beomi

/

KoAlpaca-llama-1-7b

like 28

According to his github:

Polyglot-ko 5.8B 기반 [Full Finetune] -> 🤗 https://huggingface.co/beomi/KoAlpaca-Polyglot
Meta LLAMA 7B 기반 [Full Finetune] -> 🤗 https://huggingface.co/beomi/KoAlpaca
Meta LLAMA 13B 기반 [LoRA] -> 🤗 https://huggingface.co/beomi/KoAlpaca-13B-LoRA
Meta LLAMA 30B 기반 [LoRA] -> 🤗 https://huggingface.co/beomi/KoAlpaca-30B-LoRA
Meta LLAMA 65B 기반 [LoRA] -> 🤗 https://huggingface.co/beomi/KoAlpaca-65B-LoRA

so, this is llama 7b, full finetuned.

Why 7B is so big?

it's because it is 32bit(fp32)