Blizado commited on
Commit
ec18f1c
1 Parent(s): e2e1628

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +14 -3
README.md CHANGED
@@ -3,11 +3,11 @@ base_model: []
3
  tags:
4
  - mergekit
5
  - merge
6
-
7
  ---
8
  # merge
9
 
10
- This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).
11
 
12
  ## Merge Details
13
  ### Merge Method
@@ -17,8 +17,19 @@ This model was merged using the SLERP merge method.
17
  ### Models Merged
18
 
19
  The following models were included in the merge:
20
- * [DiscoResearch/DiscoLM_German_7b_v1](https://huggingface.co/DiscoResearch/DiscoLM_German_7b_v1)
21
  * [OpenPipe/mistral-ft-optimized-1227](https://huggingface.co/OpenPipe/mistral-ft-optimized-1227)
 
 
 
 
 
 
 
 
 
 
 
 
22
 
23
  ### Configuration
24
 
 
3
  tags:
4
  - mergekit
5
  - merge
6
+ - + german
7
  ---
8
  # merge
9
 
10
+ This is a experimental merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).
11
 
12
  ## Merge Details
13
  ### Merge Method
 
17
  ### Models Merged
18
 
19
  The following models were included in the merge:
 
20
  * [OpenPipe/mistral-ft-optimized-1227](https://huggingface.co/OpenPipe/mistral-ft-optimized-1227)
21
+ * [DiscoResearch/DiscoLM_German_7b_v1](https://huggingface.co/DiscoResearch/DiscoLM_German_7b_v1)
22
+
23
+ ####Why this two models?
24
+
25
+ Because both used models are, to my knowledge, the two best models when it comes to German language generation.
26
+
27
+ DiscoLM German 7B is is up to this date (01/21/2024) by far the best German model and makes far fewer grammatical errors and his German generally sounds good. But it is finetuned on Mistral V0.2 or even V0.1.
28
+
29
+ Mistral FT Optimized 1227 is much better in German than Mistral 7B V0.2 and other German fine-tuning models that make grammar errors in almost every sentence. But even that model is a good step behind DiscoLM German 7B and creates not so well formed German sentences.
30
+
31
+ The ulterior motive was now combining this two models to get a even better German model, especially for German roleplay.
32
+
33
 
34
  ### Configuration
35