Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: other
|
3 |
+
language:
|
4 |
+
- en
|
5 |
+
library_name: transformers
|
6 |
+
pipeline_tag: text-generation
|
7 |
+
tags:
|
8 |
+
- causal-lm
|
9 |
+
- text-generation-inference
|
10 |
+
- merge
|
11 |
+
---
|
12 |
+
|
13 |
+
# FOR EXPERIMENT
|
14 |
+
|
15 |
+
## Description
|
16 |
+
|
17 |
+
[**stabilityai/stablelm-zephyr-3b**](https://huggingface.co/stabilityai/stablelm-zephyr-3b), [**StableMed-3b**](https://huggingface.co/cxllin/StableMed-3b) merged with a new, experimental implementation of "dare ties" via mergekit. See:
|
18 |
+
|
19 |
+
> [Language Models are Super Mario: Absorbing Abilities from Homologous Models as a Free Lunch](https://github.com/yule-BUAA/MergeLM)
|
20 |
+
|
21 |
+
> https://github.com/cg123/mergekit/tree/dare
|
22 |
+
|
23 |
+
|
24 |
+
## Usage
|
25 |
+
|
26 |
+
`StableLM Zephyr 3B` uses the following instruction format:
|
27 |
+
```
|
28 |
+
<|user|>
|
29 |
+
List 3 synonyms for the word "tiny"<|endoftext|>
|
30 |
+
<|assistant|>
|
31 |
+
1. Dwarf
|
32 |
+
2. Little
|
33 |
+
3. Petite<|endoftext|>
|
34 |
+
```
|
35 |
+
|
36 |
+
***
|
37 |
+
## Testing Notes
|
38 |
+
|
39 |
+
Merged in mergekit with the following config, and the tokenizer from chargoddard's Yi-Llama:
|
40 |
+
|
41 |
+
```
|
42 |
+
models:
|
43 |
+
- model: stabilityai/stablelm-zephyr-3b
|
44 |
+
# no parameters necessary for base model
|
45 |
+
- model: cxllin/StableMed-3b
|
46 |
+
parameters:
|
47 |
+
weight: 0.08
|
48 |
+
density: 0.5
|
49 |
+
merge_method: dare_ties
|
50 |
+
base_model: stabilityai/stablelm-zephyr-3b
|
51 |
+
parameters:
|
52 |
+
int8_mask: true
|
53 |
+
dtype: bfloat16
|
54 |
+
```
|
55 |
+
|
56 |
+
## Model Details
|
57 |
+
- License: [StabilityAI Non-Commercial Research Community License](https://huggingface.co/stabilityai/stablelm-zephyr-3b/raw/main/LICENSE)
|