xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-10 00:06:06 +08:00
Code Issues Packages Projects Releases Wiki Activity

[CI/Build] Fix test_defaults_with_usage_context in AMD CI (#27926)

Browse Source 
Signed-off-by: zhewenli <zhewenli@meta.com>
This commit is contained in:
Zhewen Li 2025-11-05 15:40:24 -08:00 committed by GitHub
parent 5ee93a5956
commit 0b8e871e5e
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 4 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
tests/v1/engine/test_engine_args.py
View File

@ -53,10 +53,12 @@ def test_defaults_with_usage_context():
vllm_config: VllmConfig = engine_args.create_engine_config(UsageContext.LLM_CLASS)
from vllm.platforms import current_platform
from vllm.utils.mem_constants import GiB_bytes
device_memory = current_platform.get_device_total_memory()
device_name = current_platform.get_device_name().lower()
if "h100" in device_name or "h200" in device_name:
# For H100 and H200, we use larger default values.
if device_memory >= 70 * GiB_bytes and "a100" not in device_name:
# For GPUs like H100, H200, and MI300x with >= 70GB memory
default_llm_tokens = 16384
default_server_tokens = 8192
default_max_num_seqs = 1024