---

license: mit language: - gu metrics: - glue pipeline_tag: text-generation library_name: transformers tags: - gujurati - indic-model - abirmaheshwari - A.B.I.R-ecosystem - Indian-lang - indiaai

ABIRGUJv1 — Gujarati-First Transformer Language Model

ABIRGUJv1 is a custom transformer-based causal language model designed specifically for Gujarati and English text generation. It is trained from scratch using a custom tokenizer and architecture optimized for efficient Gujarati language understanding, translation, and generation.

This model focuses on providing lightweight, efficient Gujarati NLP capabilities while maintaining strong contextual understanding.

Model Details

Model Description

ABIRGUJv1 is a decoder-only transformer model trained entirely from scratch without using pretrained weights. It uses a custom tokenizer trained on Gujarati-focused data and is optimized for efficient inference and fine-tuning.

• Developed by: Abir Maheshwari
• Funded by: Independent Research
• Shared by: Abir Maheshwari
• Model type: Causal Language Model (Decoder-only Transformer)
• Language(s): Gujarati, English
• License: MIT
• Finetuned from model: None (trained from scratch)

Model Sources

• Repository: https://huggingface.co/abirmaheshwari/abirguJv1
• Architecture: Custom Transformer
• Framework: PyTorch + HuggingFace Transformers

Uses

Direct Use

ABIRGUJv1 is suitable for:

• Gujarati text generation
• English-to-Gujarati translation
• Gujarati chatbots
• Text completion
• NLP research
• Educational purposes

Example applications:

• Gujarati AI assistants
• Content generation
• Research on small language models

Downstream Use

This model can be fine-tuned for:

• Gujarati instruction models
• Question answering
• Domain-specific NLP tasks
• Gujarati conversational agents

Out-of-Scope Use

Not recommended for:

• Critical decision systems
• Medical diagnosis
• Legal advice
• Safety-critical applications This is an early version model.

Bias, Risks, and Limitations

ABIRGUJv1 may:

• Produce incorrect information
• Reflect biases present in training data
• Generate incomplete or nonsensical outputs This is expected for models trained on limited datasets.

Recommendations

Use this model:

• For research
• For experimentation
• For fine-tuning Not recommended for production without further training.

How to Get Started

from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("abirmaheshwari/abirguJv1", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("abirmaheshwari/abirguJv1", trust_remote_code=True)

input_text = "મારું નામ"

inputs = tokenizer(input_text, return_tensors="pt")

outputs = model.generate(
    **inputs,
    max_length=100,
    temperature=0.7
)
print(tokenizer.decode(outputs[0]))