Merge pull request #166 from prphrl/docs

Update docs and add Sphinx configuration key.

docs/source/cheatsheet/model_cheatsheets.rst → docs/source/cheatsheet/model_cheatsheet.rst

@@ -15,10 +15,14 @@ GNN Cheatsheet
       - The Heterogeneous Graph Attention Network (HAN) model, as introduced in the `"Heterogeneous Graph Attention Network" <https://arxiv.org/abs/1903.07293>`__ paper.
     * - OGC
       - The OGC method from the `"From Cluster Assumption to Graph Convolution: Graph-based Semi-Supervised Learning Revisited" <https://arxiv.org/abs/2309.13599>`__ paper.
+    * - HGT
+      - The Heterogeneous Graph Transformer (HGT) model, as introduced in the `"Heterogeneous Graph Transformer" <https://arxiv.org/abs/2003.01332>`__ paper.
+    * - SAGE
+      - The SAGE model, as introduced in the `"Inductive Representation Learning on Large Graphs" <https://arxiv.org/abs/1706.02216>`__ paper.
     * - RECT_L
       - The RECT model, or more specifically its supervised part RECT-L, from the `"Network Embedding with Completely-imbalanced Labels" <https://arxiv.org/abs/2007.03545>`__ paper.
     * - Label-Free-GNN
-      - Two classical methods (Random and VertexCover) and PS-FeatProp-W from the `"Label-free Node Classification on Graphs with Large Language Models (LLMS)" <https://arxiv.org/abs/2310.04668>`__ paper.
+      - Two classic methods (Random and VertexCover) and PS-FeatProp-W from the `"Label-free Node Classification on Graphs with Large Language Models (LLMS)" <https://arxiv.org/abs/2310.04668>`__ paper.
     * - TAPE
       - The TAPE method from the `"Harnessing Explanations: LLM-to-LM Interpreter for Enhanced Text-Attributed Graph Representation Learning" <https://arxiv.org/abs/2305.19523>`__ paper.
 
@@ -36,6 +40,10 @@ TNN Cheatsheet
       - The Tab-Transformer model introduced in the `"TabTransformer: Tabular Data Modeling Using Contextual Embeddings" <https://arxiv.org/abs/2012.06678>`_ paper.
     * - TabNet
       - The TabNet model introduced in the `"TabNet: Attentive Interpretable Tabular Learning" <https://arxiv.org/abs/1908.07442>`_ paper.
+    * - Excelformer
+      - The ExcelFormerConv Layer introduced in the `"ExcelFormer: A neural network surpassing GBDTs on tabular data" <https://arxiv.org/abs/2301.02819>`_ paper.
+    * - Trompt
+      - The TromptConv Layer introduced in the `"Trompt: Towards a Better Deep Neural Network for Tabular Data" <https://arxiv.org/abs/2305.18446>`_ paper.
 
 
 RTL Cheatsheet

docs/source/index.rst

@@ -31,12 +31,12 @@ Highlight Features:
    :maxdepth: 1
    :caption: TUTORIALS
 
+   tutorial/transforms
+   tutorial/convolutions
    tutorial/gnns
    tutorial/tnns
    tutorial/rtls
    tutorial/llm_methods
-   tutorial/convolutions
-   tutorial/transforms
 
 
 .. toctree::
@@ -56,8 +56,7 @@ Highlight Features:
    :maxdepth: 1
    :caption: CHEATSHEETS
 
-   cheatsheet/model_cheatsheets
-   cheatsheet/dataset_cheatsheets
+   cheatsheet/model_cheatsheet
 
 
 .. toctree::

docs/source/introduce/table_data_handle.rst

@@ -6,8 +6,8 @@ Data Handling of Tables
 
 A table contains many different columns with many different types. Each column type in Rllm is described by a certain semantic type, i.e., ColType. Rllm supports two basic column types so far:
 
-- :obj:`ColType.CATEGORICAL`:represent categorical or discrete data, such as grade levels in a student dataset and diabetes types in a diabetes dataset.
-- :obj:`ColType.NUMERICAL`:represent numerical or continuous data, such as such as temperature in a weather dataset and income in a salary dataset.
+- :obj:`ColType.CATEGORICAL`: represent categorical or discrete data, such as grade levels in a student dataset and diabetes types in a diabetes dataset.
+- :obj:`ColType.NUMERICAL`: represent numerical or continuous data, such as such as temperature in a weather dataset and income in a salary dataset.
 
 A table in Rllm is described by an instance of :class:`~rllm.data.table_data.TableData` with many default attributes:
 
@@ -157,4 +157,4 @@ Rllm also supports a custom dataset, so that you can use Rllm for your own probl
 
     # Set "y" as the target column.
 
-    dataset = TableData(df, col_types=col_types, target_col="y")
+    dataset = TableData(df, col_types=col_types, target_col="y")

readthedocs.yml

@@ -1,5 +1,8 @@
 version: 2
 
+sphinx:
+  configuration: docs/conf.py
+
 build:
    os: ubuntu-22.04
    tools:

rllm/nn/conv/table_conv/ft_transformer_conv.py

@@ -10,9 +10,8 @@
 
 
 class FTTransformerConv(torch.nn.Module):
-    r"""The FT-Transformer backbone in the
-    `"Revisiting Deep Learning Models for Tabular Data"
-    <https://arxiv.org/abs/2106.11959>`_ paper.
+    r"""The FT-Transformer backbone in the `"Revisiting Deep Learning
+    Models for Tabular Data" <https://arxiv.org/abs/2106.11959>`_ paper.
 
     This module concatenates a learnable CLS token embedding :obj:`x_cls` to
     the input tensor :obj:`x` and applies a multi-layer Transformer on the

rllm/nn/conv/table_conv/saint_conv.py

@@ -11,10 +11,9 @@
 class SAINTConv(torch.nn.Module):
     r"""The SAINTConv Layer introduced in the
     `"SAINT: Improved Neural Networks for Tabular Data via Row Attention
-        and Contrastive Pre-Training"
-    <https://arxiv.org/abs/2106.01342>`_ paper.
+    and Contrastive Pre-Training" <https://arxiv.org/abs/2106.01342>`__ paper.
 
-    This layer applies two `TransformerEncoder` modules: one for aggregating
+    This layer applies two :obj:`TransformerEncoder` modules: one for aggregating
     information between columns, and another for aggregating information
     between samples. This dual attention mechanism allows the model to capture
     complex relationships both within the features of a single sample and