Home

Awesome

LlaSMol

This is the official code repository for the paper LlaSMol: Advancing Large Language Models for Chemistry with a Large-Scale, Comprehensive, High-Quality Instruction Tuning Dataset.

Tasks and Examples

Our models are trained on 4 types of 14 essential tasks, which, along with examples, are listed below.

Name Conversion

Name Conversion - IUPAC to Molecular Formula

Query: What is the molecular formula of the compound with this IUPAC name <IUPAC> 2,5-diphenyl-1,3-oxazole </IUPAC> ?

Response: <MOLFORMULA> C15H11NO </MOLFORMULA>

Name Conversion - IUPAC to SMILES

Query: Could you provide the SMILES for <IUPAC> 4-ethyl-4-methyloxolan-2-one </IUPAC> ?

Response: Of course. It's <SMILES> CCC1(C)COC(=O)C1 </SMILES> .

Name Conversion - SMILES to Molecular Formula

Query: Given the SMILES representation <SMILES> S=P1(N(CCCl)CCCl)NCCCO1 </SMILES>, what would be its molecular formula?

Response: It is <MOLFORMULA> C7H15Cl2N2OPS </MOLFORMULA> .

Name Conversion - SMILES to IUPAC

Query: Translate the given SMILES formula of a molecule <SMILES> CCC(C)C1CNCCCNC1 </SMILES> into its IUPAC name.

Response: <IUPAC> 3-butan-2-yl-1,5-diazocane </IUPAC>

Property Prediction

Property Prediction - ESOL

Query: How soluble is <SMILES> CC(C)Cl </SMILES> ?

Response: Its log solubility is <NUMBER> -1.41 </NUMBER> mol/L.

Property Prediction - LIPO

Query: Predict the octanol/water distribution coefficient logD under the circumstance of pH 7.4 for <SMILES> NC(=O)C1=CC=CC=C1O </SMILES> .

Response: <NUMBER> 1.090 </NUMBER>

Property Prediction - BBBP

Query: Is blood-brain barrier permeability (BBBP) a property of <SMILES> CCNC(=O)/C=C/C1=CC=CC(Br)=C1 </SMILES> ?

Response: <BOOLEAN> Yes </BOOLEAN>

Property Prediction - Clintox

Query: Is <SMILES> COC[C@@H](NC(C)=O)C(=O)NCC1=CC=CC=C1 </SMILES> toxic?

Response: <BOOLEAN> No </BOOLEAN>

Property Prediction - HIV

Query: Can <SMILES> CC1=CN(C2C=CCCC2O)C(=O)NC1=O </SMILES> serve as an inhibitor of HIV replication?

Response: <BOOLEAN> No </BOOLEAN>

Property Prediction - SIDER

Query: Are there any known side effects of <SMILES> CC1=CC(C)=C(NC(=O)CN(CC(=O)O)CC(=O)O)C(C)=C1Br </SMILES> affecting the heart?

Response: <BOOLEAN> No </BOOLEAN>

Molecule Description

Molecule Captioning

Query: Describe this molecule: <SMILES> CCOC(=O)C1=CN=CN1[C@H](C)C1=CC=CC=C1 </SMILES>

Response: The molecule is an imidazole derivative with short-acting sedative, hypnotic, and general anesthetic properties. Etomidate appears to have gamma-aminobutyric acid (GABA) like effects, mediated through GABA-A receptor. The action enhances the inhibitory effect of GABA on the central nervous system by causing chloride channel opening events which leads to membrane hyperpolarization.

Molecule Generation

It is not required to use tags to wrap up the input description.

Query: Give me a molecule that satisfies the conditions outlined in the description: The molecule is a member of the class of tripyrroles that is a red-coloured pigment with antibiotic properties produced by Serratia marcescens. It has a role as an antimicrobial agent, a biological pigment, a bacterial metabolite, an apoptosis inducer and an antineoplastic agent. It is a tripyrrole, an aromatic ether and a ring assembly.

Response: Here is a potential molecule: <SMILES> CCCCCC1=C(C)NC(/C=C2\N=C(C3=CC=CN3)C=C2OC)=C1 </SMILES>

Chemical Reaction

Forward Synthesis

Query: <SMILES> NC1=CC=C2OCOC2=C1.O=CO </SMILES> Based on the reactants and reagents given above, suggest a possible product.

Response: A possible product can be <SMILES> O=CNC1=CC=C2OCOC2=C1 </SMILES> .

Retrosynthesis

Query: Identify possible reactants that could have been used to create the specified product. <SMILES> CC1=CC=C(N)N=C1N </SMILES>

Response: <SMILES> CC(C#N)CCC#N.N </SMILES>

Training

If you need to fine-tune a base model on SMolInstruct, please first clone this repo to your machine, and cd to the folder, then use the following command.

MODELNAME=LlaSMol-Mistral-7B && CUDA_VISIBLE_DEVICES=0,1,2,3 accelerate launch finetune.py --data_path osunlp/SMolInstruct --base_model mistralai/Mistral-7B-v0.1 --wandb_project LlaSMol --wandb_run_name $MODELNAME --wandb_log_model true ---output_dir checkpoint/$MODELNAME

The above is an example command for fine-tuning Mistral with LoRA, using 4 GPUs. If with other base models, the lora settings (e.g., lora_target_modules) might need to be modified accordingly.

Usage

Clone this repo to your machine, and cd to the folder.

Generation

You could use the following code to query the models with your questions.

from generation import LlaSMolGeneration

generator = LlaSMolGeneration('osunlp/LlaSMol-Mistral-7B')
generator.generate('Can you tell me the IUPAC name of <SMILES> C1CCOC1 </SMILES> ?')

Note:

  1. In the input query, please use corresponding tags to wrap up specific content.

    • SMILES representation: <SMILES> ... </SMILES>
    • IUPAC name: <IUPAC> ... </IUPAC>

    Other tags may appear in models' responses:

    • Molecular formula: <MOLFORMULA> ... </MOLFORMULA>
    • Number: <NUMBER> ... </NUMBER>
    • Boolean: <BOOLEAN> ... </BOOLEAN>

    Please see the examples in the above section.

  2. The code would canonicalize SMILES string automatically, as long as it is wrapped in <SMILES> ... </SMILES>.

Evaluation on SMolInstruct

Step 1. Generate responses for samples

Use the following command to apply LlaSMol models to generate responses for samples in SmolInstruct.

python generate_on_dataset.py --model_name osunlp/LlaSMol-Mistral-7B --output_dir eval/LlaSMol-Mistral-7B/output

By default, it generates for all the tasks. You could also specify tasks by adding argument like --tasks "['forward_synthesis','retrosynthesis']". If not setting tasks, the script will generate for all the tasks in SMolInstruct.

Step 2. Extract predicted answer from model outputs

Use the command to extract predicted answers from model's output, and store them in the pred domains. By default, it extract the part between the corresponding tags (e.g., <SMILES> ... </SMILES>). If the tags are missing or incomplete, the extracted answer will be empty and regarded as "no answer" in metric calculation.

python extract_prediction.py --output_dir eval/LlaSMol-Mistral-7B/output --prediction_dir eval/LlaSMol-Mistral-7B/prediction

By default, it extracts predicted answers for all the tasks. It skips task if its output file is not found. You could also specify tasks like --tasks "['forward_synthesis','retrosynthesis']".

Step 3. Calculate metrics

Use the following command to compute metrics for all the tasks.

python compute_metrics.py --prediction_dir eval/LlaSMol-Mistral-7B/prediction

By default, it extracts predicted answers for all the tasks. It skips task if its output file is not found. You could also specify tasks like --tasks "['forward_synthesis','retrosynthesis']".

Citation

If our paper or related resources prove valuable to your research, we kindly ask for citation. Please feel free to contact us with any inquiries.

@article{yu2024llasmol,
    title={LlaSMol: Advancing Large Language Models for Chemistry with a Large-Scale, Comprehensive, High-Quality Instruction Tuning Dataset},
    author={Botao Yu and Frazier N. Baker and Ziqi Chen and Xia Ning and Huan Sun},
    journal={arXiv preprint arXiv:2402.09391},
    year={2024}
}