File size: 2,265 Bytes
422dbb3 00623df 422dbb3 00623df 87077ad 00623df 30d6bd5 00623df 87077ad eea521f c462825 87077ad 00623df 87077ad 00623df |
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 64 65 66 67 68 69 70 71 72 73 74 75 76 77 |
---
language:
- ko
datasets:
- kyujinpy/KOR-OpenOrca-Platypus-v3
library_name: transformers
pipeline_tag: text-generation
license: cc-by-nc-sa-4.0
---
# **Ko-PlatYi-6B-O**
<img src='./Ko-PlatYi.png' width=256>
## Model Details
**Model Developers** Kyujin Han (kyujinpy)
**Input** Models input text only.
**Output** Models generate text only.
**Model Architecture**
Ko-PlatYi-6B-O is an auto-regressive language model based on the Yi-34B transformer architecture.
**Blog Link**
Blog: [Coming soon...]
Github: [Coming soon...]
**Base Model**
[beomi/Yi-Ko-6B](https://huggingface.co./beomi/Yi-Ko-6B)
**Training Dataset**
[kyujinpy/KOR-OpenOrca-Platypus-v3](https://huggingface.co./datasets/kyujinpy/KOR-OpenOrca-Platypus-v3).
# **Model Benchmark**
## Open leaderboard
> Follow up as [link](https://huggingface.co./spaces/upstage/open-ko-llm-leaderboard).
| Model | Average | ARC | HellaSwag | MMLU | TruthfulQA | CommonGen-V2 |
| --- | --- | --- | --- | --- | --- | --- |
| **Ko-PlatYi-6B-O** | 49.00 | 43.52 | 53.59 | 47.47 | 41.01 | 59.39 |
| Ko-PlatYi-6B-kiwi | 48.75 | 41.98 | 53.61 | 46.10 | 38.30 | 63.75 |
| Ko-PlatYi-6B-gu | 48.76 | 42.75 | 54.00 | 44.66 | 41.22 | 61.16 |
| Ko-PlatYi-6B | 49.97 | 43.00 | 53.55 | 46.50 | 40.31 | 66.47 |
| Yi-Ko-6B | 48.79 | 41.04 | 53.39 | 46.28 | 41.64 | 61.63 |
---
## AI-Harness Evaluation
> AI-Harness evaluation; [link](https://github.com/Beomi/ko-lm-evaluation-harness)
| Model | BoolQ | Copa | HellaSwag | Sentineg |
| --- | --- | --- | --- | --- |
| | *Zero-shot* ||||
| **Ko-PlatYi-6B-O** | 0.3343 | 0.7687 | 0.4833 | 0.5794 |
| Ko-PlatYi-6B-kiwi | 0.3343 | 0.7665 | 0.4746 | **0.6248** |
| Ko-PlatYi-6B-gu | **0.7077** | **0.7696** | 0.4797 | 0.3979 |
| Ko-PlatYi-6B | 0.3343 | 0.7684 | **0.4917** | 0.5226 |
| Yi-Ko-6B | **0.7070** | 0.7696 | **0.5009** | 0.4044 |
---
# Implementation Code
```python
### KO-Platypus
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
repo = "kyujinpy/Ko-PlatYi-6B-O"
OpenOrca = AutoModelForCausalLM.from_pretrained(
repo,
return_dict=True,
torch_dtype=torch.float16,
device_map='auto'
)
OpenOrca_tokenizer = AutoTokenizer.from_pretrained(repo)
```
|