{% extends "base.html" %} {% block nav %}{{ spec.name }}{% endblock %} {% block content %}

Review episodes

{{ done }} of {{ total }} sampled episodes reviewed. Label each episode before you see the reward model's suggestion; the agreement in the report uses these first labels.

{% if state %}

Episode {{ state.episode_index }}

Your label: {{ "success" if state.first else "failure" }}{% if state.revised %}, changed to {{ "success" if state.final else "failure" }}{% endif %}.

{% if state.suggested is not none %}

The reward model suggested {{ "success" if state.suggested else "failure" }} (score {{ "%.2f" | format(state.score) }}). {% if state.suggested == state.final %}It agrees with your label.{% else %}It disagrees with your label.{% endif %}

{% endif %}
Change the label

Next episode

{% elif item %}

Episode {{ item.episode_index }} ({{ item.position }} of {{ item.total }})

{% if item.video %}

Camera {{ item.video.camera }}, {{ "%.1f" | format(item.video.end_s - item.video.start_s) }} s.

{% else %}

This dataset has no video for the episode; review it from its recording and label it here.

{% endif %}
{% elif total %}

Every sampled episode has been reviewed. Draw more with fieldtrial review-sample, or see the agreement in the report.

{% else %}

No episodes to review yet. Score a dataset with fieldtrial score-episodes, then draw a sample with fieldtrial review-sample.

{% endif %} {% endblock %}