85 lines
5.5 KiB
Markdown
85 lines
5.5 KiB
Markdown
Quantization made by Richard Erkhov.
|
|
|
|
[Github](https://github.com/RichardErkhov)
|
|
|
|
[Discord](https://discord.gg/pvy7H8DZMG)
|
|
|
|
[Request more models](https://github.com/RichardErkhov/quant_request)
|
|
|
|
|
|
llama-3-merged-linear - GGUF
|
|
- Model creator: https://huggingface.co/uygarkurt/
|
|
- Original model: https://huggingface.co/uygarkurt/llama-3-merged-linear/
|
|
|
|
|
|
| Name | Quant method | Size |
|
|
| ---- | ---- | ---- |
|
|
| [llama-3-merged-linear.Q2_K.gguf](https://huggingface.co/RichardErkhov/uygarkurt_-_llama-3-merged-linear-gguf/blob/main/llama-3-merged-linear.Q2_K.gguf) | Q2_K | 2.96GB |
|
|
| [llama-3-merged-linear.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/uygarkurt_-_llama-3-merged-linear-gguf/blob/main/llama-3-merged-linear.IQ3_XS.gguf) | IQ3_XS | 3.28GB |
|
|
| [llama-3-merged-linear.IQ3_S.gguf](https://huggingface.co/RichardErkhov/uygarkurt_-_llama-3-merged-linear-gguf/blob/main/llama-3-merged-linear.IQ3_S.gguf) | IQ3_S | 3.43GB |
|
|
| [llama-3-merged-linear.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/uygarkurt_-_llama-3-merged-linear-gguf/blob/main/llama-3-merged-linear.Q3_K_S.gguf) | Q3_K_S | 3.41GB |
|
|
| [llama-3-merged-linear.IQ3_M.gguf](https://huggingface.co/RichardErkhov/uygarkurt_-_llama-3-merged-linear-gguf/blob/main/llama-3-merged-linear.IQ3_M.gguf) | IQ3_M | 3.52GB |
|
|
| [llama-3-merged-linear.Q3_K.gguf](https://huggingface.co/RichardErkhov/uygarkurt_-_llama-3-merged-linear-gguf/blob/main/llama-3-merged-linear.Q3_K.gguf) | Q3_K | 3.74GB |
|
|
| [llama-3-merged-linear.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/uygarkurt_-_llama-3-merged-linear-gguf/blob/main/llama-3-merged-linear.Q3_K_M.gguf) | Q3_K_M | 3.74GB |
|
|
| [llama-3-merged-linear.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/uygarkurt_-_llama-3-merged-linear-gguf/blob/main/llama-3-merged-linear.Q3_K_L.gguf) | Q3_K_L | 4.03GB |
|
|
| [llama-3-merged-linear.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/uygarkurt_-_llama-3-merged-linear-gguf/blob/main/llama-3-merged-linear.IQ4_XS.gguf) | IQ4_XS | 4.18GB |
|
|
| [llama-3-merged-linear.Q4_0.gguf](https://huggingface.co/RichardErkhov/uygarkurt_-_llama-3-merged-linear-gguf/blob/main/llama-3-merged-linear.Q4_0.gguf) | Q4_0 | 4.34GB |
|
|
| [llama-3-merged-linear.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/uygarkurt_-_llama-3-merged-linear-gguf/blob/main/llama-3-merged-linear.IQ4_NL.gguf) | IQ4_NL | 4.38GB |
|
|
| [llama-3-merged-linear.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/uygarkurt_-_llama-3-merged-linear-gguf/blob/main/llama-3-merged-linear.Q4_K_S.gguf) | Q4_K_S | 4.37GB |
|
|
| [llama-3-merged-linear.Q4_K.gguf](https://huggingface.co/RichardErkhov/uygarkurt_-_llama-3-merged-linear-gguf/blob/main/llama-3-merged-linear.Q4_K.gguf) | Q4_K | 4.58GB |
|
|
| [llama-3-merged-linear.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/uygarkurt_-_llama-3-merged-linear-gguf/blob/main/llama-3-merged-linear.Q4_K_M.gguf) | Q4_K_M | 4.58GB |
|
|
| [llama-3-merged-linear.Q4_1.gguf](https://huggingface.co/RichardErkhov/uygarkurt_-_llama-3-merged-linear-gguf/blob/main/llama-3-merged-linear.Q4_1.gguf) | Q4_1 | 4.78GB |
|
|
| [llama-3-merged-linear.Q5_0.gguf](https://huggingface.co/RichardErkhov/uygarkurt_-_llama-3-merged-linear-gguf/blob/main/llama-3-merged-linear.Q5_0.gguf) | Q5_0 | 5.21GB |
|
|
| [llama-3-merged-linear.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/uygarkurt_-_llama-3-merged-linear-gguf/blob/main/llama-3-merged-linear.Q5_K_S.gguf) | Q5_K_S | 5.21GB |
|
|
| [llama-3-merged-linear.Q5_K.gguf](https://huggingface.co/RichardErkhov/uygarkurt_-_llama-3-merged-linear-gguf/blob/main/llama-3-merged-linear.Q5_K.gguf) | Q5_K | 5.34GB |
|
|
| [llama-3-merged-linear.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/uygarkurt_-_llama-3-merged-linear-gguf/blob/main/llama-3-merged-linear.Q5_K_M.gguf) | Q5_K_M | 5.34GB |
|
|
| [llama-3-merged-linear.Q5_1.gguf](https://huggingface.co/RichardErkhov/uygarkurt_-_llama-3-merged-linear-gguf/blob/main/llama-3-merged-linear.Q5_1.gguf) | Q5_1 | 5.65GB |
|
|
| [llama-3-merged-linear.Q6_K.gguf](https://huggingface.co/RichardErkhov/uygarkurt_-_llama-3-merged-linear-gguf/blob/main/llama-3-merged-linear.Q6_K.gguf) | Q6_K | 6.14GB |
|
|
| [llama-3-merged-linear.Q8_0.gguf](https://huggingface.co/RichardErkhov/uygarkurt_-_llama-3-merged-linear-gguf/blob/main/llama-3-merged-linear.Q8_0.gguf) | Q8_0 | 7.95GB |
|
|
|
|
|
|
|
|
|
|
Original model description:
|
|
---
|
|
library_name: transformers
|
|
license: mit
|
|
---
|
|
|
|
# LLM Model Merging
|
|
|
|
## YouTube Tutorial
|
|
<div align="center">
|
|
<a href="https://youtu.be/gNXBp3wttFU">Model Merging: Merge LLMs to Create Frankestein Models - Python, HuggingFace, Mergekit</a>
|
|
<br>
|
|
<br>
|
|
<a href="https://youtu.be/gNXBp3wttFU">
|
|
<img src="./thumbnail1-button.png" height="85%" width="85%%"/>
|
|
</a>
|
|
</div>
|
|
|
|
## GitHub
|
|
You can find the GitHub from here; https://github.com/uygarkurt/Model-Merge
|
|
|
|
In this specific case, I typed `llama-3` into the open LLM leaderboard, took the best 3 models, merged them and created
|
|
a better ranking model wihtout any training.
|
|
|
|
As the main libraries we will be using [mergekit](https://github.com/arcee-ai/mergekit).
|
|
|
|
<br/>
|
|
<div align="center">
|
|
<a href="">
|
|
<img alt="open-source-image"
|
|
src="https://img.shields.io/badge/%E2%9D%A4%EF%B8%8F_Open_Source-%2350C878?style=for-the-badge"/>
|
|
</a>
|
|
<a href="https://youtu.be/gNXBp3wttFU">
|
|
<img alt="youtube-tutorial"
|
|
src="https://img.shields.io/badge/YouTube_Tutorial-grey?style=for-the-badge&logo=YouTube&logoColor=%23FF0000"/>
|
|
</a>
|
|
<a href="https://github.com/uygarkurt/Model-Merge">
|
|
<img alt="github-image"
|
|
src="https://img.shields.io/badge/github-%23121011.svg?style=for-the-badge&logo=github&logoColor=white"
|
|
</a>
|
|
</div>
|
|
|