Fix test

tests/test_full_model.py

@@ -157,7 +157,7 @@ def test_input_ids_and_embeds(tokenizer, model, ref_model, max_new_tokens=4):
     assert inputs.keys() == {"input_ids", "attention_mask"}
 
     outputs = model.generate(**inputs, max_new_tokens=max_new_tokens)
-    ref_outputs = ref_model.generate(inputs, max_new_tokens=max_new_tokens)
+    ref_outputs = ref_model.generate(**inputs, max_new_tokens=max_new_tokens)
     assert torch.allclose(outputs, ref_outputs), f"Outputs are not identical to HF"
 
     with model.inference_session(max_length=inputs["input_ids"].shape[1] + max_new_tokens):