TEMPLATE for Ollama Modelfile?
#29
by
Ray9821
- opened
When I do ollama pull, it works fine as suggested but I would like to use unsloth quantized GGUF versions that I download myself. When I just use a model file with the PARAMETERs filled in, the output is garbled because the TEMPLATE is not set? gpt-oss uses the openai Harmony template as its noted here. Does any one have the value for the Modelfile TEMPLATE to make the unsloth GGUF ones to work thru ollama create?
The template is in the download files section. Or you can grab the full modelfile from your working container instance with 'ollama show gpt-oss:20b --modelfile'
We automatically set it now I'm pretty sure!