add tabnet

[GreatV]

• [Dy2St] PyTorch 和 Paddle 仓库模型导出成功率对比任务 PaddlePaddle/Paddle#58985

• pytorch 无法导出

torch._dynamo.exc.UserError: Dynamic control flow is not supported at the moment. Please use functorch.experimental.control_flow.cond to explicitly capture the control flow

from user code:
   File "/home/greatx/repos/tabnet/pytorch_tabnet/tab_network.py", line 615, in forward
    x = self.embedder(x)
  File "/home/greatx/repos/tabnet/venv/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1527, in _call_impl
    return forward_call(*args, **kwargs)
  File "/home/greatx/repos/tabnet/pytorch_tabnet/tab_network.py", line 886, in forward
    if is_continuous:

Set TORCH_LOGS="+dynamo" and TORCHDYNAMO_VERBOSE=1 for more information

• paddle 无法导出

paddle.jit.dy2static.error.Dy2StKeyError: In transformed code:

    File "/home/greatx/repos/tabnet.paddle/pytorch_tabnet/tab_network.py", line 605, in forward
        x = self.embedder(x)
    File "/home/greatx/repos/tabnet.paddle/pytorch_tabnet/tab_network.py", line 859, in forward
        if self.skip_embedding:
    File "/home/greatx/repos/tabnet.paddle/pytorch_tabnet/tab_network.py", line 863, in forward
        for feat_init_idx, is_continuous in enumerate(self.continuous_idx):
    File "/home/greatx/repos/tabnet.paddle/pytorch_tabnet/tab_network.py", line 864, in forward
        if is_continuous:
    File "/home/greatx/repos/tabnet.paddle/pytorch_tabnet/tab_network.py", line 867, in forward
            cols.append(x[:, feat_init_idx].astype(dtype="float32").reshape((-1, 1)))
        else:
            cols.append(
            ~~~~~~~~~~~~ <--- HERE
                self.embeddings[cat_feat_counter](
                    x[:, feat_init_idx].astype(dtype="int64")

    File "/home/greatx/repos/tabnet.paddle/venv/lib/python3.10/site-packages/paddle/nn/layer/container.py", line 438, in __getitem__
        return self._sub_layers[str(idx)]

    KeyError: 'var _generated_var_2 : LOD_TENSOR.shape().dtype(int64).stop_gradient(True)'

[SigureMo]

LGTMeow