Downloads · 30 days
2
11% of all-time downloads
wkshin89_-_Yi-Ko-6B-Instruct-v1.0-8bits
RichardErkhov/wkshin89_-_Yi-Ko-6B-Instruct-v1.0-8bits
wkshin89_-_Yi-Ko-6B-Instruct-v1.0-8bits is a machine learning model from RichardErkhov. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
All-time downloads
19
Public
Parameters
6.2B
6.8 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors6.8 GB · 100%
Parameter types
How the weights are stored.
I85.5B · 90%
Model card
From the Hugging Face model README
Quantization made by Richard Erkhov.
Yi-Ko-6B-Instruct-v1.0 - bnb 8bits
- Model creator: https://huggingface.co/wkshin89/
- Original model: https://huggingface.co/wkshin89/Yi-Ko-6B-Instruct-v1.0/
Original model description:
license: cc-by-nc-4.0 datasets:
- kyujinpy/KOR-OpenOrca-Platypus-v3
- beomi/KoAlpaca-v1.1a
- maywell/ko_wikidata_QA language:
- ko base_model: beomi/Yi-Ko-6B
Yi-Ko-6B-Instruct-v1.0
Model Details
Base Model
Training Dataset
- kyujinpy/KOR-OpenOrca-Platypus-v3 🙇
- beomi/KoAlpaca-v1.1a 🙇
- maywell/ko_wikidata_QA 🙇
- AIHub MRC 데이터 선별 후 Instruction Format 맞게 변경 후 사용
Benchmark Results
AI-Harness Evaluation
https://github.com/Beomi/ko-lm-evaluation-harness
| Model | kobest_boolq | kobest_copa | kobest_hellaswag | kobest_sentineg | korunsmile | pawsx_ko |
|---|---|---|---|---|---|---|
| Zero-shot | ||||||
| Yi-Ko-6B-Instruct-v1.0 | 0.6619 | 0.7794 | 0.4858 | 0.4589 | 0.3520 | 0.5545 |
| Yi-Ko-6B | 0.7070 | 0.7696 | 0.5009 | 0.4044 | 0.3828 | 0.5145 |
Instruction Format
### User:
{instruction}
### Assistant:
{response}
Loading the Model
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained("wkshin89/Yi-Ko-6B-Instruct-v1.0")
model = AutoModelForCausalLM.from_pretrained(
"wkshin89/Yi-Ko-6B-Instruct-v1.0",
device_map="auto",
torch_dtype=torch.bfloat16,
)
- Type
- llama
- Created
- Oct 11, 2024
- Updated
- Oct 11, 2024