xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-04-08 20:07:09 +08:00
Code Issues Packages Projects Releases Wiki Activity

[ci] set timeout for test_oot_registration.py (#7082)

Browse Source
This commit is contained in:
youkaichao 2024-08-02 10:03:24 -07:00 committed by GitHub
parent c16eaac500
commit 806949514a
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
3 changed files with 10 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
tests/entrypoints/openai/test_oot_registration.py
View File

@ -36,10 +36,12 @@ def test_oot_registration_for_api_server():
ctx = torch.multiprocessing.get_context()
server = ctx.Process(target=server_function, args=(port, ))
server.start()
MAX_SERVER_START_WAIT_S = 60
client = OpenAI(
base_url=f"http://localhost:{port}/v1",
api_key="token-abc123",
)
now = time.time()
while True:
try:
completion = client.chat.completions.create(
@ -57,6 +59,8 @@ def test_oot_registration_for_api_server():
except OpenAIError as e:
if "Connection error" in str(e):
time.sleep(3)
if time.time() - now > MAX_SERVER_START_WAIT_S:
raise RuntimeError("Server did not start in time") from e
else:
raise e
server.kill()

4
vllm/worker/worker.py
View File

@ -186,7 +186,9 @@ class Worker(LocalOrDistributedWorkerBase):
# GPU did not change their memory usage during the profiling.
peak_memory = self.init_gpu_memory - free_gpu_memory
assert peak_memory > 0, (
"Error in memory profiling. This happens when the GPU memory was "
"Error in memory profiling. "
f"Initial free memory {self.init_gpu_memory}, current free memory"
f" {free_gpu_memory}. This happens when the GPU memory was "
"not properly cleaned up before initializing the vLLM instance.")
cache_block_size = self.get_cache_block_size_bytes()

4
vllm/worker/xpu_worker.py
View File

@ -138,7 +138,9 @@ class XPUWorker(LoraNotSupportedWorkerBase, Worker):
# GPU did not change their memory usage during the profiling.
peak_memory = self.init_gpu_memory - free_gpu_memory
assert peak_memory > 0, (
"Error in memory profiling. This happens when the GPU memory was "
"Error in memory profiling. "
f"Initial free memory {self.init_gpu_memory}, current free memory"
f" {free_gpu_memory}. This happens when the GPU memory was "
"not properly cleaned up before initializing the vLLM instance.")
cache_block_size = self.get_cache_block_size_bytes()