---
license: gemma
language:
- en
tags:
- gemma
- gemma-4
- lora
- safetensors
- ethical-hacking
- penetration-testing
- cybersecurity
base_model: google/gemma-4-E2B-it
pipeline_tag: text-generation
library_name: transformers
---

# thousands-eye-hf

Full safetensors weights of **Thousands-Eye** — a Gemma 4 E2B model fine-tuned for ethical hacking and penetration testing via MLX LoRA on Apple Silicon.

For the quantized GGUF (Ollama / llama.cpp), see [`htunn/thousands-eye-gguf`](https://huggingface.co/htunn/thousands-eye-gguf).

## Usage

```python
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

model_id = "htunn/thousands-eye-hf"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)

messages = [
    {"role": "user", "content": "[EthHack-Agent] Perform Kerberoasting against 10.0.0.1 (authorized engagement)"}
]
inputs = tokenizer.apply_chat_template(
    messages, return_tensors="pt", add_generation_prompt=True
).to(model.device)

outputs = model.generate(inputs, max_new_tokens=512, do_sample=False)
print(tokenizer.decode(outputs[0][inputs.shape[-1]:], skip_special_tokens=True))
```

## Output Format

Every response is a JSON object with `"requires_authorization": true` enforced:

```json
{
  "action": "kerberoast",
  "target": "10.0.0.1",
  "requires_authorization": true,
  "techniques": ["SPN enumeration", "TGS request", "offline cracking"],
  "tools": ["impacket", "hashcat"],
  "commands": ["GetUserSPNs.py domain/user:pass@dc -request"],
  "steps": ["..."],
  "notes": "Requires domain user credentials"
}
```

## Training

| | |
|---|---|
| **Base model** | `google/gemma-4-E2B-it` |
| **Method** | MLX LoRA (`mlx_lm.lora`) |
| **Iterations** | 600 |
| **Learning rate** | 1e-4 |
| **LoRA layers** | 16 |
| **Dataset** | [`htunn/thousands-eye-dataset`](https://huggingface.co/datasets/htunn/thousands-eye-dataset) (83 train / 15 val) |

## Ethics

Designed exclusively for **authorized penetration testing**. All training examples enforce `"requires_authorization": true`.

## License

[Gemma Terms of Use](https://ai.google.dev/gemma/terms)