search

Tencent / NeuralNLP-NeuralClassifier

An Open-source Neural Hierarchical Multi-label Text Classification Toolkit
Other
1.83k stars 402 forks source link

cuda out of memory #68

Closed woiza closed 4 years ago

woiza commented 4 years ago

Hi, I have 2 GPUs with 8GB of memory each. Training your "TextVDCNN" model fails (out of memory) and only one GPU is used. Is it possible to use your toolkit with 2 GPUs (data parallelism)?

coderbyr commented 4 years ago

sorry, this toolkit does't support multi Gpus training because Pytorch distributed learning( dataparallel or distributed dataparallel) need to split input tensors in batch dimension, while in this toolkit the input is a dict, so we need to modify or wrapper the standard dataparallel to solve it. It will be added in future.