[Benchmarks] Change all args to from `no_...` to their positive form (#7075)

* Changed name to all no_... arguments and all references to them, inverting the boolean condition

* Change benchmark tests to use new Benchmark Args

* Update src/transformers/benchmark/benchmark_args_utils.py
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* Update src/transformers/benchmark/benchmark.py
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* Fix Style. Add --no options in help

* fix some part of tests

* Update src/transformers/benchmark/benchmark_args_utils.py

* Update src/transformers/benchmark/benchmark_args_utils.py

* Update src/transformers/benchmark/benchmark_args_utils.py

* fix all tests

* make style

* add backwards compability

* make backwards compatible
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
Co-authored-by: fmcurti <fcurti@DESKTOP-RRQURBM.localdomain>

examples/benchmarking/run_benchmark.py

@@ -20,7 +20,25 @@ from transformers import HfArgumentParser, PyTorchBenchmark, PyTorchBenchmarkArg
 
 def main():
     parser = HfArgumentParser(PyTorchBenchmarkArguments)
+    try:
         benchmark_args = parser.parse_args_into_dataclasses()[0]
+    except ValueError as e:
+        arg_error_msg = "Arg --no_{0} is no longer used, please use --no-{0} instead."
+        begin_error_msg = " ".join(str(e).split(" ")[:-1])
+        full_error_msg = ""
+        depreciated_args = eval(str(e).split(" ")[-1])
+        wrong_args = []
+        for arg in depreciated_args:
+            # arg[2:] removes '--'
+            if arg[2:] in PyTorchBenchmarkArguments.deprecated_args:
+                # arg[5:] removes '--no_'
+                full_error_msg += arg_error_msg.format(arg[5:])
+            else:
+                wrong_args.append(arg)
+        if len(wrong_args) > 0:
+            full_error_msg = full_error_msg + begin_error_msg + str(wrong_args)
+        raise ValueError(full_error_msg)
+
     benchmark = PyTorchBenchmark(args=benchmark_args)
     benchmark.run()
 

examples/benchmarking/run_benchmark_tf.py

@@ -22,6 +22,24 @@ def main():
     parser = HfArgumentParser(TensorFlowBenchmarkArguments)
     benchmark_args = parser.parse_args_into_dataclasses()[0]
     benchmark = TensorFlowBenchmark(args=benchmark_args)
+    try:
+        benchmark_args = parser.parse_args_into_dataclasses()[0]
+    except ValueError as e:
+        arg_error_msg = "Arg --no_{0} is no longer used, please use --no-{0} instead."
+        begin_error_msg = " ".join(str(e).split(" ")[:-1])
+        full_error_msg = ""
+        depreciated_args = eval(str(e).split(" ")[-1])
+        wrong_args = []
+        for arg in depreciated_args:
+            # arg[2:] removes '--'
+            if arg[2:] in TensorFlowBenchmark.deprecated_args:
+                # arg[5:] removes '--no_'
+                full_error_msg += arg_error_msg.format(arg[5:])
+            else:
+                wrong_args.append(arg)
+        if len(wrong_args) > 0:
+            full_error_msg = full_error_msg + begin_error_msg + str(wrong_args)
+        raise ValueError(full_error_msg)
     benchmark.run()
 
 

src/transformers/benchmark/benchmark.py

@@ -229,7 +229,7 @@ class PyTorchBenchmark(Benchmark):
             if self.args.is_tpu:
                 # tpu
                 raise NotImplementedError(
-                    "Memory Benchmarking is currently not implemented for TPU. Please disable memory benchmarking with `--no_memory` or `args.no_memory=True`"
+                    "Memory Benchmarking is currently not implemented for TPU. Please disable memory benchmarking with `--no-memory` or `args.memory=False`"
                 )
             elif self.args.is_gpu:
                 if not is_py3nvml_available():

src/transformers/benchmark/benchmark_args.py

@@ -34,6 +34,34 @@ logger = logging.get_logger(__name__)
 
 @dataclass
 class PyTorchBenchmarkArguments(BenchmarkArguments):
+
+    deprecated_args = [
+        "no_inference",
+        "no_cuda",
+        "no_tpu",
+        "no_speed",
+        "no_memory",
+        "no_env_print",
+        "no_multi_process",
+    ]
+
+    def __init__(self, **kwargs):
+        """This __init__ is there for legacy code. When removing
+        deprecated args completely, the class can simply be deleted
+        """
+        for deprecated_arg in self.deprecated_args:
+            if deprecated_arg in kwargs:
+                positive_arg = deprecated_arg[3:]
+                setattr(self, positive_arg, not kwargs.pop(deprecated_arg))
+                logger.warning(
+                    f"{deprecated_arg} is depreciated. Please use --no-{positive_arg} or {positive_arg}={kwargs[positive_arg]}"
+                )
+
+        self.torchscript = kwargs.pop("torchscript", self.torchscript)
+        self.torch_xla_tpu_print_metrics = kwargs.pop("torch_xla_tpu_print_metrics", self.torch_xla_tpu_print_metrics)
+        self.fp16_opt_level = kwargs.pop("fp16_opt_level", self.fp16_opt_level)
+        super().__init__(**kwargs)
+
     torchscript: bool = field(default=False, metadata={"help": "Trace the models using torchscript"})
     torch_xla_tpu_print_metrics: bool = field(default=False, metadata={"help": "Print Xla/PyTorch tpu metrics"})
     fp16_opt_level: str = field(
@@ -50,7 +78,7 @@ class PyTorchBenchmarkArguments(BenchmarkArguments):
     @torch_required
     def _setup_devices(self) -> Tuple["torch.device", int]:
         logger.info("PyTorch: setting up devices")
-        if self.no_cuda:
+        if not self.cuda:
             device = torch.device("cpu")
             n_gpu = 0
         elif is_torch_tpu_available():
@@ -63,7 +91,7 @@ class PyTorchBenchmarkArguments(BenchmarkArguments):
 
     @property
     def is_tpu(self):
-        return is_torch_tpu_available() and not self.no_tpu
+        return is_torch_tpu_available() and self.tpu
 
     @property
     @torch_required

src/transformers/benchmark/benchmark_args_tf.py

@@ -31,6 +31,34 @@ logger = logging.get_logger(__name__)
 
 @dataclass
 class TensorFlowBenchmarkArguments(BenchmarkArguments):
+
+    deprecated_args = [
+        "no_inference",
+        "no_cuda",
+        "no_tpu",
+        "no_speed",
+        "no_memory",
+        "no_env_print",
+        "no_multi_process",
+    ]
+
+    def __init__(self, **kwargs):
+        """This __init__ is there for legacy code. When removing
+        deprecated args completely, the class can simply be deleted
+        """
+        for deprecated_arg in self.deprecated_args:
+            if deprecated_arg in kwargs:
+                positive_arg = deprecated_arg[3:]
+                kwargs[positive_arg] = not kwargs.pop(deprecated_arg)
+                logger.warning(
+                    f"{deprecated_arg} is depreciated. Please use --no-{positive_arg} or {positive_arg}={kwargs[positive_arg]}"
+                )
+        self.tpu_name = kwargs.pop("tpu_name", self.tpu_name)
+        self.device_idx = kwargs.pop("device_idx", self.device_idx)
+        self.eager_mode = kwargs.pop("eager_mode", self.eager_mode)
+        self.use_xla = kwargs.pop("use_xla", self.use_xla)
+        super().__init__(**kwargs)
+
     tpu_name: str = field(
         default=None,
         metadata={"help": "Name of TPU"},
@@ -50,7 +78,7 @@ class TensorFlowBenchmarkArguments(BenchmarkArguments):
     @cached_property
     @tf_required
     def _setup_tpu(self) -> Tuple["tf.distribute.cluster_resolver.TPUClusterResolver"]:
-        if not self.no_tpu:
+        if self.tpu:
             try:
                 if self.tpu_name:
                     tpu = tf.distribute.cluster_resolver.TPUClusterResolver(self.tpu_name)
@@ -98,7 +126,7 @@ class TensorFlowBenchmarkArguments(BenchmarkArguments):
     @property
     @tf_required
     def n_gpu(self) -> int:
-        if not self.no_cuda:
+        if self.cuda:
             return len(self.gpu_list)
         return 0
 

src/transformers/benchmark/benchmark_args_utils.py

@@ -57,22 +57,38 @@ class BenchmarkArguments:
         metadata={"help": "List of sequence lengths for which memory and time performance will be evaluated"},
     )
 
-    no_inference: bool = field(default=False, metadata={"help": "Don't benchmark inference of model"})
-    no_cuda: bool = field(default=False, metadata={"help": "Whether to run on available cuda devices"})
-    no_tpu: bool = field(default=False, metadata={"help": "Whether to run on available tpu devices"})
+    inference: bool = field(
+        default=True,
+        metadata={"help": "Whether to benchmark inference of model. Inference can be disabled via --no-inference."},
+    )
+    cuda: bool = field(
+        default=True,
+        metadata={"help": "Whether to run on available cuda devices. Cuda can be disabled via --no-cuda."},
+    )
+    tpu: bool = field(
+        default=True, metadata={"help": "Whether to run on available tpu devices. TPU can be disabled via --no-tpu."}
+    )
     fp16: bool = field(default=False, metadata={"help": "Use FP16 to accelerate inference."})
     training: bool = field(default=False, metadata={"help": "Benchmark training of model"})
     verbose: bool = field(default=False, metadata={"help": "Verbose memory tracing"})
-    no_speed: bool = field(default=False, metadata={"help": "Don't perform speed measurements"})
-    no_memory: bool = field(default=False, metadata={"help": "Don't perform memory measurements"})
+    speed: bool = field(
+        default=True,
+        metadata={"help": "Whether to perform speed measurements. Speed measurements can be disabled via --no-speed."},
+    )
+    memory: bool = field(
+        default=True,
+        metadata={
+            "help": "Whether to perform memory measurements. Memory measurements can be disabled via --no-memory"
+        },
+    )
     trace_memory_line_by_line: bool = field(default=False, metadata={"help": "Trace memory line by line"})
     save_to_csv: bool = field(default=False, metadata={"help": "Save result to a CSV file"})
     log_print: bool = field(default=False, metadata={"help": "Save all print statements in a log file"})
-    no_env_print: bool = field(default=False, metadata={"help": "Don't print environment information"})
-    no_multi_process: bool = field(
-        default=False,
+    env_print: bool = field(default=False, metadata={"help": "Whether to print environment information"})
+    multi_process: bool = field(
+        default=True,
         metadata={
-            "help": "Don't use multiprocessing for memory and speed measurement. It is highly recommended to use multiprocessing for accurate CPU and GPU memory measurements. This option should only be used for debugging / testing and on TPU."
+            "help": "Whether to use multiprocessing for memory and speed measurement. It is highly recommended to use multiprocessing for accurate CPU and GPU memory measurements. This option should only be disabled for debugging / testing and on TPU."
         },
     )
     inference_time_csv_file: str = field(
@@ -122,7 +138,7 @@ class BenchmarkArguments:
 
     @property
     def do_multi_processing(self):
-        if self.no_multi_process:
+        if not self.multi_process:
             return False
         elif self.is_tpu:
             logger.info("Multiprocessing is currently not possible on TPU.")

src/transformers/benchmark/benchmark_tf.py

@@ -248,7 +248,7 @@ class TensorFlowBenchmark(Benchmark):
                 if self.args.is_tpu:
                     # tpu
                     raise NotImplementedError(
-                        "Memory Benchmarking is currently not implemented for TPU. Please disable memory benchmarking with `args.no_memory=True`"
+                        "Memory Benchmarking is currently not implemented for TPU. Please disable memory benchmarking with `args.memory=False`"
                     )
                 elif self.args.is_gpu:
                     # gpu

src/transformers/benchmark/benchmark_utils.py

@@ -584,9 +584,9 @@ class Benchmark(ABC):
         else:
             self.config_dict = {model_name: config for model_name, config in zip(self.args.model_names, configs)}
 
-        if not self.args.no_memory and os.getenv("TRANSFORMERS_USE_MULTIPROCESSING") == 0:
+        if self.args.memory and os.getenv("TRANSFORMERS_USE_MULTIPROCESSING") == 0:
             logger.warning(
-                "Memory consumption will not be measured accurately if `args.no_multi_process` is set to `True.` The flag 'TRANSFORMERS_USE_MULTIPROCESSING' should only be disabled for debugging / testing."
+                "Memory consumption will not be measured accurately if `args.multi_process` is set to `False.` The flag 'TRANSFORMERS_USE_MULTIPROCESSING' should only be disabled for debugging / testing."
             )
 
         self._print_fn = None
@@ -669,24 +669,24 @@ class Benchmark(ABC):
 
             for batch_size in self.args.batch_sizes:
                 for sequence_length in self.args.sequence_lengths:
-                    if not self.args.no_inference:
-                        if not self.args.no_memory:
+                    if self.args.inference:
+                        if self.args.memory:
                             memory, inference_summary = self.inference_memory(model_name, batch_size, sequence_length)
                             inference_result_memory[model_name]["result"][batch_size][sequence_length] = memory
-                        if not self.args.no_speed:
+                        if self.args.speed:
                             time = self.inference_speed(model_name, batch_size, sequence_length)
                             inference_result_time[model_name]["result"][batch_size][sequence_length] = time
 
                     if self.args.training:
-                        if not self.args.no_memory:
+                        if self.args.memory:
                             memory, train_summary = self.train_memory(model_name, batch_size, sequence_length)
                             train_result_memory[model_name]["result"][batch_size][sequence_length] = memory
-                        if not self.args.no_speed:
+                        if self.args.speed:
                             time = self.train_speed(model_name, batch_size, sequence_length)
                             train_result_time[model_name]["result"][batch_size][sequence_length] = time
 
-        if not self.args.no_inference:
-            if not self.args.no_speed:
+        if self.args.inference:
+            if self.args.speed:
                 self.print_fn("\n" + 20 * "=" + ("INFERENCE - SPEED - RESULT").center(40) + 20 * "=")
                 self.print_results(inference_result_time, type_label="Time in s")
                 self.save_to_csv(inference_result_time, self.args.inference_time_csv_file)
@@ -695,7 +695,7 @@ class Benchmark(ABC):
                         "TPU was used for inference. Note that the time after compilation stabilized (after ~10 inferences model.forward(..) calls) was measured."
                     )
 
-            if not self.args.no_memory:
+            if self.args.memory:
                 self.print_fn("\n" + 20 * "=" + ("INFERENCE - MEMORY - RESULT").center(40) + 20 * "=")
                 self.print_results(inference_result_memory, type_label="Memory in MB")
                 self.save_to_csv(inference_result_memory, self.args.inference_memory_csv_file)
@@ -705,7 +705,7 @@ class Benchmark(ABC):
                 self.print_memory_trace_statistics(inference_summary)
 
         if self.args.training:
-            if not self.args.no_speed:
+            if self.args.speed:
                 self.print_fn("\n" + 20 * "=" + ("TRAIN - SPEED - RESULTS").center(40) + 20 * "=")
                 self.print_results(train_result_time, "Time in s")
                 self.save_to_csv(train_result_time, self.args.train_time_csv_file)
@@ -714,7 +714,7 @@ class Benchmark(ABC):
                         "TPU was used for training. Note that the time after compilation stabilized (after ~10 train loss=model.forward(...) + loss.backward() calls) was measured."
                     )
 
-            if not self.args.no_memory:
+            if self.args.memory:
                 self.print_fn("\n" + 20 * "=" + ("TRAIN - MEMORY - RESULTS").center(40) + 20 * "=")
                 self.print_results(train_result_memory, type_label="Memory in MB")
                 self.save_to_csv(train_result_memory, self.args.train_memory_csv_file)
@@ -723,7 +723,7 @@ class Benchmark(ABC):
                 self.print_fn("\n" + 20 * "=" + ("TRAIN - MEMOMRY - LINE BY LINE - SUMMARY").center(40) + 20 * "=")
                 self.print_memory_trace_statistics(train_summary)
 
-        if not self.args.no_env_print:
+        if self.args.env_print:
             self.print_fn("\n" + 20 * "=" + ("ENVIRONMENT INFORMATION").center(40) + 20 * "=")
             self.print_fn(
                 "\n".join(["- {}: {}".format(prop, val) for prop, val in self.environment_info.items()]) + "\n"

tests/test_benchmark.py

@@ -24,10 +24,10 @@ class BenchmarkTest(unittest.TestCase):
         benchmark_args = PyTorchBenchmarkArguments(
             models=[MODEL_ID],
             training=False,
-            no_inference=False,
+            inference=True,
             sequence_lengths=[8],
             batch_sizes=[1],
-            no_multi_process=True,
+            multi_process=False,
         )
         benchmark = PyTorchBenchmark(benchmark_args)
         results = benchmark.run()
@@ -39,10 +39,10 @@ class BenchmarkTest(unittest.TestCase):
         benchmark_args = PyTorchBenchmarkArguments(
             models=[MODEL_ID],
             training=False,
-            no_inference=False,
+            inference=True,
             sequence_lengths=[8],
             batch_sizes=[1],
-            no_multi_process=True,
+            multi_process=False,
             only_pretrain_model=True,
         )
         benchmark = PyTorchBenchmark(benchmark_args)
@@ -55,11 +55,11 @@ class BenchmarkTest(unittest.TestCase):
         benchmark_args = PyTorchBenchmarkArguments(
             models=[MODEL_ID],
             training=False,
-            no_inference=False,
+            inference=True,
             torchscript=True,
             sequence_lengths=[8],
             batch_sizes=[1],
-            no_multi_process=True,
+            multi_process=False,
         )
         benchmark = PyTorchBenchmark(benchmark_args)
         results = benchmark.run()
@@ -72,11 +72,11 @@ class BenchmarkTest(unittest.TestCase):
         benchmark_args = PyTorchBenchmarkArguments(
             models=[MODEL_ID],
             training=False,
-            no_inference=False,
+            inference=True,
             fp16=True,
             sequence_lengths=[8],
             batch_sizes=[1],
-            no_multi_process=True,
+            multi_process=False,
         )
         benchmark = PyTorchBenchmark(benchmark_args)
         results = benchmark.run()
@@ -91,10 +91,10 @@ class BenchmarkTest(unittest.TestCase):
         benchmark_args = PyTorchBenchmarkArguments(
             models=[MODEL_ID],
             training=True,
-            no_inference=False,
+            inference=True,
             sequence_lengths=[8],
             batch_sizes=[1],
-            no_multi_process=True,
+            multi_process=False,
         )
         benchmark = PyTorchBenchmark(benchmark_args, configs=[config])
         results = benchmark.run()
@@ -106,10 +106,10 @@ class BenchmarkTest(unittest.TestCase):
         benchmark_args = PyTorchBenchmarkArguments(
             models=[MODEL_ID],
             training=True,
-            no_inference=True,
+            inference=False,
             sequence_lengths=[8],
             batch_sizes=[1],
-            no_multi_process=True,
+            multi_process=False,
         )
         benchmark = PyTorchBenchmark(benchmark_args)
         results = benchmark.run()
@@ -122,11 +122,11 @@ class BenchmarkTest(unittest.TestCase):
         benchmark_args = PyTorchBenchmarkArguments(
             models=[MODEL_ID],
             training=True,
-            no_inference=True,
+            inference=False,
             sequence_lengths=[8],
             batch_sizes=[1],
             fp16=True,
-            no_multi_process=True,
+            multi_process=False,
         )
         benchmark = PyTorchBenchmark(benchmark_args)
         results = benchmark.run()
@@ -139,10 +139,10 @@ class BenchmarkTest(unittest.TestCase):
         benchmark_args = PyTorchBenchmarkArguments(
             models=[MODEL_ID],
             training=False,
-            no_inference=False,
+            inference=True,
             sequence_lengths=[8],
             batch_sizes=[1],
-            no_multi_process=True,
+            multi_process=False,
         )
         benchmark = PyTorchBenchmark(benchmark_args, configs=[config])
         results = benchmark.run()
@@ -155,10 +155,10 @@ class BenchmarkTest(unittest.TestCase):
         benchmark_args = PyTorchBenchmarkArguments(
             models=[MODEL_ID],
             training=False,
-            no_inference=False,
+            inference=True,
             sequence_lengths=[8],
             batch_sizes=[1],
-            no_multi_process=True,
+            multi_process=False,
         )
         benchmark = PyTorchBenchmark(benchmark_args, configs=[config])
         results = benchmark.run()
@@ -171,10 +171,10 @@ class BenchmarkTest(unittest.TestCase):
         benchmark_args = PyTorchBenchmarkArguments(
             models=[MODEL_ID],
             training=True,
-            no_inference=True,
+            inference=False,
             sequence_lengths=[8],
             batch_sizes=[1],
-            no_multi_process=True,
+            multi_process=False,
         )
         benchmark = PyTorchBenchmark(benchmark_args, configs=[config])
         results = benchmark.run()
@@ -187,10 +187,10 @@ class BenchmarkTest(unittest.TestCase):
         benchmark_args = PyTorchBenchmarkArguments(
             models=[MODEL_ID],
             training=True,
-            no_inference=True,
+            inference=True,
             sequence_lengths=[8],
             batch_sizes=[1],
-            no_multi_process=True,
+            multi_process=False,
         )
         benchmark = PyTorchBenchmark(benchmark_args, configs=[config])
         results = benchmark.run()
@@ -203,7 +203,7 @@ class BenchmarkTest(unittest.TestCase):
             benchmark_args = PyTorchBenchmarkArguments(
                 models=[MODEL_ID],
                 training=True,
-                no_inference=False,
+                inference=True,
                 save_to_csv=True,
                 sequence_lengths=[8],
                 batch_sizes=[1],
@@ -212,7 +212,7 @@ class BenchmarkTest(unittest.TestCase):
                 inference_memory_csv_file=os.path.join(tmp_dir, "inf_mem.csv"),
                 train_time_csv_file=os.path.join(tmp_dir, "train_time.csv"),
                 env_info_csv_file=os.path.join(tmp_dir, "env.csv"),
-                no_multi_process=True,
+                multi_process=False,
             )
             benchmark = PyTorchBenchmark(benchmark_args)
             benchmark.run()
@@ -235,13 +235,13 @@ class BenchmarkTest(unittest.TestCase):
             benchmark_args = PyTorchBenchmarkArguments(
                 models=[MODEL_ID],
                 training=True,
-                no_inference=False,
+                inference=True,
                 sequence_lengths=[8],
                 batch_sizes=[1],
                 log_filename=os.path.join(tmp_dir, "log.txt"),
                 log_print=True,
                 trace_memory_line_by_line=True,
-                no_multi_process=True,
+                multi_process=False,
             )
             benchmark = PyTorchBenchmark(benchmark_args)
             result = benchmark.run()

tests/test_benchmark_tf.py

@@ -26,11 +26,11 @@ class TFBenchmarkTest(unittest.TestCase):
         benchmark_args = TensorFlowBenchmarkArguments(
             models=[MODEL_ID],
             training=False,
-            no_inference=False,
+            inference=True,
             sequence_lengths=[8],
             batch_sizes=[1],
             eager_mode=True,
-            no_multi_process=True,
+            multi_process=False,
         )
         benchmark = TensorFlowBenchmark(benchmark_args)
         results = benchmark.run()
@@ -42,10 +42,10 @@ class TFBenchmarkTest(unittest.TestCase):
         benchmark_args = TensorFlowBenchmarkArguments(
             models=[MODEL_ID],
             training=False,
-            no_inference=False,
+            inference=True,
             sequence_lengths=[8],
             batch_sizes=[1],
-            no_multi_process=True,
+            multi_process=False,
             only_pretrain_model=True,
         )
         benchmark = TensorFlowBenchmark(benchmark_args)
@@ -58,10 +58,10 @@ class TFBenchmarkTest(unittest.TestCase):
         benchmark_args = TensorFlowBenchmarkArguments(
             models=[MODEL_ID],
             training=False,
-            no_inference=False,
+            inference=True,
             sequence_lengths=[8],
             batch_sizes=[1],
-            no_multi_process=True,
+            multi_process=False,
         )
         benchmark = TensorFlowBenchmark(benchmark_args)
         results = benchmark.run()
@@ -74,11 +74,11 @@ class TFBenchmarkTest(unittest.TestCase):
         benchmark_args = TensorFlowBenchmarkArguments(
             models=[MODEL_ID],
             training=False,
-            no_inference=False,
+            inference=True,
             sequence_lengths=[8],
             batch_sizes=[1],
             eager_mode=True,
-            no_multi_process=True,
+            multi_process=False,
         )
         benchmark = TensorFlowBenchmark(benchmark_args, [config])
         results = benchmark.run()
@@ -91,10 +91,10 @@ class TFBenchmarkTest(unittest.TestCase):
         benchmark_args = TensorFlowBenchmarkArguments(
             models=[MODEL_ID],
             training=False,
-            no_inference=False,
+            inference=True,
             sequence_lengths=[8],
             batch_sizes=[1],
-            no_multi_process=True,
+            multi_process=False,
         )
         benchmark = TensorFlowBenchmark(benchmark_args, [config])
         results = benchmark.run()
@@ -106,10 +106,10 @@ class TFBenchmarkTest(unittest.TestCase):
         benchmark_args = TensorFlowBenchmarkArguments(
             models=[MODEL_ID],
             training=True,
-            no_inference=True,
+            inference=False,
             sequence_lengths=[8],
             batch_sizes=[1],
-            no_multi_process=True,
+            multi_process=False,
         )
         benchmark = TensorFlowBenchmark(benchmark_args)
         results = benchmark.run()
@@ -122,10 +122,10 @@ class TFBenchmarkTest(unittest.TestCase):
         benchmark_args = TensorFlowBenchmarkArguments(
             models=[MODEL_ID],
             training=True,
-            no_inference=True,
+            inference=False,
             sequence_lengths=[8],
             batch_sizes=[1],
-            no_multi_process=True,
+            multi_process=False,
         )
         benchmark = TensorFlowBenchmark(benchmark_args, [config])
         results = benchmark.run()
@@ -138,10 +138,10 @@ class TFBenchmarkTest(unittest.TestCase):
         benchmark_args = TensorFlowBenchmarkArguments(
             models=[MODEL_ID],
             training=False,
-            no_inference=False,
+            inference=True,
             sequence_lengths=[8],
             batch_sizes=[1],
-            no_multi_process=True,
+            multi_process=False,
         )
         benchmark = TensorFlowBenchmark(benchmark_args, configs=[config])
         results = benchmark.run()
@@ -154,11 +154,11 @@ class TFBenchmarkTest(unittest.TestCase):
         benchmark_args = TensorFlowBenchmarkArguments(
             models=[MODEL_ID],
             training=False,
-            no_inference=False,
+            inference=True,
             sequence_lengths=[8],
             batch_sizes=[1],
             use_xla=True,
-            no_multi_process=True,
+            multi_process=False,
         )
         benchmark = TensorFlowBenchmark(benchmark_args)
         results = benchmark.run()
@@ -170,14 +170,14 @@ class TFBenchmarkTest(unittest.TestCase):
         with tempfile.TemporaryDirectory() as tmp_dir:
             benchmark_args = TensorFlowBenchmarkArguments(
                 models=[MODEL_ID],
-                no_inference=False,
+                inference=True,
                 save_to_csv=True,
                 sequence_lengths=[8],
                 batch_sizes=[1],
                 inference_time_csv_file=os.path.join(tmp_dir, "inf_time.csv"),
                 inference_memory_csv_file=os.path.join(tmp_dir, "inf_mem.csv"),
                 env_info_csv_file=os.path.join(tmp_dir, "env.csv"),
-                no_multi_process=True,
+                multi_process=False,
             )
             benchmark = TensorFlowBenchmark(benchmark_args)
             benchmark.run()
@@ -197,14 +197,14 @@ class TFBenchmarkTest(unittest.TestCase):
         with tempfile.TemporaryDirectory() as tmp_dir:
             benchmark_args = TensorFlowBenchmarkArguments(
                 models=[MODEL_ID],
-                no_inference=False,
+                inference=True,
                 sequence_lengths=[8],
                 batch_sizes=[1],
                 log_filename=os.path.join(tmp_dir, "log.txt"),
                 log_print=True,
                 trace_memory_line_by_line=True,
                 eager_mode=True,
-                no_multi_process=True,
+                multi_process=False,
             )
             benchmark = TensorFlowBenchmark(benchmark_args)
             result = benchmark.run()