add ChemFM in leaderboard

[feiyang-cai]

@amva13

This PR adds the ChemFM model to the ADMET Leaderboard. Below are the relevant details:

A. A link to the codes for training the model and running the benchmarks.
The code for training the ChemFM model and running the benchmarks can be found at the following GitHub repository:
https://github.com/TheLuoFengLab/ChemFM/tree/master/finetuning/property_prediction

B. Instructions for replicating your results running these codes.
Instructions for replicating our results are included in the README file within the GitHub repository. Additionally, we have provided links to the model checkpoints. https://github.com/TheLuoFengLab/ChemFM/tree/master/finetuning/property_prediction

C. Description of the hardware used for training the model.
The model was trained on a single H100 GPU machine. However, it is also possible to train the model on a moderate GPU machine.

D. Link to paper to be included on the website
The full paper can be accessed here:
https://arxiv.org/abs/2410.21422

E. Paper authors and developers of the model.
Feiyang Cai, Clemson University

F. The dictionary string output of running the TDC benchmark group evaluations.

{
    'caco2_wang': [0.322, 0.026],
    'bioavailability_ma': [0.715, 0.011],
    'lipophilicity_astrazeneca': [0.460, 0.006],
    'solubility_aqsoldb': [0.725, 0.011],
    'hia_hou': [0.984, 0.004],
    'pgp_broccatelli': [0.931, 0.003],
    'bbb_martins': [0.908, 0.010],
    'ppbr_az': [7.505, 0.073],
    'vdss_lombardo': [0.662, 0.013],
    'cyp2c9_veith': [0.788, 0.005],
    'cyp2d6_veith': [0.704, 0.003],
    'cyp3a4_veith': [0.878, 0.003],
    'cyp2c9_substrate_carbonmangels': [0.414, 0.027],
    'cyp2d6_substrate_carbonmangels': [0.739, 0.024],
    'cyp3a4_substrate_carbonmangels': [0.654, 0.022],
    'half_life_obach': [0.551, 0.020],
    'clearance_hepatocyte_az': [0.495, 0.030],
    'clearance_microsome_az': [0.611, 0.016],
    'ld50_zhu': [0.541, 0.015],
    'herg': [0.848, 0.009],
    'ames': [0.854, 0.007],
    'dili': [0.920, 0.012]
}

P. S.

Fix a result typo for "BaseBoosting KyQVZ6b2" in "LD50_Zhu" dataset.

0.522 -> 0.552; based on the last line in their codes https://github.com/Oloren-AI/OCE-TDC/blob/main/submission_20221216.ipynb

[amva13]

@ada-f haven't gotten to look too much yet but just fyi. Might be relevant to you.

[amva13]

@feiyang-cai merged. Great work!

[feiyang-cai]

@amva13 Great! Thank you!