MemoryError：std :: bad_alloc：Rapids.ai Dask-cuDF

问题描述

我想加载5.9 GB CSV，并且我不使用pandas库。我有4个GPU。我使用rapids.ai来更快地加载此大型数据集，但是每次尝试时，都会显示此错误，尽管我的其他GPU内存中有空间。一开始，GPU的内存使用情况是：

GPU 0
total    : 11554717696
free     : 11126046720
used     : 428670976
GPU 1
total    : 11554717696
free     : 11542331392
used     : 12386304
GPU 2
total    : 11554717696
free     : 11542331392
used     : 12386304
GPU 3
total    : 11551440896
free     : 11113070592
used     : 438370304

，代码为：

import cudf
import pandas as pd
import time
import subprocess as sp
import os
import dask_cudf

name = 'T100'
path = '/media/mo/2438a3d1-29fe-4c6f-aafb-f906acd5140d/AIMD/c1/trajs/'+name+'.CSV'
start = time.time()


data = dask_cudf.from_cudf(cudf.read_csv(path),npartitions=4).compute()
done = time.time()
elapsed = done - start
print(elapsed)

提示：

---------------------------------------------------------------------------
MemoryError                               Traceback (most recent call last)
<ipython-input-3-1fff5fb4e9b4> in <module>
      2 
      3 
----> 4 data = dask_cudf.from_cudf(cudf.read_csv(path),5                          npartitions=4).compute()
      6 done = time.time()

~/anaconda3/envs/machineLearning/lib/python3.7/contextlib.py in inner(*args,**kwds)
     72         def inner(*args,**kwds):
     73             with self._recreate_cm():
---> 74                 return func(*args,**kwds)
     75         return inner
     76 

~/anaconda3/envs/machineLearning/lib/python3.7/site-packages/cudf/io/csv.py in read_csv(filepath_or_buffer,lineterminator,quotechar,quoting,doublequote,header,mangle_dupe_cols,usecols,sep,delimiter,delim_whitespace,skipinitialspace,names,dtype,skipfooter,skiprows,dayfirst,compression,thousands,decimal,true_values,false_values,nrows,byte_range,skip_blank_lines,parse_dates,comment,na_values,keep_default_na,na_filter,prefix,index_col,**kwargs)
     82         na_filter=na_filter,83         prefix=prefix,---> 84         index_col=index_col,85     )
     86 

cudf/_lib/csv.pyx in cudf._lib.csv.read_csv()

MemoryError: std::bad_alloc: CUDA error at: /conda/conda-bld/librmm_1591196551527/work/include/rmm/mr/device/cuda_memory_resource.hpp66: cudaErrorMemoryAllocation out of memory

解决方法

问题的答案：CUDF error processing a large number of parquet files

说明如何使用dask_cudf读取大文件：https://stackoverflow.com/a/58123478/13887495

按照答案中的说明进行操作可以帮助您解决MemoryError: std::bad_alloc: CUDA error at: /conda/conda-bld/librmm_1591196551527/work/include/rmm/mr/device/cuda_memory_resource.hpp66: cudaErrorMemoryAllocation out of memory

,

代码应为

data = dask_cudf.read_csv(path,npartitions=4)