README.md

metadata

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

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

Training Dataset
kyujinpy/KOR-OpenOrca-Platypus-v3.

Model Benchmark

Open leaderboard

Follow up as link.

Model	Average	ARC	HellaSwag	MMLU	TruthfulQA	CommonGen-V2
Ko-PlatYi-6B-O	NaN	NaN	NaN	NaN	NaN	NaN
Ko-PlatYi-6B-kiwi	NaN	NaN	NaN	NaN	NaN	NaN
Ko-PlatYi-6B-gu	NaN	NaN	NaN	NaN	NaN	NaN
Ko-PlatYi-6B	NaN	NaN	NaN	NaN	NaN	NaN
Yi-Ko-6B	48.79	41.04	53.39	46.28	41.64	61.63

---

AI-Harness Evaluation

AI-Harness evaluation; link

| Model | BoolQ | Copa | HellaSwag | Sentineg | | --- | --- | --- | --- | --- | --- | --- | --- | --- | | | 0-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

### 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)