File size: 3,667 Bytes
ef5db9b 3d4f827 ef5db9b |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 |
---
base_model:
- grimjim/llama-3-merge-virt-req-8B
library_name: transformers
pipeline_tag: text-generation
tags:
- mergekit
- merge
- facebook
- meta
- pytorch
- llama
- llama-3
license: other
license_name: llama3
license_link: LICENSE
---
# ⚡ExLlamaV2 quant of : [Llama-3-8B-Irene-v0.2](https://huggingface.co/Virt-io/Llama-3-8B-Irene-v0.2)
> [!note]
> ➡️ **Exl2 version** : [0.0.20](https://github.com/turboderp/exllamav2/releases/tag/v0.0.20)<br/>
> ➡️ **Cal. dataset** : Default.<br/>
> 📄 <a href="https://huggingface.co/Meggido/Llama-3-8B-Irene-v0.2-6.5bpw-h8-exl2/resolve/main/measurement.json" download>Measurement.json</a> file.
> [!IMPORTANT]
> Quants:<br>
> [mradermacher/Llama-3-8B-Irene-v0.2-GGUF](https://huggingface.co/mradermacher/Llama-3-8B-Irene-v0.2-GGUF)<br>
> [mradermacher/Llama-3-8B-Irene-v0.2-i1-GGUF](https://huggingface.co/mradermacher/Llama-3-8B-Irene-v0.2-i1-GGUF)<br>
<img src="https://huggingface.co/Virt-io/Llama-3-8B-Irene-v0.2/resolve/main/Gnome.png">
# Llama-3-8B-Irene-v0.2
Mergin' o' models, ye say? Well, that be a task fit fer a clever gnome like meself! When combinin' similar models, I like to use model stock tae bring 'em together. And when I'm slerpin', I makes sure tae use a gradient that tapers off at both ends. That way, the model stays mostly uncensored, ye see.
Now, if I'm mergin' two uncensored models with Slerp, I just favors the one I want more o'! But when it comes tae makin' the gradient, I likes tae get wild and fluctuate between low and high values, ye know what I mean? It's like addin' a bit o' magic tae the mix, helps keep the results from gettin' too boring.
Course, this be just one gnome's way o' doin' things. I'm sure there be other clever methods out there
## Merge Details
### Merge Method
This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).
This model was merged using the SLERP merge method.
### Models Merged
The following models were included in the merge:
* Mergekit/llama3-SOVL-v1
* [grimjim/llama-3-merge-virt-req-8B](https://huggingface.co/grimjim/llama-3-merge-virt-req-8B)
* NousResearch/Meta-Llama-3-8B-Instruct
* Locutusque/llama-3-neural-chat-v2.2-8B
* NousResearch/Hermes-2-Pro-Llama-3-8B
* rombodawg/Llama-3-8B-Instruct-Coder-v2
* aaditya/Llama3-OpenBioLLM-8B
* ResplendentAI/SOVL_Llama3_8B
* openlynn/Llama-3-Soliloquy-8B-v2
* grimjim/llama-3-merge-pp-instruct-8B
* ChaoticNeutrals/Poppy_Porpoise-0.72-L3-8B
### Configuration
The following YAML configuration was used to produce this model:
```yaml
slices:
- sources:
- model: grimjim/llama-3-merge-virt-req-8B
layer_range: [0, 32]
- model: Mergekit/llama3-SOVL-v1
layer_range: [0, 32]
merge_method: slerp
base_model: grimjim/llama-3-merge-virt-req-8B
parameters:
t:
- value: [0.5, 0.35, 0.55, 0.35, 0.75, 0.35, 0.90, 0.35, 0.75, 0.35, 0.55, 0.35, 0.5]
dtype: bfloat16
```
# llama3-SOVL-v1
```
slices:
- sources:
- model: Mergekit/SMART-CODER
layer_range: [0, 32]
- model: ResplendentAI/SOVL_Llama3_8B
layer_range: [0, 32]
merge_method: slerp
base_model: Mergekit/SMART-CODER
parameters:
t:
- value: [0.90, 0.55, 0.75, 0.35, 0.45, 0.90, 0.25, 0.90, 0.45, 0.35, 0.75, 0.55, 0.90]
dtype: bfloat16
```
# SMART-CODER
```
models:
- model: NousResearch/Meta-Llama-3-8B-Instruct
- model: Locutusque/llama-3-neural-chat-v2.2-8B
- model: NousResearch/Hermes-2-Pro-Llama-3-8B
- model: rombodawg/Llama-3-8B-Instruct-Coder-v2
- model: aaditya/Llama3-OpenBioLLM-8B
merge_method: model_stock
base_model: NousResearch/Meta-Llama-3-8B-Instruct
dtype: bfloat16
``` |