Financial Metrics: NOI, Cap Rate, DSCR, IRR, Equity Multiple, GAV, NAV, FFO, AFFO, Debt Yield
Fund Structures: GP/LP, Preferred Return, Waterfall distributions, Carried Interest, Catch-up
Debt Structures: Senior Loan, Mezzanine, Preferred Equity, A/B Notes, Cash Sweep, LTV
Investment Strategies: Core, Core+, Value-Add, Opportunistic, Distressed
Lease Terms: DTM, NTV, WALE, TI, CAM, Escalation, Base Year, Stop
Asset Classes: Office, Industrial, Retail, Multifamily, SFR, BTR, Hospitality, DataCenter

Training Data

Trained on 658 Q&A pairs covering:

CRE terminology definitions
Financial calculations and formulas
Investment analysis concepts
Debt and equity structures
Transaction types and strategies

Usage

Ollama (Recommended)

bash

# Run directly from HuggingFace (no download needed)
ollama run hf.co/pramod-thaz/finance-cre-finetuned

# Or pull locally first
ollama pull pramod-thaz/finance-cre-finetuned
ollama run finance-cre-finetuned

llama.cpp

bash

# Download GGUF and run with llama.cpp
./llama-cli -m finance-cre-finetuned-v0.1-Q4_K_M.gguf -n 256 --temp 0.3

Python (transformers)

python

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "pramod-thaz/finance-cre-finetuned"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

Version History

Version	Date	Notes
v0.1	April 2026	Initial release - 658 training pairs

Limitations

Knowledge limited to training data (658 Q&A pairs)
May not have real-time market data
Designed for Q&A, not complex financial calculations
Some specialized terms may require additional training

Disclaimer

This model is fine-tuned for educational and informational purposes. For actual investment decisions, consult with qualified professionals.

Citation

Fine-tuned Phi-3.5-mini for Commercial Real Estate
https://huggingface.co/pramod-thaz/finance-cre-finetuned

License

This model is based on Phi-3.5-mini which uses the MIT License. Check individual model cards for specific licensing information.

Note: This is v0.1 - a foundational model. Future versions will include more training data and improved accuracy on CRE-specific terminology.