Pull nomic-embed-text for Assistent memory, write ollama-roles.json, CPU Modelfile, and raise MAX_LOADED_MODELS/NUM_PARALLEL to 2.
Co-authored-by: Cursor <cursoragent@cursor.com>
A 1-token /api/chat after tags (and again if /api/ps is empty) loads VL weights before the first message. Mid KEEP_ALIVE is 15m so a short image-gen burst does not unload the model.
Co-authored-by: Cursor <cursoragent@cursor.com>
- Enhanced the `OllamaTune` class to include a new `context_length` attribute, improving the configuration for different GPU tiers.
- Updated performance tuning logic to set appropriate context lengths for low, mid, high, and ultra tiers, ensuring optimal resource allocation.
- Modified installation scripts to reflect the new context length settings, enhancing the installation process for Ollama.
- Revised documentation to include context length details in the GPU performance table, providing clearer guidance for users.
- Added tests to validate the correct context length settings in various scenarios, ensuring robustness in performance tuning.
- Introduced GPU probing functionality to gather and store GPU specifications in `/mnt/swarm_data/.gpu-rent-gpu.json`, aiding in performance tuning.
- Updated `install_ollama.sh` and `install_llamacpp.sh` to utilize GPU information for configuring optimal runtime parameters.
- Enhanced `provision.py` to include GPU probing and performance tuning logic, ensuring better resource allocation for LLM operations.
- Improved documentation in `decisions.md`, `llm.md`, and `swarmui.md` to reflect changes in GPU handling and performance tuning processes.
- Added new tests to validate the GPU probing and model resolution logic, ensuring robustness in handling various GPU configurations.
Leonid PershinandCursor