HW 202337458 Backbone Swin Replacement

Required Code Included

• register_for_auto_class
• AutoConfig
• AutoModel

Files

• modeling_custom_swin.py
• backbone_swin_replacement.ipynb
• custom_swin_model/
• test_result.json
• README.md

How to Use

import torch
from modeling_custom_swin import CustomBackboneConfig, CustomSwinImageClassifier

config = CustomBackboneConfig()
model = CustomSwinImageClassifier(config)

pixel_values = torch.randn(2, 3, 224, 224)
labels = torch.tensor([0, 1])

outputs = model(pixel_values=pixel_values, labels=labels)
print(outputs.loss)
print(outputs.logits.shape)

Expected Output

torch.Size([2, 2])

--- license: apache-2.0 tags: - backbone - swin-transformer - image-classification - automodel - homework --- # HW-202337458-backbone-swin-replacement ## Assignment 커스텀 Backbone 모델 교체 실습 과제입니다. ## Student - Student ID: 202337458 - Hugging Face ID: hf-june ## Backbone Replacement - 기존 Backbone: ViT - 교체 Backbone: Swin Transformer - 사용 모델: microsoft/swin-tiny-patch4-window7-224 ## Required Code Included - register_for_auto_class - AutoConfig - AutoModel ## Files - modeling_custom_swin.py - backbone_swin_replacement.ipynb - custom_swin_model/ - test_result.json - README.md ## How to Use ```python import torch from modeling_custom_swin import CustomBackboneConfig, CustomSwinImageClassifier config = CustomBackboneConfig() model = CustomSwinImageClassifier(config) pixel_values = torch.randn(2, 3, 224, 224) labels = torch.tensor([0, 1]) outputs = model(pixel_values=pixel_values, labels=labels) print(outputs.loss) print(outputs.logits.shape) ``` ## Expected Output ```text torch.Size([2, 2]) ```