Great, thank you!
Swimwear7679
Recent community posts
**Is your feature request related to a problem? Please describe.**
When loading an existing game, the model selection always defaults to "gpt2-large", but I expect it to default to the model I was last using with the loaded game.
**Describe the feature you'd like**
The model selector should default to the model I was last using with the loaded game.
**Describe alternatives you've considered**
Alternatively, the model selector would not be displayed on a game load. Instead, the last used model would automatically be loaded. The model could be changed from a "game settings" page.
Describe the bug
There is no (obvious) way to exit the console.
To Reproduce
- Load a game
- Click the "Console" button
- The console launches
There is no button to close the console. Esc also does not work. Issuing the "quit" command seems to exit the whole game window.
Expected behavior
There is a button to close the console. Esc also closes the console.
Screenshots
Desktop (please complete the following information):
- OS: Windows 11
I really like the idea of this game but am struggling to make it work. I've been trying to play the "black friday" scenario, but can't get past the first few prompts without running into issues. Sometimes the generated text cuts off in the middle of a sentence, sometimes it generates text that is nonsensical or inconsistent with the history, and sometimes it repeats itself within the first few sentences and gets into an endless loop.
I've tried playing around with the following options:
- models: gpt2-large, gpt-neo-2.7b, gpt-J-6B-adventure
- generation max length: 60-400
- randomness: 0.4-0.7
- max time: 60-120
- context max length: 1000-10000
- memory max length: 400-4000
- remove unfinished sentences: off/on
- offload model: off/on
System info:
Processor AMD Ryzen 5 5600 6-Core Processor, 3501 Mhz, 6 Core(s), 12 Logical Processor(s) Installed Physical Memory (RAM) 32.0 GB Name NVIDIA GeForce RTX 3080 OS Name Microsoft Windows 11 Home Version 10.0.22621 Build 22621
Can someone please recommend a setup (model and all settings) that have given good results? Thanks!
Describe the bug
The generated text ends with partial sentences even with the "Remove unfinished sentences" option checked.
To Reproduce
- Check the "Remove unfinished sentences" option
- Start a new scenario using the "gpt2-large" model
- Input any prompt
The generated text ends with a partial sentence.
Expected behavior
The generated text does not end with a partial sentence.
Screenshots
Desktop (please complete the following information):
- OS: Windows 11
Additional context
The AI does not use all the allotted time to generate the text. It only takes about 10-20 seconds but I still get partial sentences and really short output.