ollama

History

Jesse Gross c826e57475 runner.go: Better abstract vision model integration

-Update mllama to take the cross attention state as embeddings in
a batch, more similar to how Llava handles it. This improves
integration with the input cache.
-Pass locations in a prompt for embeddings using tags similar to Llava.
-Abstract interface to vision models so the main runner accesses Clip
and Mllama similarly

Co-authored-by: Michael Yang <mxyng@pm.me>

2024-10-30 14:53:43 -07:00

0001-cuda.patch

llama: Decouple patching script from submodule (#7139 )

2024-10-17 15:03:09 -07:00

0002-pretokenizer.patch

llama: Decouple patching script from submodule (#7139 )

2024-10-17 15:03:09 -07:00

0003-metal.patch

llama: Decouple patching script from submodule (#7139 )

2024-10-17 15:03:09 -07:00

0004-ggml-metal.patch

llama: Decouple patching script from submodule (#7139 )