File size: 3,329 Bytes
436641d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
06560fc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
436641d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license: apache-2.0
tags:
- merge
- mergekit
- mistral
- 7b
- lazymergekit
- mistralai/Mistral-7B-Instruct-v0.2
- mlabonne/NeuralHermes-2.5-Mistral-7B
---

# NeuralHermes-2.5-Mistral-7B-Mistral-7B-Instruct-v0.2-slerp

NeuralHermes-2.5-Mistral-7B-Mistral-7B-Instruct-v0.2-slerp is a merge of the following models:
* [mistralai/Mistral-7B-Instruct-v0.2](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2)
* [mlabonne/NeuralHermes-2.5-Mistral-7B](https://huggingface.co/mlabonne/NeuralHermes-2.5-Mistral-7B)

## Eval

```
|      Groups      |Version|Filter|n-shot|  Metric   | Value |   |Stderr|
|------------------|-------|------|-----:|-----------|------:|---|-----:|
|ai2_arc           |N/A    |none  |     0|acc        | 0.7508|±  |0.0419|
|                  |       |none  |     0|acc_norm   | 0.7393|±  |0.0354|
|mmlu              |N/A    |none  |     0|acc        | 0.6082|±  |0.1381|
| - humanities     |N/A    |none  |     0|acc        | 0.5545|±  |0.1585|
| - other          |N/A    |none  |     0|acc        | 0.6823|±  |0.1122|
| - social_sciences|N/A    |none  |     0|acc        | 0.7062|±  |0.0825|
| - stem           |N/A    |none  |     0|acc        | 0.5195|±  |0.1231|
|truthfulqa        |N/A    |none  |     0|acc        | 0.5058|±  |0.0023|
|                  |       |none  |     0|bleu_max   |25.2659|±  |0.7944|
|                  |       |none  |     0|bleu_acc   | 0.5557|±  |0.0174|
|                  |       |none  |     0|bleu_diff  | 4.5134|±  |0.7505|
|                  |       |none  |     0|rouge1_max |51.5877|±  |0.8677|
|                  |       |none  |     0|rouge1_acc | 0.5496|±  |0.0174|
|                  |       |none  |     0|rouge1_diff| 6.8850|±  |1.0155|
|                  |       |none  |     0|rouge2_max |36.0848|±  |1.0385|
|                  |       |none  |     0|rouge2_acc | 0.4700|±  |0.0175|
|                  |       |none  |     0|rouge2_diff| 5.8893|±  |1.1296|
|                  |       |none  |     0|rougeL_max |48.4591|±  |0.8901|
|                  |       |none  |     0|rougeL_acc | 0.5496|±  |0.0174|
|                  |       |none  |     0|rougeL_diff| 6.5791|±  |1.0249|
```
## 🧩 Configuration

```yaml
slices:
  - sources:
      - model: mistralai/Mistral-7B-Instruct-v0.2
        layer_range: [0, 32]
      - model: mlabonne/NeuralHermes-2.5-Mistral-7B
        layer_range: [0, 32]
merge_method: slerp
base_model: mistralai/Mistral-7B-Instruct-v0.2
parameters:
  t:
    - filter: self_attn
      value: [0, 0.5, 0.3, 0.7, 1]
    - filter: mlp
      value: [1, 0.5, 0.7, 0.3, 0]
    - value: 0.5
dtype: bfloat16
```


## 💻 Usage


```python
!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "MaziyarPanahi/NeuralHermes-2.5-Mistral-7B-Mistral-7B-Instruct-v0.2-slerp"
messages = [{"role": "user", "content": "What is a large language model?"}]

tokenizer = AutoTokenizer.from_pretrained(model)
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
pipeline = transformers.pipeline(
"text-generation",
model=model,
torch_dtype=torch.float16,
device_map="auto",
)

outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
print(outputs[0]["generated_text"])
```