ollama

• Fixed an issue where the ollama command line tool wouldn't correctly install

• Performance improvements with ollama create: it now uses less memory and will create custom models in less time
• Fixed an issue where running ollama create name -f requires an absolute file path to the model file; relative paths are now supported
• Fixed an issue where running ollama pull for a model that is already downloaded would show 0B

Early preview release

• Show performance details, such as tokens per second with ollama run --verbose
• Fixed a bug where ollama run would show an error when a model was already pulled

• Minor bug fixes

• Much faster performance
• ollama run now supports a larger prompt
• Fixed an issue where ollama run would hang after 40-50 words being output
• Fixed an issue where ollama run would fail if offline, even with a model downloaded
• • ollama run now prints better errors if a problem is encountered
• Ollama's logs can now be found under ~/.ollama/logs

• Improved reliability in cases where Ollama would encounter an error
• Fixed a bug where cancelling text generation would cause an error

This is an early preview release of Ollama