Make the server random by default (#493)

docs/hyperparameter_tuning.md

@@ -22,10 +22,13 @@ On the other hand, if you see `token usage` very high and you frequently see war
 ### Tune `--dp-size` and `--tp-size`
 Data parallelism is better for throughput. When there is enough GPU memory, always favor data parallelism for throughput.
 
+### (Minor) Tune `--max-prefill-tokens`, `--mem-fraction-static`, `--max-running-requests`.
+If you see out of memory (OOM) errors, you can decrease these parameters.
+If OOM happens during prefill, try to decrease `--max-prefill-tokens`.
+If OOM happens during decoding, try to decrease `--max-running-requests`.
+You can also try to decrease `--mem-fraction-static`, which reduces the memory usage of the KV cache memory pool and helps both prefill and decoding.
+
 ### (Minor) Tune `--schedule-heuristic`
 If you have many shared prefixes, use the default `--schedule-heuristic lpm`. `lpm` stands for longest prefix match.
 When you have no shared prefixes at all or you always send the requests with the shared prefixes together,
 you can try `--schedule-heuristic fcfs`. `fcfs` stands for first come first serve.
-
-### (Minor) Tune `--max-prefill-tokens`, `--mem-fraction-static`, `--max-running-requests`.
-If you see out of memory errors, you can decrease them. Otherwise, the default value should work well.

python/sglang/srt/server_args.py

@@ -2,6 +2,7 @@
 
 import argparse
 import dataclasses
+import random
 from typing import List, Optional, Union
 
 
@@ -32,7 +33,7 @@ class ServerArgs:
     # Other runtime options
     tp_size: int = 1
     stream_interval: int = 8
-    random_seed: int = 42
+    random_seed: Optional[int] = None
 
     # Logging
     log_level: str = "info"
@@ -72,6 +73,9 @@ class ServerArgs:
         elif self.additional_ports is None:
             self.additional_ports = []
 
+        if self.random_seed is None:
+            self.random_seed = random.randint(0, 1 << 30)
+
     @staticmethod
     def add_cli_args(parser: argparse.ArgumentParser):
         parser.add_argument(