johannhartmann commited on
Commit
90c6620
·
verified ·
1 Parent(s): 3179d3c

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +9 -7
README.md CHANGED
@@ -22,15 +22,18 @@ language:
22
 
23
  ![image/png](https://huggingface.co/mayflowergmbh/Wiedervereinigung-7b/resolve/main/Wiedervereinigung-7b.png)
24
 
25
- This is a dpo aligned merge of multiple german models scoring 7.1 on the mt-bench-de average.
26
- It is a merge of the best german 7B models with 7b parameters as a dare_ties merge.
27
  Since the original models based on mistral - three of them on the brilliant german LeoLM/leo-mistral-hessianai-7b - they are reunited in this merged model.
28
- Therefore the name, no nationalist ideas involved. To improve result quality they are dpo-trained with a german translation of intel-orca-dpo
29
- using our german fork of [LLaMA-Factory](https://github.com/mayflower/LLaMA-Factory-de).
30
 
31
- ## mt-bench-de
 
32
 
33
- Is the merged model good? Well, of course. But it is even better with the help of some dpo tuning.
 
 
 
 
34
 
35
  ```json
36
  {
@@ -57,7 +60,6 @@ Wiedervereinigung-7b itself is a [LazyMergekit](https://colab.research.google.c
57
  * [VAGOsolutions/SauerkrautLM-7b-v1-mistral](https://huggingface.co/VAGOsolutions/SauerkrautLM-7b-v1-mistral)
58
  * [malteos/hermeo-7b](https://huggingface.co/malteos/hermeo-7b)
59
 
60
- All the actual heavylifting has been done by the creators of these models.
61
 
62
  ## 🧩 Configuration
63
 
 
22
 
23
  ![image/png](https://huggingface.co/mayflowergmbh/Wiedervereinigung-7b/resolve/main/Wiedervereinigung-7b.png)
24
 
25
+ This is a dpo aligned merge of our favourite german models, scoring 7.11 on the mt-bench-de average.
 
26
  Since the original models based on mistral - three of them on the brilliant german LeoLM/leo-mistral-hessianai-7b - they are reunited in this merged model.
27
+ Therefore the name, no nationalist ideas involved :-). To improve result quality they are dpo-trained with a german translation of intel-orca-dpo we added support for 10 german datasets to [LLaMA-Factory](https://github.com/hiyouga/LLaMA-Factory).
 
28
 
29
+ Kudos to the authors of the original models at [DiscoResearch](https://huggingface.co/DiscoResearch) and [VAGOsolutions](https://huggingface.co/VAGOsolutions), [Malte Ostendorff](https://huggingface.co/malteos)
30
+ and [Matthias Uhlig](https://huggingface.co/DRXD1000). We are your fan club.
31
 
32
+ This model was brought to you and the nvidia bill was paid by [Mayflower GmbH](https://mayflower.de/).
33
+
34
+ ## Benchmark results: mt-bench-de
35
+
36
+ Is the merged model alone already good? Well, of course. But it is even better with the help of some dpo tuning.
37
 
38
  ```json
39
  {
 
60
  * [VAGOsolutions/SauerkrautLM-7b-v1-mistral](https://huggingface.co/VAGOsolutions/SauerkrautLM-7b-v1-mistral)
61
  * [malteos/hermeo-7b](https://huggingface.co/malteos/hermeo-7b)
62
 
 
63
 
64
  ## 🧩 Configuration
65