Delete module.rst

docs/source/tutorial/module.rst

-Creating Temporal GNN Models
-============================
-
-Continuous-time Temporal GNN Models
------------------------------------
-
-To create a continuous-time temporal GNN model, we first need to define a configuration file with the suffix yml to specify the model structures and parameters. Here we use the configuration file :code:`TGN.yml` for TGN model as an example:
-
-.. code-block:: yaml
-
-    sampling:
-      - layer: 1
-        neighbor: 
-          - 10
-        strategy: 'recent'
-        prop_time: False
-        history: 1
-        duration: 0
-        num_thread: 32
-    memory: 
-      - type: 'node'
-        dim_time: 100
-        deliver_to: 'self'
-        mail_combine: 'last'
-        memory_update: 'gru'
-        mailbox_size: 1
-        combine_node_feature: True
-        dim_out: 100
-    gnn:
-      - arch: 'transformer_attention'
-        use_src_emb: False
-        use_dst_emb: False
-        layer: 1
-        att_head: 2
-        dim_time: 100
-        dim_out: 100
-    train:
-      - epoch: 20
-        batch_size: 200
-        # reorder: 16
-        lr: 0.0001
-        dropout: 0.2
-        att_dropout: 0.2
-        all_on_gpu: True
-
-The configuration file is composed of four parts: :code:`sampling`, :code:`memory`, :code:`gnn` and :code:`train`. Here are their meanings:
-
-- :code:`sampling`: This part specifies the sampling strategy for the temporal graph. :code:`layer` field specifies the number of layers in the sampling strategy. The :code:`neighbor` field specifies the number of neighbors to sample for each layer. The :code:`strategy` field specifies the sampling strategy(recent or uniform). The :code:`prop_time` field specifies whether to propagate the time information. The :code:`history` field specifies the number of historical timestamps to use. The :code:`duration` field specifies the duration of the time window. The :code:`num_thread` field specifies the number of threads to use for sampling.
-- :code:`memory`: This part specifies the memory module. :code:`type` field specifies the type of memory module(node or none). :code:`dim_time` field specifies the dimension of the time embedding. :code:`deliver_to` field specifies the destination of the message. :code:`mail_combine` field specifies the way to combine the messages. :code:`memory_update` field specifies the way to update the memory. :code:`mailbox_size` field specifies the size of the mailbox. :code:`combine_node_feature` field specifies whether to combine the node features. :code:`dim_out` field specifies the dimension of the output.
-- :code:`gnn`: This part specifies the GNN module. :code:`arch` field specifies the architecture of the GNN module. :code:`use_src_emb` field specifies whether to use the source embedding. :code:`use_dst_emb` field specifies whether to use the destination embedding. :code:`layer` field specifies the number of layers in the GNN module. :code:`att_head` field specifies the number of attention heads. :code:`dim_time` field specifies the dimension of the time embedding. :code:`dim_out` field specifies the dimension of the output.
-- :code:`train`: This part specifies the training parameters. :code:`epoch` field specifies the number of epochs. :code:`batch_size` field specifies the batch size. :code:`lr` field specifies the learning rate. :code:`dropout` field specifies the dropout rate. :code:`att_dropout` field specifies the attention dropout rate. :code:`all_on_gpu` field specifies whether to put all the data on GPU.
-
-After defining the configuration file, we can firstly read the parameters from the configuration file and create the model by constructing a :code:`General Model` object:
-
-.. code-block:: python
-
-    def parse_config(f):
-        conf = yaml.safe_load(open(f, 'r'))
-        sample_param = conf['sampling'][0]
-        memory_param = conf['memory'][0]
-        gnn_param = conf['gnn'][0]
-        train_param = conf['train'][0]
-        return sample_param, memory_param, gnn_param, train_param
-    
-    sample_param, memory_param, gnn_param, train_param = parse_config('./config/{}.yml'.format(args.model))
-    model = GeneralModel(gnn_dim_node, gnn_dim_edge, sample_param, memory_param, gnn_param, train_param).cuda()
-    model = DDP(model)
-
-Then a :code:`GeneralModel` object is created. If needed, we can adjust the model's parameters by modifying the contents of the configuration file. Here we provide 5 models for continuous-time temporal GNNs:
-
-- :code:`TGN`: The TGN model proposed in `Temporal Graph Networks for Deep Learning on Dynamic Graphs <https://arxiv.org/abs/2006.10637>`__.
-- :code:`DyRep`: The DyRep model proposed in `Representation Learning and Reasoning on Temporal Knowledge Graphs <https://arxiv.org/abs/1803.04051>`__.
-- :code:`TIGER`: The TIGER model proposed in `TIGER: A Transformer-Based Framework for Temporal Knowledge Graph Completion <https://arxiv.org/abs/2302.06057>`__.
-- :code:`Jodie`: The Jodie model proposed in `JODIE: Joint Optimization of Dynamics and Importance for Online Embedding <https://arxiv.org/abs/1908.01207>`__.
-- :code:`TGAT`: The TGAT model proposed in `Temporal Graph Attention for Deep Temporal Modeling <https://arxiv.org/abs/2002.07962>`__.
\ No newline at end of file