Patrick Devine
Patrick Devine
The S3 driver currently has the "secure" option which will turn off TLS, however there is currently no option for using TLS but with an untrusted cert. This prevents alternative...
This change adds a new "Active" member to Block so that it can be used to tell whether a given widget is in focus or not. I added some examples...
This change switches the REPL to use `/api/chat` when running in interactive mode. It will still use `/api/generate` for non-interactive sessions. I've also attempted to DRY out the display response...
This change allows you to run `ollama pull --upgrade-all` which will check each of your local models and upgrade any that are out of date. It uses Etags to check...
This (admittedly very large) change converts a safetensors file in the `FROM` line of a Modelfile into a 16 bit *non-quantized* Ollama model without having to use llamacpp's `convert.py` script....
This change adds the ability to convert from a Gemma safetensors model.
This huge patch bomb just migrates from `github.com/jmorganca/ollama` to `github.com/ollama/ollama`. Fixes #3327
This change splits up the gemma/mistral conversion logic into their own files and creates a new ModelArch interface which any new converter can implement to support different model types.
When I'm using the pytorch interface and getting `HalfStorage` (i.e. 16 bit float values) I'd like to be able to get access to the underlying `[]uint16` data instead of having...
This change adds better validation to the `OLLAMA_HOST` variable when used with `ollama serve`. It should work with both IPv4 and IPv6, and includes unit tests.