wassname/vllm
1
0
Fork 0
mirror of https://github.com/wassname/vllm.git synced 2026-09-09 11:40:01 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Core] Optimize Async + Multi-step (#8050)

Browse Source
This commit is contained in:
Alexander Matveev
2024-09-03 18:50:29 +00:00
committed by GitHub
parent 95a178f861
commit 6d646d08a2
8 changed files with 325 additions and 247 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/multi_step/test_correctness_async_llm.py
+2 -2
View File
@@ -103,13 +103,13 @@ async def test_multi_step(
model,
server_args + distributed_args,
num_logprobs,
max_wait_seconds=3 * 240)
max_wait_seconds=5 * 240)
test_completions = await completions_with_server_args(
prompts,
model,
ms_server_args + distributed_args,
num_logprobs,
max_wait_seconds=3 * 240)
max_wait_seconds=5 * 240)
# Assert multi-step scheduling produces identical tokens
# to single-step scheduling.