BrahmaNet

BrahmaNet: Phi-3 SFT for Invoice Extraction

<div align="center">

</div>

Model Description

BrahmaNet is a specialized language model fine-tuned from Microsoft's Phi-3-mini-4k-instruct for extracting structured information from invoice documents. The model is optimized to understand invoice formats and convert unstructured text into well-structured JSON output.

• Developed by: Gokul Alex
• Model type: Causal Language Model
• Language(s): English
• License: MIT
• Finetuned from model: microsoft/Phi-3-mini-4k-instruct

Uses

Direct Use

This model is designed for extracting structured information from invoice documents including:

• Invoice numbers and dates
• Supplier/vendor information
• Total amounts and line items
• Customer details
• Payment terms

Downstream Use

The model can be fine-tuned further for:

• Receipt processing
• Purchase order extraction
• Financial document analysis
• Custom structured data extraction tasks

Out-of-Scope Use

• General purpose chat or conversation
• Mathematical reasoning beyond basic arithmetic
• Legal document analysis
• Medical or sensitive personal information extraction

How to Get Started with the Model

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

# Load model and tokenizer
model_name = "gokulalex/BrahmaNet"
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    device_map="auto",
    torch_dtype=torch.bfloat16,
    trust_remote_code=True
)

# Prepare prompt
prompt = """Extract invoice information as JSON:

Document: Invoice Number: INV-2023-001, Date: 2023-10-15, Supplier: ABC Corporation, Total Amount: $1,250.00

JSON:"""

# Generate response
inputs = tokenizer(prompt, return_tensors="pt", max_length=512, truncation=True)
outputs = model.generate(
    inputs.input_ids,
    max_new_tokens=150,
    do_sample=True,
    temperature=0.3,
    top_p=0.9,
    pad_token_id=tokenizer.eos_token_id
)

response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response)