File size: 1,526 Bytes
9283c2b
 
 
 
 
92fe5c7
9283c2b
 
92fe5c7
9283c2b
 
 
 
 
 
92fe5c7
9283c2b
 
 
 
 
 
 
92fe5c7
9283c2b
 
 
 
239c5ab
9283c2b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
tags:
- merge
- mergekit
- lazymergekit
- SeaLLMs/SeaLLM-7B-v2
- SeaLLMs/SeaLLM-7B-v2
base_model:
- SeaLLMs/SeaLLM-7B-v2
- SeaLLMs/SeaLLM-7B-v2
---

# SailingLLM

SailingLLM is a merge of the following models using [LazyMergekit](https://colab.research.google.com/drive/1obulZ1ROXHjYLn6PPZJwRR6GzgQogxxb?usp=sharing):
* [SeaLLMs/SeaLLM-7B-v2](https://huggingface.co/SeaLLMs/SeaLLM-7B-v2)
* [SeaLLMs/SeaLLM-7B-v2](https://huggingface.co/SeaLLMs/SeaLLM-7B-v2)

## 🧩 Configuration

```yaml
slices:
  - sources:
      - model: SeaLLMs/SeaLLM-7B-v2
        layer_range: [0, 32]
      - model: SeaLLMs/SeaLLM-7B-v2
        layer_range: [0, 32]
merge_method: slerp
base_model: SeaLLMs/SeaLLM-7B-v2
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 = "JSpergel/SailingLLM"
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"])
```