VVGONLINE/Vikas-AI

This README.md is designed to be the primary model card for your Hugging Face repository. It uses the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) license, which strictly prohibits commercial use of the model.

---
license: cc-by-nc-4.0
language:
- en
tags:
- digital business consulting
- custom-chat
- non-commercial
model_name: Vikas-AI
pipeline_tag: text-generation
---

# Vikas-AI

Vikas-AI is a lightweight, high-performance tiny language model. It has been trained on custom conversational data on a local NVIDIA GeForce RTX 2070 (8GB VRAM) setup.

This model is optimized for efficient inference on consumer hardware, mobile devices, and browser-based environments.

## ⚠️ License and Usage

This model is released under the **[CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/)** license. 

- **Non-Commercial Use Only**: You may not use this model or its derivatives for any commercial purposes or financial gain.
- **Attribution**: You must give appropriate credit if you redistribute or adapt the model.
- **Usage**: This model is made by VVG ONLINE as POC.

## Vikas AI Model Details

- **Developed by:** VVG ONLINE under supervision of our CEO,CTO, Founder Vaibhav V. Gupta
- **Training Hardware:** NVIDIA GeForce RTX 2070 with Max-Q Design (8GB VRAM)
- **Training Method:** LoRA (Low-Rank Adaptation) followed by a full weight merge.
- **Input Format:** ChatML / `<|im_start|>` prompt format.

## Training Procedure

The model was trained using the following conservative parameters to ensure stability on 8GB hardware:

- **Learning Rate:** 2e-5
- **Batch Size:** 1 (with Gradient Accumulation 16)
- **Precision:** FP16
- **Max Sequence Length:** 256 tokens
- **Epochs:** 3

### Prompt Template
```text
user
{prompt}
assistant

How to Use

Using Transformers (Python)

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_id = "VVGONLINE/Vikas-AI"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id, 
    torch_dtype=torch.float16, 
    device_map="auto"
)

prompt = "user\nWhat can you do?\nassistant\n"
inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
outputs = model.generate(**inputs, max_new_tokens=100)
print(tokenizer.decode(outputs, skip_special_tokens=True))

Using Transformers.js (Web/Browser)

This model is compatible with transformers.js for local browser inference.

import { pipeline } from '@xenova/transformers';

const generator = await pipeline('text-generation', 'VVGONLINE/Vikas-AI');
const output = await generator('user\nHi!\nassistant\n', {
    max_new_tokens: 256,
    temperature: 0.7
});

Limitations and Bias

• Size: As it is a tiny model, it may struggle with complex logic or highly technical domains compared to larger models.

Feedback and Contributions

This is an experimental project. If you find issues or have suggestions for further fine-tuning, feel free to open a discussion on the community tab!