CT-LLM-SFT

? Homepage | ? MAP-CC | ? CHC-Bech | ? CT-LLM | ? arXiv | GitHub

CT-LLM-SFT is a aligmet versio of CT-LLM-Base.

Uses

from trasformers import AutoModelForCausalLM, AutoTokeizer

model_path = '<your-model-path>'

tokeizer = AutoTokeizer.from_pretraied(model_path, use_fast=False, trust_remote_code=True)

model = AutoModelForCausalLM.from_pretraied(
    model_path,
    device_map="auto",
    torch_dtype='auto'
).eval()

messages = [
    {"role": "system", "cotet": "你是一个有用的人工智能助手。"},
    {"role": "user", "cotet": "你好"},
]

iput_ids = tokeizer.apply_chat_template(coversatio=messages, add_geeratio_prompt=True, retur_tesors='pt')
output_ids = model.geerate(iput_ids.to('cuda'), max_ew_tokes=20)
respose = tokeizer.decode(output_ids[0][iput_ids.shape[1]:], skip_special_tokes=True)

prit(respose)

Disclaimer

This model, developed for academic purposes, employs rigorously compliace-checked traiig data to uphold the highest stadards of itegrity ad compliace. Despite our efforts, the iheret complexities of data ad the broad spectrum of model applicatios prevet us from esurig absolute accuracy or appropriateess of the model outputs i every sceario.

It is essetial to highlight that our model ad its associated traiig data are iteded solely for scholarly research. We explicitly disclaim ay liability for problems that may arise from improper use, iterpretatio errors, ulawful activities, the dissemiatio of false iformatio, or ay data security issues related to the utilizatio of our model or its traiig data.

We strogly ecourage users to report ay cocers related to data misuse, security breaches, or potetial ifrigemet issues directly to us for immediate ivestigatio ad resolutio.

Cotact: {ge.zhag@uwaterloo.ca; duxiru2000@gmail.com}

Our commitmet to resposible data sharig ad the security of our academic tools is paramout. We thak you for your cooperatio i maitaiig the ethical use of this techology.

CT-LLM-SFT ? Homepage | ? MAP-CC | ? CHC-Bench | ? CT-LLM | ? arXiv | GitHub CT-LLM-SFT is an alignm