Views
No views yet
| Model Name | Stage | Multimodal | Description |
|---|---|---|---|
| InstructBioMol-base (This Model) | Pretraining | ❎ | Continual pretrained model on molecular sequences, protein sequences, and scientific literature. |
| InstructBioMol-instruct-stage1 | Instruction tuning (stage 1) | ✅ | Stage1 instruction-tuned model with biomolecular multimodal processing capabilities. (e.g., 3D molecules/proteins) |
| InstructBioMol-instruct | Instruction tuning (stage 1 and 2) | ✅ | Fully instruction-tuned model (stage1 & stage2) with biomolecular multimodal processing capabilities (e.g., 3D molecules/proteins) |
<p> (e.g., <p>M<p>A<p>L<p>W...).1from transformers import LlamaForCausalLM, LlamaTokenizer
2import torch
3
4model_name = "hicai-zju/InstructBioMol-base"
5tokenizer = LlamaTokenizer.from_pretrained(model_name)
6model = LlamaForCausalLM.from_pretrained(model_name, device_map="cuda:0")
7
8prompt = "<p>M" # protein sequence
9# prompt = "[C]" # molecule sequence
10# prompt = 'Scientific' # natural language
11inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
12
13with torch.no_grad():
14 outputs = model.generate(
15 **inputs,
16 max_new_tokens=100,
17 temperature=0.7,
18 top_p=0.9,
19 do_sample=True
20 )
21
22generated_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
23print(generated_text)1@article{zhuang2025advancing,
2 author = {Xiang Zhuang and
3 Keyan Ding and
4 Tianwen Lyu and
5 Yinuo Jiang and
6 Xiaotong Li and
7 Zhuoyi Xiang and
8 Zeyuan Wang and
9 Ming Qin and
10 Kehua Feng and
11 Jike Wang and
12 Qiang Zhang and
13 Huajun Chen},
14 title={Advancing biomolecular understanding and design following human instructions},
15 journal={Nature Machine Intelligence},
16 pages={1--14},
17 year={2025},
18 publisher={Nature Publishing Group UK London}
19}