From cf4cd53982161cb67c86b69f4a319625b231b0bd Mon Sep 17 00:00:00 2001 From: Reid <61492567+reidliu41@users.noreply.github.com> Date: Sat, 5 Jul 2025 22:24:32 +0800 Subject: [PATCH] [Misc] Add logger.exception for TPU information collection failures (#20510) Signed-off-by: reidliu41 --- vllm/usage/usage_lib.py | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/vllm/usage/usage_lib.py b/vllm/usage/usage_lib.py index c149637635b7..92245498de65 100644 --- a/vllm/usage/usage_lib.py +++ b/vllm/usage/usage_lib.py @@ -20,9 +20,12 @@ import torch import vllm.envs as envs from vllm.connections import global_http_connection +from vllm.logger import init_logger from vllm.utils import cuda_device_count_stateless, cuda_get_device_properties from vllm.version import __version__ as VLLM_VERSION +logger = init_logger(__name__) + _config_home = envs.VLLM_CONFIG_ROOT _USAGE_STATS_JSON_PATH = os.path.join(_config_home, "usage_stats.json") _USAGE_STATS_DO_NOT_TRACK_PATH = os.path.join(_config_home, "do_not_track") @@ -183,7 +186,7 @@ class UsageMessage: self.gpu_memory_per_device = ( torch_xla.core.xla_model.get_memory_info()["bytes_limit"]) except Exception: - pass + logger.exception("Failed to collect TPU information") self.provider = _detect_cloud_provider() self.architecture = platform.machine() self.platform = platform.platform()