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, 02:15 UTC
[XPU] XCCL support enabled in torch 2.8.0.dev nightly builds (#20705)
Signed-off-by: ratnampa <ratnam.parikh@intel.com>
Ratnam Parikh committed
Jul 10, 2025 at 20:39 UTC
35514b682ac737505209eca21a26df86293ed5fb
1 file changed
+2
-2
vllm/utils/__init__.py
+2
-2
@@ -2085,10 +2085,10 @@ def supports_dynamo() -> bool:
2085
return base_torch_version >= Version("2.4.0")
2086
2087
2088
-# Supports xccl with PyTorch versions >= 2.8.0 for XPU platform
2088
+# Supports xccl with PyTorch versions >= 2.8.0.dev for XPU platform
2089
def supports_xccl() -> bool:
2090
return is_torch_equal_or_newer(
2091
- "2.8.0") and torch.distributed.is_xccl_available()
2091
+ "2.8.0.dev") and torch.distributed.is_xccl_available()
2092
2093
2094
# Some backends use pytorch version < 2.4.0 which doesn't