🤏 smolified-zion

Intelligence, Distilled.

This is a Domain Specific Language Model (DSLM) generated by the Smolify Foundry.

It has been synthetically distilled from SOTA reasoning engines into a high-efficiency architecture, optimized for deployment on edge hardware (CPU/NPU) or low-VRAM environments.

📦 Asset Details

• Origin: Smolify Foundry (Job ID: 26c00716)
• Architecture: DSLM-Micro (270M Parameter Class)
• Training Method: Proprietary Neural Distillation
• Optimization: 4-bit Quantized / FP16 Mixed
• Dataset: Link to Dataset

🚀 Usage (Inference)

This model is compatible with standard inference backends like vLLM.

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "smolify/smolified-zion"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto")

messages = [
    {"role": "system", "content": '''You are a cybersecurity expert specializing in API defense and real-time incident response.'''},
    {"role": "user", "content": '''Input: {"threat": "bot_scraping", "risk": 0.65, "endpoint": "/api/prices", "ip": "104.28.32.1"}'''}
]
text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True).removeprefix('<bos>')

from transformers import TextStreamer
_ = model.generate(**tokenizer(text, return_tensors="pt").to("cuda"), max_new_tokens=1000, streamer=TextStreamer(tokenizer, skip_prompt=True))

⚖️ License & Ownership

This model weights are a sovereign asset owned by smolify. Generated via Smolify.ai.