Masahiro

I tried code like this. This didn't help. I got same OOM message. ```Elixir policy = Axon.MixedPrecision.create_policy(compute: :f16) {:ok, clip} = Bumblebee.load_model({:hf, repository_id, subdir: "text_encoder"}, log_params_diff: false ) clip =...

I set the following Policy and confirmed that the image can be generated for the host client(not cuda). ```Elixir policy = Axon.MixedPrecision.create_policy( params: {:f, 16}, compute: {:f, 32}, output: {:f,...