Llama-2-KoEn-13B 🦙🇰🇷🇺🇸

Llama-2-KoEn serves as an advanced iteration of Llama 2, benefiting from an expanded vocabulary and the inclusion of Korean + English corpus in its further pretraining. Just like its predecessor, Llama-2-KoEn operates within the broad range of generative text models that stretch from 7 billion to 70 billion parameters. This repository focuses on the 13B pretrained version, which is tailored to fit the Hugging Face Transformers format. For access to the other models, feel free to consult the index provided below.

Model Details

Model Developers Junbum Lee (Beomi), Taekyoon Choi (Taekyoon)

Variations Llama-2-KoEn will come in a range of parameter sizes — 7B, 13B, and 70B — as well as pretrained and fine-tuned variations.

Input Models input text only.

Output Models generate text only.

Model Architecture

Llama-2-KoEn is an auto-regressive language model that uses an optimized transformer architecture based on Llama-2.

	Training Data	Params	Content Length	GQA	Tokens	LR
Llama 2	A new mix of Korean + English online data	13B	4k	✗	>60B	1e<sup>-5</sup>

Vocab Expansion

Model Name	Vocabulary Size	Description
Original Llama-2	32000	Sentencepiece BPE
Expanded Llama-2-Ko	46336	Sentencepiece BPE. Added Korean vocab and merges

Tokenizing "안녕하세요, 오늘은 날씨가 좋네요."

Model	Tokens
Llama-2	`['▁', '안', '<0xEB>', '<0x85>', '<0x95>', '하', '세', '요', ',', '▁', '오', '<0xEB>', '<0x8A>', '<0x98>', '은', '▁', '<0xEB>', '<0x82>', '<0xA0>', '씨', '가', '▁', '<0xEC>', '<0xA2>', '<0x8B>', '<0xEB>', '<0x84>', '<0xA4>', '요']`
Llama-2-Ko	`['▁안녕', '하세요', ',', '▁오늘은', '▁날', '씨가', '▁좋네요']`

Tokenizing "Llama 2: Open Foundation and Fine-Tuned Chat Models"

Model	Tokens
Llama-2	`['▁L', 'l', 'ama', '▁', '2', ':', '▁Open', '▁Foundation', '▁and', '▁Fine', '-', 'T', 'un', 'ed', '▁Ch', 'at', '▁Mod', 'els']`
Llama-2-Ko	`['▁L', 'l', 'ama', '▁', '2', ':', '▁Open', '▁Foundation', '▁and', '▁Fine', '-', 'T', 'un', 'ed', '▁Ch', 'at', '▁Mod', 'els']`

Model Benchmark

LM Eval Harness - Korean (polyglot branch)

Llama-2-KoEn-13B Benchmark Score

Used EleutherAI's lm-evaluation-harness https://github.com/EleutherAI/lm-evaluation-harness/tree/polyglot

Task	0-shot	5-shot	10-shot	50-shot
kobest_boolq	0.398848	0.703795	0.752612	0.7578
kobest_copa	0.776785	0.812796	0.818724	0.853953
kobest_hellaswag	0.499922	0.512659	0.503365	0.524664
kobest_sentineg	0.586955	0.974811	0.982367	0.987405
kohatespeech	0.278224	0.378693	0.370702	0.509343
kohatespeech_apeach	0.337667	0.556898	0.581788	0.667511
kohatespeech_gen_bias	0.248404	0.484745	0.473659	0.461714
korunsmile	0.327145	0.329163	0.347889	0.395522
nsmc	0.6442	0.87702	0.89982	0.90984
pawsx_ko	0.5355	0.5455	0.5435	0.5255

Note for oobabooga/text-generation-webui

Remove ValueError at load_tokenizer function(line 109 or near), in modules/models.py.

diff --git a/modules/models.py b/modules/models.py
index 232d5fa..de5b7a0 100644
--- a/modules/models.py
+++ b/modules/models.py
@@ -106,7 +106,7 @@ def load_tokenizer(model_name, model):
                 trust_remote_code=shared.args.trust_remote_code,
                 use_fast=False
             )
-        except ValueError:
+        except:
             tokenizer = AutoTokenizer.from_pretrained(
                 path_to_model,
                 trust_remote_code=shared.args.trust_remote_code,

Since Llama-2-Ko uses FastTokenizer provided by HF tokenizers NOT sentencepiece package, it is required to use use_fast=True option when initialize tokenizer.

Apple Sillicon does not support BF16 computing, use CPU instead. (BF16 is supported when using NVIDIA GPU)

LICENSE

Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International Public License, under LLAMA 2 COMMUNITY LICENSE AGREEMENT
Full License available at: https://huggingface.co/beomi/llama-2-koen-13b/blob/main/LICENSE
For Commercial Usage, contact Author.

Citation

@misc {l._junbum_2023,
	author       = { {L. Junbum, Taekyoon Choi} },
	title        = { llama-2-koen-13b },
	year         = 2023,
	url          = { https://huggingface.co/beomi/llama-2-koen-13b },
	doi          = { 10.57967/hf/1280 },
	publisher    = { Hugging Face }
}

Acknowledgement

The training is supported by TPU Research Cloud program.