Upload AgentRank model

README.md

@@ -9,55 +9,186 @@ tags:
 - memory
 - retrieval
 - rag
+- transformers
 library_name: transformers
 pipeline_tag: text-classification
 ---
 
-# AgentRank Reranker
+# 🧠 AgentRank Reranker
 
-Cross-encoder reranker for AI agent memory retrieval.
+**Cross-encoder reranker for AI agent memory retrieval — the second stage of a two-stage retrieval pipeline.**
 
-## Overview
+[![Model](https://img.shields.io/badge/🤗_HuggingFace-Model-yellow)](https://huggingface.co/vrushket/agentrank-reranker)
+[![GitHub](https://img.shields.io/badge/GitHub-Repository-black)](https://github.com/vmore2/AgentRank-base)
+[![License](https://img.shields.io/badge/License-Apache_2.0-blue.svg)](https://opensource.org/licenses/Apache-2.0)
 
-Use in a two-stage pipeline:
-1. AgentRank embedder retrieves top-50 (fast)
-2. This reranker scores and reorders to top-10 (accurate)
+---
+
+## 🎯 What is This?
+
+AgentRank Reranker is a **cross-encoder** that scores query-memory pairs for relevance. Use it after fast vector retrieval to **rerank** candidates for higher accuracy.
+
+### Two-Stage Pipeline
+
+| Stage | Model | Speed | Accuracy |
+|-------|-------|-------|----------|
+| **1. Retrieve** | [agentrank-base](https://huggingface.co/vrushket/agentrank-base) | ⚡ Fast | Good |
+| **2. Rerank** | **This model** | 🐢 Slower | **Best** |
+
+---
 
-## Model Details
+## 📊 Performance
 
-| Property | Value |
-|----------|-------|
-| Base Model | ModernBERT-base |
-| Parameters | ~149M |
-| Validation Accuracy | 89.11% |
+| Metric | Value |
+|--------|-------|
+| **Validation Accuracy** | 89.11% |
+| **Best Val Loss** | 0.2554 |
+| **Base Model** | ModernBERT-base |
+| **Parameters** | ~149M |
+| **Training Samples** | 500K |
 
-## Quick Start
+---
+
+## 🚀 Quick Start
+
+### Installation
+
+`ash
+pip install transformers torch
+`
+
+### Score a Query-Memory Pair
 
 `python
 from transformers import AutoModelForSequenceClassification, AutoTokenizer
 import torch
 
+# Load model
 model = AutoModelForSequenceClassification.from_pretrained("vrushket/agentrank-reranker")
 tokenizer = AutoTokenizer.from_pretrained("vrushket/agentrank-reranker")
+model.eval()
 
+# Score relevance
 query = "What did we discuss about Python?"
-memory = "User prefers Python for backend"
+memory = "User mentioned they prefer Python for backend development"
 
-inputs = tokenizer(query, memory, return_tensors="pt", truncation=True)
+inputs = tokenizer(query, memory, return_tensors="pt", truncation=True, max_length=512)
 with torch.no_grad():
     score = torch.sigmoid(model(**inputs).logits).item()
-print(f"Score: {score:.4f}")
+
+print(f"Relevance: {score:.2%}")  # e.g., "Relevance: 94.23%"
 `
 
-## Links
+### Rerank Top-K Candidates
 
-- GitHub: github.com/vmore2/AgentRank-base
-- PyPI: pip install agentrank
+`python
+def rerank(query: str, candidates: list[str], top_k: int = 10) -> list[tuple[float, str]]:
+    scores = []
+    for memory in candidates:
+        inputs = tokenizer(query, memory, return_tensors="pt", truncation=True)
+        with torch.no_grad():
+            score = torch.sigmoid(model(**inputs).logits).item()
+        scores.append((score, memory))
+    
+    # Sort by score descending
+    return sorted(scores, reverse=True)[:top_k]
+
+# Example
+top_50_from_vector_db = [...]  # Retrieved via agentrank-base
+top_10 = rerank("What's my Python preference?", top_50_from_vector_db)
+`
+
+---
+
+## 🔧 Full Two-Stage Pipeline
+
+`python
+from agentrank import AgentRankEmbedder
+from transformers import AutoModelForSequenceClassification, AutoTokenizer
+import torch
 
-## Author
+# Stage 1: Fast retrieval with embedder
+embedder = AgentRankEmbedder.from_pretrained("vrushket/agentrank-base")
+query_embedding = embedder.encode(["What's my Python preference?"])
+# ... search vector DB for top-50 candidates ...
+
+# Stage 2: Accurate reranking
+reranker = AutoModelForSequenceClassification.from_pretrained("vrushket/agentrank-reranker")
+tokenizer = AutoTokenizer.from_pretrained("vrushket/agentrank-reranker")
+
+scored = []
+for memory in top_50_candidates:
+    inputs = tokenizer(query, memory, return_tensors="pt", truncation=True)
+    with torch.no_grad():
+        score = torch.sigmoid(reranker(**inputs).logits).item()
+    scored.append((score, memory))
+
+# Return top-10
+final_results = sorted(scored, reverse=True)[:10]
+`
 
-Vrushket More (vrushket2604@gmail.com)
+---
+
+## 📁 Related Models
+
+| Model | Type | HuggingFace | Use Case |
+|-------|------|-------------|----------|
+| **AgentRank-Base** | Embedder | [vrushket/agentrank-base](https://huggingface.co/vrushket/agentrank-base) | Fast retrieval (768-dim) |
+| **AgentRank-Small** | Embedder | [vrushket/agentrank-small](https://huggingface.co/vrushket/agentrank-small) | Lightweight (384-dim) |
+| **AgentRank-Reranker** | Cross-encoder | This model | Accurate reranking |
+
+---
+
+## 🏋️ Training Details
+
+- **Base Model**: ModernBERT-base
+- **Epochs**: 3
+- **Batch Size**: 16
+- **Learning Rate**: 2e-5
+- **Optimizer**: AdamW
+- **Loss**: Binary Cross-Entropy
+- **Hardware**: 2x RTX 6000 Ada (48GB)
+
+---
+
+## 📦 Installation via PyPI
+
+`ash
+pip install agentrank
+`
+
+---
+
+## 🔗 Links
+
+- **GitHub**: [github.com/vmore2/AgentRank-base](https://github.com/vmore2/AgentRank-base)
+- **PyPI**: [pypi.org/project/agentrank](https://pypi.org/project/agentrank)
+- **CogniHive** (Multi-agent memory): [pypi.org/project/cognihive](https://pypi.org/project/cognihive)
+
+---
+
+## 📧 Contact
+
+- **Author**: Vrushket More
+- **Email**: vrushket2604@gmail.com
+- **LinkedIn**: [linkedin.com/in/vrushket-more](https://linkedin.com/in/vrushket-more)
+
+---
+
+## 📄 Citation
+
+`ibtex
+@misc{agentrank2024,
+  author = {More, Vrushket},
+  title = {AgentRank: Temporal-Aware Embeddings for AI Agent Memory},
+  year = {2024},
+  publisher = {HuggingFace},
+  url = {https://huggingface.co/vrushket/agentrank-reranker}
+}
+`
+
+---
 
-## License
+## 📜 License
 
 Apache 2.0