[AMD] Enable more passing fp8 downcast clamping tests (#7363)

With triton-lang/triton#7361,
these tests are now passing.

Author: antiagainst

python/test/unit/language/test_conversions.py

@@ -376,16 +376,9 @@ def test_typeconvert_downcast_clamping(src_dtype, dst_dtype, mode, rounding="rtn
 
         if dst_dtype in ('float8e5', 'float8e4nv') and rounding == 'rtne' and torch.cuda.get_device_capability(0) < (9, 0):
             pytest.skip(f"{dst_dtype} downcast with RTNE rounding tests only supported on NVGPU with compute capability 9.0+")
-    elif is_hip():
-        if is_hip_cdna2():
-            pytest.skip(f"{dst_dtype} downcast to {dst_dtype} with clamping is not fully tested on AMDGPU CDNA2")
-
-        if is_hip_cdna3():
-            if src_dtype == 'bfloat16' and dst_dtype == 'float8e4nv':
-                pytest.skip(f"{src_dtype} downcast to {dst_dtype} with clamping is not fully tested on AMDGPU CDNA3")
-            if dst_dtype == 'float8e5' and mode in ('inf', '-inf'):
-                pytest.skip(f"Downcast to {dst_dtype} with clamping for `inf` or `-inf` "
-                             "is not fully tested on AMDGPU CDNA3")
+    elif is_hip_cdna2() or is_hip_cdna3():
+        if src_dtype == 'bfloat16' and dst_dtype == 'float8e4nv':
+            pytest.skip(f"{src_dtype} downcast to {dst_dtype} with clamping is not fully tested on AMDGPU CDNA2/3")
 
     converter = {
         tl.float8e4nv: torch.float8_e4m3fn,