Text Generation
Transformers
English
phi3
finance
entity-extraction
ner
phi-3
production
indian-banking
custom_code
4-bit precision
Instructions to use Ranjit0034/finance-entity-extractor with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Ranjit0034/finance-entity-extractor with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="Ranjit0034/finance-entity-extractor", trust_remote_code=True)# pip install -U transformers accelerate # Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("Ranjit0034/finance-entity-extractor", trust_remote_code=True) model = AutoModelForCausalLM.from_pretrained("Ranjit0034/finance-entity-extractor", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use Ranjit0034/finance-entity-extractor with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Ranjit0034/finance-entity-extractor" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Ranjit0034/finance-entity-extractor", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/Ranjit0034/finance-entity-extractor
- SGLang
How to use Ranjit0034/finance-entity-extractor with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "Ranjit0034/finance-entity-extractor" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Ranjit0034/finance-entity-extractor", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "Ranjit0034/finance-entity-extractor" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Ranjit0034/finance-entity-extractor", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use Ranjit0034/finance-entity-extractor with Docker Model Runner:
docker model run hf.co/Ranjit0034/finance-entity-extractor
|
Download CHANGELOG.md from Ranjit0034/finance-entity-extractor: direct link, hf CLI and curl.
- Browser
- Download file 3.4 kB
-
https://huggingface.co/Ranjit0034/finance-entity-extractor/resolve/main/CHANGELOG.md
- Command line
-
hf download hf://Ranjit0034/finance-entity-extractor/CHANGELOG.md
-
curl -L -o CHANGELOG.md https://huggingface.co/Ranjit0034/finance-entity-extractor/resolve/main/CHANGELOG.md
3.4 kB
Changelog
All notable changes to FinEE will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
[Unreleased]
Added
- (Next features go here)
[1.1.0] - 2026-01-12
Added
Complete Data Pipeline (
scripts/data_pipeline/)step1_unify.py: Unifies MBOX, JSON, CSV, XML sourcesstep2_filter.py: Removes OTPs, spam, marketing messagesstep3_baseline.py: Tests regex extractor accuracystep4_label.py: Creates labeled training data with ground truth
Synthetic Data Generator
generate_synthetic.py: Production-grade grammar-based generator- 100K+ realistic Indian bank transactions
- All major banks (HDFC, ICICI, SBI, Axis, Kotak, PNB, BOB, etc.)
- Brokerages (Zerodha, Groww, Upstox, Angel One, 5Paisa, etc.)
- E-commerce, food, travel, utilities, entertainment categories
generate_advanced.py: Advanced features- Markov Chain for realistic message flow
- Real data calibration from actual samples
- Multilingual support (Hindi, Tamil, Telugu, Bengali, Kannada)
- Data augmentation and edge case oversampling
LLM Fine-tuning Pipeline (
scripts/finetune.py)- Supports MLX (Apple Silicon) and PyTorch backends
- LoRA fine-tuning with automatic data preparation
- Model fusion and evaluation utilities
Performance
- Trained on 152,519 records (2,419 real + 100K synthetic + 50K multilingual)
- Val loss: 2.42 → 0.46 (81% reduction)
- 100% JSON parsing accuracy on test cases
- Multilingual extraction working (Hindi, Tamil, Telugu, Bengali, Kannada)
- Fine-tuned model: 7.6GB (Phi-3-mini + LoRA fused)
Models
- Fine-tuned model:
finetuned-v1/on Hugging Face - LoRA adapters:
lora-adapters/on Hugging Face
[1.0.3] - 2025-01-11
Added
- Lakhs notation support (e.g., "1.5 Lakh" → 150000)
benchmark.pyscript for accuracy verification- Torture test suite for edge cases
CONTRIBUTING.mdwith Git Flow guidelines- Professional branching: main, develop, feature/*
Changed
- Moved Jupyter notebooks to
experiments/folder - Default
use_llm=Falsefor instant usage (no model download) - Updated README with edge case examples
Fixed
- Double-escaped regex patterns in amount extraction
[1.0.2] - 2025-01-11
Changed
- Default to regex-only mode (
use_llm=False) - Package works instantly without downloading 5GB model
Fixed
- Package build configuration (hatch sources)
[1.0.1] - 2025-01-11
Fixed
- Package did not include source files (only 5KB)
- Fixed
pyproject.tomlbuild configuration
[1.0.0] - 2025-01-11
Added
- Initial PyPI release
- 5-tier additive extraction pipeline (Cache/Regex/Rules/LLM/Validate)
- Multi-backend support (MLX, PyTorch, GGUF)
- CLI with
fineecommand - 88 unit tests
- Colab demo notebook
- JSON schema contract documentation
- Support for 5 banks: HDFC, ICICI, SBI, Axis, Kotak
Performance
- 94.5% field accuracy on multi-bank benchmark
- <1ms latency in regex-only mode
- 50,000+ messages/second throughput