_confidence / len(lines)
positive_pct = (positive_count / len(lines)) * 100
negative_pct = (negative_count / len(lines)) * 100
neutral_pct = (neutral_count / len(lines)) * 100
# Get average sentiment
avg_sentiment = get_average_sentiment(positive_count, negative_count, neutral_count)
# Prepare the output
output = {
"total_lines_analyzed": len(lines),
"average_confidence": avg_confidence,
"average_sentiment": avg_sentiment,
"sentiment_distribution": {
"positive": positive_pct,
"negative": negative_pct,
"neutral": neutral_pct
},
"line_results": line_results
}
return output
def postprocess(self, output):
if "error" in output:
return [{"error": output["error"]}]
# Return only the line-level results as a list
return output["line_results"]
def __call__(self, data):
# Main method to handle the request
text = self.preprocess(data)
output = self.inference(text)
return self.postprocess(output) |