search

timeseriesAI / tsai

Time series Timeseries Deep Learning Machine Learning Python Pytorch fastai | State-of-the-art Deep Learning library for Time Series and Sequences in Pytorch / fastai
https://timeseriesai.github.io/tsai/
Apache License 2.0
5.16k stars 644 forks source link

CUDA cannot allocate memory error #8

Closed AnoopRKulkarni closed 4 years ago

AnoopRKulkarni commented 4 years ago

I am trying to use timeserieAI with my own dataset (1D sequences) and learn.fit_one_cycle() gives an error "Can not allocate memory".

Couple of questions?

1) Is the timeserieAI databunch (using TSList) memory heavy? My np array size is only about 150M, but if I load databunch with the suggested approach almost 6-7GB of RAM is taken away.

2) Is there anything I can do?

best regards, ~anoop

oguiza commented 4 years ago

Please, try to use the updated timeseriesAI library (based on fastai v2). It should be less memory intensive.