Adding Custom Models to Ollama

---

---

AI Summary

Summary: Creating and Sharing an Emoji-Generating AI Model

• Goal: Convert and quantize an AI model to respond in emojis.
• Finding the Model:
  • Use Hugging Face to search for emoji text generation models.
  • Select “Open Hermes Emoji Tron 001” as the suitable model.
• Downloading the Model:
  • Use Hugging Face Downloader tool for efficient downloading.
  • Copy the model repository link and use the downloader command to fetch files.
• Conversion and Quantization:
  • Ensure Docker is installed and running.
  • Use Docker to pull the “ol/quantise” image.
  • Run Docker commands to convert and quantize the model, typically using q40 for a balance of performance and quality.
• Configuring the Model:
  • Identify the correct quantization parameter.
  • Download the converted and quantized model files.
  • Determine the system prompt and template from the Hugging Face documentation or by investigating related models.
• Testing the Model:
  • Use a code editor to set up the model configuration.
  • Run the model with test phrases to generate emoji responses.
  • Adjust parameters as needed to refine output.
• Sharing the Model:
  • Create an account on Alama AI for a namespace.
  • Rename the model to include your namespace.
  • Use the “ol push” command to upload the model.
  • Edit the model’s descriptions and share it with the community.
• Note: The process may involve troubleshooting and manual adjustments.

For more detailed instructions or troubleshooting, refer to the original AI model repository and documentation.