renshanhf/weighted-triplet-finetuned-model

Weighted Triplet Fine-Tuned Model

This model is a SentenceTransformer fine-tuned with weighted triplet loss and metadata injection for finance-related semantic search.

Training Data

• Triplet data (anchor, positive, negative) with metadata (domain, year)
• Example: [domain:finance] [year:2024] Quarterly report shows revenue growth.

Sample Data

You can download the sample dataset used for demonstration here:
sample_data.json

Intended Use

• Semantic search and retrieval-augmented generation (RAG) in finance and similar domains.

Limitations

• Trained on synthetic/small dataset for demonstration.
• Metadata format must match training (e.g., [domain:finance] [year:2024] ...).

Example

from sentence_transformers import SentenceTransformer
model = SentenceTransformer("renshanhf/weighted-triplet-finetuned-model")
embedding = model.encode("[domain:finance] [year:2024] Quarterly report shows revenue growth.")