# Design for UI model ## Tasks: **NOTE:** I ordered the options imo, best is the first one. - [ ] Define which type of input from users (Model_ID, path, model.json) - [ ] Create a table of recommendation for users (which hardware for which size) - [ ] Create UI (@ashley) - [ ] Place for user input - [ ] Place for user to adjust params -> We can opt out this since we have `right panel`. - [ ] Adding Docs - [ ] Adding QA checklist for this ## Teamates to involved - Ashley: design UI - James: App pod - Rex: QA :::warning This will be decapriated soon. See below. ::: # How to use custom models on Jan As Jan can run any GGUF models on Huggingface hub, please follow step by step # 1. Choose GGUF model You can use any models you like. In this guide we will use our latest model [Mistral-7B-Intruct-SLERP](janhq/Mistral-7B-Instruct-v0.2-SLERP-GGUF). ```title=Model_ID janhq/Mistral-7B-Instruct-v0.2-SLERP-GGUF ``` # 2. Create model configuration file You will need to navigate to your `jan/models` folder on your computer. Windows ```title=Path to jan/models C:/Users/<your_user_name>/jan/models ``` MacOS/ Linux ```title=Path to jan/models jan/models ``` Create a folder with the name of the model ```Example mistral-7B-intruct-slerp ``` Create a `model.json` inside the new created folder. The following is an example tempalate for `model.json` ``` { "source_url": "https://huggingface.co/janhq/Mistral-7B-Instruct-v0.2-SLERP-GGUF/resolve/main/mistral-7b-instruct-v0.2-slerp.Q4_K_M.gguf", "id": "mistral-7B-intruct-slerp", "object": "model", "name": "Mistral-7B-Intruct-SLERP", "version": "1.0", "description": "This is a guide for you to run custom model on Jan. Have fun.", "format": "gguf", "settings": { "ctx_len": 2048, "prompt_template": "<|im_start|>system\n{system_message}<|im_end|>\n<|im_start|>user\n{prompt}<|im_end|>\n<|im_start|>assistant" }, "parameters": { "max_tokens": 2048 }, "metadata": { "author": "Jan", "tags": ["7B", "Finetuned"] }, "engine": "nitro" } ``` :::info 3 Key notes Ensure the `source_url` is the link to download model Ensure the `id` is the same with the new created folder Ensure to choose right `prompt_template` ::: # 3. Use custom model You can restart Jan app and look for the your new custom model in the Hub. ![image](https://hackmd.io/_uploads/HJLAqvwI6.png) There you go. If you have any questions or request GGUF version of any models, please message us on [Jan Discord](https://discord.gg/Dt7MxDyNNZ).