>>102053954
I was trying to convert from any tensor type to another. I'm guessing something implicit is happening that function.
I was about to give a bit more context, but then realized the author of the code actually had what I was trying to do commented out, so I just uncommented it...
final_weight = ggml_new_tensor(compute_ctx, GGML_TYPE_F32, ggml_n_dims(weight), weight->ne);
final_weight = ggml_cpy(compute_ctx, weight, final_weight);
// final_weight = to_f32(compute_ctx, weight);
// final_weight = ggml_add_inplace(compute_ctx, final_weight, updown);
// final_weight = ggml_cpy(compute_ctx, final_weight, weight);