Updates to fused epilogue (#383)

* Enhancements and fixes to fused GEMM and Convolution epilogue.
* Need to explicitly list cudart as unit test library dependency.

tools/library/scripts/generator.py

@@ -1056,7 +1056,6 @@ def GenerateSM75_TensorOp_8832_TN(manifest, args):
 
   min_cc = 75
   max_cc = 1024
-
   alignment_constraints = [32,]
 
   for math_inst in math_instructions:
@@ -1136,7 +1135,6 @@ def GenerateSM75_TensorOp_8832_Interleaved(manifest, args):
 
   min_cc = 75
   max_cc = 1024
-
   alignment_constraints = [32,]
 
   for math_inst in math_instructions:
@@ -1907,7 +1905,6 @@ def GenerateSM80_TensorOp_16864_Interleaved(manifest, args):
 
   min_cc = 80
   max_cc = 1024
-
   alignment_constraints = [32,]
 
   for math_inst in math_instructions:

tools/profiler/src/cublas_helpers.cu

@@ -102,6 +102,12 @@ bool get_cublas_datatype(cublasDataType_t &data_type, library::NumericTypeID ele
     data_type = CUDA_R_16F;
     return true;
     
+  case library::NumericTypeID::kBF16:
+    break;
+  
+  case library::NumericTypeID::kTF32: 
+    break;
+  
   case library::NumericTypeID::kF32:
     data_type = CUDA_R_32F;
     return true;