minor

Author: Your Name

examples/llm_qat/README.md

@@ -82,7 +82,6 @@ def forward_loop(model):
 
 
 # Quantize the model in-place; The model should be unwrapped from any distributed wrapper
-# The model may be wrapped in a DataParallel or DistributedDataParallel after `mtq.quantize`
 model = mtq.quantize(model, mtq.INT8_DEFAULT_CFG, forward_loop)
 
 # Save the modelopt quantizer states

modelopt/torch/quantization/plugins/transformers_trainer.py

@@ -195,6 +195,7 @@ def _restore_modelopt_state_with_weights(self):
     def _quantize_model(self):
         """Quantize the model. Restore the quantization state if it exists."""
         dataset = self.train_dataset if self.train_dataset is not None else self.eval_dataset
+        assert dataset is not None, "Calibration requires either eval or train dataset."
         num_samples = min(self.quant_args.calib_size, len(dataset))  # type: ignore [union-attr]
         dataset = torch.utils.data.Subset(dataset, list(range(num_samples)))
         data_loader = self.get_eval_dataloader(dataset)