Remove shell=True calls across the repo.

compliance/TEST01/run_verification.py

@@ -104,24 +104,24 @@ def main():
     unixmode_str = unixmode if unixmode == "" else unixmode + " "
 
     # run verify accuracy
-    verify_accuracy_command = (
-        sys.executable + " "
-        + verify_accuracy_binary
-        + " --dtype "
-        + args.dtype
-        + unixmode_str
-        + " -r "
-        + os.path.join(results_dir, "accuracy", "mlperf_log_accuracy.json")
-        + " -t "
-        + os.path.join(compliance_dir, "mlperf_log_accuracy.json")
-    )
+    verify_accuracy_command = [
+        sys.executable 
+        , verify_accuracy_binary
+        , "--dtype"
+        , dtype
+        , unixmode_str
+        , "--reference_accuracy"
+        , os.path.join(results_dir, "accuracy", "mlperf_log_accuracy.json")
+        , "--test_accuracy"
+        , os.path.join(compliance_dir, "mlperf_log_accuracy.json")]
+    verify_accuracy_command = [arg.strip() for arg in verify_accuracy_command if arg.strip() != ""]
+
     try:
         with open("verify_accuracy.txt", "w") as f:
             process = subprocess.Popen(
                 verify_accuracy_command,
                 stdout=subprocess.PIPE,
                 stderr=subprocess.STDOUT,
-                shell=True,
                 text=True
             )
             # Write output to both console and file
@@ -145,24 +145,21 @@ def main():
     verify_performance_binary = os.path.join(
         os.path.dirname(__file__), "verify_performance.py"
     )
-    verify_performance_command = (
-        sys.executable + " "
-        + verify_performance_binary
-        + " -r"
-        + os.path.join(results_dir, "performance",
-                       "run_1", "mlperf_log_detail.txt")
-        + " -t"
-        + os.path.join(compliance_dir, "mlperf_log_detail.txt")
-    )
-
+    verify_performance_command = [
+        sys.executable,
+        verify_performance_binary,
+        "-r",
+        os.path.join(results_dir, "performance", "run_1", "mlperf_log_detail.txt"),
+        "-t",
+        os.path.join(compliance_dir, "mlperf_log_detail.txt")
+    ]
     try:
         with open("verify_performance.txt", "w") as f:
             process = subprocess.Popen(
                 verify_performance_command,
                 stdout=subprocess.PIPE,
                 stderr=subprocess.STDOUT,
                 text=True,
-                shell=True,
             )
             # Write output to both console and file
             for line in process.stdout:

compliance/TEST04/run_verification.py

@@ -57,24 +57,21 @@ def main():
     verify_performance_binary = os.path.join(
         os.path.dirname(__file__), "verify_performance.py"
     )
-    verify_performance_command = (
-        sys.executable + " "
-        + verify_performance_binary
-        + " -r"
-        + os.path.join(results_dir, "performance",
-                       "run_1", "mlperf_log_summary.txt")
-        + " -t"
-        + os.path.join(compliance_dir, "mlperf_log_summary.txt")
-    )
-
+    verify_performance_command = [
+        sys.executable,
+        verify_performance_binary,
+        "-r",
+        os.path.join(results_dir, "performance", "run_1", "mlperf_log_summary.txt"),
+        "-t",
+        os.path.join(compliance_dir, "mlperf_log_summary.txt")
+    ]
     try:
         with open("verify_performance.txt", "w") as f:
             process = subprocess.Popen(
                 verify_performance_command,
                 stdout=subprocess.PIPE,  # capture output
                 stderr=subprocess.STDOUT,
                 text=True,  # decode output as text
-                shell=True,
             )
             # Write output to both console and file
             for line in process.stdout:

language/bert/accuracy-squad.py

@@ -514,14 +514,19 @@ def append_feature(feature):
     )
 
     print("Evaluating predictions...")
-    cmd = "python3 {:}/evaluate_v1.1.py {:} {:} {}".format(
-        os.path.dirname(os.path.abspath(__file__)),
+    cmd = [
+        "python3",
+        os.path.join(
+            os.path.dirname(os.path.abspath(__file__)),
+            "evaluate_v1.1.py"
+        ),
         args.val_data,
         args.out_file,
-        "--max_examples {}".format(
-            args.max_examples) if args.max_examples else "",
-    )
-    subprocess.check_call(cmd, shell=True)
+        "--max_examples"if args.max_examples else "",
+        str(args.max_examples) if args.max_examples else ""
+    ]
+    cmd = [arg for arg in cmd if arg.strip() != ""]
+    subprocess.check_call(cmd)
 
 
 if __name__ == "__main__":

language/bert/run.py

@@ -99,7 +99,6 @@ def get_args():
 
 def main():
     args = get_args()
-
     sut = None
 
     if not args.network or args.network == "sut":
@@ -199,15 +198,19 @@ def main():
             sut.sut, sut.qsl.qsl, settings, log_settings, args.audit_conf
         )
         if args.accuracy and not os.environ.get("SKIP_VERIFY_ACCURACY"):
-            cmd = "python3 {:}/accuracy-squad.py {}".format(
-                os.path.dirname(os.path.abspath(__file__)),
-                (
-                    "--max_examples {}".format(args.max_examples)
-                    if args.max_examples
-                    else ""
+            cmd = [
+                "python3",
+                os.path.join(
+                    os.path.dirname(os.path.abspath(__file__)), "accuracy-squad.py"
                 ),
-            )
-            subprocess.check_call(cmd, shell=True)
+                "--max_examples" if args.max_examples
+                else "",
+                args.max_examples
+                if args.max_examples
+                else "",
+            ]
+            cmd = [arg for arg in cmd if arg.strip() != ""]
+            subprocess.check_call(cmd)
 
     print("Done!")
 

retired_benchmarks/speech_recognition/rnnt/run.py

@@ -115,9 +115,18 @@ def main():
     )
 
     if args.accuracy:
-        cmd = f"python3 accuracy_eval.py --log_dir {log_path} --dataset_dir {args.dataset_dir} --manifest {args.manifest}"
+        cmd = [
+            "python3",
+            "accuracy_eval.py",
+            "--log_dir",
+            log_path,
+            "--dataset_dir",
+            args.dataset_dir,
+            "--manifest",
+            args.manifest,
+        ]
         print(f"Running accuracy script: {cmd}")
-        subprocess.check_call(cmd, shell=True)
+        subprocess.check_call(cmd)
 
     print("Done!")
 

retired_benchmarks/translation/gnmt/tensorflow/run_task.py

@@ -88,44 +88,25 @@
 
 outpath = os.path.join(args.output_path, "output", "console_out_gnmt.txt")
 
-cmd = (
-    "python -m nmt.nmt \
-    --src=en --tgt=de \
-    --ckpt="
-    + cpk_path
-    + " \
-    --hparams_path="
-    + haparams_path
-    + " \
-    --out_dir="
-    + out_dir
-    + " \
-    --vocab_prefix="
-    + vocab_prefix
-    + " \
-    --inference_input_file="
-    + inference_input_file
-    + " \
-    --inference_output_file="
-    + inference_output_file
-    + " \
-    --inference_ref_file="
-    + inference_ref_file
-    + " \
-    --infer_batch_size="
-    + args.batch_size
-    + " \
-    --num_inter_threads="
-    + args.num_inter_threads
-    + " \
-    --num_intra_threads="
-    + args.num_intra_threads
-    + " \
-    --iterations="
-    + str(iterations)
-    + " \
-    --run="
-    + args.run
-)
 
-return_code = subprocess.call(cmd, shell=True)
+cmd = [
+    "python",
+    "-m",
+    "nmt.nmt",      
+    "--src=en",
+    "--tgt=de",
+    "--ckpt={}".format(cpk_path),
+    "--hparams_path={}".format(haparams_path),
+    "--out_dir={}".format(out_dir),
+    "--vocab_prefix={}".format(vocab_prefix),
+    "--inference_input_file={}".format(inference_input_file),
+    "--inference_output_file={}".format(inference_output_file),
+    "--inference_ref_file={}".format(inference_ref_file),
+    "--infer_batch_size={}".format(args.batch_size),
+    "--num_inter_threads={}".format(args.num_inter_threads),
+    "--num_intra_threads={}".format(args.num_intra_threads),
+    "--iterations={}".format(iterations),
+    "--run={}".format(args.run)
+]
+
+return_code = subprocess.call(cmd)

text_to_video/wan-2.2-t2v-a14b/run_mlperf.py

@@ -46,7 +46,8 @@ def load_prompts(dataset_path):
 
 
 class Model:
-    def __init__(self, model_path, device, config, prompts, fixed_latent=None, rank=0):
+    def __init__(self, model_path, device, config,
+                 prompts, fixed_latent=None, rank=0):
         self.device = device
         self.rank = rank
         self.height = config["height"]
@@ -106,7 +107,8 @@ def flush_queries(self):
 
 
 class DebugModel:
-    def __init__(self, model_path, device, config, prompts, fixed_latent=None, rank=0):
+    def __init__(self, model_path, device, config,
+                 prompts, fixed_latent=None, rank=0):
         self.prompts = prompts
 
     def issue_queries(self, query_samples):
@@ -186,7 +188,8 @@ def get_args():
     parser.add_argument(
         "--scenario",
         default="SingleStream",
-        help="mlperf benchmark scenario, one of " + str(list(SCENARIO_MAP.keys())),
+        help="mlperf benchmark scenario, one of " +
+        str(list(SCENARIO_MAP.keys())),
     )
     parser.add_argument(
         "--user_conf",
@@ -202,7 +205,10 @@ def get_args():
         help="performance sample count",
         default=5000,
     )
-    parser.add_argument("--accuracy", action="store_true", help="enable accuracy pass")
+    parser.add_argument(
+        "--accuracy",
+        action="store_true",
+        help="enable accuracy pass")
     # Dont overwrite these for official submission
     parser.add_argument("--count", type=int, help="dataset items to use")
     parser.add_argument("--time", type=int, help="time to scan in seconds")
@@ -271,7 +277,10 @@ def run_mlperf(args, config):
 
         audit_config = os.path.abspath(args.audit_conf)
         if os.path.exists(audit_config):
-            settings.FromConfig(audit_config, "wan-2.2-t2v-a14b", args.scenario)
+            settings.FromConfig(
+                audit_config,
+                "wan-2.2-t2v-a14b",
+                args.scenario)
         settings.scenario = SCENARIO_MAP[args.scenario]
 
         settings.mode = lg.TestMode.PerformanceOnly
@@ -297,8 +306,10 @@ def run_mlperf(args, config):
         if args.samples_per_query:
             settings.multi_stream_samples_per_query = args.samples_per_query
         if args.max_latency:
-            settings.server_target_latency_ns = int(args.max_latency * NANO_SEC)
-            settings.multi_stream_expected_latency_ns = int(args.max_latency * NANO_SEC)
+            settings.server_target_latency_ns = int(
+                args.max_latency * NANO_SEC)
+            settings.multi_stream_expected_latency_ns = int(
+                args.max_latency * NANO_SEC)
 
         performance_sample_count = (
             args.performance_sample_count
@@ -311,7 +322,8 @@ def run_mlperf(args, config):
             count, performance_sample_count, load_query_samples, unload_query_samples
         )
 
-        lg.StartTestWithLogSettings(sut, qsl, settings, log_settings, audit_config)
+        lg.StartTestWithLogSettings(
+            sut, qsl, settings, log_settings, audit_config)
 
         lg.DestroyQSL(qsl)
         lg.DestroySUT(sut)

tools/submission/generate_final_report.py

@@ -101,11 +101,11 @@ def main():
         "singlestream": "SingleStream",
         "multistream": "MultiStream",
         "server": "Server",
-        "interactive":"Interactive",
+        "interactive": "Interactive",
         "offline": "Offline",
     }
-    df["Scenario"] = df["Scenario"].apply(lambda x: scenario_map.get(str(x).lower(), x))
-
+    df["Scenario"] = df["Scenario"].apply(
+        lambda x: scenario_map.get(str(x).lower(), x))
 
     output = args.input[:-4]
     writer = pd.ExcelWriter(output + ".xlsx", engine="xlsxwriter")