Model
SE(3)-Transformer checkpoint trained on QM9 homo task
Use the NGC CLI to download:
Copied!
1 Version
21.07.0_amp_bs-240_epochs-300Selected09/22/2022 5:26 PM UTC106.36 MBAccuracy: 0.03424300 EpochsBatch Size: 240GPU: A100 Copied!
21.07.0_amp_bs-240_epochs-300Selected
09/22/2022 5:26 PM UTC106.36 MBAccuracy: 0.03424300 EpochsBatch Size: 240GPU: A100
Copied!
architecture
| Key | Value |
|---|---|
| task | training |
| type | base |
performance
| Key | Value |
|---|---|
| accuracy | 0.03424 |
| training_loss | 0.03307 |
training
| Key | Value |
|---|---|
| batch_size | 240 |
| LR_phase2 | 0.0001 |
| LR_phase1 | 0.004 |
| training_precision | AMP |
| bs_phase2 | 240 |
| bs_phase1 | 240 |
| dataset | QM9 |
| epochs | 300 |
| iterations | 300 |