NVIDIA Deep Learning Examples
Use the NGC CLI to download:
Copied!
Model
NVIDIA Deep Learning Examples
ResNet50 checkpoints (PaddlePaddle, AMP, ImageNet)ResNet50 checkpoint trained with AMP on ImageNet
Use the NGC CLI to download:
Copied!
1 Version
22.05.0_ampSelected06/07/2022 12:36 PM UTC197.47 MBAccuracy: 090 EpochsBatch Size: 256GPU: A100Copied!
22.05.0_ampSelected
06/07/2022 12:36 PM UTC197.47 MBAccuracy: 090 EpochsBatch Size: 256GPU: A100
Copied!
architecture
| Key | Value |
|---|---|
| task | classification |
| type | resnet50 |
performance
| Key | Value |
|---|---|
| val_top1 | 77.1185427 |
| val_top5 | 93.2955994 |
training
| Key | Value |
|---|---|
| batch_size | 256 |
| weight_decay | 3.0517578125e-05 |
| num_classes | 1000 |
| lr | 2.048 |
| optimizer_batch_size | 2048 |
| arch | resnet50 |
| lr_schedule | cosine |
| dataset | ImageNet |
| epochs | 90 |
| warmup | 5 |
| label_smoothing | 0.1 |
| momentum | 0.875 |