Use the NGC CLI to download:
Copied!
Model
BERT Distilled 6L-768D PyTorch checkpoint finetuned on SST-2 dataset using AMP
Use the NGC CLI to download:
Copied!
1 Version
20.12.0_ampSelected07/07/2022 2:26 PM UTC255.7 MBAccuracy: 0.9220 EpochsBatch Size: 0GPU: V100Copied!
20.12.0_ampSelected
07/07/2022 2:26 PM UTC255.7 MBAccuracy: 0.9220 EpochsBatch Size: 0GPU: V100
Copied!
architecture
| Key | Value |
|---|---|
| attn_heads | 12 |
| task | SST-2 |
| hidden_size | 768 |
| di | 3072 |
| layers | 6 |
| type | Dist-6L-768D |
performance
| Key | Value |
|---|---|
| accuracy | 0.922 |
training
| Key | Value |
|---|---|
| training_precision | AMP |
| dataset | SST-2 |