coliseum034/coliseum-attacker-wild

This model is a fine-tuned version of unsloth/qwen2.5-0.5b-instruct-unsloth-bnb-4bit. It was trained up to 2x faster utilizing Unsloth and Hugging Face's TRL library.

This model is structurally geared toward advanced security operations, multi-agent system simulations, and red-teaming applications in the wild.

⚙️ Model Details

• License: Apache 2.0
• Base Model: unsloth/qwen2.5-0.5b-instruct-unsloth-bnb-4bit
• Architecture: Qwen2 (0.5B parameters)
• Language: English
• Quantization: 4-bit (bitsandbytes)

📊 Training & Evaluation Metrics

The model was trained over 3 epochs for a total of 921 global steps. The training procedure demonstrated consistent learning, achieving a final validation perplexity of ~5.168.

Per-Epoch Results

Epoch	Training Loss	Validation Loss	Perplexity (PPL)
1.0	1.6638	1.6605	5.262
2.0	1.5345	1.6314	5.111
3.0	1.4212	1.6425	5.168

Final Held-Out Metrics

• Final Training Loss: 1.4212
• Final Evaluation Loss: 1.6425
• Final Perplexity: 5.168

Training Hyperparameters & Performance

• Global Steps: 921
• Total Training Runtime: ~36 minutes, 48 seconds (2207.98 seconds)
• Training Samples per Second: 6.658
• Training Steps per Second: 0.417
• Total FLOPs: 8.527 x 10^15

💻 Framework Versions

• PEFT
• Transformers
• Unsloth
• TRL
• Safetensors
• PyTorch

🚀 Usage

This model uses the standard transformers library pipeline or text-generation-inference.

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "coliseum034/coliseum-attacker-wild"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id)

prompt = "Analyze this sequence for potential exploitation vectors:"
inputs = tokenizer(prompt, return_tensors="pt")

outputs = model.generate(**inputs, max_new_tokens=100)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))