dongqifong

numpy==1.21.5
pandas==1.3.5
matplotlib==3.5.1
scipy==1.7.3
scikit-learn==1.0.2

dongqifong

FROM pytorch/pytorch:1.9.0-cuda10.2-cudnn7-runtime

RUN mkdir /app

COPY requirements.txt /app

WORKDIR /app

RUN python -m pip install - upgrade pip && pip install -r requirements.txt && rm -rf requirements.txt

dongqifong

Run docker image and execute run.py

$ docker run --rm -it --init --gpus=all -v <絕對路徑>:/app --name <container_name> <image_name:tag> python run.py

e.g.

$ docker run --rm -it --init --gpus=all -v C:\Users\user\Desktop\docker\pytorch:/app --name container1 torch-gpu-1.9:0.1.0 python run.py

dongqifong

model.eval() # 取消dropout or batch normalization
with torch.no_grad(): # 不計算梯度
  predict = []
  for data, labels in test_loader:
    # Forward
    out = model(data)
    predict.append(out.numpy())
model.train() # 重新開啟dropout or batch normalization

dongqifong

# model object
model = MyModel()

# define lr, optimizer, loss function
lr = 0.001
optimizer = torch.optim.Adam(model.parameters(), lr=lr)
loss_function = nn.CrossEntropyLoss()

# Train
epochs = 100

dongqifong

class MyModel(nn.Module):
    def __init__(self):
        super(MyModel, self).__init__()
        # Encoder
        self.block1 = nn.Sequential(
            nn.Linear(128,64),
            nn.Tanh(),
            nn.Linear(64, 32),
            nn.Tanh(),
            nn.Linear(32, 16),