File size: 4,301 Bytes
1c83592 9456956 1c83592 0cf32e7 1c83592 e9efe6d |
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 |
---
base_model:
- anthracite-org/magnum-v3-9b-customgemma2
- nbeerbower/gemma2-gutenberg-9B
- grimjim/Magnolia-v1-Gemma2-8k-9B
- UCLA-AGI/Gemma-2-9B-It-SPPO-Iter3
- BeaverLegacy/Smegmma-Deluxe-9B-v1
- ifable/gemma-2-Ifable-9B
library_name: transformers
tags:
- mergekit
- merge
---
# merge
This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).
## Merge Details
### Merge Method
This model was merged using the SLERP method to create an intermediate model. I used the
[Model Stock](https://arxiv.org/abs/2403.19522) merge method after, using the SLERP model as a base.
The idea was to make a nice and smart base model and add in a few pinches of spice.
For some reason it wouldn't let me use any other merge method- it gave me ModelReference errors about my
intermediary model for every method except Model Stock for some reason. I'll see if I can fix it and
upload my intended task-arithmetic version as a v2.
This is the only one of my like 700 merges that I think uses something novel/interesting
enough in its creation to merit an upload.
Named after the **aster**, a purple-violet star-shaped perennial flower. It's pretty and has a huge family, much like this model.
### Models Merged
The following models were included in the merge:
* [anthracite-org/magnum-v3-9b-customgemma2](https://huggingface.co/anthracite-org/magnum-v3-9b-customgemma2)
* [nbeerbower/gemma2-gutenberg-9B](https://huggingface.co/nbeerbower/gemma2-gutenberg-9B)
* [grimjim/Magnolia-v1-Gemma2-8k-9B](https://huggingface.co/grimjim/Magnolia-v1-Gemma2-8k-9B)
* [UCLA-AGI/Gemma-2-9B-It-SPPO-Iter3](https://huggingface.co/UCLA-AGI/Gemma-2-9B-It-SPPO-Iter3)
* [BeaverLegacy/Smegmma-Deluxe-9B-v1](https://huggingface.co/BeaverLegacy/Smegmma-Deluxe-9B-v1)
* [ifable/gemma-2-Ifable-9B](https://huggingface.co/ifable/gemma-2-Ifable-9B)
### Configuration
The following YAML configuration was used to produce this model:
```yaml
# THIS YAML CONFIGURATION WAS USED TO CREATE THE INTERMEDIARY MODEL.
# slices:
# - sources:
# - model: anthracite-org/magnum-v3-9b-customgemma2
# layer_range: [0, 42]
# - model: nbeerbower/gemma2-gutenberg-9B
# layer_range: [0, 42]
# merge_method: slerp
# base_model: nbeerbower/gemma2-gutenberg-9B
# parameters:
# t:
# - filter: self_attn
# value: [0.2, 0.5, 0.4, 0.7, 1]
# - filter: mlp
# value: [1, 0.5, 0.3, 0.4, 0.2]
# - value: 0.5
# dtype: float16
# THIS YAML CONFIGURATION WAS USED TO CREATE ASTER. The E: model is the intermediate
# model created in the previous config.
models:
- model: E:/models/mergekit/output/intermediate/
- model: BeaverLegacy/Smegmma-Deluxe-9B-v1
parameters:
weight: 0.3
- model: ifable/gemma-2-Ifable-9B
parameters:
weight: 0.3
- model: UCLA-AGI/Gemma-2-9B-It-SPPO-Iter3
parameters:
weight: 0.15
- model: grimjim/Magnolia-v1-Gemma2-8k-9B
parameters:
weight: 0.25
merge_method: model_stock
base_model: E:/models/mergekit/output/intermediate/
dtype: float16
```
Alright, now back to smashing models together and seeing what happens... |