From 268c3250787729198a0d86378a36e81d4606a2af Mon Sep 17 00:00:00 2001
From: WWW <jadewangcn@outlook.com>
Date: Fri, 11 Apr 2025 06:31:17 +0800
Subject: [PATCH] Fix range_ratio Bug in RandomDataset (#16126)

Signed-off-by: jadewang21 <jadewangcn@outlook.com>
---
 benchmarks/benchmark_dataset.py    | 23 ++++++++++++++++++-----
 benchmarks/benchmark_serving.py    | 19 ++++++++++++-------
 benchmarks/benchmark_throughput.py | 20 ++++++++++++--------
 3 files changed, 42 insertions(+), 20 deletions(-)

diff --git a/benchmarks/benchmark_dataset.py b/benchmarks/benchmark_dataset.py
index c6630800..8bb7cfe0 100644
--- a/benchmarks/benchmark_dataset.py
+++ b/benchmarks/benchmark_dataset.py
@@ -288,7 +288,7 @@ def process_image(image: Any) -> Mapping[str, Any]:
 class RandomDataset(BenchmarkDataset):
     # Default values copied from benchmark_serving.py for the random dataset.
     DEFAULT_PREFIX_LEN = 0
-    DEFAULT_RANGE_RATIO = 1.0
+    DEFAULT_RANGE_RATIO = 0.0
     DEFAULT_INPUT_LEN = 1024
     DEFAULT_OUTPUT_LEN = 128
 
@@ -308,19 +308,32 @@ class RandomDataset(BenchmarkDataset):
         output_len: int = DEFAULT_OUTPUT_LEN,
         **kwargs,
     ) -> list[SampleRequest]:
+        # Enforce range_ratio < 1
+        assert range_ratio < 1.0, (
+            "random_range_ratio must be < 1.0 to ensure a valid sampling range"
+        )
+
         vocab_size = tokenizer.vocab_size
 
         prefix_token_ids = (np.random.randint(
             0, vocab_size, size=prefix_len).tolist() if prefix_len > 0 else [])
 
-        input_low = int(input_len * range_ratio)
-        output_low = int(output_len * range_ratio)
+        # New sampling logic: [X * (1 - b), X * (1 + b)]
+        input_low = int(input_len * (1 - range_ratio))
+        input_high = int(input_len * (1 + range_ratio))
+        output_low = int(output_len * (1 - range_ratio))
+        output_high = int(output_len * (1 + range_ratio))
+
+        # Add logging for debugging
+        logger.info("Sampling input_len from [%s, %s]", input_low, input_high)
+        logger.info("Sampling output_len from [%s, %s]", output_low,
+                    output_high)
 
         input_lens = np.random.randint(input_low,
-                                       input_len + 1,
+                                       input_high + 1,
                                        size=num_requests)
         output_lens = np.random.randint(output_low,
-                                        output_len + 1,
+                                        output_high + 1,
                                         size=num_requests)
         offsets = np.random.randint(0, vocab_size, size=num_requests)
 
diff --git a/benchmarks/benchmark_serving.py b/benchmarks/benchmark_serving.py
index 0f061f89..b5bd840d 100644
--- a/benchmarks/benchmark_serving.py
+++ b/benchmarks/benchmark_serving.py
@@ -996,18 +996,23 @@ if __name__ == "__main__":
     random_group.add_argument(
         "--random-range-ratio",
         type=float,
-        default=1.0,
-        help="Range of sampled ratio of input/output length, "
-        "used only for random sampling.",
+        default=0.0,
+        help="Range ratio for sampling input/output length, "
+        "used only for random sampling. Must be in the range [0, 1) to define "
+        "a symmetric sampling range"
+        "[length * (1 - range_ratio), length * (1 + range_ratio)].",
     )
     random_group.add_argument(
         "--random-prefix-len",
         type=int,
         default=0,
-        help="Number of fixed prefix tokens before random "
-        " context. The length range of context in a random "
-        " request is [random-prefix-len, "
-        " random-prefix-len + random-prefix-len * random-range-ratio).")
+        help=("Number of fixed prefix tokens before the random context "
+              "in a request. "
+              "The total input length is the sum of `random-prefix-len` and "
+              "a random "
+              "context length sampled from [input_len * (1 - range_ratio), "
+              "input_len * (1 + range_ratio)]."),
+    )
 
     hf_group = parser.add_argument_group("hf dataset options")
     hf_group.add_argument("--hf-subset",
diff --git a/benchmarks/benchmark_throughput.py b/benchmarks/benchmark_throughput.py
index 33a31e62..39e4a34b 100644
--- a/benchmarks/benchmark_throughput.py
+++ b/benchmarks/benchmark_throughput.py
@@ -594,18 +594,22 @@ if __name__ == "__main__":
         default=None,
         help="Path to the lora adapters to use. This can be an absolute path, "
         "a relative path, or a Hugging Face model identifier.")
-    parser.add_argument("--prefix-len",
-                        type=int,
-                        default=None,
-                        help="Number of prefix tokens per request."
-                        "This is for the RandomDataset and SonnetDataset")
+    parser.add_argument(
+        "--prefix-len",
+        type=int,
+        default=0,
+        help="Number of fixed prefix tokens before the random "
+        "context in a request (default: 0).",
+    )
     # random dataset
     parser.add_argument(
         "--random-range-ratio",
         type=float,
-        default=None,
-        help="Range of sampled ratio of input/output length, "
-        "used only for RandomDataSet.",
+        default=0.0,
+        help="Range ratio for sampling input/output length, "
+        "used only for RandomDataset. Must be in the range [0, 1) to define "
+        "a symmetric sampling range "
+        "[length * (1 - range_ratio), length * (1 + range_ratio)].",
     )
 
     # hf dtaset