Enable gptqmodel

src/transformers/quantizers/quantizer_gptq.py

@@ -22,7 +22,7 @@
 if TYPE_CHECKING:
     from ..modeling_utils import PreTrainedModel
 
-from ..utils import is_auto_gptq_available, is_optimum_available, is_torch_available, logging
+from ..utils import is_auto_gptq_available, is_gptqmodel_available, is_optimum_available, is_torch_available, logging
 from ..utils.quantization_config import GPTQConfig, QuantizationConfigMixin
 
 
@@ -35,11 +35,11 @@
 class GptqHfQuantizer(HfQuantizer):
     """
     Quantizer of the GPTQ method - for GPTQ the quantizer support calibration of the model through
-    `auto_gptq` package. Quantization is done under the hood for users if they load a non-prequantized model.
+    `auto_gptq` or `gptqmodel` package. Quantization is done under the hood for users if they load a non-prequantized model.
     """
 
     requires_calibration = False
-    required_packages = ["optimum", "auto_gptq"]
+    required_packages = ["optimum", "gptqmodel"]
     optimum_quantizer = None
 
     def __init__(self, quantization_config: QuantizationConfigMixin, **kwargs):
@@ -49,16 +49,21 @@ def __init__(self, quantization_config: QuantizationConfigMixin, **kwargs):
         self.optimum_quantizer = GPTQQuantizer.from_dict(self.quantization_config.to_dict_optimum())
 
     def validate_environment(self, *args, **kwargs):
-        gptq_supports_cpu = version.parse(importlib.metadata.version("auto-gptq")) > version.parse("0.4.2")
+        gptq_supports_cpu = (
+            is_auto_gptq_available()
+            and version.parse(importlib.metadata.version("auto-gptq")) > version.parse("0.4.2")
+        ) or is_gptqmodel_available()
         if not gptq_supports_cpu and not torch.cuda.is_available():
             raise RuntimeError("GPU is required to quantize or run quantize model.")
-        elif not (is_optimum_available() and is_auto_gptq_available()):
+        elif not (is_optimum_available() and (is_auto_gptq_available() or is_gptqmodel_available())):
             raise ImportError(
-                "Loading a GPTQ quantized model requires optimum (`pip install optimum`) and auto-gptq library (`pip install auto-gptq`)"
+                "Loading a GPTQ quantized model requires optimum (`pip install optimum`) and auto-gptq or gptqmodel library (`pip install auto-gptq` or `pip install gptqmodel`)"
             )
-        elif version.parse(importlib.metadata.version("auto_gptq")) < version.parse("0.4.2"):
+        elif is_auto_gptq_available() and version.parse(importlib.metadata.version("auto_gptq")) < version.parse(
+            "0.4.2"
+        ):
             raise ImportError(
-                "You need a version of auto_gptq >= 0.4.2 to use GPTQ: `pip install --upgrade auto-gptq`"
+                "You need a version of auto_gptq >= 0.4.2 to use GPTQ: `pip install --upgrade auto-gptq` or use gptqmodel by `pip install gptqmodel`"
             )
 
     def update_torch_dtype(self, torch_dtype: "torch.dtype") -> "torch.dtype":

src/transformers/utils/__init__.py

@@ -143,6 +143,7 @@
     is_g2p_en_available,
     is_galore_torch_available,
     is_gguf_available,
+    is_gptqmodel_available,
     is_grokadamw_available,
     is_hqq_available,
     is_in_notebook,

src/transformers/utils/import_utils.py

@@ -139,6 +139,7 @@ def _is_package_available(pkg_name: str, return_version: bool = False) -> Union[
 _openai_available = _is_package_available("openai")
 _optimum_available = _is_package_available("optimum")
 _auto_gptq_available = _is_package_available("auto_gptq")
+_gptqmodel_available = _is_package_available("gptqmodel")
 # `importlib.metadata.version` doesn't work with `awq`
 _auto_awq_available = importlib.util.find_spec("awq") is not None
 _quanto_available = _is_package_available("quanto")
@@ -1005,6 +1006,10 @@ def is_auto_gptq_available():
     return _auto_gptq_available
 
 
+def is_gptqmodel_available():
+    return _gptqmodel_available
+
+
 def is_eetq_available():
     return _eetq_available