Address reviewer feedback

- min torchao version
- remove TODO

docs/source/developer_guides/quantization.md

@@ -204,7 +204,7 @@ model = get_peft_model(base_model, peft_config)
 
 ### Caveats:
 
-- Use the most recent versions of torchao and transformers for best results.
+- Use the most recent versions of torchao (>= v0.4.0) and transformers (> 4.42).
 - Only linear layers are currently supported.
 - `quant_type = "int4_weight_only"` is currently not supported.
 - `NF4` is not implemented in transformers as of yet and is thus also not supported.

src/peft/import_utils.py

@@ -91,4 +91,15 @@ def is_hqq_available():
 
 @lru_cache
 def is_torchao_available():
-    return importlib.util.find_spec("torchao") is not None
+    if importlib.util.find_spec("torchao") is None:
+        return False
+
+    TORCHAO_MINIMUM_VERSION = packaging.version.parse("0.4.0")
+    torchao_version = packaging.version.parse(importlib_metadata.version("torchao"))
+
+    if torchao_version < TORCHAO_MINIMUM_VERSION:
+        raise ImportError(
+            f"Found an incompatible version of auto-gptq. Found version {torchao_version}, "
+            f"but only versions above {TORCHAO_MINIMUM_VERSION} are supported"
+        )
+    return True

src/peft/tuners/lora/torchao.py

@@ -114,7 +114,6 @@ def unmerge(self) -> None:
             del weight
 
     def __repr__(self) -> str:
-        # TODO
         rep = super().__repr__()
         return rep.replace("lora.Linear", f"lora.{self.__class__.__name__}")