Merge branch 'main' into sm_image_dgl_2.4

.github/workflow_scripts/e2e_gb_check.sh

@@ -8,6 +8,6 @@ GS_HOME=$(pwd)
 # Install graphstorm from checked out code
 pip3 install "$GS_HOME" --upgrade
 
-bash ./tests/end2end-tests/setup.sh
 bash ./tests/end2end-tests/create_data.sh
 bash ./tests/end2end-tests/graphbolt-gs-integration/graphbolt-graph-construction.sh
+bash ./tests/end2end-tests/graphbolt-gs-integration/graphbolt-training-inference.sh

.github/workflows/continuous-integration.yml

@@ -191,6 +191,7 @@ jobs:
       uses: aws-actions/configure-aws-credentials@v1
       with:
         role-to-assume: arn:aws:iam::698571788627:role/github-oidc-role
+        role-duration-seconds: 14400
         aws-region: us-east-1
     - name: Checkout repository
       uses: actions/checkout@v3

docs/source/advanced/link-prediction.rst

@@ -12,8 +12,8 @@ Optimizing model performance
 ----------------------------
 GraphStorm incorporates three ways of improving model performance of link
 prediction. Firstly, GraphStorm avoids information leak in model training.
-Secondly, to better handle heterogeneous graphs, GraphStorm provides three ways
-to compute link prediction scores: dot product, DistMult and RotatE.
+Secondly, to better handle heterogeneous graphs, GraphStorm provides four ways
+to compute link prediction scores: dot product, DistMult, TransE, and RotatE.
 Thirdly, GraphStorm provides two options to compute training losses, i.e.,
 cross entropy loss and contrastive loss. The following sub-sections provide more details.
 
@@ -32,7 +32,7 @@ GraphStorm provides supports to avoid theses problems:
 
 Computing Link Prediction Scores
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
-GraphStorm provides three ways to compute link prediction scores: Dot Product, DistMult and RotatE.
+GraphStorm provides four ways to compute link prediction scores: Dot Product, DistMult, TransE, and RotatE.
 
 * **Dot Product**: The Dot Product score function is as:
 
@@ -53,7 +53,21 @@ GraphStorm provides three ways to compute link prediction scores: Dot Product, D
     The ``relation_emb`` values are initialized from a uniform distribution
     within the range of ``(-gamma/hidden_size, gamma/hidden_size)``,
     where ``gamma`` and ``hidden_size`` are hyperparameters defined in
-    :ref:`Model Configurations<configurations-model>`。
+    :ref:`Model Configurations<configurations-model>`.
+
+* **TransE**: The TransE score function is as:
+
+    .. math::
+        score = gamma - \|h+r-t\|^{frac{1}{2}} \text{or} gamma - \|h+r-t\|
+
+    where the ``head_emb`` is the node embedding of the head node,
+    the ``tail_emb`` is the node embedding of the tail node,
+    the ``relation_emb`` is the relation embedding of the specific edge type.
+    The ``relation_emb`` values are initialized from a uniform distribution
+    within the range of ``(-gamma/(hidden_size/2), gamma/(hidden_size/2))``,
+    where ``gamma`` and ``hidden_size`` are hyperparameters defined in
+    :ref:`Model Configurations<configurations-model>`.
+    To learn more information about TransE, please refer to `the DGLKE doc <https://dglke.dgl.ai/doc/kg.html#transe>`__.
 
 * **RotatE**: The RotatE score function is as:
 

docs/source/api/references/graphstorm.eval.rst

@@ -42,8 +42,6 @@ Evaluators
 
     GSgnnClassificationEvaluator
     GSgnnRegressionEvaluator
-    GSgnnMrrLPEvaluator
-    GSgnnPerEtypeMrrLPEvaluator
-    GSgnnHitsLPEvaluator
-    GSgnnPerEtypeHitsLPEvaluator
+    GSgnnLPEvaluator
+    GSgnnPerEtypeLPEvaluator
     GSgnnRconstructFeatRegScoreEvaluator

docs/source/api/references/graphstorm.model.rst

@@ -101,3 +101,7 @@ Decoder Layer
     LinkPredictContrastiveDistMultDecoder
     LinkPredictRotatEDecoder
     LinkPredictContrastiveRotatEDecoder
+    LinkPredictWeightedRotatEDecoder
+    LinkPredictTransEDecoder
+    LinkPredictContrastiveTransEDecoder
+    LinkPredictWeightedTransEDecoder

docs/source/cli/graph-construction/single-machine-gconstruct.rst

@@ -33,7 +33,7 @@ Full argument list of the ``gconstruct.construct_graph`` command
 * **-\-num-processes-for-nodes**: the number of processes to process node data simultaneously. Increase this number can speed up node data processing.
 * **-\-num-processes-for-edges**: the number of processes to process edge data simultaneously. Increase this number can speed up edge data processing.
 * **-\-output-dir**: (**Required**) the path of the output data files.
-* **-\-graph-name**: (**Required**) the name assigned for the graph.
+* **-\-graph-name**: (**Required**) the name assigned for the graph. The graph name must adhere to the `Python identifier naming rules<https://docs.python.org/3/reference/lexical_analysis.html#identifiers>`_ with the exception that hyphens (``-``) are permitted and the name can start with numbers.
 * **-\-remap-node-id**: boolean value to decide whether to rename node IDs or not. Adding this argument will set it to be true, otherwise false.
 * **-\-add-reverse-edges**: boolean value to decide whether to add reverse edges for the given graph. Adding this argument sets it to true; otherwise, it defaults to false. It is **strongly** suggested to include this argument for graph construction, as some nodes in the original data may not have in-degrees, and thus cannot update their presentations by aggregating messages from their neighbors. Adding this arugment helps prevent this issue.
 * **-\-output-format**: the format of constructed graph, options are ``DGL``,  ``DistDGL``.  Default is ``DistDGL``. It also accepts multiple graph formats at the same time separated by an space, for example ``--output-format "DGL DistDGL"``. The output format is explained in the :ref:`Output <gcon-output-format>` section above.

docs/source/cli/model-training-inference/configuration-run.rst

@@ -482,12 +482,12 @@ Link Prediction Task
     - Yaml: ``num_negative_edges_eval: 1000``
     - Argument: ``--num-negative-edges-eval 1000``
     - Default value: ``1000``
-- **lp_decoder_type**: Set the decoder type for loss function in Link Prediction tasks. Currently GraphStorm support  ``dot_product``, ``distmult`` and ``rotate``.
+- **lp_decoder_type**: Set the decoder type for loss function in Link Prediction tasks. Currently GraphStorm support  ``dot_product``, ``distmult``, ``rotate``, ``transe_l1``, and ``transe_l2``.
 
     - Yaml: ``lp_decoder_type: dot_product``
     - Argument: ``--lp-decoder-type dot_product``
     - Default value: ``distmult``
-- **gamma**: Set the value of the hyperparameter denoted by the symbol gamma. Gamma is used in the following cases: i/ focal loss for binary classification ii/ DistMult score function for link prediction and iii/ RotatE score function for link prediction.
+- **gamma**: Set the value of the hyperparameter denoted by the symbol gamma. Gamma is used in the following cases: i/ focal loss for binary classification ii/ DistMult score function for link prediction, iii/ TransE score function for link prediction, and iv/ RotatE score function for link prediction.
 
     - Yaml: ``gamma: 10.0``
     - Argument: ``--gamma 10.0``
@@ -586,4 +586,4 @@ GraphStorm provides a set of parameters to control GNN distillation.
 
     - Yaml: ``max_seq_len: 1024``
     - Argument: ``--max-seq-len 1024``
-    - Default value: ``1024``
+    - Default value: ``1024``

docs/source/cli/model-training-inference/distributed/sagemaker.rst

@@ -388,7 +388,7 @@ The rest of the arguments are passed on to ``sagemaker_train.py`` or ``sagemaker
 
 * **--task-type**: Task type.
 * **--graph-data-s3**: S3 location of the input graph.
-* **--graph-name**: Name of the input graph.
+* **--graph-name**: Name of the input graph. The graph name must adhere to the `Python identifier naming rules<https://docs.python.org/3/reference/lexical_analysis.html#identifiers>`_ with the exception that hyphens (``-``) are permitted and the name can start with numbers.
 * **--yaml-s3**: S3 location of yaml file for training and inference.
 * **--custom-script**: Custom training script provided by customers to run customer training logic. This should be a path to the Python script within the Docker image.
 * **--output-emb-s3**: S3 location to store GraphStorm generated node embeddings. This is an inference only argument.

graphstorm-processing/docker/push_gsprocessing_image.sh

@@ -42,7 +42,7 @@ parse_params() {
   IMAGE='graphstorm-processing'
   VERSION=`poetry version --short`
   LATEST_VERSION=${VERSION}
-  REGION=$(aws configure get region)
+  REGION=$(aws configure get region) || REGION=""
   REGION=${REGION:-us-west-2}
   ACCOUNT=$(aws sts get-caller-identity --query Account --output text)
   ARCH='x86_64'

graphstorm-processing/graphstorm_processing/distributed_executor.py

@@ -54,6 +54,7 @@
 import json
 import logging
 import os
+import re
 from pathlib import Path
 import tempfile
 import time
@@ -540,7 +541,11 @@ def parse_args() -> argparse.Namespace:
     parser.add_argument(
         "--graph-name",
         type=str,
-        help="Name for the graph being processed.",
+        help="Name for the graph being processed."
+        "The graph name must adhere to the Python "
+        "identifier naming rules with the exception "
+        "that hyphens (-) are permitted and the name "
+        "can start with numbers",
         required=False,
         default=None,
     )
@@ -564,6 +569,33 @@ def parse_args() -> argparse.Namespace:
     return parser.parse_args()
 
 
+def check_graph_name(graph_name):
+    """Check whether the graph name is a valid graph name.
+
+        We enforce that the graph name adheres to the Python
+        identifier naming rules as in
+        https://docs.python.org/3/reference/lexical_analysis.html#identifiers,
+        with the exception that hyphens (-) are permitted
+        and the name can start with numbers.
+        This helps avoid the cases when an invalid graph name,
+        such as `/graph`, causes unexpected errors.
+
+        Note: Same as graphstorm.utils.check_graph_name.
+
+    Parameter
+    ---------
+    graph_name: str
+        Graph Name.
+    """
+    gname = re.sub(r"^\d+", "", graph_name)
+    assert gname.replace("-", "_").isidentifier(), (
+        "GraphStorm expects the graph name adheres to the Python"
+        "identifier naming rules with the exception that hyphens "
+        "(-) are permitted and the name can start with numbers. "
+        f"Got: {graph_name}"
+    )
+
+
 def main():
     """Main entry point for GSProcessing"""
     # Allows us to get typed arguments from the command line
@@ -572,6 +604,7 @@ def main():
         level=gsprocessing_args.log_level,
         format="[GSPROCESSING] %(asctime)s %(levelname)-8s %(message)s",
     )
+    check_graph_name(gsprocessing_args.graph_name)
 
     # Determine execution environment
     if os.path.exists("/opt/ml/config/processingjobconfig.json"):

python/graphstorm/__init__.py

@@ -29,7 +29,8 @@
 from .gsf import create_builtin_edge_model
 from .gsf import create_builtin_node_model
 from .gsf import (create_task_decoder,
-                  create_evaluator)
+                  create_evaluator,
+                  create_lp_evaluator)
 
 from .gsf import (create_builtin_node_decoder,
                   create_builtin_edge_decoder,

python/graphstorm/config/__init__.py

@@ -31,7 +31,9 @@
 
 from .config import (BUILTIN_LP_DOT_DECODER,
                      BUILTIN_LP_DISTMULT_DECODER,
-                     BUILTIN_LP_ROTATE_DECODER)
+                     BUILTIN_LP_ROTATE_DECODER,
+                     BUILTIN_LP_TRANSE_L1_DECODER,
+                     BUILTIN_LP_TRANSE_L2_DECODER)
 from .config import SUPPORTED_LP_DECODER
 
 from .config import (GRAPHSTORM_MODEL_EMBED_LAYER,