Improve RulePolicy prediction timings

[wochinge]

Description of Problem: RulePolicy predictions are slow. This is especially noticeable during training when we check for contradictions. The reason for this is that we call _rule_key_to_state a lot when checking if a rule is applicable. For each time we are doing this we are loading the json which is encoding the rule.

This is a followup from this issue.

Examples (if relevant):

• You can find a profiling of contradiction check here (viewable with tools like snakeviz).
• quick fix was done here: speed up rule policy contraction check #8851

Definition of Done:

• Determine a quicker way to encode / load the rule states (e.g. by unwrapping the jsons just once or use a faster mechanism to load the jsons)

[stale]

Exalate commented:

stale[bot] commented:

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[danielenricocahall]

Exalate commented:

danielenricocahall commented:

Now that caching has been added as an enhancement, could another possible solution be using a more performance optimized JSON library (e.g; orjson, rapidjson)? And/or would it make sense to make the cache size configurable? Similar issue has been raised here

[sync-by-unito]

➤ Maxime Verger commented:

💡 Heads up! We're moving issues to Jira: https://rasa-open-source.atlassian.net/browse/OSS.

From now on, this Jira board is the place where you can browse (without an account) and create issues (you'll need a free Jira account for that). This GitHub issue has already been migrated to Jira and will be closed on January 9th, 2023. Do not forget to subscribe to the corresponding Jira issue!

➡️ More information in the forum: https://forum.rasa.com/t/migration-of-rasa-oss-issues-to-jira/56569.