Benchmarks: Code Revision - Fix some issue for BERT benchmark. (#58)

Benchmarks: Code Revision - Fix some issue for BERT benchmark. (#58)

examples/benchmarks/pytorch_bert_large.py

 # Copyright (c) Microsoft Corporation.
 # Licensed under the MIT license.
 
-"""Model benchmark example for bert-large.
+"""Model benchmark example for bert-large (24-layer, 1024-hidden, 16-heads, 340M parameters).
 
 Commands to run:
   python3 examples/benchmarks/pytorch_bert_large.py (Single GPU)
-  python3 -m torch.distributed.launch --nproc_per_node=8 examples/benchmarks/pytorch_bert_large.py (Distributed)
+  python3 -m torch.distributed.launch --use_env --nproc_per_node=8 examples/benchmarks/pytorch_bert_large.py \
+      --distributed (Distributed)
 """
 
-from superbench.benchmarks import Framework, BenchmarkRegistry
+import argparse
+
+from superbench.benchmarks import Platform, Framework, BenchmarkRegistry
 from superbench.common.utils import logger
 
 if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        '--distributed', action='store_true', default=False, help='Whether to enable distributed training.'
+    )
+    args = parser.parse_args()
+
+    # Specify the model name and benchmark parameters.
+    model_name = 'bert-large'
+    parameters = '--batch_size 1 --duration 120 --seq_len 128 --precision float32 --run_count 2'
+    if args.distributed:
+        parameters += ' --distributed_impl ddp --distributed_backend nccl'
+
     # Create context for bert-large benchmark and run it for 120 * 2 seconds.
     context = BenchmarkRegistry.create_benchmark_context(
-        'bert-large',
-        parameters='--batch_size 1 --duration 120 --seq_len 8 --precision float32 --run_count 2',
-        framework=Framework.PYTORCH
+        model_name, platform=Platform.CUDA, parameters=parameters, framework=Framework.PYTORCH
     )
 
     benchmark = BenchmarkRegistry.launch_benchmark(context)

superbench/benchmarks/model_benchmarks/pytorch_bert.py

@@ -171,10 +171,11 @@ def _inference_step(self, precision):
             self._model.eval()
             while True:
                 for idx, sample in enumerate(self._dataloader):
-                    torch.cuda.synchronize()
                     start = time.time()
+                    if self._gpu_available:
                         sample = sample.cuda()
                     self._model(sample)
+                    if self._gpu_available:
                         torch.cuda.synchronize()
                     end = time.time()
                     curr_step += 1

tests/benchmarks/model_benchmarks/test_pytorch_bert.py

@@ -4,8 +4,8 @@
 """Tests for BERT model benchmarks."""
 
 from tests.helper import decorator
-from superbench.benchmarks import BenchmarkRegistry, Precision, Platform, Framework
-import superbench.benchmarks.model_benchmarks.pytorch_bert as pybert
+from superbench.benchmarks import BenchmarkRegistry, Platform, Framework, BenchmarkType, ReturnCode
+from superbench.benchmarks.model_benchmarks.pytorch_bert import PytorchBERT
 
 
 @decorator.cuda_test
@@ -15,86 +15,45 @@ def test_pytorch_bert_base():
     context = BenchmarkRegistry.create_benchmark_context(
         'bert-base',
         platform=Platform.CUDA,
-        parameters='--batch_size 32 --num_classes 5 --seq_len 512',
+        parameters='--batch_size 1 --num_classes 5 --seq_len 8 --num_warmup 2 --num_steps 4 \
+            --model_action train inference',
         framework=Framework.PYTORCH
     )
 
     assert (BenchmarkRegistry.is_benchmark_context_valid(context))
 
-    benchmark_name = BenchmarkRegistry._BenchmarkRegistry__get_benchmark_name(context)
-    assert (benchmark_name == 'pytorch-bert-base')
+    benchmark = BenchmarkRegistry.launch_benchmark(context)
 
-    (benchmark_class,
-     predefine_params) = BenchmarkRegistry._BenchmarkRegistry__select_benchmark(benchmark_name, context.platform)
-    assert (benchmark_class == pybert.PytorchBERT)
+    # Check basic information.
+    assert (benchmark)
+    assert (isinstance(benchmark, PytorchBERT))
+    assert (benchmark.name == 'pytorch-bert-base')
+    assert (benchmark.type == BenchmarkType.MODEL)
 
-    parameters = context.parameters
-    if predefine_params:
-        parameters = predefine_params + ' ' + parameters
-
-    benchmark = benchmark_class(benchmark_name, parameters)
-    assert (benchmark._preprocess() is True)
-
-    # Predefined parameters of bert-base model.
+    # Check predefined parameters of resnet101 model.
     assert (benchmark._args.hidden_size == 768)
     assert (benchmark._args.num_hidden_layers == 12)
     assert (benchmark._args.num_attention_heads == 12)
     assert (benchmark._args.intermediate_size == 3072)
 
-    # Parameters from BenchmarkContext.
-    assert (benchmark._args.batch_size == 32)
-    assert (benchmark._args.num_classes == 5)
-    assert (benchmark._args.seq_len == 512)
-
-    # Test Dataset.
-    assert (len(benchmark._dataset) == benchmark._args.sample_count * benchmark._world_size)
-
-    # Test _create_model().
-    assert (benchmark._create_model(Precision.FLOAT32) is True)
-    assert (isinstance(benchmark._model, pybert.BertBenchmarkModel))
-
-
-@decorator.cuda_test
-@decorator.pytorch_test
-def test_pytorch_bert_large():
-    """Test pytorch-bert-large benchmark."""
-    context = BenchmarkRegistry.create_benchmark_context(
-        'bert-large',
-        platform=Platform.CUDA,
-        parameters='--batch_size 32 --num_classes 5 --seq_len 512',
-        framework=Framework.PYTORCH
-    )
-
-    assert (BenchmarkRegistry.is_benchmark_context_valid(context))
-
-    benchmark_name = BenchmarkRegistry._BenchmarkRegistry__get_benchmark_name(context)
-    assert (benchmark_name == 'pytorch-bert-large')
-
-    (benchmark_class,
-     predefine_params) = BenchmarkRegistry._BenchmarkRegistry__select_benchmark(benchmark_name, context.platform)
-    assert (benchmark_class is pybert.PytorchBERT)
-
-    parameters = context.parameters
-    if predefine_params:
-        parameters = predefine_params + ' ' + parameters
-
-    benchmark = benchmark_class(benchmark_name, parameters)
-    assert (benchmark._preprocess() is True)
-
-    # Predefined parameters of bert-large model.
-    assert (benchmark._args.hidden_size == 1024)
-    assert (benchmark._args.num_hidden_layers == 24)
-    assert (benchmark._args.num_attention_heads == 16)
-    assert (benchmark._args.intermediate_size == 4096)
-
-    # Parameters from BenchmarkContext.
-    assert (benchmark._args.batch_size == 32)
+    # Check parameters specified in BenchmarkContext.
+    assert (benchmark._args.batch_size == 1)
     assert (benchmark._args.num_classes == 5)
-    assert (benchmark._args.seq_len == 512)
+    assert (benchmark._args.seq_len == 8)
+    assert (benchmark._args.num_warmup == 2)
+    assert (benchmark._args.num_steps == 4)
 
-    # Test Dataset.
+    # Check dataset scale.
     assert (len(benchmark._dataset) == benchmark._args.sample_count * benchmark._world_size)
 
-    # Test _create_model().
-    assert (benchmark._create_model(Precision.FLOAT32) is True)
-    assert (isinstance(benchmark._model, pybert.BertBenchmarkModel))
+    # Check results and metrics.
+    assert (benchmark.run_count == 1)
+    assert (benchmark.return_code == ReturnCode.SUCCESS)
+    for metric in [
+        'steptime_train_float32', 'throughput_train_float32', 'steptime_train_float16', 'throughput_train_float16',
+        'steptime_inference_float32', 'throughput_inference_float32', 'steptime_inference_float16',
+        'throughput_inference_float16'
+    ]:
+        assert (len(benchmark.raw_data[metric]) == benchmark.run_count)
+        assert (len(benchmark.raw_data[metric][0]) == benchmark._args.num_steps)
+        assert (len(benchmark.result[metric]) == benchmark.run_count)