burakaktna/listybox-etsy-listing-json

Etsy Product Listing Dataset (JSON SFT Format)

🎯 Optimized for Fine-tuning

Clean, minimal dataset for training vision-language models to generate Etsy product listings.

Key Features

• JSON SFT Format: Ready for training with standard SFT trainers
• Minimal Instructions: ~200 chars average (vs 2000+ in verbose versions)
• Token Efficient: 90% reduction in instruction tokens
• Production Ready: Model learns the task, not prompt engineering

Dataset Structure

{
  "messages": [
    {
      "role": "system",
      "content": [{"type": "text", "text": "You are an Etsy listing expert..."}]
    },
    {
      "role": "user",
      "content": [
        {"type": "text", "text": "Create Etsy listing for [product]"}, 
        {"type": "image", "image_url": {"url": "https://..."}} 
      ]
    },
    {
      "role": "assistant",
      "content": [{"type": "text", "text": "{"title": "...", "description": "...", "tags": [...]}"}]
    }
  ]
}

Files

• sft.jsonl: Training data (536 examples)
• eval.jsonl: Evaluation data (3283 examples)

Usage

import json

# Load training data
with open('sft.jsonl', 'r') as f:
    data = [json.loads(line) for line in f]

# Each example has minimal instruction
for example in data:
    messages = example['messages']
    # Process for your training framework

Benefits

✅ Before (Verbose):

⚠️ CRITICAL WARNING: 21 CHARACTER TAGS ARE INVALID! 
MAXIMUM IS 20! ⚠️ ABSOLUTE CHARACTER RULES...
[2000+ characters of instructions]

✅ After (Clean):

Create Etsy listing for Infant Bodysuit: A charming design...

Training Tips

1. Model learns actual task, not prompt tricks
2. Use with Qwen2.5-VL or similar vision models
3. Minimal prompts = faster convergence
4. Production: Just send image + "Create Etsy listing"

License

Apache 2.0