Form Deactivation Followed by Fallback Policy

I am using Rasa 1.10.x , Having problems when with the policies chosen.

I built a form to collect data from users and validate it, on the happy path the forms and follow-up actions work fine. while on the un-happy path if any validation is broken it continues correctly and deactivates the form then Fallback Policy is activated.

policies:
  - name: FallbackPolicy
    nlu_threshold: 0.6
    core_threshold: 0.5
    fallback_action_name: 'action_my_default_fallback'
  - name: AugmentedMemoizationPolicy
  - name: MemoizationPolicy
  - name: FormPolicy

Happy Path :

botfront-rasa   | 2021-03-29 16:19:29 DEBUG    rasa.core.policies.form_policy  - There is no active form
botfront-rasa   | 2021-03-29 16:19:29 DEBUG    rasa.core.policies.ensemble  - best policy prediction policy_1_AugmentedMemoizationPolicy with confidence (1.0, 3)             vs form policy policy_3_FormPolicy with confidence (0.0, 5) 
botfront-rasa   | 2021-03-29 16:19:29 DEBUG    rasa.core.policies.ensemble  - Best policy prediction policy_1_AugmentedMemoizationPolicy
botfront-rasa   | 2021-03-29 16:19:29 DEBUG    rasa.core.policies.ensemble  - Predicted next action using policy_1_AugmentedMemoizationPolicy
botfront-rasa   | 2021-03-29 16:19:29 DEBUG    rasa.core.processor  - Predicted next action 'action_listen' with confidence 1.00.

Un-happy Path:

botfront-rasa   | 2021-03-29 16:08:43 DEBUG    rasa.core.policies.memoization  - There is no memorised next action
botfront-rasa   | 2021-03-29 16:08:43 DEBUG    rasa.core.policies.form_policy  - There is no active form
botfront-rasa   | 2021-03-29 16:08:43 DEBUG    rasa.core.policies.ensemble  - best policy prediction policy_0_FallbackPolicy with confidence (0.5, 4)             vs form policy policy_3_FormPolicy with confidence (0.0, 5) 
botfront-rasa   | 2021-03-29 16:08:43 DEBUG    rasa.core.policies.ensemble  - Best policy prediction policy_0_FallbackPolicy
botfront-rasa   | 2021-03-29 16:08:43 DEBUG    rasa.core.policies.ensemble  - Predicted next action using policy_0_FallbackPolicy
botfront-rasa   | 2021-03-29 16:08:43 DEBUG    rasa.core.processor  - Predicted next action 'action_my_default_fallback' with confidence 0.50.
botfront-rasa   | 2021-03-29 16:08:43 DEBUG    rasa.core.actions.action  - Calling action endpoint to run action 'action_my_default_fallback'.
botfront-actions | 2021-03-29 16:08:43 DEBUG    rasa_sdk.executor  - Received request to run 'action_my_default_fallback'
botfront-actions | 2021-03-29 16:08:43 DEBUG    rasa_sdk.executor  - Finished running 'action_my_default_fallback'

How I should approach this . Thanks in advance

When I changed to TEDPolicy, it workes fine in both cases (Happy Path and Un-Happy Path)

Any explains for this. Also , How to choose the correct Policy

Can you post the stories for your happy and unhappy paths? Can’t really see what’s happening without it.

Thanks @mloubser for you reply.

Here is story example

* inform_checkBalance
- action_slot_reset
- utter_check_help
- check_balance_form
- form {"name":"check_balance_form"}
- form {"name": null}
- slot{"name": "ahmed"}
- slot{"age":"25"}
- slot{"account_id": "0123654789654123"}
- slot{"email": "test.test@test.com"}
- action_balance
- slot{"balance":"balance"}
- action_check_balance 
- action_send_email_checkbalance
- utter_creat_Submit
- utter_ask_for_anotherProcess

Attached also project files bf_RasaExport.zip (41.9 KB)

I can’t see in your stories where is this unhappy path - can you provide them separately, as minimal reproducible examples?