Ollama Settings

Ollama Settings Interface

This section covers the Ollama-related configuration options in WordOllama to enhance your model usage experience.

EndPoint Settings

• Default Address: http://localhost:11434
• Custom Scenarios:
  • After modifying environment variables: Use the newly specified address.
  • LAN Mode: Enter a local network address (e.g., http://192.168.1.100:11434).
  • Remote Server: Enter the full server address.

Model Management

Select Model

• Choose an installed model from the dropdown menu.
• Click the “Refresh” button to update the list of available models.
• View the status of the currently selected model in real-time.

Download New Model

1. Enter the model name:
   • Short form: qwen2.5
   • Full form: qwen2.5:latest
2. Click the “Download” button to start the process.
3. Monitor real-time download progress at the bottom of the interface.

Delete Model

1. Select the model to delete from the list (click any cell).
2. Click the “Delete Selected Model” button.
3. Confirm the deletion.
4. Wait for the deletion to complete.

Performance Parameters

Tokens Limit

• Purpose: Controls the maximum length of model output.
• Impact: Higher limits require more memory or GPU VRAM.
• Recommendation: Adjust based on your hardware capabilities.

Tokens and Resource Usage Reference Table

7B Model Resource Usage Estimate

Tokens Limit	CPU Memory Usage	GPU VRAM Usage	Use Case
512	~4GB	~3GB	Short text processing, simple Q&A
1024	~6GB	~4GB	General document processing, translation
2048	~8GB	~5GB	Long document processing, in-depth analysis
4096	~12GB	~7GB	Complex document processing, large translations
8192	~16GB	~9GB	Ultra-long documents, research analysis

13B Model Resource Usage Estimate

Tokens Limit	CPU Memory Usage	GPU VRAM Usage	Use Case
512	~8GB	~6GB	Short text processing, simple Q&A
1024	~12GB	~8GB	General document processing, translation
2048	~16GB	~10GB	Long document processing, in-depth analysis
4096	~24GB	~14GB	Complex document processing, large translations
8192	~32GB	~18GB	Ultra-long documents, research analysis

Tip

• Actual resource usage may vary depending on model version, optimization, or runtime environment.
• In CPU mode, reserve at least 50% extra memory for system stability.
• In GPU mode, reserve at least 2GB of VRAM for system overhead.
• If memory or VRAM is insufficient, consider lowering the Tokens limit.

Temperature Settings

• Range: 0.0 - 1.0
• Purpose: Controls output randomness.
  • Lower values (near 0): More deterministic and conservative output.
  • Higher values (near 1): More random and creative output.
• Recommendation: Adjust based on your use case.