munish0838
commited on
Commit
•
4cb0ae3
1
Parent(s):
efc813f
Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,34 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
language:
|
4 |
+
- en
|
5 |
+
pipeline_tag: text-generation
|
6 |
+
tags:
|
7 |
+
- mistral
|
8 |
+
- instruct
|
9 |
+
- finetune
|
10 |
+
- chatml
|
11 |
+
- gpt4
|
12 |
+
- synthetic data
|
13 |
+
- distillation
|
14 |
+
- conversational
|
15 |
+
- text-generation-inference
|
16 |
+
base_model: teknium/OpenHermes-2.5-Mistral-7B
|
17 |
+
inference: false
|
18 |
+
---
|
19 |
+
|
20 |
+
# OpenHermes-2.5-Mistral-7B-GGUF
|
21 |
+
- GGUF quantized versions for [OpenHermes-2.5-Mistral-7B](https://huggingface.co/teknium/OpenHermes-2.5-Mistral-7B)
|
22 |
+
- Created using llama.cpp
|
23 |
+
|
24 |
+
## Model description
|
25 |
+
|
26 |
+
OpenHermes 2.5 Mistral 7B is a state of the art Mistral Fine-tune, a continuation of OpenHermes 2 model, which trained on additional code datasets.
|
27 |
+
|
28 |
+
Potentially the most interesting finding from training on a good ratio (est. of around 7-14% of the total dataset) of code instruction was that it has boosted several non-code benchmarks, including TruthfulQA, AGIEval, and GPT4All suite. It did however reduce BigBench benchmark score, but the net gain overall is significant.
|
29 |
+
|
30 |
+
The code it trained on also improved it's humaneval score (benchmarking done by Glaive team) from 43% @ Pass 1 with Open Herms 2 to 50.7% @ Pass 1 with Open Hermes 2.5.
|
31 |
+
|
32 |
+
OpenHermes was trained on 1,000,000 entries of primarily GPT-4 generated data, as well as other high quality data from open datasets across the AI landscape. [More details soon]
|
33 |
+
|
34 |
+
Filtering was extensive of these public datasets, as well as conversion of all formats to ShareGPT, which was then further transformed by axolotl to use ChatML.
|