A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 23:30 UTC
[MISC] tiny fixes (#13378)
Mengqing Cao committed
Feb 17, 2025 at 16:57 UTC
238dfc8ac3f9befb594e1ad2f249616afc68d484
2 files changed
+2
-2
vllm/executor/executor_base.py
+1
-1
@@ -109,7 +109,7 @@ class ExecutorBase(ABC):
109
"""
110
# NOTE: This is logged in the executor because there can be >1 workers.
111
logger.info("# %s blocks: %d, # CPU blocks: %d",
112
- vllm.platforms.current_platform.dispatch_key,
112
+ vllm.platforms.current_platform.device_name,
113
num_gpu_blocks, num_cpu_blocks)
114
max_concurrency = (num_gpu_blocks * self.cache_config.block_size /
115
self.model_config.max_model_len)
vllm/platforms/interface.py
+1
-1
@@ -327,7 +327,7 @@ class Platform:
327
"""
328
Get device specific communicator class for distributed communication.
329
"""
330
- return "vllm.distributed.device_communicator.base_device_communicator.DeviceCommunicatorBase" # noqa
330
+ return "vllm.distributed.device_communicators.base_device_communicator.DeviceCommunicatorBase" # noqa
331
332
333
class UnspecifiedPlatform(Platform):