Disable flaky eagle tests (#5753)

README.md

@@ -71,5 +71,5 @@ It is supported by the following institutions: AMD, Atlas Cloud, Baseten, Cursor
 
 For enterprises interested in adopting or deploying SGLang at scale, including technical consulting, sponsorship opportunities, or partnership inquiries, please contact us at contact@sglang.ai.
 
-## Acknowledgment and Citation
-We learned the design and reused code from the following projects: [Guidance](https://github.com/guidance-ai/guidance), [vLLM](https://github.com/vllm-project/vllm), [LightLLM](https://github.com/ModelTC/lightllm), [FlashInfer](https://github.com/flashinfer-ai/flashinfer), [Outlines](https://github.com/outlines-dev/outlines), and [LMQL](https://github.com/eth-sri/lmql). Please cite the paper, [SGLang: Efficient Execution of Structured Language Model Programs](https://arxiv.org/abs/2312.07104), if you find the project useful.
+## Acknowledgment
+We learned the design and reused code from the following projects: [Guidance](https://github.com/guidance-ai/guidance), [vLLM](https://github.com/vllm-project/vllm), [LightLLM](https://github.com/ModelTC/lightllm), [FlashInfer](https://github.com/flashinfer-ai/flashinfer), [Outlines](https://github.com/outlines-dev/outlines), and [LMQL](https://github.com/eth-sri/lmql).

python/sglang/srt/model_executor/cuda_graph_runner.py

@@ -279,9 +279,9 @@ class CudaGraphRunner:
                 f"Capture cuda graph failed: {e}\n"
                 "Possible solutions:\n"
                 "1. set --mem-fraction-static to a smaller value (e.g., 0.8 or 0.7)\n"
-                "2. set --cuda-graph-max-bs to a smaller value (e.g., 32)\n"
+                "2. set --cuda-graph-max-bs to a smaller value (e.g., 16)\n"
                 "3. disable torch compile by not using --enable-torch-compile\n"
-                "4. disable cuda graph by --disable-cuda-graph\n"
+                "4. disable cuda graph by --disable-cuda-graph. (Not recommonded. Huge perf loss)\n"
                 "Open an issue on GitHub https://github.com/sgl-project/sglang/issues/new/choose \n"
             )
 

python/sglang/srt/model_executor/model_runner.py

@@ -955,12 +955,6 @@ class ModelRunner:
             return
 
         if self.server_args.disable_cuda_graph:
-            logger.warning(
-                "\n\nCUDA Graph is DISABLED.\n"
-                "This will cause significant performance degradation.\n"
-                "CUDA Graph should almost never be disabled in most usage scenarios.\n"
-                "If you encounter OOM issues, please try setting --mem-fraction-static to a lower value (such as 0.8 or 0.7) instead of disabling CUDA Graph.\n"
-            )
             return
 
         tic = time.time()

python/sglang/srt/speculative/eagle_draft_cuda_graph_runner.py

@@ -85,9 +85,9 @@ class EAGLEDraftCudaGraphRunner:
                 f"Capture cuda graph failed: {e}\n"
                 "Possible solutions:\n"
                 "1. set --mem-fraction-static to a smaller value (e.g., 0.8 or 0.7)\n"
-                "2. disable torch compile by not using --enable-torch-compile\n"
-                "3. specify --dtype to the same dtype (e.g. bfloat16)\n"
-                "4. disable cuda graph by --disable-cuda-graph\n"
+                "2. set --cuda-graph-max-bs to a smaller value (e.g., 16)\n"
+                "3. disable torch compile by not using --enable-torch-compile\n"
+                "4. disable cuda graph by --disable-cuda-graph. (Not recommonded. Huge perf loss)\n"
                 "Open an issue on GitHub https://github.com/sgl-project/sglang/issues/new/choose \n"
             )
 

test/srt/test_eagle_infer.py

@@ -1,5 +1,4 @@
 import json
-import multiprocessing as mp
 import os
 import random
 import threading
@@ -8,7 +7,6 @@ import unittest
 from concurrent.futures import ThreadPoolExecutor
 from functools import partial
 from types import SimpleNamespace
-from typing import List, Optional
 
 import numpy as np
 import requests
@@ -18,7 +16,6 @@ import sglang as sgl
 from sglang.srt.hf_transformers_utils import get_tokenizer
 from sglang.srt.utils import kill_process_tree
 from sglang.test.few_shot_gsm8k import run_eval
-from sglang.test.runners import DEFAULT_PROMPTS, SRTRunner
 from sglang.test.test_utils import (
     DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST,
     DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST,
@@ -541,36 +538,5 @@ class TestEAGLEServerTriton(TestEAGLEServer):
         )
 
 
-class TestEAGLEServerPageSize(TestEAGLEServer):
-    @classmethod
-    def setUpClass(cls):
-        cls.base_url = DEFAULT_URL_FOR_TEST
-        cls.process = popen_launch_server(
-            DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST,
-            cls.base_url,
-            timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
-            other_args=[
-                "--speculative-algorithm",
-                "EAGLE",
-                "--speculative-draft-model-path",
-                DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST,
-                "--speculative-num-steps",
-                5,
-                "--speculative-eagle-topk",
-                1,
-                "--speculative-num-draft-tokens",
-                6,
-                "--mem-fraction-static",
-                0.7,
-                "--chunked-prefill-size",
-                128,
-                "--max-running-requests",
-                8,
-                "--page-size",
-                8,
-            ],
-        )
-
-
 if __name__ == "__main__":
     unittest.main()