Model
Tranformer-XL Large PyTorch checkpoint trained with AMP
Use the NGC CLI to download:
Copied!
1 Version
19.11.0_ampSelected09/22/2022 5:10 PM UTC3.2 GBAccuracy: 00 EpochsBatch Size: 128GPU: V100 Copied!
19.11.0_ampSelected
09/22/2022 5:10 PM UTC3.2 GBAccuracy: 00 EpochsBatch Size: 128GPU: V100
Copied!
architecture
| Key | Value |
|---|---|
| type | Base |
performance
| Key | Value |
|---|---|
| perplexity | 18.24 |
training
| Key | Value |
|---|---|
| batch_size | 128 |
| learning rate | 0.01 |
| gpus | 16x V100 32G |
| training_precision | AMP |
| steps | 100000 |
| dataset | WikiText-103 |