Fix build error of userbuffer.cu

Signed-off-by: wenjh <wenjh@sugon.com>

transformer_engine/common/comm_gemm_overlap/userbuffers/userbuffers.cu

@@ -2067,7 +2067,7 @@ template void reducescatter2_userbuff_strided_atomic_fp8<te_hip_fp8_e4m3>(
     void *output, float *scale, const int handler, const int offset, const int rowelements,
     const int colelements, const int strideelements_out, const int strideelements_in,
     const int numchunks, void *counters, communicator *comm, cudaStream_t stream);
-template void reducescatter2_userbuff_strided_atomic_fp8<hip_f8<te_hip_fp8_e5m2>(
+template void reducescatter2_userbuff_strided_atomic_fp8<te_hip_fp8_e5m2>(
     void *output, float *scale, const int handler, const int offset, const int rowelements,
     const int colelements, const int strideelements_out, const int strideelements_in,
     const int numchunks, void *counters, communicator *comm, cudaStream_t stream);
@@ -2075,7 +2075,7 @@ template void reducescatter2_userbuff_strided_multiatomic_fp8<te_hip_fp8_e4m3>(
     void *output, float *scale, const int handler, const int offset, const int rowelements,
     const int colelements, const int strideelements_out, const int strideelements_in,
     const int numchunks, void *counters, communicator *comm, cudaStream_t stream);
-template void reducescatter2_userbuff_strided_multiatomic_fp8<hip_f8<te_hip_fp8_e5m2>(
+template void reducescatter2_userbuff_strided_multiatomic_fp8<te_hip_fp8_e5m2>(
     void *output, float *scale, const int handler, const int offset, const int rowelements,
     const int colelements, const int strideelements_out, const int strideelements_in,
     const int numchunks, void *counters, communicator *comm, cudaStream_t stream);