Let’s park this case since accuracy is rising and I do not have SFML stats.
I understand that alternative would be probably remove trained model and start again.
Maybe later if you release TFS in docker I can connect with SFML and see if it helps
TFS HA is already released! Stats is included in SFML you just need to add it
TFS for HAOS is released but as I have HA in docker I cannot use add-ons and I need to wait for TFS for docker - I understand it is still learning?
I will check SFML stats
That’s funny that you closed the GitHub ticket claiming that all the users that experienced discrepancy after the update are lying just because I was using custom chart (taking directly values from sensor attribute) and not sfml stats that is not even available for raspberry pi that I was using when creating that ticket
SFML Stats x86_64 only
Ehhh…. Big shame, we can also close this topic