File size: 4,422 Bytes
45444a8
 
801cf6d
 
 
 
 
 
 
 
45444a8
 
2385765
b7577f8
2385765
b7577f8
250c782
801cf6d
 
812c543
2385765
812c543
2385765
812c543
2385765
812c543
2385765
812c543
801cf6d
 
 
 
b7577f8
 
 
4aebaf7
 
 
 
 
 
 
d385b66
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b7577f8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5d748fc
 
ac4bde9
5d748fc
 
 
 
 
 
 
 
 
 
 
 
801cf6d
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
---
license: apache-2.0
datasets:
- LeoLM/OpenSchnabeltier
language:
- de
pipeline_tag: conversational
tags:
- hermeo
- laser
---

(Evaluation WIP)

## Hermes + Leo + German Laser = Germeo

## Germeo-7B-Laser
A German-English understanding, but German-only speaking model merged from Hermeo-7B.

### Model details

**Merged from**: leo-mistral-hessianai-7b-chat and DPOpenHermes-7B-v2

**Model type**: Causal decoder-only transformer language model

**Languages**: German replies with English Understanding Capabilities

**Laser-Data**: LeoLM/OpenSchnabeltier


This is an early experiment on laser and its influence on language understanding. It generally improves the language understanding capabilities.
The hypothesis is that it degrades the probability of English replies and increasing those of German replies. The models internal German capabilities are boosted. 

Will keep you updated..

### Acknowledgements:

I would like to thank everyone that participated in making this model and its training possible:
To [@malteos](https://huggingface.co/malteos) for hermeo
To [@cognitivecomputations](https://huggingface.co/cognitivecomputations) and Fernando Fernandes Neto for their implementation of LASER
To [@LeoLM](https://huggingface.co/LeoLM) and Björn for the OpenSchnabeltier dataset.


### Prompt format:

```python
streamer = TextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
# Convert prompt to tokens
prompt_template = """<|im_start|>system
Du bist ein hilfreicher Assistent.<|im_end|>
<|im_start|>user
{prompt}<|im_end|>
<|im_start|>assistant"""

prompt = "Schreibe eine Stellenanzeige für Data Scientist bei AXA!"

final_prompt = prompt_template.format(prompt=prompt)
```

### German benchmarks

| **German tasks:**             | **MMLU-DE**    | **Hellaswag-DE** | **ARC-DE**      |**Average**      |
|-------------------------------|-------------|---------------|--------------|--------------|
| **Models / Few-shots:**       | _(5 shots)_ | _(10 shots)_  | _(24 shots)_ | |
| _7B parameters_      |  | |  | |
| llama-2-7b                    | 0.400       | 0.513         | 0.381        | 0.431  |
| leo-hessianai-7b              | 0.400       | 0.609         | 0.429        | 0.479 |
| bloom-6b4-clp-german          | 0.274       | 0.550         | 0.351        | 0.392 |
| mistral-7b                    | **0.524**       | 0.588         | 0.473        | 0.528 |
| leo-mistral-hessianai-7b      | 0.481       | 0.663         | 0.485        | 0.543 |
| leo-mistral-hessianai-7b-chat | 0.458       | 0.617         | 0.465        | 0.513 |
| DPOpenHermes-7B-v2            | 0.517         | 0.603         | 0.515        | 0.545 |
| hermeo-7b                     | 0.511       | **0.668**         | **0.528**        | **0.569** |
| **germeo-7b-laser (this model)**| ?       | ?        | ?      | ? |
| _13B parameters_      |  | |  | |
| llama-2-13b                    | 0.469       | 0.581        | 0.468        | 0.506 |
| leo-hessianai-13b              | **0.486**       | **0.658**         | **0.509**       | **0.551** |
| _70B parameters_      |  | |  | |
| llama-2-70b                    | 0.597       | 0.674       | 0.561       | 0.611 |
| leo-hessianai-70b              | **0.653**       | **0.721**         | **0.600**       | **0.658** |


Even though the model does not generate English text without being explicitly asked, performance on English Benchmarks is still up:

### English benchmarks

| **English tasks:**                 | **MMLU**    | **Hellaswag** | **ARC**      | **Average** |
|------------------------------------|-------------|---------------|--------------|-------------|
| **Models / Few-shots:**            | _(5 shots)_ | _(10 shots)_  | _(24 shots)_ |             |
| llama-2-7b                         |       0.466 |         0.786 |        0.530 |       0.594 |
| leolm-hessianai-7b                 |       0.423 |         0.759 |        0.522 |       0.568 |
| bloom-6b4-clp-german               |       0.264 |         0.525 |        0.328 |       0.372 |
| mistral-7b                         |   **0.635** |     **0.832** |        0.607 |   **0.691** |
| leolm-mistral-hessianai-7b         |       0.550 |         0.777 |        0.518 |       0.615 |
| hermeo-7b                          |       0.601 |         0.821 |    **0.620** |       0.681 |
| germeo-7b-laser (this model)       |       0.601 |         0.828 |        0.608 |       0.679 |