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, 19:45 UTC
OpenVINO: added CPU-like conditions (#14338)
Signed-off-by: Ilya Lavrenov <ilya.lavrenov@intel.com>
Ilya Lavrenov committed
Mar 7, 2025 at 10:24 UTC
8ca7a71df787ad711ad3ac70a5bd2eb2bb398938
2 files changed
+3
-2
tests/conftest.py
+2
-1
@@ -257,7 +257,8 @@ class HfRunner:
257
return x
258
259
if device is None:
260
- device = "cpu" if current_platform.is_cpu() else "cuda"
260
+ device = "cpu" if current_platform.is_cpu(
261
+ ) or current_platform.is_openvino() else "cuda"
262
263
if isinstance(x, dict):
264
return {k: self.wrap_device(v, device) for k, v in x.items()}
tests/utils.py
+1
-1
@@ -698,7 +698,7 @@ def large_gpu_mark(min_gb: int) -> pytest.MarkDecorator:
698
without enough resources, or called when filtering tests to run directly.
699
"""
700
try:
701
- if current_platform.is_cpu():
701
+ if current_platform.is_cpu() or current_platform.is_openvino():
702
memory_gb = 0
703
else:
704
memory_gb = current_platform.get_device_total_memory() / GB_bytes