Benchmarks: Add Benchmark - Add FAMBench based on docker benchmark (#338)

**Description**
Integrate FAMBench into superbench based on docker implementation:
https://github.com/facebookresearch/FAMBench

The script to run all benchmarks is:
https://github.com/facebookresearch/FAMBench/blob/main/benchmarks/run_all.sh

examples/benchmarks/fambench.py

@@ -0,0 +1,22 @@
+# Copyright (c) Microsoft Corporation.
+# Licensed under the MIT license.
+
+"""Docker benchmark example for FAMBench.
+
+Commands to run:
+  python3 examples/benchmarks/fambench.py
+"""
+
+from superbench.benchmarks import BenchmarkRegistry, Platform
+from superbench.common.utils import logger
+
+if __name__ == '__main__':
+    context = BenchmarkRegistry.create_benchmark_context('fambench', platform=Platform.CUDA)
+
+    benchmark = BenchmarkRegistry.launch_benchmark(context)
+    if benchmark:
+        logger.info(
+            'benchmark: {}, return code: {}, result: {}'.format(
+                benchmark.name, benchmark.return_code, benchmark.result
+            )
+        )

superbench/benchmarks/docker_benchmarks/__init__.py

@@ -6,5 +6,9 @@
 from superbench.benchmarks.docker_benchmarks.docker_base import DockerBenchmark, CudaDockerBenchmark, \
     RocmDockerBenchmark
 from superbench.benchmarks.docker_benchmarks.rocm_onnxruntime_performance import RocmOnnxRuntimeModelBenchmark
+from superbench.benchmarks.docker_benchmarks.fambench import FAMBenchBenchmark
 
-__all__ = ['DockerBenchmark', 'CudaDockerBenchmark', 'RocmDockerBenchmark', 'RocmOnnxRuntimeModelBenchmark']
+__all__ = [
+    'DockerBenchmark', 'CudaDockerBenchmark', 'RocmDockerBenchmark', 'RocmOnnxRuntimeModelBenchmark',
+    'FAMBenchBenchmark'
+]

superbench/benchmarks/docker_benchmarks/fambench.py

@@ -0,0 +1,82 @@
+# Copyright (c) Microsoft Corporation.
+# Licensed under the MIT license.
+
+"""Module of the FAMBench benchmarks.
+
+Including:
+  DLRM
+  XLMR
+"""
+
+from superbench.common.utils import logger
+from superbench.benchmarks import BenchmarkRegistry, Platform
+from superbench.benchmarks.docker_benchmarks.docker_base import CudaDockerBenchmark
+
+
+class FAMBenchBenchmark(CudaDockerBenchmark):
+    """The FAMBench E2E model benchmark class."""
+    def __init__(self, name, parameters=''):
+        """Constructor.
+
+        Args:
+            name (str): benchmark name.
+            parameters (str): benchmark parameters.
+        """
+        super().__init__(name, parameters)
+
+        # Image uri of the current docker-benchmark.
+        self._image_uri = 'superbench/benchmark:cuda11.1.1-fambench'
+
+        # Image digest of the current docker-benchmark.
+        self._digest = 'b7b0d07270055287129e8b4b32be0863cbc3cc061610fcfaccf3a7450906e36f'
+
+        # Container name of the current docker-benchmark.
+        self._container_name = 'fambench-benchmarks'
+
+        # Entrypoint option of the current docker-benchmark.
+        self._entrypoint = '/workspace/FAMBench/benchmarks/run_all_benchmarks.sh'
+
+        # CMD option of the current docker-benchmark.
+        self._cmd = None
+
+    def _process_raw_result(self, cmd_idx, raw_output):
+        """Function to parse raw results and save the summarized results.
+
+          self._result.add_raw_data() and self._result.add_result() need to be called to save the results.
+
+        Args:
+            cmd_idx (int): the index of command corresponding with the raw_output.
+            raw_output (str): raw output string of the micro-benchmark.
+
+        Return:
+            True if the raw output string is valid and result can be extracted.
+        """
+        self._result.add_raw_data('raw_output', raw_output, self._args.log_raw_data)
+
+        content = raw_output.splitlines(False)
+        try:
+            result_header = 'benchmark implementation mode config score'
+            found = False
+            for line in content:
+                if result_header in line:
+                    found = True
+                elif found:
+                    items = line.split(' ')
+                    if len(items) == 7:
+                        name = '_'.join(items[0:4] + [items[5]])
+                        for char in ['-', ' ', '=', '/']:
+                            name = name.replace(char, '_')
+                        score = float(items[4])
+                        self._result.add_result(name.lower(), score)
+        except BaseException as e:
+            logger.error(
+                'The result format is invalid - round: {}, benchmark: {}, message: {}.'.format(
+                    self._curr_run_index, self._name, str(e)
+                )
+            )
+            return False
+
+        return True
+
+
+BenchmarkRegistry.register_benchmark('fambench', FAMBenchBenchmark, platform=Platform.CUDA)

tests/benchmarks/docker_benchmarks/test_fambench.py

@@ -0,0 +1,38 @@
+# Copyright (c) Microsoft Corporation.
+# Licensed under the MIT License.
+
+"""Tests for FAMBench modules."""
+
+from types import SimpleNamespace
+
+from superbench.benchmarks import BenchmarkRegistry, BenchmarkType, Platform, ReturnCode
+from superbench.benchmarks.result import BenchmarkResult
+
+
+def test_fambench():
+    """Test FAMBench benchmarks."""
+    benchmark_name = 'fambench'
+    (benchmark_class,
+     predefine_params) = BenchmarkRegistry._BenchmarkRegistry__select_benchmark(benchmark_name, Platform.CUDA)
+    assert (benchmark_class)
+    benchmark = benchmark_class(benchmark_name)
+    assert (benchmark._benchmark_type == BenchmarkType.DOCKER)
+    assert (benchmark._image_uri == 'superbench/benchmark:cuda11.1.1-fambench')
+    assert (benchmark._container_name == 'fambench-benchmarks')
+    assert (benchmark._entrypoint == '/workspace/FAMBench/benchmarks/run_all_benchmarks.sh')
+    assert (benchmark._cmd is None)
+    benchmark._result = BenchmarkResult(benchmark._name, benchmark._benchmark_type, ReturnCode.SUCCESS)
+    benchmark._args = SimpleNamespace(log_raw_data=False)
+
+    raw_output = """
+benchmark implementation mode config score units batch_latency_95_sec
+DLRM OOTB eval tiny 152.800399 ex/s 0.515052
+DLRM OOTB train tiny 35.483686 ex/s None
+DLRM UBENCH train linear_[(2,2,2,2,2)] 3.679281e-07 TF/s None
+XLMR OOTB eval default-config 1.015586 ex/s 16.463461
+"""
+    assert (benchmark._process_raw_result(0, raw_output))
+    assert (benchmark.result['dlrm_ootb_eval_tiny_ex_s'][0] == 152.800399)
+    assert (benchmark.result['dlrm_ootb_train_tiny_ex_s'][0] == 35.483686)
+    assert (benchmark.result['dlrm_ubench_train_linear_[(2,2,2,2,2)]_tf_s'][0] == 3.679281e-07)
+    assert (benchmark.result['xlmr_ootb_eval_default_config_ex_s'][0] == 1.015586)