---
license: apache-2.0
library_name: transformers
tags:
  - financial
  - transactions
  - llama
  - foundation-model
  - embeddings
pipeline_tag: feature-extraction
---

# NVIDIA Transaction Decoder Foundation Model (~29M)

Pretrained checkpoint from the [NVIDIA Build Your Own Transaction Foundation Model](https://github.com/NVIDIA-AI-Blueprints/transaction-foundation-model) developer example, as described in the [NVIDIA Technical Blog](https://developer.nvidia.com/blog/build-your-own-transaction-foundation-model-for-financial-intelligence/).

- Architecture: Llama decoder-only (~29M params, hidden 512, 8 layers, GQA)
- Vocabulary: ~6,251 domain-specific financial transaction tokens
- Trained ~3,000 steps on IBM TabFormer with NeMo AutoModel
- Load with `AutoModelForCausalLM.from_pretrained(...)` and extract last-token embeddings

This Hub mirror exists so Spaces / Transformers can load the Git LFS checkpoint published under `models/decoder-foundation-model/` in the upstream GitHub repo.