markrodrigo
commited on
Commit
•
87cf2f4
1
Parent(s):
d98cee2
Update Hypers
Browse files
README.md
CHANGED
@@ -173,6 +173,18 @@ More information needed
|
|
173 |
### Training data
|
174 |
Custom synthetic
|
175 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
176 |
### Training results
|
177 |
|
178 |
| Training Loss | Epoch | Step | Validation Loss |
|
|
|
173 |
### Training data
|
174 |
Custom synthetic
|
175 |
|
176 |
+
### Training hyperparameters
|
177 |
+
|
178 |
+
The following hyperparameters were used during training:
|
179 |
+
- learning_rate: 3e-05
|
180 |
+
- train_batch_size: 10
|
181 |
+
- eval_batch_size: 3
|
182 |
+
- distributed_type: multi-GPU
|
183 |
+
- num_devices: 2
|
184 |
+
- optimizer: Adam 8bit
|
185 |
+
- lr_scheduler_type: linear
|
186 |
+
- num_epochs: 4
|
187 |
+
|
188 |
### Training results
|
189 |
|
190 |
| Training Loss | Epoch | Step | Validation Loss |
|