ImageNet Classification Notebook Development

I need a well-structured Python notebook that walks through the full image-classification workflow on the ImageNet dataset. The notebook should start with downloading or mounting ImageNet, move through data preprocessing and augmentation, define and train a modern CNN (transfer learning with ResNet, EfficientNet, or a comparable backbone), and finish with clear validation metrics and loss/accuracy plots. Please include explanatory markdown so the code is easy to follow, plus comments on any hyper-parameters you tune. You are free to use PyTorch or TensorFlow—whichever you feel gives the quickest, most reproducible pipeline on a single-GPU setup. The notebook must execute top-to-bottom without manual tweaks once paths and batch size are set. Deliverables: • Jupyter notebook (.ipynb) runnable end-to-end on ImageNet • Helper script or cell that handles data download/mounting • Trained model saved to disk and re-load demo cell • README section inside the notebook with environment setup and expected training time Acceptance criteria: the notebook runs without errors, achieves a reasonable top-1 accuracy on the validation split within a small proof-of-concept training window (e.g., a few epochs), and all steps—data loading, augmentation, model definition, training loop, evaluation, and inference demo—are clearly documented. Budget: INR 12500–37500