- Add abstract func for HFP8QuantizedToFloat (pytorch#3000)

Summary:
X-link: facebookresearch/FBGEMM#93

Pull Request resolved: pytorch#3000

as title

Reviewed By: jianyuh

Differential Revision: D61340643

fbgemm_gpu/fbgemm_gpu/sparse_ops.py

@@ -972,6 +972,10 @@ def float_to_hfp8_quantized(
     return torch.empty_like(input, dtype=torch.uint8)
 
 
+def hfp8_quantized_to_float(input: Tensor, ebits: int, exponent_bias: int) -> Tensor:
+    return torch.empty_like(input, dtype=torch.float32)
+
+
 def _setup() -> None:
     # pyre-ignore[16]
     _setup.done = getattr(_setup, "done", False)
@@ -1102,6 +1106,11 @@ def impl_autograd(op_name, fn, setup_context: Optional[Callable] = None) -> None
             "fbgemm::FloatToHFP8Quantized",
             float_to_hfp8_quantized,
         )
+        impl_abstract(
+            "fbgemm::HFP8QuantizedToFloat",
+            hfp8_quantized_to_float,
+        )
+
         _setup.done = True
 
 

fbgemm_gpu/test/quantize/failures_dict_fast.json

@@ -53,19 +53,6 @@
           "status": "xfail"
         }
       },
-      "fbgemm::HFP8QuantizedToFloat": {
-        "SplitTableBatchedEmbeddingsTest.test_faketensor__test_nbit_forward_cpu": {
-          "comment": "",
-          "status": "xfail"
-        },
-        "SplitTableBatchedEmbeddingsTest.test_faketensor__test_nbit_forward_gpu_no_cache": {
-          "comment": "",
-          "status": "xfail"
-        },
-        "SplitTableBatchedEmbeddingsTest.test_faketensor__test_nbit_forward_gpu_no_cache_fp8_2048": {
-          "comment": "",
-          "status": "xfail"
-        }
-      }
+      "fbgemm::HFP8QuantizedToFloat": {}
     }
   }