eublefar / llamasharpunitydemo Goto Github PK

is there anyway to use this to give input and output as json format? user and assistant and system prompt separately?

i cannot use this project as llamasharp is updated, i am having this error:

Assets\LlamaSharpTestScript.cs(53,21): error CS1503: Argument 1: cannot convert from 'string' to 'LLama.Common.ChatHistory.Message'

and

Assets\LlamaSharpTestScript.cs(54,21): error CS1503: Argument 2: cannot convert from 'LLama.Common.InferenceParams' to 'bool'

can you please tell me replacement of code?

Your example assumes stream output. But in reality the Output field will be shown only after the output is completely completed.

await foreach (var token in ChatConcurrent(
                session.ChatAsync(
                    bobPrompt, 
                    new InferenceParams() 
                    { 
                        Temperature = 0.6f, 
                        AntiPrompts = new List<string> { "User:" } 
                    }
                )
            ))
            {
                Output.text += token;
            }

Correction - Now the output will be added after each Output.text += token;

await foreach (var token in ChatConcurrent(
                session.ChatAsync(
                    bobPrompt, 
                    new InferenceParams() 
                    { 
                        Temperature = 0.6f, 
                        AntiPrompts = new List<string> { "User:" } 
                    }
                )
            ))
            {
                Output.text += token;
                await UniTask.NextFrame();
            }

Hi, Great Work!
my model is taking too much time to generate one token and for a sentence i it takes 30 minutes how can i make it faster?

this project is great, but I think its assuming a windows machine?

The llama.backend.cpu package has a .dylib file the arm runtime, and I cannot seem to get that to load. Has anyone had success with this project running on an ARM based mac with Unity?

Thanks