xiang song(charlie.song)
xiang song(charlie.song)
Add doc about saving configuration files when processing graph input data (e.g. --output-conf-file). The saved configuration file will record how data are transformed (numerical features, categorical features, etc.). For example:...
GraphStorm should support multi-task learning as its built-in capability. * Unsupervised multi-task learning: Other than using link prediction as the sole supervision signal for unsupervised learning, one can add other...
https://github.com/awslabs/graphstorm/blob/main/python/graphstorm/gconstruct/file_io.py#L153 The default behavior of numpy.array(string) will allocate too much memory than required, which is not efficient.
We need to support following user experience: - [ ] #513 - [ ] Resuming a distillation task from a saved checkpoint - [ ] Provide an end2end experience of...
GraphStorm's edge sampler does not guarantee that for each edge type in training set it will sample more than 1 edges. This will cause an error in when batch-norm is...
Currently, the implementation of computing losses for link prediction training (in GSgnnLinkPredictionModel.forward) is not friendly to loss functions like contrastive loss and triple loss which requires that a positive edge...
*Issue #, if available:* *Description of changes:* By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.
Rename training_scripts and inference_scripts folders to training_examples and inference_examples respectively.
Currently, GraphStorm does not save/load optimizer states of learnable sparse embeddings. (https://github.com/awslabs/graphstorm/blob/main/python/graphstorm/model/utils.py#L336-L337). We need to support it.
*Issue #, if available:* #959 *Description of changes:* By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your...