search

ant-research / Parameter_Inference_Efficient_PIE

Apache License 2.0
8 stars 3 forks source link

Concerning the code efficiency #1

Open anyezhiya368 opened 2 years ago

anyezhiya368 commented 2 years ago

Hi, I am reproducing the result of the FTAI entity typing model and I am very interested in the expected time to achieve the reported result, for the training process seems to take way too long time and cannot reach the reported performance at my end. Many thanks!

chocolate9624 commented 2 years ago

Yes, the training process will take long time. The WikiKG90Mv2 is a huge dataset and our implemention is a single GPU version. Thus, a mutl-gpu implemention is needed to accelerate the network tunning and experiment process.