---
base_model: SanjiWatsuki/Loyal-Macaroni-Maid-7B
inference: false
license: cc-by-nc-4.0
model_creator: Sanji Watsuki
model_name: Loyal Macaroni Maid 7B
model_type: mistral
prompt_template: 'Below is an instruction that describes a task. Write a response
that appropriately completes the request.
### Instruction:
{prompt}
### Response:
'
quantized_by: TheBloke
tags:
- merge
- not-for-all-audiences
- nsfw
---
# Loyal Macaroni Maid 7B - GGUF
- Model creator: [Sanji Watsuki](https://huggingface.co/SanjiWatsuki)
- Original model: [Loyal Macaroni Maid 7B](https://huggingface.co/SanjiWatsuki/Loyal-Macaroni-Maid-7B)
## Description
This repo contains GGUF format model files for [Sanji Watsuki's Loyal Macaroni Maid 7B](https://huggingface.co/SanjiWatsuki/Loyal-Macaroni-Maid-7B).
These files were quantised using hardware kindly provided by [Massed Compute](https://massedcompute.com/).
### About GGUF
GGUF is a new format introduced by the llama.cpp team on August 21st 2023. It is a replacement for GGML, which is no longer supported by llama.cpp.
Here is an incomplete list of clients and libraries that are known to support GGUF:
* [llama.cpp](https://github.com/ggerganov/llama.cpp). The source project for GGUF. Offers a CLI and a server option.
* [text-generation-webui](https://github.com/oobabooga/text-generation-webui), the most widely used web UI, with many features and powerful extensions. Supports GPU acceleration.
* [KoboldCpp](https://github.com/LostRuins/koboldcpp), a fully featured web UI, with GPU accel across all platforms and GPU architectures. Especially good for story telling.
* [GPT4All](https://gpt4all.io/index.html), a free and open source local running GUI, supporting Windows, Linux and macOS with full GPU accel.
* [LM Studio](https://lmstudio.ai/), an easy-to-use and powerful local GUI for Windows and macOS (Silicon), with GPU acceleration. Linux available, in beta as of 27/11/2023.
* [LoLLMS Web UI](https://github.com/ParisNeo/lollms-webui), a great web UI with many interesting and unique features, including a full model library for easy model selection.
* [Faraday.dev](https://faraday.dev/), an attractive and easy to use character-based chat GUI for Windows and macOS (both Silicon and Intel), with GPU acceleration.
* [llama-cpp-python](https://github.com/abetlen/llama-cpp-python), a Python library with GPU accel, LangChain support, and OpenAI-compatible API server.
* [candle](https://github.com/huggingface/candle), a Rust ML framework with a focus on performance, including GPU support, and ease of use.
* [ctransformers](https://github.com/marella/ctransformers), a Python library with GPU accel, LangChain support, and OpenAI-compatible AI server. Note, as of time of writing (November 27th 2023), ctransformers has not been updated in a long time and does not support many recent models.
## Repositories available
* [AWQ model(s) for GPU inference.](https://huggingface.co/TheBloke/Loyal-Macaroni-Maid-7B-AWQ)
* [GPTQ models for GPU inference, with multiple quantisation parameter options.](https://huggingface.co/TheBloke/Loyal-Macaroni-Maid-7B-GPTQ)
* [2, 3, 4, 5, 6 and 8-bit GGUF models for CPU+GPU inference](https://huggingface.co/TheBloke/Loyal-Macaroni-Maid-7B-GGUF)
* [Sanji Watsuki's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co/SanjiWatsuki/Loyal-Macaroni-Maid-7B)
## Prompt template: Alpaca
```
Below is an instruction that describes a task. Write a response that appropriately completes the request.
### Instruction:
{prompt}
### Response:
```
## Compatibility
These quantised GGUFv2 files are compatible with llama.cpp from August 27th onwards, as of commit [d0cee0d](https://github.com/ggerganov/llama.cpp/commit/d0cee0d36d5be95a0d9088b674dbb27354107221)
They are also compatible with many third party UIs and libraries - please see the list at the top of this README.
## Explanation of quantisation methods
Click to see details
The new methods available are:
* GGML_TYPE_Q2_K - "type-1" 2-bit quantization in super-blocks containing 16 blocks, each block having 16 weight. Block scales and mins are quantized with 4 bits. This ends up effectively using 2.5625 bits per weight (bpw)
* GGML_TYPE_Q3_K - "type-0" 3-bit quantization in super-blocks containing 16 blocks, each block having 16 weights. Scales are quantized with 6 bits. This end up using 3.4375 bpw.
* GGML_TYPE_Q4_K - "type-1" 4-bit quantization in super-blocks containing 8 blocks, each block having 32 weights. Scales and mins are quantized with 6 bits. This ends up using 4.5 bpw.
* GGML_TYPE_Q5_K - "type-1" 5-bit quantization. Same super-block structure as GGML_TYPE_Q4_K resulting in 5.5 bpw
* GGML_TYPE_Q6_K - "type-0" 6-bit quantization. Super-blocks with 16 blocks, each block having 16 weights. Scales are quantized with 8 bits. This ends up using 6.5625 bpw
Refer to the Provided Files table below to see what files use which methods, and how.
## Provided files
| Name | Quant method | Bits | Size | Max RAM required | Use case |
| ---- | ---- | ---- | ---- | ---- | ----- |
| [loyal-macaroni-maid-7b.Q2_K.gguf](https://huggingface.co/TheBloke/Loyal-Macaroni-Maid-7B-GGUF/blob/main/loyal-macaroni-maid-7b.Q2_K.gguf) | Q2_K | 2 | 3.08 GB| 5.58 GB | smallest, significant quality loss - not recommended for most purposes |
| [loyal-macaroni-maid-7b.Q3_K_S.gguf](https://huggingface.co/TheBloke/Loyal-Macaroni-Maid-7B-GGUF/blob/main/loyal-macaroni-maid-7b.Q3_K_S.gguf) | Q3_K_S | 3 | 3.17 GB| 5.67 GB | very small, high quality loss |
| [loyal-macaroni-maid-7b.Q3_K_M.gguf](https://huggingface.co/TheBloke/Loyal-Macaroni-Maid-7B-GGUF/blob/main/loyal-macaroni-maid-7b.Q3_K_M.gguf) | Q3_K_M | 3 | 3.52 GB| 6.02 GB | very small, high quality loss |
| [loyal-macaroni-maid-7b.Q3_K_L.gguf](https://huggingface.co/TheBloke/Loyal-Macaroni-Maid-7B-GGUF/blob/main/loyal-macaroni-maid-7b.Q3_K_L.gguf) | Q3_K_L | 3 | 3.82 GB| 6.32 GB | small, substantial quality loss |
| [loyal-macaroni-maid-7b.Q4_0.gguf](https://huggingface.co/TheBloke/Loyal-Macaroni-Maid-7B-GGUF/blob/main/loyal-macaroni-maid-7b.Q4_0.gguf) | Q4_0 | 4 | 4.11 GB| 6.61 GB | legacy; small, very high quality loss - prefer using Q3_K_M |
| [loyal-macaroni-maid-7b.Q4_K_S.gguf](https://huggingface.co/TheBloke/Loyal-Macaroni-Maid-7B-GGUF/blob/main/loyal-macaroni-maid-7b.Q4_K_S.gguf) | Q4_K_S | 4 | 4.14 GB| 6.64 GB | small, greater quality loss |
| [loyal-macaroni-maid-7b.Q4_K_M.gguf](https://huggingface.co/TheBloke/Loyal-Macaroni-Maid-7B-GGUF/blob/main/loyal-macaroni-maid-7b.Q4_K_M.gguf) | Q4_K_M | 4 | 4.37 GB| 6.87 GB | medium, balanced quality - recommended |
| [loyal-macaroni-maid-7b.Q5_0.gguf](https://huggingface.co/TheBloke/Loyal-Macaroni-Maid-7B-GGUF/blob/main/loyal-macaroni-maid-7b.Q5_0.gguf) | Q5_0 | 5 | 5.00 GB| 7.50 GB | legacy; medium, balanced quality - prefer using Q4_K_M |
| [loyal-macaroni-maid-7b.Q5_K_S.gguf](https://huggingface.co/TheBloke/Loyal-Macaroni-Maid-7B-GGUF/blob/main/loyal-macaroni-maid-7b.Q5_K_S.gguf) | Q5_K_S | 5 | 5.00 GB| 7.50 GB | large, low quality loss - recommended |
| [loyal-macaroni-maid-7b.Q5_K_M.gguf](https://huggingface.co/TheBloke/Loyal-Macaroni-Maid-7B-GGUF/blob/main/loyal-macaroni-maid-7b.Q5_K_M.gguf) | Q5_K_M | 5 | 5.13 GB| 7.63 GB | large, very low quality loss - recommended |
| [loyal-macaroni-maid-7b.Q6_K.gguf](https://huggingface.co/TheBloke/Loyal-Macaroni-Maid-7B-GGUF/blob/main/loyal-macaroni-maid-7b.Q6_K.gguf) | Q6_K | 6 | 5.94 GB| 8.44 GB | very large, extremely low quality loss |
| [loyal-macaroni-maid-7b.Q8_0.gguf](https://huggingface.co/TheBloke/Loyal-Macaroni-Maid-7B-GGUF/blob/main/loyal-macaroni-maid-7b.Q8_0.gguf) | Q8_0 | 8 | 7.70 GB| 10.20 GB | very large, extremely low quality loss - not recommended |
**Note**: the above RAM figures assume no GPU offloading. If layers are offloaded to the GPU, this will reduce RAM usage and use VRAM instead.
## How to download GGUF files
**Note for manual downloaders:** You almost never want to clone the entire repo! Multiple different quantisation formats are provided, and most users only want to pick and download a single file.
The following clients/libraries will automatically download models for you, providing a list of available models to choose from:
* LM Studio
* LoLLMS Web UI
* Faraday.dev
### In `text-generation-webui`
Under Download Model, you can enter the model repo: TheBloke/Loyal-Macaroni-Maid-7B-GGUF and below it, a specific filename to download, such as: loyal-macaroni-maid-7b.Q4_K_M.gguf.
Then click Download.
### On the command line, including multiple files at once
I recommend using the `huggingface-hub` Python library:
```shell
pip3 install huggingface-hub
```
Then you can download any individual model file to the current directory, at high speed, with a command like this:
```shell
huggingface-cli download TheBloke/Loyal-Macaroni-Maid-7B-GGUF loyal-macaroni-maid-7b.Q4_K_M.gguf --local-dir . --local-dir-use-symlinks False
```
More advanced huggingface-cli download usage (click to read)
You can also download multiple files at once with a pattern:
```shell
huggingface-cli download TheBloke/Loyal-Macaroni-Maid-7B-GGUF --local-dir . --local-dir-use-symlinks False --include='*Q4_K*gguf'
```
For more documentation on downloading with `huggingface-cli`, please see: [HF -> Hub Python Library -> Download files -> Download from the CLI](https://huggingface.co/docs/huggingface_hub/guides/download#download-from-the-cli).
To accelerate downloads on fast connections (1Gbit/s or higher), install `hf_transfer`:
```shell
pip3 install hf_transfer
```
And set environment variable `HF_HUB_ENABLE_HF_TRANSFER` to `1`:
```shell
HF_HUB_ENABLE_HF_TRANSFER=1 huggingface-cli download TheBloke/Loyal-Macaroni-Maid-7B-GGUF loyal-macaroni-maid-7b.Q4_K_M.gguf --local-dir . --local-dir-use-symlinks False
```
Windows Command Line users: You can set the environment variable by running `set HF_HUB_ENABLE_HF_TRANSFER=1` before the download command.
## Example `llama.cpp` command
Make sure you are using `llama.cpp` from commit [d0cee0d](https://github.com/ggerganov/llama.cpp/commit/d0cee0d36d5be95a0d9088b674dbb27354107221) or later.
```shell
./main -ngl 35 -m loyal-macaroni-maid-7b.Q4_K_M.gguf --color -c 32768 --temp 0.7 --repeat_penalty 1.1 -n -1 -p "Below is an instruction that describes a task. Write a response that appropriately completes the request.\n\n### Instruction:\n{prompt}\n\n### Response:"
```
Change `-ngl 32` to the number of layers to offload to GPU. Remove it if you don't have GPU acceleration.
Change `-c 32768` to the desired sequence length. For extended sequence models - eg 8K, 16K, 32K - the necessary RoPE scaling parameters are read from the GGUF file and set by llama.cpp automatically. Note that longer sequence lengths require much more resources, so you may need to reduce this value.
If you want to have a chat-style conversation, replace the `-p ` argument with `-i -ins`
For other parameters and how to use them, please refer to [the llama.cpp documentation](https://github.com/ggerganov/llama.cpp/blob/master/examples/main/README.md)
## How to run in `text-generation-webui`
Further instructions can be found in the text-generation-webui documentation, here: [text-generation-webui/docs/04 ‐ Model Tab.md](https://github.com/oobabooga/text-generation-webui/blob/main/docs/04%20%E2%80%90%20Model%20Tab.md#llamacpp).
## How to run from Python code
You can use GGUF models from Python using the [llama-cpp-python](https://github.com/abetlen/llama-cpp-python) or [ctransformers](https://github.com/marella/ctransformers) libraries. Note that at the time of writing (Nov 27th 2023), ctransformers has not been updated for some time and is not compatible with some recent models. Therefore I recommend you use llama-cpp-python.
### How to load this model in Python code, using llama-cpp-python
For full documentation, please see: [llama-cpp-python docs](https://abetlen.github.io/llama-cpp-python/).
#### First install the package
Run one of the following commands, according to your system:
```shell
# Base ctransformers with no GPU acceleration
pip install llama-cpp-python
# With NVidia CUDA acceleration
CMAKE_ARGS="-DLLAMA_CUBLAS=on" pip install llama-cpp-python
# Or with OpenBLAS acceleration
CMAKE_ARGS="-DLLAMA_BLAS=ON -DLLAMA_BLAS_VENDOR=OpenBLAS" pip install llama-cpp-python
# Or with CLBLast acceleration
CMAKE_ARGS="-DLLAMA_CLBLAST=on" pip install llama-cpp-python
# Or with AMD ROCm GPU acceleration (Linux only)
CMAKE_ARGS="-DLLAMA_HIPBLAS=on" pip install llama-cpp-python
# Or with Metal GPU acceleration for macOS systems only
CMAKE_ARGS="-DLLAMA_METAL=on" pip install llama-cpp-python
# In windows, to set the variables CMAKE_ARGS in PowerShell, follow this format; eg for NVidia CUDA:
$env:CMAKE_ARGS = "-DLLAMA_OPENBLAS=on"
pip install llama-cpp-python
```
#### Simple llama-cpp-python example code
```python
from llama_cpp import Llama
# Set gpu_layers to the number of layers to offload to GPU. Set to 0 if no GPU acceleration is available on your system.
llm = Llama(
model_path="./loyal-macaroni-maid-7b.Q4_K_M.gguf", # Download the model file first
n_ctx=32768, # The max sequence length to use - note that longer sequence lengths require much more resources
n_threads=8, # The number of CPU threads to use, tailor to your system and the resulting performance
n_gpu_layers=35 # The number of layers to offload to GPU, if you have GPU acceleration available
)
# Simple inference example
output = llm(
"Below is an instruction that describes a task. Write a response that appropriately completes the request.\n\n### Instruction:\n{prompt}\n\n### Response:", # Prompt
max_tokens=512, # Generate up to 512 tokens
stop=[""], # Example stop token - not necessarily correct for this specific model! Please check before using.
echo=True # Whether to echo the prompt
)
# Chat Completion API
llm = Llama(model_path="./loyal-macaroni-maid-7b.Q4_K_M.gguf", chat_format="llama-2") # Set chat_format according to the model you are using
llm.create_chat_completion(
messages = [
{"role": "system", "content": "You are a story writing assistant."},
{
"role": "user",
"content": "Write a story about llamas."
}
]
)
```
## How to use with LangChain
Here are guides on using llama-cpp-python and ctransformers with LangChain:
* [LangChain + llama-cpp-python](https://python.langchain.com/docs/integrations/llms/llamacpp)
* [LangChain + ctransformers](https://python.langchain.com/docs/integrations/providers/ctransformers)
## Discord
For further support, and discussions on these models and AI in general, join us at:
[TheBloke AI's Discord server](https://discord.gg/theblokeai)
## Thanks, and how to contribute
Thanks to the [chirper.ai](https://chirper.ai) team!
Thanks to Clay from [gpus.llm-utils.org](llm-utils)!
I've had a lot of people ask if they can contribute. I enjoy providing models and helping people, and would love to be able to spend even more time doing it, as well as expanding into new projects like fine tuning/training.
If you're able and willing to contribute it will be most gratefully received and will help me to keep providing more models, and to start work on new AI projects.
Donaters will get priority support on any and all AI/LLM/model questions and requests, access to a private Discord room, plus other benefits.
* Patreon: https://patreon.com/TheBlokeAI
* Ko-Fi: https://ko-fi.com/TheBlokeAI
**Special thanks to**: Aemon Algiz.
**Patreon special mentions**: Michael Levine, 阿明, Trailburnt, Nikolai Manek, John Detwiler, Randy H, Will Dee, Sebastain Graf, NimbleBox.ai, Eugene Pentland, Emad Mostaque, Ai Maven, Jim Angel, Jeff Scroggin, Michael Davis, Manuel Alberto Morcote, Stephen Murray, Robert, Justin Joy, Luke @flexchar, Brandon Frisco, Elijah Stavena, S_X, Dan Guido, Undi ., Komninos Chatzipapas, Shadi, theTransient, Lone Striker, Raven Klaugh, jjj, Cap'n Zoog, Michel-Marie MAUDET (LINAGORA), Matthew Berman, David, Fen Risland, Omer Bin Jawed, Luke Pendergrass, Kalila, OG, Erik Bjäreholt, Rooh Singh, Joseph William Delisle, Dan Lewis, TL, John Villwock, AzureBlack, Brad, Pedro Madruga, Caitlyn Gatomon, K, jinyuan sun, Mano Prime, Alex, Jeffrey Morgan, Alicia Loh, Illia Dulskyi, Chadd, transmissions 11, fincy, Rainer Wilmers, ReadyPlayerEmma, knownsqashed, Mandus, biorpg, Deo Leter, Brandon Phillips, SuperWojo, Sean Connelly, Iucharbius, Jack West, Harry Royden McLaughlin, Nicholas, terasurfer, Vitor Caleffi, Duane Dunston, Johann-Peter Hartmann, David Ziegler, Olakabola, Ken Nordquist, Trenton Dambrowitz, Tom X Nguyen, Vadim, Ajan Kanaga, Leonard Tan, Clay Pascal, Alexandros Triantafyllidis, JM33133, Xule, vamX, ya boyyy, subjectnull, Talal Aujan, Alps Aficionado, wassieverse, Ari Malik, James Bentley, Woland, Spencer Kim, Michael Dempsey, Fred von Graf, Elle, zynix, William Richards, Stanislav Ovsiannikov, Edmond Seymore, Jonathan Leane, Martin Kemka, usrbinkat, Enrico Ros
Thank you to all my generous patrons and donaters!
And thank you again to a16z for their generous grant.
# Original model card: Sanji Watsuki's Loyal Macaroni Maid 7B
## Description
This repository hosts quantized GGUF files for **Loyal-Macaroni-Maid-7B**, a 7B model aimed at having engaging RP with solid character card adherence and being a smart cookie at the same time.
In my limited testing, it's a great RP model suitable for RP/ERP with sharp reasoning skills for a 7B. I expect it to both benchmark well and be a very suitable model for general use.
## Prompt template: Custom format, or Alpaca
### Custom format:
I found the best SillyTavern results from using the Noromaid template.
SillyTavern config files: [Context](https://files.catbox.moe/ifmhai.json), [Instruct](https://files.catbox.moe/ttw1l9.json). Additionally, here is my [Text Completion preset](https://huggingface.co/SanjiWatsuki/Loyal-Macaroni-Maid-7B/blob/main/Characters/MinP.json)
Otherwise, I tried to ensure that most of the underlying merged models were Alpaca-ish.
### Alpaca:
```
Below is an instruction that describes a task. Write a response that appropriately completes the request.
### Instruction:
{prompt}
### Response:
```
## Helpful Tips
For SFW RP, I found that I got the most use out of this model when I had an RPG Narrator in a group chat with the characters I wanted to RP with. Here is an importable character card for the best RPG Narrator I found thus far.
For basic ChatGPT tasks, here is the basic Assistant card that I use. I found it works best with Default context template / Alpaca instruct template in Silly Tavern.
## Frankenstein's Merger
**tl;dr: This is a bunch of model merger slop with a bunch of RP cherries on top.**
I'll keep it a buck - I'm not a fan of this model's composition. Based on my testing, it seemed like models that were built from a merger of OpenChat-3.5/Starling and NeuralChat v3.1 had surprisingly good character card coherence for a 7B model - better than either one in isolation. This is validated both in my personal benchmarks as well as the [Ayumi NSFW ERP ALC-IQ3 metric](http://ayumi.m8geil.de/ayumi_bench_v3_results.html) which rates character card coherence and is dominated by OpenNeuralChat mergers for small models.
The issue is... prompt format.
OpenChat-3.5 uses an abomination of a prompt format with "GPT4 Correct User/Assistant" all over it in a ChatML-style prompt with extra tokens for padding and end-of-turn. NeuralChat v3.1 uses a weird Alpaca-like format with "### System/User/Assistant" all over it. Almost every RP finetune standardized on Alpaca or an expanded Alpaca with janky multi-turn prompting (since Alpaca doesn't have multi-turn prompting).
Most model mergers like [Q-bert/MetaMath-Cybertron-Starling](https://huggingface.co/Q-bert/MetaMath-Cybertron-Starling) just slam them together and toss the extra ChatML tokens, resulting in a half-Alpaca-like half-ChatML-like Frankenstein's monster. For the most part, using Alpaca as the lingua franca just kinda works but [there are exceptions that can make a generation go off the rails](https://huggingface.co/AIDC-ai-business/Marcoroni-7B-v3/discussions/6). I found this to be a bit of an issue in certain SillyTavern test cases.
Regardless, the strong Chat Arena performances from 7B models continues to lead me to believe they're the strongest base for an all-purpose model.
### The sauce (All You Need is DARE)
**tl;dr: It's an OpenChat/NeuralChat merger with as much RP as possible stuffed in using the DARE TIES merger method.**
This model is a DARE TIES merger between Toten5/Marcoroni-neural-chat-7B-v2, chargoddard/loyal-piano-m7, Undi95/Toppy-M-7B, NeverSleep/Noromaid-7b-v0.2, and athirdpath/NSFW_DPO_vmgb-7b on top of a mistralai/Mistral-7B-v0.1 base.
```
models:
- model: mistralai/Mistral-7B-v0.1
# no parameters necessary for base model
- model: Toten5/Marcoroni-neural-chat-7B-v2
parameters:
weight: 0.3
density: 0.8
- model: chargoddard/loyal-piano-m7
parameters:
weight: 0.4
density: 0.8
- model: Undi95/Toppy-M-7B
parameters:
weight: 0.2
density: 0.4
- model: NeverSleep/Noromaid-7b-v0.2
parameters:
weight: 0.2
density: 0.4
- model: athirdpath/NSFW_DPO_vmgb-7b
parameters:
weight: 0.2
density: 0.4
merge_method: dare_ties
base_model: mistralai/Mistral-7B-v0.1
parameters:
int8_mask: true
dtype: bfloat16
```
There's a lot to unpack here. I went with DARE TIES because it appeared to be a viable way to combine information into models without losing smarts. Directly SLERPing a smart cookie model with an ERP brained model will often dilute both the model's smarts and RPing ability. This is an attempt to have my cookie and eat it, too.
First, there are two high density high weight models:
[chargoddard/loyal-piano-m7](https://huggingface.co/chargoddard/loyal-piano-m7) is the easy primary model choice. It's an Alpaca prompt format model that scores highly, is very creative for a 7B, and is primarily trained on RP data.
[Toten5/Marcoroni-neural-chat-7B-v2](https://huggingface.co/Toten5/Marcoroni-neural-chat-7B-v2) is the unintuitive second model pick. It is a merger of mergers that chains back to being an OpenChat/NeuralChat merger being SLERPed back into NeuralChat a second time. Despite SLERPing NeuralChat in multiple times, it retains its high benchmark scores. I opted to pick this model as my base because I believed it was the OpenChat/NeuralChat model that benchmarked well that was closest to the O.G. NeuralChat which has the most Alpaca-like prompt.
By picking a density of 0.8, these models have a 96% chance of showing up for any TIE merger. This should ensure that there is a solid "base" of deltas from the base Mistral model that captures most of what makes these models good. High density with 0.3-0.4 weights have been shown in mergers like [jan-hq/supermario-v2](https://huggingface.co/jan-hq/supermario-v2)
Next, there are 3 RP models merged in with medium density. [Undi95/Toppy-M-7B](https://huggingface.co/Undi95/Toppy-M-7B), [NeverSleep/Noromaid-7b-v0.2](https://huggingface.co/NeverSleep/Noromaid-7b-v0.2), and [athirdpath/NSFW_DPO_vmgb-7b](https://huggingface.co/athirdpath/NSFW_DPO_vmgb-7b). Toppy-M-7B is an easy pick for being a well regarded 7B RP model - although, it is a merger of many mergers which might dilute its effectiveness as a lower density merge. NeverSleep/Noromaid-7b-v0.2 pulls in the unique private Noromaid RP dataset. Finally, athirdpath/NSFW_DPO_vmgb-7b is another Frankenstein OpenNeuralChat merger that happens to be DPOed on athirdpath's NSFW Alpaca pairs which seemed like another good RP addition to the model (plus, maybe it tilts it to being more Alpaca-flavored, idk).
By picking a density of 0.4, these models should *largely* impart some of their flavor onto the merger. I suspect the density could go even lower and the models could be used even more like a LoRA-like merger on top.
The DARE TIES merger is intentionally overweight and non-normalized at 1.3 total weight. I intentionally went overweight to try and better capture the individual characteristics from the various models. With wide mergers, a weight of 1.0 can often become incoherent like [jan-hq/supermario-v1](https://huggingface.co/jan-hq/supermario-v1).
Putting it all together, ~60% of the model is "base models" like OpenChat/NeuralChat/Loyal-Piano-M7. ~40% of the model is effectively me trying to extract RP information from existing RP models. The only non-RP model is the Marcoroni base which means that almost 80% of this model is intended for RP.
Not that the benchmarks matter, but if this merger works right, it'll be a high benchmarking 7B that is both smart and strong at RP.