U4RASD/dalla-model-training

Dalla (دلّة) - Arabic Language Models

Dalla is a family of Arabic language models optimized for Arabic text processing through advanced tokenization techniques.

Available Models

Instruction-Tuned Variants

• Dalla-Gemma-IT - Instruction-tuned variant based on Gemma architecture
• Dalla-Llama-IT - Instruction-tuned variant based on Llama architecture

Usage

from transformers import AutoTokenizer, AutoModelForCausalLM

# Load model and tokenizer
model_name = "U4RASD/dalla-gemma-it"  # or "U4RASD/dalla-llama-it"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)