r"""Custom fully connected layer that supports basis function combinations defined by symbolic expressions. Args: in_features (int): Numb

r"""Build a neural network model of a custom architecture (GPU‑efficient version).

r"""TransformerDecoderLayer with TNLinear.

r"""TransformerEncoderLayer with TNLinear. Args: d_model: number of expected features (required). nhead: number of heads (require

r"""Dataset class to handle data pairs.

r"""Stack of N decoder layers with TNLinear.

r"""Stack of N encoder layers with TNLinear.

BLT(柏拉图)网关,OpenAI Chat Completions 兼容接口。 默认基址含 /v1,路径将拼接为 /chat/completions。

CSTCloud(科技云)提供商,OpenAI Chat Completions 兼容接口。 默认基址:https://uni-api.cstcloud.cn/v1 使用示例:model="CSTCloud/gpt-oss-120b" 或 "CSTCloud/qwen3:235b"

Policy network that outputs a probability distribution over the basis functions.

A parsed Python program.

Parse code to collect all required information and produce a `Program`.

Reinforcement Learning symbolic regressor that discovers a vectorized (homogeneous) expression of the form: f(x) = sum_k c_k * sum_j

A multi‑layer GRU with custom neuron aggregation. This is the equivalent of :class:`torch.nn.GRU` where the input is first augmented by

A GRU cell with custom neuron aggregation. This is the equivalent of :class:`torch.nn.GRUCell` where both linear transformations use :cl

A multi‑layer LSTM with custom neuron aggregation. This is the equivalent of :class:`torch.nn.LSTM` where the input is first augmented b

An LSTM cell with custom neuron aggregation. This is the equivalent of :class:`torch.nn.LSTMCell` where both linear transformations use

A multi‑layer RNN with custom neuron aggregation. This is the equivalent of :class:`torch.nn.RNN` where the input is first augmented by

An RNN cell with custom neuron aggregation. This is the equivalent of :class:`torch.nn.RNNCell` where both linear transformations use :c

r"""Transformer model with TNLinear in encoder/decoder layers. Args: d_model: number of expected features (default=512). nhead: n

r"""Class for plotting and visualizing training progress.

r"""Class for plotting and visualizing training progress.

r"""Basic module used to ensure the normal operation of MLPClassifier.

r"""Basic module used to ensure the normal operation of MLPRegressor.

Basic Block for resnet 18 and resnet 34

Residual block for resnet over 50 layers

Base class for multi‑layer RNNs with custom neuron aggregation. This class augments the input by applying the basis functions from the s

Base class for RNN cells with custom neuron aggregation. This class uses :class:`TNLinear` for both input‑to‑hidden and hidden‑to‑hidden

Base class: N-dimensional convolution layer supporting symbolic expressions.

Base class: N-dimensional transposed convolution layer supporting symbolic expressions.