Update README.md
Browse files
README.md
CHANGED
@@ -1,4 +1,33 @@
|
|
1 |
---
|
|
|
2 |
datasets:
|
3 |
- gbstox/agronomy-resources
|
4 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
+
base_model: mistralai/Mistral-7B-v0.1
|
3 |
datasets:
|
4 |
- gbstox/agronomy-resources
|
5 |
+
tags:
|
6 |
+
- mistral
|
7 |
+
- instruct
|
8 |
+
- finetune
|
9 |
+
- chatml
|
10 |
+
- gpt4
|
11 |
+
- synthetic data
|
12 |
+
- distillation
|
13 |
+
language:
|
14 |
+
- en
|
15 |
+
---
|
16 |
+
|
17 |
+
# AgronoMistral
|
18 |
+
<img src="https://cdn-uploads.huggingface.co/production/uploads/63042a3d7373aacccd896484/UqOsPoPc__ytlGDHYfI9S.jpeg" width="800" >
|
19 |
+
|
20 |
+
# About
|
21 |
+
Agronomistral is fine tuned with agronomy information from textbooks, university extension websites, government websites, and agronomic studies.
|
22 |
+
<br>Agronomistral consistently outperforms the base model, mistralai/Mistral-7B-v0.1 on the [agronomy benchmark](https://github.com/gbstox/agronomy_llm_benchmarking).
|
23 |
+
|
24 |
+
# Benchmark comparison
|
25 |
+
| Model Name | Score | Date Tested |
|
26 |
+
|----------------|-------|-----------------|
|
27 |
+
| gpt-4-1106-preview | 83.84% | 2024-01-02 |
|
28 |
+
| Mixtral-8x7B-Instruct-v0.1 | 75.76% | 2024-01-02 |
|
29 |
+
| fbn-norm | 74.75% | 2024-01-02 |
|
30 |
+
| gpt-3.5-turbo | 73.74% | 2024-01-02 |
|
31 |
+
| OpenHermes-2p5-Mistral-7B | 65.66% | 2024-01-02 |
|
32 |
+
| AgronoMistral | 58.59% | 2024-01-08 |
|
33 |
+
| mistral-7b-instruct-v0.1_Q5_K_M | 47.47% | 2024-01-08 |
|