|
@@ -33,7 +33,7 @@ def train(
|
|
|
micro_batch_size: int = 4,
|
|
micro_batch_size: int = 4,
|
|
|
num_epochs: int = 3,
|
|
num_epochs: int = 3,
|
|
|
learning_rate: float = 3e-4,
|
|
learning_rate: float = 3e-4,
|
|
|
- cutoff_len: int = 512,
|
|
|
|
|
|
|
+ cutoff_len: int = 256,
|
|
|
val_set_size: int = 2000,
|
|
val_set_size: int = 2000,
|
|
|
# lora hyperparams
|
|
# lora hyperparams
|
|
|
lora_r: int = 8,
|
|
lora_r: int = 8,
|
Eric Wang