FIX Prevent CUDA context initialization due to AWQ (#2230)

Importing from AWQ triggers CUDA context initialization, which can be
problematic in some circumstances (see #1877). This PR moves the import
so that it's local, preventing this issue.

src/peft/tuners/lora/awq.py

@@ -22,10 +22,6 @@
 from peft.tuners.tuners_utils import BaseTunerLayer
 
 
-if is_auto_awq_available():
-    from awq.modules.linear import WQLinear_GEMM
-
-
 class AwqLoraLinear(torch.nn.Module, LoraLayer):
     def __init__(
         self,
@@ -105,18 +101,21 @@ def dispatch_awq(
     else:
         target_base_layer = target
 
-    if is_auto_awq_available() and isinstance(target_base_layer, WQLinear_GEMM):
-        # Raise the error only at the dispatch level
-        AUTOAWQ_MINIMUM_VERSION = packaging.version.parse("0.2.0")
-        version_autoawq = packaging.version.parse(importlib_metadata.version("autoawq"))
+    if is_auto_awq_available():
+        from awq.modules.linear import WQLinear_GEMM
+
+        if isinstance(target_base_layer, WQLinear_GEMM):
+            # Raise the error only at the dispatch level
+            AUTOAWQ_MINIMUM_VERSION = packaging.version.parse("0.2.0")
+            version_autoawq = packaging.version.parse(importlib_metadata.version("autoawq"))
 
-        if AUTOAWQ_MINIMUM_VERSION > version_autoawq:
-            raise ImportError(
-                f"Found an incompatible version of auto-awq. Found version {version_autoawq}, "
-                f"but only versions above {AUTOAWQ_MINIMUM_VERSION} are supported for PEFT."
-            )
+            if AUTOAWQ_MINIMUM_VERSION > version_autoawq:
+                raise ImportError(
+                    f"Found an incompatible version of auto-awq. Found version {version_autoawq}, "
+                    f"but only versions above {AUTOAWQ_MINIMUM_VERSION} are supported for PEFT."
+                )
 
-        new_module = AwqLoraLinear(target, adapter_name, **kwargs)
-        target.qweight = target_base_layer.qweight
+            new_module = AwqLoraLinear(target, adapter_name, **kwargs)
+            target.qweight = target_base_layer.qweight
 
     return new_module