Quantized Gorilla (#160)

Resolved #77 , demo displaying local inference with textwebui.

K-quantized gorilla models can be found on
[Huggingface](https://huggingface.co/gorilla-llm):
[Llama-based](https://huggingface.co/gorilla-llm/gorilla-7b-hf-v1-gguf),
[MPT-Based](https://huggingface.co/gorilla-llm/gorilla-mpt-7b-hf-v0-gguf),
[Falcon-Based](https://huggingface.co/gorilla-llm/gorilla-falcon-7b-hf-v0-gguf),
[`gorilla-openfunctions-v0-gguf`](https://huggingface.co/gorilla-llm/gorilla-openfunctions-v0-gguf),
[`gorilla-openfunctions-v1-gguf`](https://huggingface.co/gorilla-llm/gorilla-openfunctions-v1-gguf)

A tutorial walkthrough on how to quantize model using llama.cpp with
different quantization methods documented in
[![Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1JP_MN-J1rODo9k_-dR_9c9EnZRCfcVNe?usp=sharing).

Running local inference with Gorilla on a clean interface is simple.
Demoed using
[text-generation-webui](https://github.com/oobabooga/text-generation-webui),
add your desired models, and run inference.

More details in `/inference` README

Co-authored-by: Pranav Ramesh <[email protected]>

---------

Co-authored-by: Pranav Ramesh <[email protected]>
Co-authored-by: Pranav Ramesh <[email protected]>

Author: CharlieJCJ

inference/Presentation1_final.gif

@@ -58,6 +58,51 @@ python3 serve/gorilla_falcon_cli.py --model-path path/to/gorilla-falcon-7b-hf-v0
 
 > Add "--device mps" if you are running on your Mac with Apple silicon (M1, M2, etc)
 
+### Inference Gorilla locally
+
+K-quantized gorilla models can be found on [Huggingface](https://huggingface.co/gorilla-llm): [Llama-based](https://huggingface.co/gorilla-llm/gorilla-7b-hf-v1-gguf), [MPT-Based](https://huggingface.co/gorilla-llm/gorilla-mpt-7b-hf-v0-gguf), [Falcon-Based](https://huggingface.co/gorilla-llm/gorilla-falcon-7b-hf-v0-gguf)
+
+K-quantized `gorilla-openfunctions-v0` and `gorilla-openfunctions-v1` models can be found on [Huggingface](https://huggingface.co/gorilla-llm): [`gorilla-openfunctions-v0-gguf`](https://huggingface.co/gorilla-llm/gorilla-openfunctions-v0-gguf), [`gorilla-openfunctions-v1-gguf`](https://huggingface.co/gorilla-llm/gorilla-openfunctions-v1-gguf)
+
+For an in depth walkthrough on how this quantization was done, follow the tutorial in 
+this [![Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1JP_MN-J1rODo9k_-dR_9c9EnZRCfcVNe?usp=sharing). This tutorial is a fully self-contained space to see an under-the-hood walkthrough of the quantization pipeline (using llama.cpp) and to test out your own prompts with different quantized versions of Gorilla. The models don't take up local space and utilize a CPU runtime. 
+
+Running local inference with Gorilla on a clean interface is simple. Follow the instructions below to set up [text-generation-webui](https://github.com/oobabooga/text-generation-webui), add your desired models, and run inference. 
+
+
+My specs, M1 Macbook Air 2020
+```
+  Model Name: MacBook Air
+  Model Identifier:	MacBookAir10,1
+  Model Number:	Z125000NMCH/A
+  Chip:	Apple M1
+  Total Number of Cores: 8 (4 performance and 4 efficiency)
+  Memory: 16 GB
+  System Firmware Version: 10151.61.4
+  OS Loader Version: 10151.61.4
+```
+
+Step 1: Clone [text-generation-webui](https://github.com/oobabooga/text-generation-webui), a Gradio web UI for Large Language Models. It supports transformers, GPTQ, AWQ, EXL2, llama.cpp (GGUF), and Llama models. It hides many complexities of llama.cpp and has a well defined interface that is easy to use.
+
+`git clone https://github.com/oobabooga/text-generation-webui.git`
+
+Step 2: Follow [text-generation-webui](https://github.com/oobabooga/text-generation-webui) instructions to run the application locally.
+1. Go to the cloned folder
+2. `./start_macos.sh` and it will output the following ![Alt text](image.png)
+3. Open a browser and go to url `http://127.0.0.1:7860/` as an example. 
+
+
+Step 3: Select the quantization method you want to use, download the quantized model and run the inference on the quantized Gorilla models. 
+
+1. Here, we can go to `Model` and there is `Download model or LoRA`. For example, we want to get the q3_K_M gguf quantized model for `gorilla-7b-hf-v1`, you would input `gorilla-llm/gorilla-7b-hf-v1` and filename as `gorilla-7b-hf-v1-q3_K_M` and click `Download`. It would say Downloading file to `models/`. ![Alt text](image-2.png)
+2. After downloading the model, you select the Model, `gorilla-7b-hf-v1-q3_K_M` for demonstration, and click `Load`. For settings, if you have laptop GPU available, increasing `n-gpu-layers` accelerates inference. ![Alt text](image-1.png)
+3. After loading, it will give a confirmation message as following. ![Alt text](image-3.png)
+4. Then go to `Chat` page, use default setting for llama based quantized models, ![Alt text](image-4.png)
+5. *Real-time inference* video demo
+![Alt text](Presentation1_final.gif)
+
+Integration with Gorilla-CLI coming soon ...
+
 ### [Optional] Batch Inference on a Prompt File
 
 After downloading the model, you need to make a jsonl file containing all the question you want to inference through Gorilla. Here is [one example](https://github.com/ShishirPatil/gorilla/blob/main/inference/example_questions/example_questions.jsonl):