Gnostic-Ai/Llama-3-1-8B-Instruct-GST-GnosticAI-Q4

Model Card for Model ID

🚀 Gnostic AI Legal Announces Revolutionary LLM Model for GST Compliance! 🚀

Model Details

Model Description

We at Gnostic AI Legal are thrilled to introduce our cutting-edge GST-focused Large Language Model (LLM)— an AI-driven solution designed to streamline compliance, enhance accuracy, and reduce manual workload for taxation professionals and businesses. With advanced legal reasoning and real-time updates on GST regulations, this model empowers firms to navigate tax complexities effortlessly.

• Developed by: Suvir Misra, ex-Principal Commissioner, CBIC, India.
• Funded by [optional]: Self Funded
• Shared by [optional]:
• Model type: Transformer-based LLM optimized for taxation and compliance
• Language(s) (NLP):
• License: Apache
• Finetuned from model [optional]: Llama-3.1-B-Instruct

Model Sources [optional]

Proprietory Datasets based upon Domain Name Knowledge on GST and Indian Taxation

• Repository: Proprietory
• Paper [optional]: [To be released]
• Demo [optional]:

Uses

Direct Use

The model is designed for: Automated GST Computation & Filing, Legal Compliance Analysis, AI-driven Legal Interpretation, Automated Tax Audit Assistance.

Downstream Use [optional]

Custom Fine-tuning for Specialized Tax Scenarios, Integration with Accounting Software, RAG implementations

Out-of-Scope Use

Non-taxation legal advice, Non-financial AI decision-making

Bias, Risks, and Limitations

Potential bias in tax interpretations, Errors in unverified data sources, Legal compliance limitations in niche tax cases.

Recommendations

Users should validate model responses with certified tax professionals.

How to Get Started with the Model

"from gst_llm import GSTModel model = GSTModel.load('gnostic-llm/gst') response = model.generate("Calculate GST for a turnover of ₹5 crores in Maharashtra.") print(response)"

Training Details

Training Data

The model is fine-tuned on Indian Taxation Datasets, Legal Case Studies, and GST Rules & Regulations.

Training Procedure

Trained using MLM_MX libraries provided by Apple, NVIDIA A100 GPUs.

Preprocessing [optional]

Tokenization, domain adaptation.

Training Hyperparameters

• Training regime: Mixed precision (fp16), batch size optimization.

Speeds, Sizes, Times [optional]

Evaluation

Testing Data, Factors & Metrics

Testing Data

GST compliance datasets created by Suvir Misra comprising of Legal text benchmarking and Financial document validation.

Factors

[More Information Needed]

Metrics

F1 Score (Legal Interpretation), Accuracy (GST computation), Precision (Regulatory compliance)

Results

To be published

Summary

Model Examination [optional]

Transformer-based deep learning model Optimized for taxation-related NLP tasks

Environmental Impact

Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).

• Hardware Type: Apple Silicon M4 126 GB, NVIDIA A100 GPU.
• Hours used: 18 hours
• Cloud Provider: Azure
• Compute Region: India
• Carbon Emitted: Not computed

Technical Specifications [optional]

Model Architecture and Objective

Compute Infrastructure

Hardware

Apple Silicon M4 126 GB, NVIDIA A100 GPU-s.

Software

PyTorch, Hugging Face Transformers, MLX-LM, Llama.cpp.

Citation [optional]

BibTeX:

[More Information Needed]

APA:

[More Information Needed]

Glossary [optional]

[More Information Needed]

More Information [optional]

SAS-Enabled Taxation LLM recomended pricing plans- Software-as-a-Service (SAS) pricing can be made available for MSME firms: Basic Plan: ₹1,999/month – Online Real-time GST advisory , Enterprise Plan: ₹9,999/month – AI-powered taxation consultation with Human Expert Advice in loop for 2 hours. Custom Solutions: Tailored pricing available for advanced integrations.

Model Card Authors [optional]

Suvir Misra, ex Principal Commissioner, CBIC, India

Model Card Contact

[email protected]