Edit model card

18禁小说生成模型

数据来源自h-corpus-2023
模型来源自Qwen1.5-0.5B

本模型仅用于学术研究,严禁商用。

from transformers import AutoModelForCausalLM, AutoTokenizer

repo_id = 'Monor/Qwen1.5-0.5B-h-world'
tokenizer = AutoTokenizer.from_pretrained(repo_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(repo_id, device_map="auto", trust_remote_code=True).eval()

inputs = tokenizer('小青突然来到了我的房间', return_tensors='pt')
inputs = inputs.to(model.device)
pred = model.generate(**inputs, max_new_tokens=32, num_beams=4, do_sample=True, pad_token_id=tokenizer.eos_token_id)
print(tokenizer.decode(pred.cpu()[0], skip_special_tokens=True))

输出:小青突然来到了我的房间,吓了我一跳。她见我睡着了,就脱了衣服钻进了被窝里,然后搂着我睡着了。


WebUI已开发完成

18禁小说生成器,点击跳转到GitHub


评测得分

根据h-eval评测方法,要求模型正确续写下一个单词,该模型获得57.9分。
评测过程使用h-eval中提供数据集,仅修改benchmark代码以适用Qwen1.5模型

Model Score
Human 80.2
rwkv-5-h-world-7B 60.3
rwkv-5-h-world-3B 59.4
rwkv-5-h-world-1b5 59.1
Qwen1.5-0.5B-h-world 57.9
Yi-34B 54.7
rwkv-h-world-1b5 54.1
rwkv-v4-7b-dengh 50.0
Yi-6B 48.7
Yi-34B-Chat-4bits 48.1
rwkv-h-world-0.4b 46.8
deepsex-34b 45.9
NSFW_13B_sft 44.3
CausalLM-14B-GPTQ 43.4
Baichuan2-7B-Base 42.7
RWKV-5-World-3B-v2-20231113-ctx4096 42.5
rwkv-h-1b5 42.1
RWKV-v5-12B-one-state-chat-16k 41.3
chatglm3-6b-base 41.2
RWKV-claude-4-World-7B-20230805-ctx65k 40.2
Baichuan2-13B-Base 39.9
RWKV-4-World-CHNtuned-7B-v1-20230709-ctx4096 39.3
Baichuan2-13B-Chat-4bits 37.4
RWKV-5-World-1B5-v2-20231025-ctx4096 36.1
Qwen-7B 33.0
chatglm3-6b 30.5
RWKV-4-World-CHNtuned-1.5B-v1-20230620-ctx4096 28.9
RWKV-4-World-CHNtuned-0.4B-v1-20230618-ctx4096 22.9
RWKV-4-Novel-3B-v1-Chn-20230412-ctx4096 20.4

Doing:

✌️基于qwen1.5-7B和qwen1.5-14B模型已训练到57%✌️

模型训练动机:科学研究

Downloads last month
313
Safetensors
Model size
464M params
Tensor type
F32
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.