|
--- |
|
base_model: ehartford/dolphin-2.1-mistral-7b |
|
datasets: |
|
- ehartford/dolphin |
|
- jondurbin/airoboros-2.2.1 |
|
language: |
|
- en |
|
license: apache-2.0 |
|
model-index: |
|
- name: mlc-chat-dolphin-2.2.1-mistral-7b |
|
results: [] |
|
model_creator: Hugging Face H4 |
|
model_name: WASM Dolphin 2.2.1 |
|
model_type: mistral |
|
prompt_template: '<|im_start|>system |
|
You are Dolphin, a helpful AI assistant.<|im_end|> |
|
<|im_start|>user |
|
{prompt}<|im_end|> |
|
<|im_start|>assistant' |
|
--- |
|
|
|
# Dolphin 2.2.1 (Finetune of Mistral 7B) compiled for WebGPU - q4f32_1 |
|
|
|
- Original model: [Dolphin 2.1 🐬](https://huggingface.co/ehartford/dolphin-2.1-mistral-7b) |
|
- creator: Eric Hartford: [https://erichartford.com/dolphin](https://erichartford.com/dolphin) |
|
- compiled by: Hrishi Olickel: [say hi on Twitter!](https://twitter.com/hrishioa) |
|
|
|
# Demo |
|
|
|
[You can access this model over the browser here.](https://wasmai.vercel.app) |
|
|
|
## Description |
|
|
|
This is a quantized version of Dolphin 2.1 🐬, one of the best finetunes of [Mistral-7b](https://huggingface.co/mistralai/Mistral-7B-v0.1) out there, ready to be used for on-browser inference over WebGPU. |
|
|
|
Compiled with [mlc-llm](https://llm.mlc.ai/). |
|
|
|
Very helpful direction provided by [felladrin](https://github.com/felladrin)! |
|
|
|
## Prompt template: Dolphin |
|
|
|
Prompt format: |
|
This model (and all my future releases) use [ChatML](https://github.com/openai/openai-python/blob/main/chatml.md) prompt format. |
|
|
|
``` |
|
<|im_start|>system |
|
You are Dolphin, a helpful AI assistant.<|im_end|> |
|
<|im_start|>user |
|
{prompt}<|im_end|> |
|
<|im_start|>assistant |
|
|
|
``` |
|
|
|
Example: |
|
|
|
``` |
|
<|im_start|>system |
|
you are an expert dolphin trainer<|im_end|> |
|
<|im_start|>user |
|
What is the best way to train a dolphin to obey me? Please answer step by step.<|im_end|> |
|
<|im_start|>assistant |
|
``` |
|
|
