Adding support for hpu device (#2595)

Summary:
Adding a new device support named 'hpu'

Pull Request resolved: #2595

Reviewed By: xuzhao9

Differential Revision: D73366654

Pulled By: atalman

fbshipit-source-id: 4f925560da9c24b5bc6bf2437b997f4dd89bdd3e

Author: arathi-hlab

conftest.py

@@ -38,6 +38,11 @@ def pytest_addoption(parser):
         action="store_true",
         help="Run benchmarks on mps only and ignore machine configuration checks",
     )
+    parser.addoption(
+        "--hpu_only",
+        action="store_true",
+        help="Run benchmarks on hpu only and ignore machine configuration checks",
+    )
 
 
 def set_fuser(fuser):

test.py

@@ -18,7 +18,6 @@
 )
 from torchbenchmark.util.metadata_utils import skip_by_metadata
 
-
 # Some of the models have very heavyweight setup, so we have to set a very
 # generous limit. That said, we don't want the entire test suite to hang if
 # a single test encounters an extreme failure, so we give up after a test is
@@ -175,6 +174,8 @@ def _load_tests():
     model_paths = _list_model_paths()
     if os.getenv("USE_CANARY_MODELS"):
         model_paths.extend(_list_canary_model_paths())
+    if hasattr(torch, "hpu") and torch.hpu.is_available():
+        devices.append("hpu")
     for path in model_paths:
         # TODO: skipping quantized tests for now due to BC-breaking changes for prepare
         # api, enable after PyTorch 1.13 release

test_bench.py

@@ -26,7 +26,7 @@
 def pytest_generate_tests(metafunc):
     # This is where the list of models to test can be configured
     # e.g. by using info in metafunc.config
-    devices = ["cpu", "cuda"]
+    devices = ["cpu", "cuda", "hpu"]
 
     if hasattr(torch.backends, "mps") and torch.backends.mps.is_available():
         devices.append("mps")
@@ -40,6 +40,9 @@ def pytest_generate_tests(metafunc):
     if metafunc.config.option.mps_only:
         devices = ["mps"]
 
+    if metafunc.config.option.hpu_only:
+        devices = ["hpu"]
+
     if metafunc.cls and metafunc.cls.__name__ == "TestBenchNetwork":
         paths = _list_model_paths()
         metafunc.parametrize(