ISANG-v1.0-8B 🌐

ISANG-v1.0-8B is a multilingual large language model designed to understand and generate text in Korean, Persian, and English. Created by Hossein Mohseni, this model is tailored for conversational tasks and is optimized for answering questions, storytelling, and assisting with creative and practical tasks in these three languages.

Model Details

• Model Type: Causal Language Model
• Architecture: 8 Billion Parameters
• Languages: Korean, Persian, English
• Framework: Hugging Face Transformers
• Device Compatibility: Optimized for inference on CPU and GPU (9GB+ VRAM recommended for GPU use).

Example Use Case

This model is ideal for:

• Multilingual Chatbots
• Translation (Korean/Persian/English)
• Creative Writing
• Language Tutoring and Assistance
• General-purpose Question Answering

How to Use

To load and use ISANG-v1.0-8B, follow the example code below:

from transformers import AutoTokenizer, AutoModelForCausalLM

# Load the model and tokenizer
model_name = "hosseinhimself/ISANG-v1.0-8B"
tokenizer = AutoTokenizer.from_pretrained(model_name, use_fast=False)
model = AutoModelForCausalLM.from_pretrained(model_name)

# Prepare a prompt
prompt = "Translate the following sentence to Persian:\n\nHello, how are you?"

# Tokenize the input
inputs = tokenizer(prompt, return_tensors="pt")

# Generate a response
output = model.generate(inputs.input_ids, max_new_tokens=100)
response = tokenizer.decode(output[0], skip_special_tokens=True)

print(response)

Chatbot Example

You can integrate ISANG into a chatbot application using Gradio:

import gradio as gr

def chat(input_text):
    inputs = tokenizer(input_text, return_tensors="pt")
    output = model.generate(inputs.input_ids, max_new_tokens=100)
    response = tokenizer.decode(output[0], skip_special_tokens=True)
    return response

iface = gr.Interface(fn=chat, inputs="text", outputs="text")
iface.launch()

Model Performance

Supported Languages

Language	Use Case Examples	Notes
Korean	Chat, Translation, Summarization	Native-level fluency
Persian	Chat, Translation, Summarization	Tailored for cultural context
English	Chat, QA, Summarization	Strong general-purpose support

Limitations

• Cultural Nuances: While ISANG is trained to handle cultural contexts, it may not always capture subtle nuances perfectly.
• Language Restriction: Only supports Korean, Persian, and English. Input in other languages may not yield coherent results.

Hyperparameters

The following parameters can be customized for generation:

• Max Tokens: 1–2048 (default: 1024)
• Temperature: 0.0–2.0 (default: 0.7)
• Top-p: Default: 0.9 (not directly exposed but can be modified in code)