update

Author: dsikka

src/compressed_tensors/compressors/model_compressors/model_compressor.py

@@ -608,7 +608,7 @@ def decompress(self, model_path: str, model: Module):
             self.sparsity_compressor is not None
             and self.sparsity_config.format != CompressionFormat.dense.value
         ):
-            # note - decompress only support one compressor so far
+            # note - decompress only supports one compressor atm
             quant_compressor = next(iter(self.quantization_compressor))
             params_to_ignore = None
             if self.quantization_compressor is not None:

src/compressed_tensors/quantization/quant_config.py

@@ -162,7 +162,7 @@ def to_dict(self):
 
     @staticmethod
     def from_pretrained(
-        model: Module, format: Optional[Union[List[str], str]] = None
+        model: Module, format: Optional[str] = None
     ) -> Optional["QuantizationConfig"]:
         """
         Converts a model into its associated QuantizationConfig based on the

tests/test_compressors/model_compressors/test_model_compressor.py

@@ -395,7 +395,7 @@ def _get_combined_config(s_config, q_config):
 )
 def test_compress_model(model_stub, q_format, s_config, tmpdir):
     model = AutoModelForCausalLM.from_pretrained(model_stub, torch_dtype=torch.float32)
-    compressor = ModelCompressor.from_pretrained_model(model, s_config, q_format)
+    compressor = ModelCompressor.from_pretrained_model(model, s_config, [q_format])
 
     # compress model by eagerly compressing state dict
     true_compressed = dict(compressor.compress(model))