Spaces:
Sleeping
Sleeping
Update config/model_config.py
Browse files- config/model_config.py +23 -23
config/model_config.py
CHANGED
@@ -1,23 +1,23 @@
|
|
1 |
-
from dataclasses import dataclass
|
2 |
-
|
3 |
-
|
4 |
-
@dataclass
|
5 |
-
class ModelConfig:
|
6 |
-
# Model architecture
|
7 |
-
n_embeds: int = 384
|
8 |
-
n_heads: int = 6
|
9 |
-
n_layers: int = 6
|
10 |
-
dropout: float = 0.3
|
11 |
-
|
12 |
-
# Training
|
13 |
-
batch_size: int = 64
|
14 |
-
block_size: int = 128
|
15 |
-
max_iters: int = 20000
|
16 |
-
eval_interval: int = 250
|
17 |
-
eval_iters: int = 200
|
18 |
-
learning_rate: float = 3e-4
|
19 |
-
weight_decay: float = 0.1
|
20 |
-
|
21 |
-
# Paths
|
22 |
-
checkpoint_path: str = "checkpoints/model.pth"
|
23 |
-
data_path: str = "/data/
|
|
|
1 |
+
from dataclasses import dataclass
|
2 |
+
|
3 |
+
|
4 |
+
@dataclass
|
5 |
+
class ModelConfig:
|
6 |
+
# Model architecture
|
7 |
+
n_embeds: int = 384
|
8 |
+
n_heads: int = 6
|
9 |
+
n_layers: int = 6
|
10 |
+
dropout: float = 0.3
|
11 |
+
|
12 |
+
# Training
|
13 |
+
batch_size: int = 64
|
14 |
+
block_size: int = 128
|
15 |
+
max_iters: int = 20000
|
16 |
+
eval_interval: int = 250
|
17 |
+
eval_iters: int = 200
|
18 |
+
learning_rate: float = 3e-4
|
19 |
+
weight_decay: float = 0.1
|
20 |
+
|
21 |
+
# Paths
|
22 |
+
checkpoint_path: str = "checkpoints/model.pth"
|
23 |
+
data_path: str = "/data/input.txt"
|