Jan
commited on
Commit
•
fbd72f9
1
Parent(s):
b2e1c77
Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,72 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
---
|
| 6 |
+
<!-- header start -->
|
| 7 |
+
<!-- 200823 -->
|
| 8 |
+
<div style="width: auto; margin-left: auto; margin-right: auto">
|
| 9 |
+
<img src="https://github.com/janhq/jan/assets/89722390/35daac7d-b895-487c-a6ac-6663daaad78e" alt="Jan banner" style="width: 100%; min-width: 400px; display: block; margin: auto;">
|
| 10 |
+
</div>
|
| 11 |
+
|
| 12 |
+
<p align="center">
|
| 13 |
+
<a href="https://jan.ai/">Jan</a>
|
| 14 |
+
- <a href="https://discord.gg/AsJ8krTT3N">Discord</a>
|
| 15 |
+
</p>
|
| 16 |
+
<!-- header end -->
|
| 17 |
+
|
| 18 |
+
# Model Description
|
| 19 |
+
This model uses the `SLERP` method to merge [Mistral-7B-Instruct-v0.2](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2) with 2 best models in 12th Dec on [OpenLLM Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard):
|
| 20 |
+
1. [v1olet_marcoroni-go-bruins-merge-7B](https://huggingface.co/v1olet/v1olet_marcoroni-go-bruins-merge-7B)
|
| 21 |
+
2. [go-bruins-v2](https://huggingface.co/rwitz/go-bruins-v2)
|
| 22 |
+
|
| 23 |
+
- base model: [Mistral-7B-Instruct-v0.2](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2)
|
| 24 |
+
|
| 25 |
+
The yaml config file for this model is here:
|
| 26 |
+
|
| 27 |
+
```yaml
|
| 28 |
+
slices:
|
| 29 |
+
- sources:
|
| 30 |
+
- model: AIDC-ai-business/Marcoroni-7B-v3
|
| 31 |
+
layer_range: [0, 32]
|
| 32 |
+
- model: rwitz/go-bruins-v2
|
| 33 |
+
layer_range: [0, 32]
|
| 34 |
+
merge_method: slerp
|
| 35 |
+
base_model: mistralai/Mistral-7B-Instruct-v0.2
|
| 36 |
+
parameters:
|
| 37 |
+
t:
|
| 38 |
+
- filter: self_attn
|
| 39 |
+
value: [0, 0.5, 0.3, 0.7, 1]
|
| 40 |
+
- filter: mlp
|
| 41 |
+
value: [1, 0.5, 0.7, 0.3, 0]
|
| 42 |
+
- value: 0.5
|
| 43 |
+
dtype: bfloat16
|
| 44 |
+
```
|
| 45 |
+
|
| 46 |
+
# About Jan
|
| 47 |
+
Jan believes in the need for an open-source AI ecosystem and is building the infra and tooling to allow open-source AIs to compete on a level playing field with proprietary ones.
|
| 48 |
+
|
| 49 |
+
Jan's long-term vision is to build a cognitive framework for future robots, who are practical, useful assistants for humans and businesses in everyday life.
|
| 50 |
+
|
| 51 |
+
# Jan Model Merger
|
| 52 |
+
This is a test project for merging models.
|
| 53 |
+
|
| 54 |
+
# Open LLM Leaderboard Evaluation Results
|
| 55 |
+
|
| 56 |
+
Detailed results can be found here.
|
| 57 |
+
|
| 58 |
+
| Metric | Value |
|
| 59 |
+
|-----------------------|---------------------------|
|
| 60 |
+
| Avg. | ?|
|
| 61 |
+
| ARC (25-shot) | ? |
|
| 62 |
+
| HellaSwag (10-shot) | ? |
|
| 63 |
+
| MMLU (5-shot) | ?|
|
| 64 |
+
| TruthfulQA (0-shot) | ? |
|
| 65 |
+
| Winogrande (5-shot) | ? |
|
| 66 |
+
| GSM8K (5-shot) | ? |
|
| 67 |
+
|
| 68 |
+
# Acknowlegement
|
| 69 |
+
- [mergekit](https://github.com/cg123/mergekit)
|
| 70 |
+
- [DARE](https://github.com/yule-BUAA/MergeLM/blob/main/README.md)
|
| 71 |
+
- [SLERP](https://github.com/Digitous/LLM-SLERP-Merge)
|
| 72 |
+
- [lm-evaluation-harness](https://github.com/EleutherAI/lm-evaluation-harness)
|