feat: add GIN results for `Peptides-func` - Githubissues

vijaydwivedi75 / lrgb

Long Range Graph Benchmark, NeurIPS 2022 Track on D&B

MIT License

149 stars 18 forks source link

feat: add GIN results for `Peptides-func` #12

Closed SauravMaheshkar closed 9 months ago

SauravMaheshkar commented 1 year ago

This PR aims to add benchmark performance for a simple Graph Isomorphism Network trained on the Peptides-func dataset (5 fold).

The model had the following configuration:

Number of parameters: 235k
Batch Size: 128
Learning Rate: 0.001 (No Scheduler)
Optimizer: AdamW
Epochs: 250
Number of Layers: 5
Hidden Dimension: 128
Jumping Knowledge Used
GIN-ε variant
No Dropout

The corresponding Weights & Biases run can be found here and the code used for training the model can be found here

Request for Review: @vijaydwivedi75 @migalkin @rampasek

SauravMaheshkar commented 9 months ago

Closing for now, happy to reopen if there is interest from the maintainers