Upload inferences.py

inferences.py

@@ -0,0 +1,53 @@
+import gradio as gr
+import torch
+from torch import nn
+from torchvision import transforms
+
+# Load the model cause it is necessary for pytorch 
+
+class MyMnist_ModelV0(nn.Module):
+  def __init__(self,  input_shape: int, hidden_units: int, hidden_units2: int, output_shape: int):
+        super().__init__()
+        self.layer_stack = nn.Sequential(
+            nn.Flatten(), # neural networks like their inputs in vector form
+            nn.Linear(in_features=input_shape, out_features=hidden_units),  # in_features = number of features in a data sample (784 pixels)
+            nn.ReLU(),
+            nn.Linear(in_features=hidden_units, out_features=hidden_units2),
+            nn.ReLU(),
+            nn.Linear(in_features=hidden_units2, out_features=output_shape)
+        )
+
+  def forward(self, x):
+      return self.layer_stack(x)
+
+# instance of the model
+load_model = MyMnist_ModelV0(input_shape=784,
+                          hidden_units=256,
+                          hidden_units2=128,
+                          output_shape=10
+)
+
+PATH = "C:/Users/Hp Pavilion/Downloads/state_dict_model.pth" # PATH where you load the model trained
+
+load_model.load_state_dict(torch.load(PATH))
+load_model.eval()
+def recognize_digit(image):
+    if image is not None:
+        # Preprocess of the image
+        transform = transforms.Compose([
+                        transforms.ToTensor(),
+                        transforms.Normalize((0.5,), (0.5,))
+                    ])
+        image = transform(image)
+        with torch.inference_mode(): # inference mode of pytoroch
+            prediction = load_model(image)
+        prediction = torch.softmax(prediction, dim=1)
+        return {str(i): float(prediction[0][i]) for i in range(10)}
+    else:
+        return ""
+    
+
+demo = gr.Interface(fn=recognize_digit, 
+                    inputs=gr.Image(shape=(28,28), image_mode="L", invert_colors=True, source="canvas"), 
+                    outputs=gr.Label(num_top_classes=1))
+demo.launch(True)