Hugging Face's logo

zarakiquemparte
/
zarablend-l2-7b

Text Generation
Transformers
PyTorch
llama
llama2
text-generation-inference
Inference Endpoints
Model card Files Community
1
zarablend-l2-7b
File size: 2,238 Bytes 
66db51f
e788e90
66db51f
 
 
 
 
 
 
 
 
 
8b14e71
 
 
 
66db51f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
df84a3a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
 
---
license: other
tags:
- llama2
---
# Model Card: Zarablend L2 7b
This model uses [Nous Hermes Llama2 7b](https://huggingface.co/NousResearch/Nous-Hermes-llama-2-7b) (66%) as a base with [Airoboros L2 7B GPT4 2.0](https://huggingface.co/jondurbin/airoboros-l2-7b-gpt4-2.0) (34%) and the result of this merge was merged with [LimaRP LLama2 7B Lora](https://huggingface.co/lemonilia/limarp-llama2).

This merge of models(hermes and airoboros) was done with this [script](https://github.com/zarakiquemparte/zaraki-tools/blob/main/merge-cli.py)

This merge of Lora with Model was done with this [script](https://github.com/zarakiquemparte/zaraki-tools/blob/main/apply-lora.py)

Quantized Model by @TheBloke:
- [GGML](https://huggingface.co/TheBloke/Zarablend-L2-7B-GGML)
- [GPTQ](https://huggingface.co/TheBloke/Zarablend-L2-7B-GPTQ)

Merge illustration:

![illustration](zarablend-merge-illustration.png)

## Usage:

Since this is a merge between Nous Hermes, Airoboros and LimaRP, the following instruction formats should work:

Alpaca 2:

```
### Instruction:
<prompt>

### Response:
<leave a newline blank for model to respond>
```

LimaRP instruction format:

```
<<SYSTEM>>
<character card and system prompt>

<<USER>>
<prompt>

<<AIBOT>>
<leave a newline blank for model to respond>
```

## Bias, Risks, and Limitations

This model is not intended for supplying factual information or advice in any form

## Training Details

This model is merged and can be reproduced using the tools mentioned above. Please refer to all provided links for extra model-specific details.
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_zarakiquemparte__zarablend-l2-7b)

| Metric                | Value                     |
|-----------------------|---------------------------|
| Avg.                  | 49.03   |
| ARC (25-shot)         | 54.44          |
| HellaSwag (10-shot)   | 78.62    |
| MMLU (5-shot)         | 47.61         |
| TruthfulQA (0-shot)   | 49.38   |
| Winogrande (5-shot)   | 73.32   |
| GSM8K (5-shot)        | 4.4        |
| DROP (3-shot)         | 35.45         |