search

YerevaNN / mimic3-benchmarks

Python suite to construct benchmark machine learning datasets from the MIMIC-III 💊 clinical database.
https://arxiv.org/abs/1703.07771
MIT License
805 stars 329 forks source link

The program runs too slowly #107

Open wushangbin opened 3 years ago

wushangbin commented 3 years ago

Hello, when I run the "python -um mimic3models.in_hospital_mortality.main --network mimic3models/keras_models/lstm.py --dim 16 --timestep 1.0 --depth 2 --dropout 0.3 --mode train --batch_size 8 --output_dir mimic3models/in_hospital_mortality", The program runs too slowly, an epoch may take more than an hour, and when the program is running, the utilization of the gpu is very low. May I ask if this is normal？ Thank you very much!!!

wayne0320 commented 3 years ago

a

hello,may I know the version of python and extension package you were using？

alanflanders commented 3 years ago

I implemented multi-processing on the data generator which has sped up the process a great deal. Have also implemented other performance improvements. I see there are pending pull requests. If this project is still active please reach out and I can submit a pull request. Note that some changes were also required to make the benchmark code compatible with Tensorflow 2.0. Also note that the performance improvements has broken the deep supervision model type.

tingwei161803 commented 1 year ago

I had the same problem. The code seems to train on CPU, not GPU. the requirement package contain tensorflow==1.15.0, which is a CPU version How do I switch to a newer version of TensorFlow or use a GPU?