flax.linen.apply takes state as the first param, followed by the input (#12510)

examples/research_projects/jax-projects/README.md

@@ -621,7 +621,7 @@ state = model_flax.init(rng, dummy_input_ids)
 and then we can do the forward pass.
 
 ```python
-sequences = model_flax.apply(input_ids, state)
+sequences = model_flax.apply(state, input_ids)
 ```
 
 Visually, the forward pass would now be represented as passing all tensors required for the computation to the model's object: