refactor mode -> scenario

Author: v-shobhit

language/gpt-oss/README.md

@@ -39,7 +39,7 @@ Then, run a benchmark script that uses the client to send/recv requests.
 ### Run the inference
 ```bash
 python3 run_mlperf.py --help
-usage: run_mlperf.py [-h] [--mode {offline,server}] --input-file INPUT_FILE [--max-samples MAX_SAMPLES] [--mlperf-conf MLPERF_CONF]
+usage: run_mlperf.py [-h] [--scenario {offline,server}] --input-file INPUT_FILE [--max-samples MAX_SAMPLES] [--mlperf-conf MLPERF_CONF]
                      [--user-conf USER_CONF] [--accuracy] [--output-dir OUTPUT_DIR] [--backend {sglang}] [--server-url SERVER_URL]
                      [--generation-config GENERATION_CONFIG] [--max-new-tokens MAX_NEW_TOKENS] [--num-workers NUM_WORKERS]
                      [--max-concurrency MAX_CONCURRENCY]
@@ -48,7 +48,7 @@ Run MLPerf inference benchmarks for gpt-oss
 
 options:
   -h, --help            show this help message and exit
-  --mode {offline,server}
+  --scenario {offline,server}
                         MLPerf scenario mode
   --input-file INPUT_FILE
                         Path to tokenized dataset (pickle file)

language/gpt-oss/run_mlperf.py

@@ -6,13 +6,13 @@
 
 Usage:
     # Offline scenario (performance)
-    python run_mlperf.py --mode offline --input-file data/accuracy_eval_tokenized.pkl
+    python run_mlperf.py --scenario offline --input-file data/accuracy_eval_tokenized.pkl
 
     # Server scenario (performance)
-    python run_mlperf.py --mode server --input-file data/accuracy_eval_tokenized.pkl
+    python run_mlperf.py --scenario server --input-file data/accuracy_eval_tokenized.pkl
 
     # Accuracy mode
-    python run_mlperf.py --mode offline --accuracy --input-file data/accuracy_eval_tokenized.pkl
+    python run_mlperf.py --scenario offline --accuracy --input-file data/accuracy_eval_tokenized.pkl
 """
 
 import argparse
@@ -71,11 +71,11 @@ def create_argument_parser() -> argparse.ArgumentParser:
 
     # Scenario selection
     parser.add_argument(
-        "--mode",
+        "--scenario",
         type=str,
         default="offline",
         choices=["offline", "server"],
-        help="MLPerf scenario mode"
+        help="MLPerf scenario (offline or server)"
     )
 
     # Dataset
@@ -296,15 +296,15 @@ def do_cleanup():
     try:
         # Create output directories
         output_dir = Path(args.output_dir)
-        log_dir = output_dir / args.mode / \
+        log_dir = output_dir / args.scenario / \
             ("accuracy" if args.accuracy else "performance")
         log_dir.mkdir(parents=True, exist_ok=True)
 
         logger.info("=" * 80)
         logger.info("MLPerf Inference Benchmark Runner for GPT-OSS")
         logger.info("=" * 80)
         logger.info(f"Backend: {args.backend}")
-        logger.info(f"Mode: {args.mode}")
+        logger.info(f"Scenario: {args.scenario}")
         logger.info(f"Accuracy: {args.accuracy}")
         logger.info(f"Input file: {args.input_file}")
         logger.info(f"Output directory: {log_dir}")
@@ -370,7 +370,7 @@ def do_cleanup():
         # - Server: Incremented as queries arrive
         pbar = tqdm(
             total=0,  # Will be updated dynamically by SUT
-            desc=f"MLPerf {args.mode}",
+            desc=f"MLPerf {args.scenario}",
             unit="query",
             leave=True,
             position=0,
@@ -381,8 +381,8 @@ def do_cleanup():
         )
 
         # Create SUT with progress bar
-        logger.debug(f"Creating {args.mode} SUT...")
-        if args.mode == "offline":
+        logger.debug(f"Creating {args.scenario} SUT...")
+        if args.scenario == "offline":
             sut = OfflineSUT(
                 backend=backend,
                 dataset=prompts,
@@ -419,7 +419,7 @@ def do_cleanup():
 
         # Configure LoadGen
         settings = configure_loadgen(
-            scenario=args.mode,
+            scenario=args.scenario,
             accuracy_mode=args.accuracy,
             mlperf_conf=args.mlperf_conf,
             user_conf=args.user_conf,