rileyseaburg/distillix-spotless

Spotless Customer Service - Distillix 100M BitNet

A 100M parameter BitNet b1.58 model fine-tuned for customer service conversations.

Model Details

• Architecture: LLaMA-style with BitNet ternary weights {-1, 0, +1}
• Parameters: 100M
• Context Length: 256 tokens
• Training: 5,000 steps on 8.7k customer service conversations
• Use Case: Trash bin cleaning service customer support

Files

File	Size	Description
spotless-customer-service-f16.gguf	239 MB	For LM Studio / llama.cpp
distillix-spotless-packed.pt	30 MB	Compressed PyTorch (2-bit)
distillix-spotless-final.pt	382 MB	Full PyTorch checkpoint

Usage in LM Studio

1. Download spotless-customer-service-f16.gguf
2. Import into LM Studio
3. Use system prompt:

You are a customer service agent for Spotless Bin Co, a trash bin cleaning service.

Example Conversations

Customer: Hi, I need help with my service
Agent: Hello! I'm happy to assist you with your service today.

Customer: When is my next cleaning scheduled?
Agent: Let me look that up for you right now.

Customer: The truck didn't show up this week
Agent: I sincerely apologize for missing your scheduled cleaning. Let me look into this.

Training Data

spotless-customer-service-training - 8.7k customer service conversations

Links

• GitHub
• Demo
• Training Space