File size: 3,118 Bytes
543df77 012cba4 6ed50c5 453e5aa 7aa6481 543df77 |
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 |
---
tags:
- mergekit
- merge
library_name: transformers
---
<img src="https://huggingface.co/Virt-io/FuseChat-Kunoichi-10.7B/resolve/main/README.png">
---
# Quantizations
[GGUF by Lewdiculous](https://huggingface.co/Lewdiculous/FuseChat-Kunoichi-10.7B-GGUF-IQ-Imatrix)
---
Merged [Kunoichi-DPO-v2-7B](https://huggingface.co/SanjiWatsuki/Kunoichi-DPO-v2-7B) into [FuseChat-7B-Varm](https://huggingface.co/FuseAI/FuseChat-7B-VaRM) to fix the GPTism
The idea was to keep FuseChat's smarts since from my testing it was amazing, just a little stubborn for RP
Silly tavern [preset](https://huggingface.co/Virt-io/FuseChat-7B-VaRM-GGUF/tree/main/presets)
Merge template copied from [TheProfessor](https://huggingface.co/abacusai/TheProfessor-155b)
---
base_model:
- FuseAI/FuseChat-7B-VaRM
- SanjiWatsuki/Kunoichi-DPO-v2-7B
---
# FUSECHAT-VaRM-Kunoichi-10.7b.v1
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 [linear](https://arxiv.org/abs/2203.05482) merge method.
### Models Merged
The following models were included in the merge:
* [FuseAI/FuseChat-7B-VaRM](https://huggingface.co/FuseAI/FuseChat-7B-VaRM)
* [SanjiWatsuki/Kunoichi-DPO-v2-7B](https://huggingface.co/SanjiWatsuki/Kunoichi-DPO-v2-7B)
### Configuration
The following YAML configuration was used to produce this model:
```yaml
merge_method: linear # use linear so we can include multiple models, albeit at a zero weight
parameters:
weight: 1.0 # weight everything as 1 unless specified otherwise - linear with one model weighted at 1 is a no-op like passthrough
slices:
- sources:
- model: FuseAI/FuseChat-7B-VaRM # embed_tokens comes along with the ride with whatever is the first layer
layer_range: [0, 1]
- model: SanjiWatsuki/Kunoichi-DPO-v2-7B # add dummy second model with 0 weight so tokenizer-based merge routine is invoked for embed_tokens
layer_range: [0, 1]
parameters:
weight: 0
- sources:
- model: FuseAI/FuseChat-7B-VaRM
layer_range: [1, 5]
- sources:
- model: SanjiWatsuki/Kunoichi-DPO-v2-7B
layer_range: [5, 7] # 2 layers
- sources:
- model: FuseAI/FuseChat-7B-VaRM
layer_range: [5, 15]
- sources:
- model: SanjiWatsuki/Kunoichi-DPO-v2-7B
layer_range: [15, 27] # 12 layers
- sources:
- model: FuseAI/FuseChat-7B-VaRM
layer_range: [15, 27]
- sources:
- model: SanjiWatsuki/Kunoichi-DPO-v2-7B
layer_range: [27, 29] # 2 layers
- sources:
- model: FuseAI/FuseChat-7B-VaRM
layer_range: [27, 31]
- sources: # same as above, but for lm_head with the last layer
- model: FuseAI/FuseChat-7B-VaRM
layer_range: [31, 32]
- model: SanjiWatsuki/Kunoichi-DPO-v2-7B
layer_range: [31, 32]
parameters:
weight: 0
dtype: float16
tokenizer_source: model:FuseAI/FuseChat-7B-VaRM # keep exact tokenizer used by dolphin - or you could use `union` if you add all of the input models to the first/last slice
``` |