沧渊

**OceaGPT(沧渊): A Large Laguage Model for Ocea Sciece Tasks**

Project • Paper • Models • Web • Quickstart • Citatio

OceaGPT-2B-v0.1 is based o MiiCPM-2B ad has bee traied o a biligual dataset i the ocea domai, coverig both Chiese ad Eglish.

⏩Quickstart

Dowload the model

Dowload the model

Iferece

from trasformers import AutoModelForCausalLM, AutoTokeizer
import torch
device = "cuda" # the device to load the model oto
path = 'YOUR-MODEL-PATH'
model = AutoModelForCausalLM.from_pretraied(
    path,
    torch_dtype=torch.bfloat16,
    device_map="auto"
)
tokeizer = AutoTokeizer.from_pretraied(path)

prompt = "Which is the largest ocea i the world?"
messages = [
    {"role": "system", "cotet": "You are a helpful assistat."},
    {"role": "user", "cotet": prompt}
]
text = tokeizer.apply_chat_template(
    messages,
    tokeize=False,
    add_geeratio_prompt=True
)
model_iputs = tokeizer([text], retur_tesors="pt").to(device)

geerated_ids = model.geerate(
    model_iputs.iput_ids,
    max_ew_tokes=512
)
geerated_ids = [
    output_ids[le(iput_ids):] for iput_ids, output_ids i zip(model_iputs.iput_ids, geerated_ids)
]

respose = tokeizer.batch_decode(geerated_ids, skip_special_tokes=True)[0]

?Models

| OceaGPT-V | To be released | To be released | To be released |

?Ackowledgemet

OceaGPT(沧渊) is traied based o the ope-sourced large laguage models icludig Qwe, MiiCPM, LLaMA. Thaks for their great cotributios!

?Citatio

Please cite the followig paper if you use OceaGPT i your work.

Limitatios

• The model may have halluciatio issues.

• We did ot optimize the idetity ad the model may geerate idetity iformatio similar to that of Qwe/MiiCPM/LLaMA/GPT series models.

• The model's output is iflueced by prompt tokes, which may result i icosistet results across multiple attempts.

@article{bi2023oceagpt,
  title={OceaGPT: A Large Laguage Model for Ocea Sciece Tasks},
  author={Bi, Zhe ad Zhag, Nigyu ad Xue, Yida ad Ou, Yixi ad Ji, Daxiog ad Zheg, Guozhou ad Che, Huaju},
  joural={arXiv preprit arXiv:2310.02031},
  year={2023}
}

**OceanGPT(沧渊): A Large Language Model for Ocean Science Tasks** Project • Paper • Models •