File size: 2,017 Bytes
499de81 | 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 | from typing import Dict, List, Any
import torch
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch.nn.functional as F
class EndpointHandler:
def __init__(self, path: str = "BAAI/bge-reranker-v2-m3"):
# Load tokenizer and model
self.tokenizer = AutoTokenizer.from_pretrained(path)
self.model = AutoModelForSequenceClassification.from_pretrained(path)
self.model.eval()
# Determine the computation device
self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
self.model.to(self.device)
def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]:
"""
Expected input format:
{
"query": "Your query here",
"texts": ["Document 1", "Document 2", ...],
"normalize": true # Optional; defaults to False
}
"""
query = data.get("query")
texts = data.get("texts", [])
normalize = data.get("normalize", False)
if not query or not texts:
return [{"error": "Both 'query' and 'texts' fields are required."}]
# Prepare input pairs
pairs = [[query, text] for text in texts]
# Tokenize input pairs
inputs = self.tokenizer(
pairs,
padding=True,
truncation=True,
return_tensors="pt",
max_length=512
).to(self.device)
with torch.no_grad():
# Get model logits
outputs = self.model(**inputs)
scores = outputs.logits.view(-1)
# Apply sigmoid normalization if requested
if normalize:
scores = torch.sigmoid(scores)
# Prepare the response
results = [
{"index": idx, "score": score.item()}
for idx, score in enumerate(scores)
]
# Sort results by descending score
results.sort(key=lambda x: x["score"], reverse=True)
return results
|