Bechmarking Ollama models installed on my machine
- Ollama models are downloaded from https://ollama.ai/
- The output will be in the log file in the markdown format: model_test.md
- Here are three sample queries:
| query | output |
|---|---|
| "Write a python code to make the simulation of the solid on solid (SOS) model on lattice of size 500" | model_test_1 |
| "Please give a precise workflow for git branching and merging" | model_test_2 |
| "Give some suggestions to become a better supervisor for Ph.D. students" | model_test_3 |
| "Give some suggestions for Ph.D. students in mathematics." | model_test_4 |
| "Write a haiku for stochastic heat eqaution." | model_test.md |
- The following models are tested:
- "mistral:latest"
- "orca2:latest"
- "llama2:latest"
- "llama2:13b"
- "llama2:70b"
- "llama2-uncensored:latest"
- "codellama:latest"
- "codellama:13b"
- "codellama:34b"
- "vicuna:latest"
- "wizardcoder:latest"
- You can edit models.conf file to add or remove models.