style: fix pre-existing formatting drift across codebase
This commit is contained in:
@@ -155,7 +155,9 @@ defmodule BDS.Embeddings.Backends.Neural do
|
||||
# Place model params/tensors on the Apple GPU (Metal) when accelerating with
|
||||
# EMLX so the compiled inference pass actually runs on-device. EXLA manages
|
||||
# its own device placement, so nothing to do there.
|
||||
defp maybe_set_default_backend(:emlx), do: Nx.global_default_backend({EMLX.Backend, device: :gpu})
|
||||
defp maybe_set_default_backend(:emlx),
|
||||
do: Nx.global_default_backend({EMLX.Backend, device: :gpu})
|
||||
|
||||
defp maybe_set_default_backend(:exla), do: :ok
|
||||
|
||||
@doc false
|
||||
|
||||
Reference in New Issue
Block a user