diff --git a/docs/references/performance_analysis_and_optimization.rst b/docs/references/performance_analysis_and_optimization.rst index 1d70fb51d..76db62df7 100644 --- a/docs/references/performance_analysis_and_optimization.rst +++ b/docs/references/performance_analysis_and_optimization.rst @@ -4,4 +4,4 @@ Performance Analysis & Optimization :maxdepth: 1 benchmark_and_profiling.md - accuracy_evaluation.md \ No newline at end of file + accuracy_evaluation.md diff --git a/python/sglang/bench_serving.py b/python/sglang/bench_serving.py index eea4d5384..343af051a 100644 --- a/python/sglang/bench_serving.py +++ b/python/sglang/bench_serving.py @@ -1304,14 +1304,12 @@ async def benchmark( if "sglang" in backend: server_info = requests.get(base_url + "/get_server_info") if server_info.status_code == 200: - if pd_separated: - accept_length = server_info.json()["decode"][0]["internal_states"][ - 0 - ].get("avg_spec_accept_length", None) - else: - accept_length = server_info.json()["internal_states"][0].get( - "avg_spec_accept_length", None - ) + server_info_json = server_info.json() + if "decode" in server_info_json: + server_info_json = server_info_json["decode"][0] + accept_length = server_info_json["internal_states"][0].get( + "avg_spec_accept_length", None + ) else: accept_length = None else: diff --git a/scripts/playground/disaggregation/cli-logprob.py b/scripts/playground/disaggregation/cli-logprob.py index 2dcfd3d4e..4c69a055b 100644 --- a/scripts/playground/disaggregation/cli-logprob.py +++ b/scripts/playground/disaggregation/cli-logprob.py @@ -1,4 +1,4 @@ -prompt = "The capital of taiwan is " +prompt = "The capital of france is " import json