[Multi-task Learning] Add support for multi-task learning

.github/workflow_scripts/e2e_mgpu_check.sh

@@ -10,4 +10,5 @@ sh ./tests/end2end-tests/create_data.sh
 bash ./tests/end2end-tests/graphstorm-lp/mgpu_test.sh
 bash ./tests/end2end-tests/graphstorm-nc/mgpu_test.sh
 bash ./tests/end2end-tests/graphstorm-ec/mgpu_test.sh
+bash ./tests/end2end-tests/graphstorm-mt/mgpu_test.sh
 

python/graphstorm/__init__.py

@@ -28,5 +28,10 @@
 from .gsf import create_builtin_lp_model
 from .gsf import create_builtin_edge_model
 from .gsf import create_builtin_node_model
+from .gsf import create_task_decoder
+
+from .gsf import (create_builtin_node_decoder,
+                  create_builtin_edge_decoder,
+                  create_builtin_lp_decoder)
 from .gsf import (get_builtin_lp_train_dataloader_class,
                   get_builtin_lp_eval_dataloader_class)

python/graphstorm/config/__init__.py

@@ -46,3 +46,5 @@
                      BUILTIN_LP_LOSS_CONTRASTIVELOSS)
 from .config import (GRAPHSTORM_LP_EMB_L2_NORMALIZATION,
                      GRAPHSTORM_LP_EMB_NORMALIZATION_METHODS)
+
+from .config import TaskInfo