关于docker pull 超时的问题，我的有效的解决方案：

sudo vi /etc/docker/daemon.json

配置如下：

{
    "runtimes": {
        "nvidia": {
            "path": "nvidia-container-runtime",
            "runtimeArgs": []
        }
    },
...

github双重身份认证：浏览器插件，身份验证器 https://chromewebstore.google.com/detail/%E8%BA%AB%E4%BB%BD%E9%AA%8C%E8%AF%81%E5%99%A8/bhghoamapcdpbohphigoooaddinpkbai

Hfmirror国内镜像下载大模型的数据及代码

download_data.sh

#pip install -U huggingface_hub
export HF_ENDPOINT=https://hf-mirror.com

data_online_path="ystemsrx/Erotic_Literature_Collection"
local_path="ystemsrx/Erotic_Literature_Collection"

huggingface-cli download \
--token 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 \
--repo-type dataset \
--resume-download $data_online_path  \
--local-dir $local_path \
...

时序二分类模型示例

class LSTMModel(nn.Module):
    def __init__(self, input_size, hidden_size, output_size, num_layers=1):
        super(LSTMModel, self).__init__()
        self.lstm = nn.LSTM(input_size, hidden_size, num_layers, batch_first=True)
        self.fc = nn.Linear(hidden_size, output_size)

    def forward(self, x):
        out, (hn, cn) = self.lstm(x)
        out = self.fc(hn[-1])
        return torch.sigmoid(out)