[GDSC-ML] Practice: Improve accuracy of ImageNet Classification project

- Models

1. Initial “mymodel.yaml”

```powershell
epochs: 10
resume: None
learning_rate: **0.001**
weight_decay: 10e-6
inference_device: "cuda"

train:
  batch_size: 4
  num_workers: 4
  valid_size: 0.3
  train_path: "./2_data/train"

test:
  batch_size: 4
  num_workers: 4
  test_path: "./2_data/val"

model:
  base_model: "convnext_base"
```

2. “convnext_base.yaml”

I follwed the hyperparamters same as convnext-base-224_finetuned_on_ImageIn_annotations
Learning rate decreased.

   epochs: 10
   resume: None
   learning_rate: **2e-05**
   weight_decay: 10e-6
   inference_device: "cuda"

   train:
     batch_size: 16
     num_workers: 4
     valid_size: 0.3
     train_path: "./2_data/train"

   test:
     batch_size: 16
     num_workers: 4
     test_path: "./2_data/val"

   model:
     base_model: "convnext_base"

valid_loss: 2.0629

3. “convnext_base_1.yaml”

epochs: 10
resume: None
learning_rate: **4e-3**
weight_decay: 10e-6
inference_device: "cuda"

train:
  batch_size: 16
  num_workers: 4
  valid_size: 0.3
  train_path: "./2_data/train"

test:
  batch_size: 16
  num_workers: 4
  test_path: "./2_data/val"

model:
  base_model: "convnext_base"

4. “RestNet50.yaml”

learning_rate: 2e-05, batch size: 16

5. “ResNet50_batch256.yaml”

learning_rate: 0.001, batch size: 256