[Mlir-commits] [mlir] Support float8_e3m4 and float8_e4m3 in np_to_memref (PR #186453)

[llvmlistbot at llvm.org]

================
@@ -77,6 +94,16 @@ def to_numpy(array):
     ), f"float8_e5m2 requires the ml_dtypes package, please run:\n\npip install ml_dtypes\n"
     if array.dtype == F8E5M2:
         return array.view("float8_e5m2")
+    assert not (
+        array.dtype == F8E3M4 and ml_dtypes is None
+    ), f"float8_e3m4 requires the ml_dtypes package, please run:\n\npip install ml_dtypes\n"
+    if array.dtype == F8E3M4:
+        return array.view("float8_e3m4")
+    assert not (
+        array.dtype == F8E4M3 and ml_dtypes is None
+    ), f"float8_e4m3 requires the ml_dtypes package, please run:\n\npip install ml_dtypes\n"
+    if array.dtype == F8E4M3:
+        return array.view("float8_e4m3")
----------------
srcarroll wrote:

> ah i see the above uses the same pattern - okay nm leave as is

i don't mind refactoring if you think it should be done. just because there's precedence, doesn't necessarily mean it's a good one. i'll defer to you or others on this as i don't have a strong opinion about it.

https://github.com/llvm/llvm-project/pull/186453