-
-
Notifications
You must be signed in to change notification settings - Fork 71
FAQ
At the moment the application is available in TestFlight. If for some reason you are not able to access testflight email me and i will build an ipa for you.
Models in ggjtv3 (must have a .bin extension) and gguf (must have a .gguf extension) formats are supported.
Make sure the correct inference is selected in the chat settings.
If the inference is correct, but the error persists, it is possible that the model is in an old format and needs to be converted and requatized from the original model.
Scripts for inference and quantization you can find here:
LLaMA, LLaMA 2, Falcon, Starcoder Alpaca, GPT4All, Chinese LLaMA, Vigogne, Vicuna, Koala, OpenBuddy, Pygmalion 7B, WizardLM, Baichuan-7B, Aquila-7B, Mistral
If the output starts normally but after a while the eval error occurs, try increasing the size of the context in chat settings.
The most common reason for an application crash is lack of memory. You may be using a model that is too large.
Extended Virtual Addressing and increased-memory-limit entitlements is required to run models larger than 3B, otherwise the application may crash. If for some reason you cannot build an application with this entitlement, you can simply turn it off.