Model
TFT Base PyTorch checkpoint trained with AMP on Electricity dataset
Use the NGC CLI to download:
Copied!
2 Versions
04/06/2023 8:02 PM UTC6.44 MBAccuracy: 030 EpochsBatch Size: 8192GPU: A100 Copied!
04/06/2023 8:02 PM UTC6.44 MBAccuracy: 030 EpochsBatch Size: 8192GPU: A100
Copied!
architecture
| Key | Value |
|---|---|
| type | base |
performance
| Key | Value |
|---|---|
| P90 | 0.03 |
| P50 | 0.058 |
| P10 | 0.028 |
training
| Key | Value |
|---|---|
| batch_size | 8192 |
| clip_grad | 0.0 |
| lr | 0.001 |
| dropout | 0.1 |
| training_precision | AMP |
| dataset | Electricity |
| bs_per_gpu | 1024 |
| epochs | 30 |
21.06.0_ampSelected09/22/2022 5:08 PM UTC6.44 MBAccuracy: 040 EpochsBatch Size: 16384GPU: V100 Copied!
21.06.0_ampSelected
09/22/2022 5:08 PM UTC6.44 MBAccuracy: 040 EpochsBatch Size: 16384GPU: V100
Copied!
architecture
| Key | Value |
|---|---|
| type | base |
performance
| Key | Value |
|---|---|
| P90 | 0.028 |
| P50 | 0.054 |
| P10 | 0.027 |
training
| Key | Value |
|---|---|
| batch_size | 16384 |
| clip_grad | 0.0 |
| lr | 0.003 |
| dropout | 0.1 |
| training_precision | AMP |
| dataset | Electricity |
| bs_per_gpu | 2048 |
| epochs | 40 |