search

Zuricho / ParallelFold

Modified version of Alphafold to divide CPU part (MSA and template searching) and GPU part. This can accelerate Alphafold when predicting multiple structures
https://parafold.sjtu.edu.cn
136 stars 45 forks source link

GPU利用率问题 #6

Closed zhoujingyu13687306871 closed 2 years ago

zhoujingyu13687306871 commented 2 years ago

博士好！我昨天进行多次尝试后，现在可以运行了，但是我发现运行run_alphafold.sh脚本的时候，涉及GPU计算部分，在相当长的一段时间处于CPU运行状态，GPU利用率长时间为0，我尝试计算一条序列长为2000的蛋白质，用了4个V100的卡，计算了9天，这个速度和情况这个是否正常呢？另外前面在安装tensorflow阶段，是否有必要安装GPU版的tensorflow呢？

@Zuricho

Zuricho commented 2 years ago

This is not right. Protein with 2000 should have less than 8 hours MSA and around 2 hour neural network inference (using GPU). I highly suspect your program didn't find your GPU and you should check whether it by using print(jax.devices())