Ko-PlatYi-6B-kiwi

Model Details

Model Developers Kyujin Han (kyujinpy)

Input Models input text only.

Output Models generate text only.

Model Architecture
Ko-PlatYi-6B-kiwi 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-Orca-Platypus-kiwi.

Model Benchmark

Open leaderboard

Follow up as link.

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

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

### KO-Platypus
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

repo = "kyujinpy/Ko-PlatYi-6B-kiwi"
OpenOrca = AutoModelForCausalLM.from_pretrained(
        repo,
        return_dict=True,
        torch_dtype=torch.float16,
        device_map='auto'
)
OpenOrca_tokenizer = AutoTokenizer.from_pretrained(repo)

---