[AMD][GLUON] Fix wmma_scaled instr shape verfication (#8425)

Compare instr shape against list instead of tuple. Also refresh the docs
for more data types.

Author: borontion

python/triton/experimental/gluon/language/amd/gfx1250/__init__.py

@@ -38,27 +38,26 @@ def wmma_scaled(a, a_scale, a_format, b, b_scale, b_format, acc, _semantic=None)
     Args:
         a (tensor): The operand A to be multiplied.
         a_scale (tensor): Scale factor for operand A.
-        a_format (str): Format of the operand A. Available formats: `e2m1'.
+        a_format (str): Format of the operand A. Available formats: `e2m1`, `e4m3`, `e5m2`.
         b (tensor): The operand B to be multiplied.
         b_scale (tensor): Scale factor for operand B.
-        b_format (str): Format of the operand B. Available formats: `e2m1'.
+        b_format (str): Format of the operand B. Available formats: `e2m1`, `e4m3`, `e5m2`.
         acc (tensor): Accumulator tensor.
     """
     _verify_wmma(3, a, b, acc)
     if a_format.value == "e2m1":
         wmma_layout = a.type.layout.parent
-        assert isinstance(wmma_layout, AMDWMMALayout) and wmma_layout.instr_shape == (16, 16, 64), \
-            "e2m1 format expects instr_shape to be (16, 16, 64)"
+        assert isinstance(wmma_layout, AMDWMMALayout) and wmma_layout.instr_shape == [16, 16, 64], \
+            "e2m1 format expects instr_shape to be [16, 16, 64]"
     if b_format.value == "e2m1":
         wmma_layout = b.type.layout.parent
-        assert isinstance(wmma_layout, AMDWMMALayout) and wmma_layout.instr_shape == (16, 16, 64), \
-            "e2m1 format expects instr_shape to be (16, 16, 64)"
+        assert isinstance(wmma_layout, AMDWMMALayout) and wmma_layout.instr_shape == [16, 16, 64], \
+            "e2m1 format expects instr_shape to be [16, 16, 64]"
 
     acc_layout = acc.type.layout
-    assert isinstance(acc_layout, AMDWMMALayout) and acc_layout.instr_shape == (16, 16, 128), \
-    "accumulator tensor's layout must be (16, 16, 128)"
+    assert isinstance(acc_layout, AMDWMMALayout) and acc_layout.instr_shape == [16, 16, 128], \
+    "accumulator tensor's layout must be [16, 16, 128]"
 
-    # TODO: Add more formats
     assert a_format.value in {"e2m1", "e4m3", "e5m2"}, f"Unsupported lhs_format: {a_format.value}"
     assert b_format.value in {"e2m1", "e4m3", "e5m2"}, f"Unsupported rhs_format: {b_format.value}"