Fix misleading CUDA device specification by replacing 'cuda:0' with 'cuda'

automotive/3d-object-detection/backend_deploy.py

@@ -54,7 +54,7 @@ def name(self):
         return "python-SUT"
 
     def load(self):
-        device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
+        device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
         PaintArgs = namedtuple(
             'PaintArgs', [
                 'training_path', 'model_path', 'cam_sync'])
@@ -84,7 +84,7 @@ def predict(self, inputs):
         ], [], [], [], [], [], []
         with torch.inference_mode():
             device = torch.device(
-                "cuda:0" if torch.cuda.is_available() else "cpu")
+                "cuda" if torch.cuda.is_available() else "cpu")
             model_input = inputs[0]
             batched_pts = model_input['pts']
             scores_from_cam = []

language/bert/pytorch_SUT.py

@@ -61,7 +61,7 @@ def __init__(self, args):
 
         self.network = args.network
         self.dev = (
-            torch.device("cuda:0") if torch.cuda.is_available(
+            torch.device("cuda") if torch.cuda.is_available(
             ) else torch.device("cpu")
         )
         self.version = transformers.__version__

language/gpt-j/backend_PyTorch.py

@@ -81,7 +81,7 @@ def __init__(
         if self.use_gpu:
             print(f"Casting models to GPU...")
             assert torch.cuda.is_available(), "torch gpu is not available, exiting..."
-            self.device = torch.device("cuda:0")
+            self.device = torch.device("cuda")
             self.model.to(self.device)
 
         self.model.eval()

language/llama2-70b/README.md

@@ -185,7 +185,7 @@ python3 -u main.py --scenario Offline \
         --dataset-path ${DATASET_PATH} \
         --output-log-dir offline-logs \
         --dtype float32 \
-        --device cuda:0 2>&1 | tee offline_performance_log.log
+        --device cuda 2>&1 | tee offline_performance_log.log
 ```
 
 ### Server

language/llama2-70b/main.py

@@ -63,7 +63,7 @@ def get_args():
     parser.add_argument(
         "--device",
         type=str,
-        choices=["cpu", "cuda:0"],
+        choices=["cpu", "cuda"],
         default="cpu",
         help="device to use",
     )

language/llama2-70b/run_accuracy.sh

@@ -12,7 +12,7 @@ python3 -u main.py --scenario Offline \
         --dataset-path ${DATASET_PATH} \
         --output-log-dir offline_accuracy_loadgen_logs \
         --dtype float32 \
-        --device cuda:0 2>&1 | tee offline_accuracy_log.log
+        --device cuda 2>&1 | tee offline_accuracy_log.log
 
 python3 evaluate-accuracy.py --checkpoint-path ${CHECKPOINT_PATH} \
         --mlperf-accuracy-file offline_accuracy_loadgen_logs/mlperf_log_accuracy.json \

language/mixtral-8x7b/README.md

@@ -174,7 +174,7 @@ python3 -u main.py --scenario Offline \
         --dataset-path ${DATASET_PATH} \
         --output-log-dir offline-logs \
         --dtype float32 \
-        --device cuda:0 2>&1 | tee offline_performance_log.log
+        --device cuda 2>&1 | tee offline_performance_log.log
 ```
 
 ### Server

language/mixtral-8x7b/main.py

@@ -46,7 +46,7 @@ def get_args():
     parser.add_argument(
         "--device",
         type=str,
-        choices=["cpu", "cuda:0"],
+        choices=["cpu", "cuda"],
         default="cpu",
         help="device to use",
     )

language/mixtral-8x7b/run_accuracy.sh

@@ -11,7 +11,7 @@ python3 -u main.py --scenario Offline \
         --dataset-path ${DATASET_PATH} \
         --output-log-dir offline_accuracy_loadgen_logs \
         --dtype float32 \
-        --device cuda:0 2>&1 | tee offline_accuracy_log.log
+        --device cuda 2>&1 | tee offline_accuracy_log.log
 
 python3 evaluate-accuracy.py --checkpoint-path ${CHECKPOINT_PATH} \
         --mlperf-accuracy-file offline_accuracy_loadgen_logs/mlperf_log_accuracy.json \

recommendation/dlrm_v2/pytorch/python/backend_pytorch_native.py

@@ -65,7 +65,7 @@ def __init__(
         os.environ["MASTER_ADDR"] = "localhost"
         os.environ["MASTER_PORT"] = "29500"
         if self.use_gpu:
-            self.device: torch.device = torch.device(f"cuda:0")
+            self.device: torch.device = torch.device(f"cuda")
             self.dist_backend = "nccl"
             # torch.cuda.set_device(self.device)
         else:

retired_benchmarks/recommendation/dlrm/pytorch/python/backend_pytorch_native.py

@@ -22,7 +22,7 @@ def __init__(self, m_spa, ln_emb, ln_bot, ln_top,
         self.ln_top = ln_top
 
         self.use_gpu = use_gpu and torch.cuda.is_available()
-        self.device = "cuda:0" if self.use_gpu else "cpu"
+        self.device = "cuda" if self.use_gpu else "cpu"
 
         ngpus = torch.cuda.device_count() if self.use_gpu else -1
         self.ndevices = min(ngpus, mini_batch_size, ln_emb.size)

retired_benchmarks/speech_recognition/rnnt/pytorch/decoders.py

@@ -43,7 +43,7 @@ def __init__(self, blank_index, model, max_symbols_per_step=30):
         self._blank_id = blank_index
         self._SOS = -1
         self.dev = (
-            torch.device("cuda:0")
+            torch.device("cuda")
             if torch.cuda.is_available()
             and os.environ.get("USE_GPU", "").lower() not in ["no", "false"]
             else torch.device("cpu")

retired_benchmarks/speech_recognition/rnnt/pytorch/model_separable_rnnt.py

@@ -70,7 +70,7 @@ def __init__(
     ):
         super().__init__()
         self.dev = (
-            torch.device("cuda:0")
+            torch.device("cuda")
             if torch.cuda.is_available()
             and os.environ.get("USE_GPU", "").lower() not in ["no", "false"]
             else torch.device("cpu")
@@ -122,7 +122,7 @@ def __init__(
     ):
         super().__init__()
         self.dev = (
-            torch.device("cuda:0")
+            torch.device("cuda")
             if torch.cuda.is_available()
             and os.environ.get("USE_GPU", "").lower() not in ["no", "false"]
             else torch.device("cpu")
@@ -195,7 +195,7 @@ def __init__(
     ):
         super().__init__()
         self.dev = (
-            torch.device("cuda:0")
+            torch.device("cuda")
             if torch.cuda.is_available()
             and os.environ.get("USE_GPU", "").lower() not in ["no", "false"]
             else torch.device("cpu")

retired_benchmarks/speech_recognition/rnnt/pytorch/rnn.py

@@ -65,7 +65,7 @@ def __init__(
         """
         super(LstmDrop, self).__init__()
         self.dev = (
-            torch.device("cuda:0")
+            torch.device("cuda")
             if torch.cuda.is_available()
             and os.environ.get("USE_GPU", "").lower() not in ["no", "false"]
             else torch.device("cpu")
@@ -111,7 +111,7 @@ def __init__(self, factor):
         super().__init__()
         self.factor = int(factor)
         self.dev = (
-            torch.device("cuda:0")
+            torch.device("cuda")
             if torch.cuda.is_available()
             and os.environ.get("USE_GPU", "").lower() not in ["no", "false"]
             else torch.device("cpu")

retired_benchmarks/speech_recognition/rnnt/pytorch_SUT.py

@@ -55,7 +55,7 @@ def __init__(
         featurizer_config = config["input_eval"]
 
         self.dev = (
-            torch.device("cuda:0")
+            torch.device("cuda")
             if torch.cuda.is_available()
             and os.environ.get("USE_GPU", "").lower() not in ["no", "false"]
             else torch.device("cpu")

retired_benchmarks/vision/classification_and_detection/python/backend_pytorch.py

@@ -53,7 +53,7 @@ def load(self, model_path, inputs=None, outputs=None):
                 self.outputs.append(i.name)
 
         # prepare the backend
-        device = "CUDA:0" if torch.cuda.is_available() else "CPU"
+        device = "cuda" if torch.cuda.is_available() else "CPU"
         self.sess = caffe2.python.onnx.backend.prepare(self.model, device)
         return self
 

retired_benchmarks/vision/classification_and_detection/python/backend_pytorch_native.py

@@ -12,7 +12,7 @@ def __init__(self):
         super(BackendPytorchNative, self).__init__()
         self.sess = None
         self.model = None
-        self.device = "cuda:0" if torch.cuda.is_available() else "cpu"
+        self.device = "cuda" if torch.cuda.is_available() else "cpu"
 
     def version(self):
         return torch.__version__

retired_benchmarks/vision/medical_imaging/3d-unet-brats19/pytorch_SUT.py

@@ -55,7 +55,7 @@ def __init__(
         )
         self.trainer.load_checkpoint_ram(params[0], False)
         self.device = torch.device(
-            "cuda:0" if torch.cuda.is_available() else "cpu")
+            "cuda" if torch.cuda.is_available() else "cpu")
 
         print("Constructing SUT...")
         self.sut = lg.ConstructSUT(self.issue_queries, self.flush_queries)

retired_benchmarks/vision/medical_imaging/3d-unet-brats19/unet_pytorch_to_onnx.py

@@ -77,7 +77,7 @@ def main():
     width = 224
     depth = 160
     channels = 4
-    device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
+    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
     dummy_input = torch.rand(
         [1, channels, height, width, depth]).float().to(device)
     torch.onnx.export(

vision/classification_and_detection/python/backend_pytorch.py

@@ -53,7 +53,7 @@ def load(self, model_path, inputs=None, outputs=None):
                 self.outputs.append(i.name)
 
         # prepare the backend
-        device = "CUDA:0" if torch.cuda.is_available() else "CPU"
+        device = "cuda" if torch.cuda.is_available() else "CPU"
         self.sess = caffe2.python.onnx.backend.prepare(self.model, device)
         return self
 

vision/classification_and_detection/python/backend_pytorch_native.py

@@ -13,7 +13,7 @@ def __init__(self):
         super(BackendPytorchNative, self).__init__()
         self.sess = None
         self.model = None
-        self.device = "cuda:0" if torch.cuda.is_available() else "cpu"
+        self.device = "cuda" if torch.cuda.is_available() else "cpu"
 
     def version(self):
         return torch.__version__

vision/medical_imaging/3d-unet-kits19/pytorch_SUT.py

@@ -71,7 +71,7 @@ def __init__(self, model_path, preprocessed_data_dir, performance_count):
             model_path
         )
         self.device = torch.device(
-            "cuda:0" if torch.cuda.is_available() else "cpu")
+            "cuda" if torch.cuda.is_available() else "cpu")
         self.model = torch.jit.load(model_path, map_location=self.device)
         self.model.eval()
 

vision/medical_imaging/3d-unet-kits19/pytorch_checkpoint_SUT.py

@@ -302,7 +302,7 @@ def __init__(self, model_path, preprocessed_data_dir, performance_count):
             model_path
         )
         self.device = torch.device(
-            "cuda:0" if torch.cuda.is_available() else "cpu")
+            "cuda" if torch.cuda.is_available() else "cpu")
         self.model = Unet3D()
         self.model.to(self.device)
         checkpoint = torch.load(model_path, map_location=self.device)

vision/medical_imaging/3d-unet-kits19/unet_pytorch_to_onnx.py

@@ -96,7 +96,7 @@ def main():
     assert Path(model_path).is_file(), "Cannot find the model file {:}!".format(
         model_path
     )
-    device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
+    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 
     model = torch.jit.load(model_path, map_location=device)
     model.eval()