The Phi-3-mini-mango-1-llamafied is an instruct finetune of
Phi-3-mini-4k-instruct with 4K context and 3.8B parameters.
It is a first cut of finetuning Phi-3 (which is a great model!) to explore its properties and behaviour. More to follow.
This version of the model has had its weight layers converted to Llama format using @vonjack's
vonjack/Phi-3-mini-4k-instruct-LLaMAfied conversion script,
allowing it to run within a llama toolset ecosystem without change or trust_remote_code. It's also interesting to see how resilient the model still is after conversion.
The original Phi-3 format weights of this model are available at
rhysjones/Phi-3-mini-mango-1.
