CUTLASS 3.3.0 (#1167)

* Release 3.3.0

Adds support for mixed precision GEMMs On Hopper and Ampere
Adds support for < 16B aligned GEMMs on Hopper
Enhancements to EVT
Enhancements to Python interface
Enhancements to Sub-byte type handling in CuTe
Several other bug-fixes and performance improvements.

* minor doc update

examples/python/00_basic_gemm.ipynb

@@ -269,6 +269,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "tiles = [td for td in tiles if td.threadblock_shape[0] >= 128]\n",
     "idx = random.randint(0, len(tiles)-1)\n",
     "td = tiles[idx]\n",
     "print('Tile description {} is: {}'.format(idx, td))\n",

examples/python/04_epilogue_visitor.ipynb

@@ -29,7 +29,7 @@
     "import cutlass\n",
     "from cutlass.epilogue import relu\n",
     "from cutlass import Tensor as FakeTensor\n",
-    "from cutlass.profiler import CUDAEventProfiler\n",
+    "from cutlass.utils.profiler import CUDAEventProfiler\n",
     "\n",
     "# This controls whether ther C++ GEMM declaration will be printed at each step. Set to `false` to\n",
     "# omit this information.\n",
@@ -160,10 +160,6 @@
     "        return example_epilogue(accum, alpha, C, beta, aux, bias)\n",
     "\n",
     "torch_reference = TorchReference()\n",
-    "if hasattr(torch, \"compile\"):\n",
-    "    # If the torch.compile feature is available\n",
-    "    torch_reference = torch.compile(torch_reference)\n",
-    "\n",
     "tensor_D_ref, tensor_F_ref = torch_reference(tensor_A, tensor_B, alpha, tensor_C, beta, aux, bias)\n",
     "\n",
     "assert torch.equal(tensor_D, tensor_D_ref)\n",