[RL Unplugged] - Trained policies for finger_turn_hard do not match the datasets

Hi, thanks for the good work.

I am running custom offline evaluation on the tasks from DeepMind Control Suite Dataset. I am taking the data from Tensorflow Datasets [here](https://www.tensorflow.org/datasets/catalog/rlu_control_suite) and downloading the maching policies from the GCP bucket [here](https://console.cloud.google.com/storage/browser/gresearch/deep-ope/rlunplugged?pageState=(%22StorageObjectListTable%22:(%22f%22:%22%255B%255D%22))&prefix=&forceOnObjectsSortingFiltering=false).

Most tasks work fine, but when I get to finger_turn_hard I get the following error while inferring the policies on the data:

It looks like the policy is expecting one of the keys to be "touch", however, this key is not in the observations. According to the documentation [here](https://www.tensorflow.org/datasets/catalog/rlu_control_suite#rlu_control_suitefinger_turn_hard), there is no dimension "touch" in this dataset.

For context, I am loading the policy using:
```python
tf.saved_model.load(policy_path)
```

And inferring using:
```python
if hasattr(policy, 'initial_state'):
action = policy(observation, ((),))[0]
else:
action = policy(observation)
```

Am I missing something here? How can I infer the finger_turn_hard policies on the provided datasets?