AssetStore Price down information

beta

Llama.cpp inference async wrapper with demo LLM model for text stylization compiled for Mobile, Web and PC platforms. It will assist you in transforming generic texts into stylized, tailored content.The package is focused on text generation, so there are no dependencies on the renderer type. Tested with iOS, WebGL, Android, Windows and MacOS.This is a set of mobile friendly LLAMA and GPT2-based GenAI LLM models for text line rewriting with iOS and WebGL wrappers beased on LLama.cpp.We ship it with three model quantised resolutions, try it online entirely in your browser with the quantised (simplified) one :- Automated NPC Dialogue stylisation demo (LLAMA based Q4_K_M ~110mb)- Single Prompt Debug demo (LLAMA based Q8_0 ~170mb)Inside the package, you will also find Q8_0, Q4 and original bf16 (~321mb) resolutions of our models ready to run inside a cross-platform game engine!You give it:``` How are you today? Pirate's Poetic Question <styleEnds><output>```It prints out: Hey there, how fares ye today?We hope that this model and wrapper can help you save a day! (or, more realistically, 3 months with a dedicated team of 3 people in our estimates).Model finetuned on a corpus of more than 0.5+ million dialogue lines (64343812 tokens).All our training data was purely syntactical and generated.Tools for output filtering are provided with release in CSharp - thus easely editable!All generative LLM models can make mistakes in their output. In some cases, they can turn questions into statements and add minor hallucinated info based on the input text and style data.The bounded model inference wrapper is based on open-source model runtime and supports the latest 405b models, so you can try them out on platforms with enough RAM/time available.The bundled wrapper allows:Build and tested for: Web, Mobile, PC.To execute model setup and execution in asynchronous mode based on threads in C++ code. (this is needed if one wants parallel code run when building targeting the Web platform)It allows tapping into Logs, output Tokens, and completion callbacks from generation.Setup model update callbacks in editor UI using events and helper scripts. Also, you can configure most model configuration/run parameters from the editor UI and at runtime.Cross-Platform: Compatible with WebGL, iOS, Android, Windows x64, and MacOS (ARM).Mobile ahead-of-time compilation friendliness.Async mode uses cross-platform C++20 threads and is compatible with Web platformThe training was done from the open-source model and weights in ORPO mode.The model is fast and slim. Weights are provided in fp16, int8 .gguf format.Use recommendationsPlease try out to free form style line - be creative, yet to get the most from them think about combining in your style line:Mood styles to try out:Inquisitive (Questioning, Curious, Skeptical, Intrigued)Emotional (Joyful, Melancholic, Indignant, Compassionate, Euphoric, Grieving, Impassioned, Jovial)Intellectual (Reflective, Pensive, Cynical)Dynamic (Confident, Resigned, Agitated, Hopeful, Fearful, Optimistic, Defiant, Adventurous, Bewildered, Determined, Hesitant, Mischievous, Overwhelmed, Melodramatic)Noble (Regal, Dignified)Light (Lighthearted, Whimsical, Nostalgic)Foreboding (Sarcastic, Foreboding)Writing styles to try out:Historical (Victorian, Gothic, Classic Literature, Folkloric, Pirates)Modern and Contemporary (Modern, Minimalist, Journalistic, Futuristic Sci-Fi, Dark Futurism, Post-Apocalyptic)Genre-Specific (Noir, Magical Realism, Dystopian, Epic, Hard-boiled, Pulp Adventure, Steampunk, Romantic Comedy, Surrealist)Expressive and Creative (Poetic, Lyrical, Beat Generation, Inspirational, Absurdist, Satirical, Whimsical, Mythological)Keep styles short. They can be phrases or simple words. Using definitions like "Simple" may help debug lines rewritten in too LLMish style.Integration notes:Web platform is RAM limited up to 4GB (x86), so only smallish models fit into it.Web published pages do not work if ram is limmited to 0.5-1GB making them fail to run on some mobile devices.Instructions for building cpp wrapper code from scratch are provided.AI LLM model may output tokens unparsable by selected FontIn demo scenes restart the postprocessing layer on camera if you get NullReferenceException PostProcessing AmbientOcclusion IsEnabledAndSupported exceptionLLM models have various sizes, that impacts ram requirements on devices you run them onLLM model was trained on one-liners and single sentences. This is when it works the best way possible.Online-Docs - see motivation-usage example:CSharp docsCPP wrapper docsModel training data was synthetically.Package artwork design were AI generated:images,3d models,rigging and animations,cubemaps.