That could do it too, I did not know that. Thank you. I'll poke at it this weekend and report back.
- Jan 18, 2019
- 182
- 361
Updated download mirrors to latest version v1.1.3
Changelog:
v1.1.0 v1.1.2 v1.1.3
Changelog:
v1.1.0 v1.1.2 v1.1.3
- Added quality TTS (only available on Chrome and Edge)
- Added language selection
- Improved responsiveness by 20x times! You now see narration text nearly instantly rather than waiting up to minutes for the AI to finish writing the entire response.
- Removed the complicated JSON formatting, the AI can now respond in a much simpler format, this should greatly increase the success rate of parsing AI responses and allow you to use even regular role play AI models!
- AI can now see player traits information and more accurate stats data (including current and max stat values along with stat descriptors)
- Note that the AI still cannot give you or remove traits
- The Stats panel now show +5 or -2 in front of the stat names to indicate change
- Added default setting to limit AI event to only one paragraph (this prevent AI from writing too much and going off-road, eating up precious memory tokens)
- Added the ability to edit the AI responses and player stats
- AI can modify the maximum value of stats
- Fixed bugs:
- Entities list now work
- 3D character viewer often didn’t load the customization data
Really rough guide as I am new to all this AI stuff.
1)
2) Search and install model - mistral-nemo-instruct-2407 6.63GB
3) on Developer tab start the server with above model.
4) Copy the "Reachable at" url into the Formamorph settings > Endpoint > URL (keeping the /v1/chat/completions at the end).
5) on right you should see a Load tap to tune model. I bumped Batch size to 2096.
6) *This is where I am still testing. I have 9800x3D + a 3080TI + 32g Ram*
I personnally bumped Context length/Tokens to 10000. Still instant on my PC, but most likely will need to be dropped for older systems, or possibly increased on newer ones like 4000/5000 RTX
7)click the "Reload to apply changes) and it should all work.
As I said I am new to this. Game/AI still forgets actions sometimes. I am currently trying to understand why that happens.
Like for example - You climb down the pipe > Get AI response you climbed down the pipe > Game offers again to climb down the pipe and repeat >.< (So easy solution it make sure you do not click same option afterwards unless it specificly and clearly says to 'continue climbing down pipe' for example).
1)
You must be registered to see the links
2) Search and install model - mistral-nemo-instruct-2407 6.63GB
3) on Developer tab start the server with above model.
4) Copy the "Reachable at" url into the Formamorph settings > Endpoint > URL (keeping the /v1/chat/completions at the end).
5) on right you should see a Load tap to tune model. I bumped Batch size to 2096.
6) *This is where I am still testing. I have 9800x3D + a 3080TI + 32g Ram*
I personnally bumped Context length/Tokens to 10000. Still instant on my PC, but most likely will need to be dropped for older systems, or possibly increased on newer ones like 4000/5000 RTX
7)click the "Reload to apply changes) and it should all work.
As I said I am new to this. Game/AI still forgets actions sometimes. I am currently trying to understand why that happens.
Like for example - You climb down the pipe > Get AI response you climbed down the pipe > Game offers again to climb down the pipe and repeat >.< (So easy solution it make sure you do not click same option afterwards unless it specificly and clearly says to 'continue climbing down pipe' for example).
Last edited:
