diff --git a/mlx_vlm/models/idefics2/idefics2.py b/mlx_vlm/models/idefics2/idefics2.py index b78f6c2..5331fc6 100644 --- a/mlx_vlm/models/idefics2/idefics2.py +++ b/mlx_vlm/models/idefics2/idefics2.py @@ -250,9 +250,9 @@ def _prepare_inputs_for_multimodal(self, image_features, inputs_embeds, input_id return inputs_embeds def __call__(self, input_ids: mx.array, pixel_values: mx.array, cache=None): - # input_embeddings = self.get_input_embeddings(input_ids, pixel_values) + input_embeddings = self.get_input_embeddings(input_ids, pixel_values) logits, cache = self.language_model( - inputs=input_ids, cache=cache, inputs_embeds=None + inputs=input_ids, cache=cache, inputs_embeds=input_embeddings ) return logits, cache