Create ft_trainer_config.json
Browse files- ft_trainer_config.json +23 -0
ft_trainer_config.json
ADDED
|
@@ -0,0 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"batch_size": 64,
|
| 3 |
+
"learning_rate": 0.0001,
|
| 4 |
+
"weight_decay": 0.01,
|
| 5 |
+
"max_epochs": 60,
|
| 6 |
+
"tasks": {
|
| 7 |
+
"prediction": 0.6,
|
| 8 |
+
"lm": 0.4
|
| 9 |
+
},
|
| 10 |
+
"compile": true,
|
| 11 |
+
"enable_ddp": false,
|
| 12 |
+
"dtype": "float32",
|
| 13 |
+
"num_workers": 16,
|
| 14 |
+
"beta1": 0.9,
|
| 15 |
+
"beta2": 0.95,
|
| 16 |
+
"gradient_accumulation_steps": 1,
|
| 17 |
+
"grad_clip": 1.0,
|
| 18 |
+
"decay_lr": true,
|
| 19 |
+
"log_interval": 10,
|
| 20 |
+
"save_interval": 5,
|
| 21 |
+
"min_lr": 1e-06,
|
| 22 |
+
"warmup_iters": 54
|
| 23 |
+
}
|