leffff/graphormer-pyg

Time-cost

Opened this issue · 2 comments

Hey,I find it will be veryyyyyyyyyyyy cost time if i set a slightly larger batch_num for this model in graphormer calculation phase for a total batch graph . One simple method to solve, i do , its use a loop do every single graph for graphormer calculation in a batch ,then stack the results for final result for the batch.

I am a beginner in Torch , I would like to inquire if this method will affect the accuracy of the model ? its actually faster in my task.

Thank you!

This method will for sure NOT affect the accuracy

This method will for sure NOT affect the accuracy

Though i donnot keep in this question, thx for your reply very much!