search

huggingface / swift-chat

Mac app to demonstrate swift-transformers
Apache License 2.0
492 stars 37 forks source link

Can't load Llama-2-7b-chat-coreml model #4

Closed awmartin closed 1 year ago

awmartin commented 1 year ago

After getting SwiftChat to run with falcon-7b (but very very slowly, perhaps 45 minutes per prompt), I tried to load the supported Llama-2-7b-chat-coreml model by selecting the .mlpackage file from the SwiftChat dialog.

SwiftChat throws this error:

2023-08-16 08:56:47.315342-0400 SwiftChat[72086:2980822] +[CATransaction synchronize] called within transaction
2023-08-16 08:56:49.270422-0400 SwiftChat[72086:2980822] [default] CGSWindowShmemCreateWithPort failed on port 0

…and the model does not load, but keeps the falcon-7b model loaded.

awmartin commented 1 year ago

Hmm, this seems to have resolved itself. Weird.