botp/zxbsmk-NSFW_13B_sft
713
A instruction-tuned model of https://huggingface.co/baichuan-inc/Baichuan-13B-Base
- Instruction-following datasets used: instructnsfwcn
- Training framework: https://github.com/hiyouga/LLaMA-Efficient-Tuning
Usage:
from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer
tokenizer = AutoTokenizer.from_pretrained("zxbsmk/NSFW_13B_sft", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("zxbsmk/NSFW_13B_sft", trust_remote_code=True).cuda()
streamer = TextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
query = "续写。白洁已经动情了,肩带滑落露出香肩。"
inputs = tokenizer([query], return_tensors="pt")
inputs = inputs.to("cuda")
generate_ids = model.generate(**inputs, max_new_tokens=1024, streamer=streamer)You could also alternatively launch a CLI demo by using the script in https://github.com/hiyouga/LLaMA-Efficient-Tuning
python src/cli_demo.py --template baichuan --model_name_or_path zxbsmk/NSFW_13B_sftExamples:
Omit part of results for better visualization.
Evaluation
Comparision between Baichuan-13B-Chat and NSFW13Bsft. (Zero-shot CMMLU)
(By the way, Baichuan-13B-Chat gets 50.43 with one-shot which seems much better than 48.86 with zero-shot.)
Contact Us
Join group via https://t.me/+JbovpBG6-gBiNDI1
