Model: Fix NCCL detection

backends/exllamav3/model.py

@@ -177,7 +177,11 @@ class ExllamaV3Container(BaseModelContainer):
                 self.use_tp = True
                 tp_backend = unwrap(kwargs.get("tensor_parallel_backend"), "native")
 
-                if not exllama_supports_nccl():
+                if tp_backend == "nccl" and not exllama_supports_nccl():
+                    unsupported_message = (
+                        "NCCL is not available. Falling back to native backend."
+                    )
+                    logger.warning(unsupported_message)
                     tp_backend = "native"
 
                 self.tp_backend = tp_backend

backends/exllamav3/utils.py

@@ -1,15 +1,13 @@
 import platform
 from loguru import logger
 
-
 def exllama_supports_nccl():
-    if platform.system() != "Windows":
+    if platform.system() == "Windows":
+        unsupported_message = (
+            "The NCCL tensor parallel backend is not supported on Windows."
+        )
+        logger.warning(unsupported_message)
         return False
 
-    unsupported_message = (
-        "The NCCL tensor parallel backend is not supported on Windows. \n"
-        "Switching to native backend."
-    )
-    logger.warning(unsupported_message)
-
-    return True
+    import torch
+    return torch.cuda.is_available() and torch.distributed.is_nccl_available()