get it working mode
Browse files
train.py
CHANGED
|
@@ -33,13 +33,13 @@ def train(rank, world_size):
|
|
| 33 |
|
| 34 |
# os.system("conda install -y -c conda-forge cudatoolkit")
|
| 35 |
|
| 36 |
-
MICRO_BATCH_SIZE =
|
| 37 |
-
BATCH_SIZE =
|
| 38 |
GRADIENT_ACCUMULATION_STEPS = BATCH_SIZE // MICRO_BATCH_SIZE
|
| 39 |
EPOCHS = 2
|
| 40 |
-
LEARNING_RATE = 2e-
|
| 41 |
LORA_R = 4
|
| 42 |
-
LORA_ALPHA =
|
| 43 |
LORA_DROPOUT = 0.05
|
| 44 |
|
| 45 |
device = torch.device(f"cuda:{rank}")
|
|
|
|
| 33 |
|
| 34 |
# os.system("conda install -y -c conda-forge cudatoolkit")
|
| 35 |
|
| 36 |
+
MICRO_BATCH_SIZE = 1
|
| 37 |
+
BATCH_SIZE = 16
|
| 38 |
GRADIENT_ACCUMULATION_STEPS = BATCH_SIZE // MICRO_BATCH_SIZE
|
| 39 |
EPOCHS = 2
|
| 40 |
+
LEARNING_RATE = 2e-10
|
| 41 |
LORA_R = 4
|
| 42 |
+
LORA_ALPHA = 8
|
| 43 |
LORA_DROPOUT = 0.05
|
| 44 |
|
| 45 |
device = torch.device(f"cuda:{rank}")
|