Upload 4 files

app.py

@@ -0,0 +1,74 @@
+
+### 1. Imports and class names setup ###
+
+# Imports
+import gradio as gr
+import os
+import torch
+
+from model import create_vit_model
+from timeit import default_timer as timer
+from typing import Tuple, Dict
+
+# Setup class names
+class_names = ['airplane', 'automobile', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck']
+
+### 2. Model and Transforms Preparation ###
+vit_model, vit_transforms = create_vit_model(num_classes = 10)
+
+# Load save weights
+vit_model.load_state_dict(
+    torch.load(
+        f="vit_cifar10_state_dict.pth",
+        map_location=torch.device("cpu") # load the model to the cpu
+    )
+)
+
+### 3. Predict function ###
+def predict(img) -> Tuple[Dict, float]:
+  # Timer
+  start_time = timer()
+
+  # Transform the input image to work with ViT
+  img = vit_transforms(img).unsqueeze(0) # unsqueeze = add batch dimension on 0th index
+
+  # Eval mode and torch inference mode on
+  vit_model.eval()
+  with torch.inference_mode():
+    # Pass transformed image through the model and turn prediction logits into probabilities
+    pred_probs = torch.softmax(vit_model(img), dim = 1)
+
+  # Create prediction label and prediction probability dictionary
+  pred_labels_and_probs = {class_names[i]: float(pred_probs[0][i]) for i in range(len(class_names))}
+
+  # Calculate prediction time
+  end_time = timer()
+  pred_time = round(end_time - start_time, 3)
+
+  # Return pred dict and pred time
+  return pred_labels_and_probs, pred_time
+
+### 4. Gradio app ###
+
+# Create title for the gradio
+title = "Object Classifier - Erdem Atak Version"
+description = "ViT computer vision model to classify CIFAR-10 objects"
+article = "PyTorch Model Deployment"
+
+# Create example list
+example_list = [["examples/" + example] for example in os.listdir("examples")]
+
+# Create the gradio demo
+demo = gr.Interface(fn = predict, # it maps inputs to outputs
+                    inputs = gr.Image(type = "pil"),
+                    outputs = [gr.Label(num_top_classes = 3,
+                                       label = "Predictions"),
+                              gr.Number(label = "Prediction Time (s)")],
+                    examples = example_list, # example list above
+                    title = title,
+                    description = description,
+                    article = article)
+
+# launch the demo
+demo.launch(debug = False,
+            share = True ) # public shareable URL

model.py

@@ -0,0 +1,24 @@
+
+# import the needed dependencies for this py file
+import torch
+import torchvision
+
+from torch import nn
+
+#Turning train as a function as 1. setup, 2 getr transforms 3. setup model instance 4. Freeze base layers and adjust output layers
+def create_vit_model(num_classes: int = 10,
+                     seed: int = 24):
+  # 1, 2, 3, Create ViT weights, transform and model
+  weights = torchvision.models.ViT_B_16_Weights.DEFAULT
+  transforms = weights.transforms()
+  model = torchvision.models.vit_b_16(weights=weights)
+
+  # 4. Freeze the base layers
+  for param in model.parameters()
+    param.requires_grad = False
+
+  # 5. Adjust the number of heads(output)
+  torch.manual_seed(seed)
+  model.heads = nn.Sequential(nn.Linear(in_features=768, out_features=num_classes, bias=True))
+
+  return model, transforms

requirements.txt

@@ -0,0 +1,3 @@
+torch
+torchvision
+gradio

vit_cifar10_state_dict.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b92b218af10764547dbcb75b3bf32681970fc48bf5dc14339a3537811c5403c
+size 343288529