updates

Author: Programmer-RD-AI

ML/__init__.py

@@ -37,7 +37,7 @@
 # os.environ["WANDB_SILENT"] = "true"
 PROJECT_NAME = "NLP-Disaster Tweets"
 device = torch.device("cuda")
-BATCH_SIZE = 32
+BATCH_SIZE = 16
 torch.backends.cudnn.benchmark = True
 torch.cuda.empty_cache()
 torch.manual_seed(42)

ML/helper_functions/test.py

@@ -26,7 +26,7 @@ def test(self) -> Dict:
         with torch.inference_mode():
             for X, y in self.test_dataloader:
                 y = y[0]
-                X = F.to_tensor(X, padding_value=1).to("cuda")
+                X = F.to_tensor(X, padding_value=1).to(device)
                 y = torch.tensor(y).to("cuda")
                 preds = torch.argmax(torch.softmax(self.model(X), dim=1), dim=1)
                 loss = self.criterion(preds.float(), y.view(-1, 1).squeeze(1).float())

ML/helper_functions/train.py

@@ -43,8 +43,6 @@ def train(self, run_name: str) -> None:
                 loss.backward()
                 self.optimizer.step()
                 iterator.set_description(f"{i}/{len(self.train_dataloader)}")
-            if self.lr_schedular:
-                self.lr_schedular.step()
             iterator.set_description(f"Testing...")
             self.model.eval()
             wandb.log(

run.py

@@ -1,94 +1,39 @@
 from ML import *
 
+print(ROBERTA_BASE_ENCODER)
 
-def train(
-    batch_size: int = 32,
-    lr: float = 0.01,
-    test_split: float = 0.25,
-    optimizer=optim.Adam,
-    epochs: int = 5,
-    name: str = "",
-    lr_schedular=None,
-    transforms=None,
-):
-    train_data_loader, test_data_loader, valid_data_loader = Load_Data(
-        Main_DL,
-        Valid_Loader,
-        [
-            "/media/user/Main/Programmer-RD-AI/Programming/Learning/JS/NLP-Disaster-Tweets/ML/data/train.csv",
-            batch_size,
-            transforms,
-        ],
-        [
-            "/media/user/Main/Programmer-RD-AI/Programming/Learning/JS/NLP-Disaster-Tweets/ML/data/test.csv",
-            1,
-        ],
-        test_split,
-        42,
-    ).ld()
-    model = TL().to(device)
-    optimizer = optimizer(model.parameters(), lr=lr)
-    criterion = nn.CrossEntropyLoss()
-    config = {
-        "model": model,
-        "criterion": criterion,
-        "optimizer": optimizer,
-        "learning_rate": lr,
-    }
-    Train(
-        model,
-        epochs,
-        config,
-        train_data_loader,
-        test_data_loader,
-        valid_data_loader,
-        criterion,
-        optimizer,
-    ).train(f"{name}")
 
-
-train(
-    transforms=Transformer().transform(),
-    batch_size=16,
-    lr=1e-3,
-    test_split=0.25,
-    optimizer=optim.Adam,
-    lr_schedular=None,
-    name=f"1e-3",
-)
-train(
-    transforms=Transformer().transform(),
-    batch_size=16,
-    lr=1e-4,
-    test_split=0.25,
-    optimizer=optim.Adam,
-    lr_schedular=None,
-    name=f"1e-4",
-)
-train(
-    transforms=Transformer().transform(),
-    batch_size=16,
-    lr=1e-5,
-    test_split=0.25,
-    optimizer=optim.Adam,
-    lr_schedular=None,
-    name=f"1e-5",
-)
-train(
-    transforms=Transformer().transform(),
-    batch_size=16,
-    lr=1e-6,
-    test_split=0.25,
-    optimizer=optim.Adam,
-    lr_schedular=None,
-    name=f"1e-6",
-)
-train(
-    transforms=Transformer().transform(),
-    batch_size=16,
-    lr=1e-7,
-    test_split=0.25,
-    optimizer=optim.Adam,
-    lr_schedular=None,
-    name=f"1e-7",
-)
+train_data_loader, test_data_loader, valid_data_loader = Load_Data(
+    Main_DL,
+    Valid_Loader,
+    [
+        "/media/user/Main/Programmer-RD-AI/Programming/Learning/JS/NLP-Disaster-Tweets/ML/data/train.csv",
+        16,
+        Transformer().transform(),
+    ],
+    [
+        "/media/user/Main/Programmer-RD-AI/Programming/Learning/JS/NLP-Disaster-Tweets/ML/data/test.csv",
+        1,
+    ],
+    0.125,
+    42,
+).ld()
+model = TL().to(device)
+optimizer = optim.Adam(model.parameters(), lr=1e-5)
+criterion = nn.CrossEntropyLoss()
+config = {
+    "model": model,
+    "criterion": criterion,
+    "optimizer": optimizer,
+    "learning_rate": 1e-5,
+}
+Train(
+    model,
+    25,
+    config,
+    train_data_loader,
+    test_data_loader,
+    valid_data_loader,
+    criterion,
+    optimizer,
+).train(f"final")

wandb/latest-run

@@ -1 +1 @@
-run-20230728_100733-nffptso0
+run-20230728_113045-bthz2tjx