diff --git a/TODO_DML.md b/TODO_DML.md
index af6559ac1..f359ba1b3 100644
--- a/TODO_DML.md
+++ b/TODO_DML.md
@@ -20,3 +20,14 @@ What should we do?
 1. Use any fixed value as the available memory capacity.
 2. Use `atiadlxx`(AMD/ATI GPU driver library) to infer vram information as similar as possible to the actual value. (works for AMDGPUs)
 3. or another better way.
+
+- Half mode does not work well.
+
+In half precision, we get an error like `mat1 and mat2 must have the same dtype`.
+
+I solved this problem by overriding forward of GroupNorm / LayerNorm / Conv2d / Linear to convert input tensor to full precision (and convert to float16 before return).
+
+What should we do?
+
+1. Override forwards (same solution)
+2. When using DirectML, set the default to full precision and notify the user.