adding 5 different models and switching effnet with swin transformer

Vit_20%Food101.pth

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:cbdb15e1e44c857322f027b28da05df6dadb78a6514b5ca525021d9a4dd31da0
-size 343566279

app.py

@@ -13,17 +13,17 @@ from pathlib import Path
 ###############################
 
 def predict(img):
-    effnetb2, transform = create_model()
+    swinb, transform = create_model()
     class_names = []
     with open('classes.txt', 'r') as f:
         class_names = [foodname.strip() for foodname in f.readlines()] 
-    effnetb2.load_state_dict(torch.load(f = 'effnet_b2_food101.pth', map_location = torch.device('cpu'))) 
+    swinb.load_state_dict(torch.load(f = 'models/swin40%newer.pth', map_location = torch.device('cpu'))) 
     
     img = transform(img).unsqueeze(0)
 
-    effnetb2.eval()
+    swinb.eval()
     with torch.inference_mode():
-        pred_label = class_names[effnetb2(img).softmax(dim = 1).argmax(dim = 1)]
+        pred_label = class_names[swinb(img).softmax(dim = 1).argmax(dim = 1)]
     print(pred_label)
     return pred_label
 

model.py

@@ -8,7 +8,7 @@ from torchvision import models, transforms
 ###############################
 
 def create_model():
-    weights = models.EfficientNet_B2_Weights.DEFAULT
+    weights = models.Swin_B_Weights.DEFAULT
     transform = weights.transforms()
 
     model = models.efficientnet_b2(weights = weights)

Vit_40%lre-4DataAugFood101.pth → models/Vit_40%lre-4DataAugFood101.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:513cc3617bf2988d9a3bf2361bfa1df5e619e078cdb7e2ebafe934b70e3fb6a9
-size 343568191
+oid sha256:17f355e8cea1d64a2a99e7a293de89b54928c1e3f83f306572ffa8cf6dbff2ba
+size 226492416

swin60%.pth

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:2e66e4a7c0b16980324566f0e8f81e1133942308e4dfe91dfabab2af89c62241
-size 347973311