Qwen3-8B-GAE

A fine-tuned version of Qwen3-8B optimized for Arabic language tasks.

Model Details

• Model type: Causal Language Model
• Language(s): Arabic, English
• Base model: Qwen/Qwen3-8B

Uses

Direct Use

General Arabic language generation and understanding tasks.

Out-of-Scope Use

Not intended for generating harmful, misleading, or factually incorrect content.

How to Get Started

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "Ocelotr/Qwen3-8B-GAE"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

messages = [{"role": "user", "content": "مرحبا، كيف حالك؟"}]
inputs = tokenizer.apply_chat_template(messages, return_tensors="pt")
outputs = model.generate(inputs, max_new_tokens=512)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Training Details

Training Procedure

• Training regime: bf16 mixed precision
• Fine-tuning method: PPO with GAE (Generalized Advantage Estimation)