Instructions to use Junping0645/vtuber with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use Junping0645/vtuber with PEFT:
from peft import PeftModel from transformers import AutoModelForCausalLM base_model = AutoModelForCausalLM.from_pretrained("EleutherAI/polyglot-ko-3.8b") model = PeftModel.from_pretrained(base_model, "Junping0645/vtuber") - Transformers
How to use Junping0645/vtuber with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="Junping0645/vtuber")# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("Junping0645/vtuber", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use Junping0645/vtuber with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Junping0645/vtuber" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Junping0645/vtuber", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/Junping0645/vtuber
- SGLang
How to use Junping0645/vtuber with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "Junping0645/vtuber" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Junping0645/vtuber", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "Junping0645/vtuber" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Junping0645/vtuber", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use Junping0645/vtuber with Docker Model Runner:
docker model run hf.co/Junping0645/vtuber
|
Download README.md from Junping0645/vtuber: direct link, hf CLI and curl.
- Browser
- Download file 3.27 kB
-
https://huggingface.co/Junping0645/vtuber/resolve/main/README.md
- Command line
-
hf download hf://Junping0645/vtuber/README.md
-
curl -L -o README.md https://huggingface.co/Junping0645/vtuber/resolve/main/README.md
3.27 kB
metadata
base_model: EleutherAI/polyglot-ko-3.8b
library_name: peft
pipeline_tag: text-generation
language:
- ko
tags:
- base_model:adapter:EleutherAI/polyglot-ko-3.8b
- lora
- transformers
- korean
- chatbot
๋ง๋๋ฌด โ ๋ ๊ฑฐ ์ด๋ฅด์ ์ฉ AI ๋ํ ์ฑ๋ด (QLoRA ์ด๋ํฐ)
EleutherAI/polyglot-ko-3.8b๋ฅผ QLoRA(4bit NF4)๋ก ํ์ธํ๋ํ LoRA ์ด๋ํฐ์
๋๋ค.
๋
๊ฑฐ ์ด๋ฅด์ ๊ณผ ์งง๊ณ ๋ฐ๋ปํ ์กด๋๋ง๋ก ๋ํํ๋ฉฐ, ๋จผ์ ๊ณต๊ฐํ๊ณ ํ์์ ์๋ถ๋ฅผ ์ฑ๊ธฐ๋ ํ๋ฅด์๋๋ก ํ์ตํ์ต๋๋ค.
ํ์ต ๋ฐ์ดํฐ
- ๋จ๋ฐ ๋ํ 11,266๊ฐ: 16๊ฐ ์นดํ ๊ณ ๋ฆฌ(๊ฑด๊ฐ/์ธ๋ก์/์์ฃผ/๋ช ์ ๋ฑ) ร 8๊ฐ ๊ฐ์ , GPT-4o-mini๋ก ์์ฑ + ์ค์ ์ด๋ฅด์ ์ธํฐ๋ทฐ ์ ์ฌ์์ ์ถ์ถํ ์ฌํฌ๋ฆฌ/๊ตฌ์ด์ฒด ๋ฐํ 1,266๊ฐ(์ฌํฌ๋ฆฌ ์ ๋ ฅ ์ดํด ๋ณด๊ฐ์ฉ)
- ๋ฉํฐํด ๋ํ 1,000ํธ(๋ํ๋น ํ๊ท 7.1ํด) โ ํด๋ง๋ค ์ฌ๋ผ์ด๋ฉ ์๋์ฐ๋ก ํ์ฅํด 3,546๊ฐ ํ์ต ์ํ๋ก ๋ณํ
- ์ต์ข train 14,072 / val 740
ํ์ต ์ค์
- LoRA: r=8, alpha=16, dropout=0.1, target_modules=["query_key_value", "dense"] (GPT-NeoX ์ดํ ์ ์์ฃผ๋ก ์ข๊ฒ ํ๊น โ ๊ณผ์ ํฉ ์ต์ )
- 4bit NF4 ์์ํ(bitsandbytes), MAX_LENGTH=768
- batch_size=2, gradient_accumulation_steps=8 (์ ํจ ๋ฐฐ์น 16), 5 epoch ์บก + EarlyStoppingCallback(patience=3)
- ๊ฒฐ๊ณผ: epoch 3.296์์ early stopping, eval_loss 1.366, RTX 3060 Ti(8GB)์์ 141.9๋ถ ์์, peak VRAM 3.4GB
์ฌ์ฉ๋ฒ
๋ฒ ์ด์ค ๋ชจ๋ธ(7.6GB)์ 4bit๋ก ๋ก๋ํ ๋ค ์ด ์ด๋ํฐ๋ฅผ ์น์ต๋๋ค.
import torch
from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
BASE_MODEL = "EleutherAI/polyglot-ko-3.8b"
ADAPTER = "Junping0645/Vtuber" # ์ด ๋ฆฌํฌ
tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL)
bnb = BitsAndBytesConfig(
load_in_4bit=True,
bnb_4bit_quant_type="nf4",
bnb_4bit_compute_dtype=torch.float16,
bnb_4bit_use_double_quant=True,
)
base = AutoModelForCausalLM.from_pretrained(BASE_MODEL, quantization_config=bnb, device_map={"": 0})
model = PeftModel.from_pretrained(base, ADAPTER)
model.eval()
prompt = "### ์ด๋ฅด์ : ์์ฆ ๋ง์ด ์ธ๋กญ์ต๋๋ค.\n### ๋ง๋๋ฌด:"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
out = model.generate(**inputs, max_new_tokens=80, do_sample=True, temperature=0.8,
top_p=0.9, repetition_penalty=1.15,
pad_token_id=tokenizer.eos_token_id, eos_token_id=tokenizer.eos_token_id)
print(tokenizer.decode(out[0], skip_special_tokens=True))
๋ฉํฐํด(๋ํ ์ด๋ ฅ ์ ์ง)
์ด์ ํด๋ค์ ๊ฐ์ ํฌ๋งท์ผ๋ก ์ด์ด๋ถ์ด๋ฉด ๋ฌธ๋งฅ์ ๋ฐ์ํ ๋ต๋ณ์ด ๋์ต๋๋ค(ํ์ต ์ ์ฌ์ฉํ ํฌ๋งท๊ณผ ๋์ผํด์ผ ํจ):
### ์ด๋ฅด์ : ์์ฃผ๊ฐ ๋ค์ ์ฃผ์ ๋๋ฌ ์จ๋์.
### ๋ง๋๋ฌด: ์ ๋ง์? ๊ธฐ๋๋์๊ฒ ์ด์! ์ฆ๊ฑฐ์ด ์๊ฐ ๋ณด๋ด์๊ธธ ๋ฐ๋์.
### ์ด๋ฅด์ : ๋ช ์ด์ธ์ง ์ ๋ฌผ์ด๋ณด์
จ๋ค์, ์ด๋ฒ์ ์ด๋ฑํ๊ต ๋ค์ด๊ฐ์.
### ๋ง๋๋ฌด: ๊ทธ๋ ๊ตฐ์! ์์ฆ ์์ด๋ค์ ๋น ๋ฅด์ฃ .
### ์ด๋ฅด์ : ๊ฑ๊ฐ ์ค๋ฉด ๋ญ ํด์ฃผ๋ฉด ์ข์ํ ๊น์?
### ๋ง๋๋ฌด:
์ฐธ๊ณ
- VRAM: ํ์ต 3.4GB / ์ถ๋ก ์ฝ 3.3GB (RTX 3060 Ti 8GB ๊ธฐ์ค)
- Framework: PEFT 0.19.1, transformers 5.13.0, bitsandbytes 0.49.2, torch 2.11.0+cu128