[dnn] added base template class for mutualized auto-tuning

2019-07-09 16:09:34 -07:00
parent 066ae338f1
commit 88675fa01a
9 changed files with 181 additions and 106 deletions
--- a/examples/python/tensorflow/batchnorm.cpp
+++ b/examples/python/tensorflow/batchnorm.cpp
@@ -133,7 +133,6 @@ public:
    triton::driver::cu_buffer tdx(ctx, (CUdeviceptr)dx->flat<float>().data(), false);
    triton::driver::cu_buffer tdg(ctx, (CUdeviceptr)dg->flat<float>().data(), false);
    triton::driver::cu_buffer tdb(ctx, (CUdeviceptr)db->flat<float>().data(), false);
-
    // create config
    triton::dnn::batchnorm_backward batchnorm(C, 1, H, W, B, "fp32");
    std::ostringstream oss;