1movmedia/pyworker
2
0
Fork 0
Code Pull Requests Activity

Lowered concurrency of vLLM and TGI benchmarks

Browse Source
This commit is contained in:
Lucas Armand
2025-12-17 11:55:33 -08:00
parent bcb04b9a32
commit e02f4bc943
2 changed files with 4 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
workers/openai/worker.py
+2 -2
View File
@@ -64,8 +64,8 @@ worker_config = WorkerConfig(
max_queue_time=600.0, max_queue_time=600.0,
benchmark_config=BenchmarkConfig( benchmark_config=BenchmarkConfig(
generator=completions_benchmark_generator, generator=completions_benchmark_generator,
concurrency=100, concurrency=10,
runs=2 runs=3
) )
), ),
HandlerConfig( HandlerConfig(
workers/tgi/worker.py
+2 -1
View File
@@ -55,7 +55,8 @@ worker_config = WorkerConfig(
max_queue_time=600.0, max_queue_time=600.0,
benchmark_config=BenchmarkConfig( benchmark_config=BenchmarkConfig(
generator=benchmark_generator, generator=benchmark_generator,
concurrency=50 concurrency=10,
runs=3
), ),
workload_calculator= lambda x: x["parameters"]["max_new_tokens"] workload_calculator= lambda x: x["parameters"]["max_new_tokens"]
), ),