File size: 1,748 Bytes
0b15696
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
base_model:
- grimjim/zephyr-beta-wizardLM-2-merge-7B
- alpindale/Mistral-7B-v0.2-hf
library_name: transformers
tags:
- mergekit
- merge
license: cc-by-nc-4.0
pipeline_tag: text-generation
---
# madwind-wizard-7B

This is a merge of pre-trained 7B language models created using [mergekit](https://github.com/cg123/mergekit).

The intended goal of this merge was to combine the 32K context window of Mistral v0.2 base with the richness and strength of the Zephyr Beta and WizardLM 2 models. This was a mixed-precision merge, promoting Mistral v0.2 base from fp16 to bf16.

The result can be used for text generation. Note that Zephyr Beta training removed in-built alignment from datasets, resulting in a model more likely to generate problematic text when prompted. This merge appears to have inherited that feature.

- Full weights: [grimjim/madwind-wizard-7B](https://huggingface.co/grimjim/madwind-wizard-7B)
- GGUF quants: [grimjim/madwind-wizard-7B-GGUF](https://huggingface.co/grimjim/madwind-wizard-7B-GGUF)

## Merge Details
### Merge Method

This model was merged using the SLERP merge method.

### Models Merged

The following models were included in the merge:
* [grimjim/zephyr-beta-wizardLM-2-merge-7B](https://huggingface.co/grimjim/zephyr-beta-wizardLM-2-merge-7B)
* [alpindale/Mistral-7B-v0.2-hf](https://huggingface.co/alpindale/Mistral-7B-v0.2-hf)

### Configuration

The following YAML configuration was used to produce this model:

```yaml
slices:
  - sources:
    - model: alpindale/Mistral-7B-v0.2-hf
      layer_range: [0,32]
    - model: grimjim/zephyr-beta-wizardLM-2-merge-7B
      layer_range: [0,32]
merge_method: slerp
base_model: alpindale/Mistral-7B-v0.2-hf
parameters:
  t:
    - value: 0.5
dtype: bfloat16

```