ARIA Qwen2.5-0.5B v1
A LoRA fine-tuned version of Qwen2.5-0.5B, optimized for ARIA (Adaptive Reasoning Intelligence Architecture).
Model Details
- Base Model: Qwen2.5-0.5B
- Fine-tuning Method: LoRA (rank 8, alpha 16)
- Training Data: 2,433 examples from ARIA's training dataset
- Training Loops: 10 completed
- Final Loss: 0.8087
- Hardware: AMD RX 6700 XT, Ryzen 7 5800X3D
Training Details
- Framework: Transformers + PEFT + Unsloth
- Quantization: 4-bit (bitsandbytes)
- Batch Size: 8
- Max Length: 256 tokens
- Learning Rate: Cosine annealing
- Epochs: 15 per loop
Usage
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("YOUR_USERNAME/aria-qwen-0.5b-v1")
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-0.5B")
inputs = tokenizer("What is ARIA?", return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=100)
print(tokenizer.decode(outputs[0]))
License
Apache 2.0
About ARIA
ARIA is a personal AI assistant project focusing on chaotic but aligned reasoning, streaming responses, and PC control integration.
- Downloads last month
- 15
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support