diff --git a/exla/lib/exla/defn.ex b/exla/lib/exla/defn.ex
index b06f139418..d3f67763a5 100644
--- a/exla/lib/exla/defn.ex
+++ b/exla/lib/exla/defn.ex
@@ -647,12 +647,12 @@ defmodule EXLA.Defn do
          %{client: %EXLA.Client{platform: :host}, builder: %Function{}} = state,
          cache
        ) do
-    dbg({type_kind})
     # We match only on platform: :host for MLIR, as we want to support
     # eigh-on-cpu as a custom call only in this case
     {tensor, cache} = recur_operator(tensor, state, cache) |> unwrap_single_tensor!()
 
-    # convert to float and ensure that we're either using f32 or f64
+    # convert to float and ensure that we're either using f32 or f64, because Eigen
+    # only supports f32 and f64 easily.
     out_type = Nx.Type.merge(Nx.Type.to_floating(eigenvecs_expr.type), {:f, 32})
 
     tensor =
@@ -669,8 +669,6 @@ defmodule EXLA.Defn do
         expr_to_typespec(%{eigenvals_expr | type: out_type})
       )
 
-    dbg(eigenvecs)
-
     {[to_type(eigenvecs, eigenvecs_expr.type), to_type(eigenvals, eigenvals_expr.type)], cache}
   end