fix config format for transformers (#120)

pyproject.toml

@@ -31,7 +31,7 @@ requires = [
 build-backend = "setuptools.build_meta"
 
 [tool.ruff]
-# Allow lines to be as long as 80.
+# Allow lines to be as long as 120.
 line-length = 120
 exclude = [
     # External file, leaving license intact

vptq/__init__.py

@@ -4,4 +4,6 @@
 # --------------------------------------------------------------------------
 
 __version__ = "0.0.3"
-from vptq.layers import AutoModelForCausalLM as AutoModelForCausalLM
+from vptq.layers import AutoModelForCausalLM, VQuantLinear
+
+__all__ = ["AutoModelForCausalLM", "VQuantLinear"]

vptq/layers/__init__.py

@@ -3,4 +3,6 @@
 # Licensed under the MIT License.
 # --------------------------------------------------------------------------
 
-from vptq.layers.model_base import AutoModelForCausalLM as AutoModelForCausalLM
+from vptq.layers.model_base import AutoModelForCausalLM, VQuantLinear
+
+__all__ = ["AutoModelForCausalLM", "VQuantLinear"]

vptq/layers/model_base.py

@@ -90,11 +90,11 @@ def from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs):
             model = cls.from_config(auto_conf, *model_args, **cls_kwargs)
 
         target_layer = VQuantLinear
-        quant_config = auto_conf.quant_config
+        quantization_config = auto_conf.quantization_config
 
         # replace linear layers with quantized linear layers
         with transformers.utils.generic.ContextManagers([accelerate.init_empty_weights()]):
-            make_quant_linear(model, quant_config, target_layer=target_layer)
+            make_quant_linear(model, quantization_config, target_layer=target_layer)
 
         no_split_module_classes = [i[1].__class__.__name__ for i in model.named_modules() if i[0].endswith(".0")]