Theodore Vasiloudis
Theodore Vasiloudis
GSF tries to download the models into /opt/ml/gsgnn_model, as seen here https://github.com/thvasilo/graphstorm/blob/8e7c4c2e10accb114f2beccaa36ec3094d01241c/python/graphstorm/sagemaker/sagemaker_infer.py#L173 One a job with large model (learnable embeddings included) we see this in the logs in terms of...
Observed the following today: ```wrap subprocess.CalledProcessError: Command 'ssh -o StrictHostKeyChecking=no -p 2222 172.31.93.146 '(export DGL_IP_CONFIG=/ip_list.txt DGL_NUM_SERVER=1 PYTHONPATH=/graphstorm/python/:/root/dgl/tools/: RANK=16 MASTER_ADDR=172.31.95.143 MASTER_PORT=12345; /opt/gs-venv/bin/python /root/dgl/tools/distpartitioning/data_proc_pipeline.py --world-size 20 --partitions-dir /mount/gpartition/spear-local-graph-04-23/dgl-2.3a240609/range/20-parts/partition_assignment --input-dir /mount/gsprocessing/spear-local-graph-04-23 --graph-name spear-global-graph-0423-features...
Our run configuration docs are currently at a high level. We should have more detailed documentation that explain the different models, loss functions etc. in a form of user guide...
Our tutorials like https://graphstorm.readthedocs.io/en/v0.3.1/tutorials/quick-start.html are using absolute paths, assuming they are being executed from within the GS Docker container. When referring to files in the repository we should be using...
We should document the expected schema for the configuration YAML and verify it during parsing, to provide better errors if the user has wrong indentation in their config.
Hello, There's a consideration at [XGBoost](https://github.com/dmlc/xgboost/issues/5746) about potentially using the KLL sketch to represent feature value histograms. One potential blocker is the need for a weighted version of the sketch,...
*Issue #, if available:* Fixes https://github.com/awslabs/graphstorm/issues/1107 *Description of changes:* * Add handling for `*` in file paths, for local and S3 input in GSProcessing * For local we glob files,...
For classification tasks, cross entropy will have a prediction shape with number of columns that's equal to the number of classes. So for binary tasks we have 2 columns in...
*Issue #, if available:* *Description of changes:* * Add instructions and support to launch SageMaker jobs locally use SageMaker's own local mode, instead of the customer Docker compose solution. *...