metadata

license: apache-2.0
language:
  - ko

KULLM project

base model: mistralai/Mistral-7B-Instruct-v0.2

datasets

KULLM dataset
hand-crafted instruction data

Implementation Code

from transformers import (
    AutoModelForCausalLM,
    AutoTokenizer
)
import torch

repo = "heavytail/kullm-mistral"
model = AutoModelForCausalLM.from_pretrained(
        repo,
        torch_dtype=torch.float16,
        device_map='auto'
)
tokenizer = AutoTokenizer.from_pretrained(repo)

Initial upload: 2024/01/28 20:30