feat: model --> ready to test train

CNN-model/cnn.py

@@ -14,7 +14,6 @@ class CNNPredictor(nn.Module):
     def __init__(
             self,
             vocab_size=256,
-            context_length=128,
             num_layers=3,
             hidden_dim=128,
             kernel_size=3,

CNN-model/data_utils.py

@@ -1,11 +0,0 @@
-import torch
-from torch.utils.data import TensorDataset
-
-
-def make_context_pairs(data: bytes, context_length: int) -> TensorDataset:
-    data = torch.tensor(list(data), dtype=torch.uint8)
-    sample_count = data.shape[0] - context_length
-    x = data.unfold(0, context_length, 1)[:sample_count]
-    y = data[context_length:]
-    return TensorDataset(x, y)
-

CNN-model/main_cnn.py

@@ -4,9 +4,10 @@ import torch.nn.functional as F
 import optuna.trial as tr
 from torch.utils.data import DataLoader
 from tqdm import tqdm
+import argparse
 
 from optuna_trial import create_model
-from data_utils import make_context_pairs
+from utils import make_context_pairs, load_data
 import optuna
 
 # hyper parameters
@@ -71,10 +72,22 @@ def objective_function(trial: tr.Trial, train_data: bytes, validation_data: byte
     return result["best_validation_loss"]
 
 if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--train-data", type=str, required=True)
+    parser.add_argument("--validation-data", type=str, required=True)
+    parser.add_argument("--batch-size", type=int, default=128)
+
+    args = parser.parse_args()
+    print(args)
+
     device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-    train_data = b""
-    validation_data = b""
-    batch_size = 0
+    train_data = load_data(args.train_data)
+    validation_data = load_data(args.validation_data)
+    batch_size = args.batch_size
+
+    print(f"training data length: {len(train_data)}")
+    print(f"validation data length: {len(validation_data)}")
+    print(f"batch size: {batch_size}")
 
     study = optuna.create_study(study_name="CNN network",direction="minimize")
     study.optimize(lambda trial: objective_function(trial, train_data, validation_data, batch_size), n_trials=10)

CNN-model/optuna_trial.py

@@ -10,7 +10,6 @@ def create_model(trial: tr.Trial, vocab_size: int = 256, context_length: int = 1
 
     return CNNPredictor(
         vocab_size=vocab_size,
-        context_length=context_length,
         num_layers=num_layers,
         hidden_dim=hidden_dim,
         kernel_size=kernel_size,

CNN-model/utils.py

@@ -0,0 +1,20 @@
+import torch
+from torch.utils.data import TensorDataset
+import matplotlib.pyplot as plt
+
+
+def make_context_pairs(data: bytes, context_length: int) -> TensorDataset:
+    data = torch.tensor(list(data), dtype=torch.long)
+    sample_count = data.shape[0] - context_length
+    x = data.unfold(0, context_length, 1)[:sample_count]
+    y = data[context_length:]
+    return TensorDataset(x, y)
+
+def print_distribution(from_to: tuple[int, int], probabilities: list[float]):
+    plt.hist(range(from_to[0], from_to[1]), weights=probabilities)
+    plt.show()
+
+
+def load_data(path: str) -> bytes:
+    with open(path, "rb") as f:
+        return f.read()