Skip to main content

h5py and multiprocessing

The HDF5 format has been working awesome for me, but I ran into danger when I started to mix it with multiprocessing. It was the worst kind of danger: the intermittent error.

Here are the dangers/issues in order of escalation

(TL;DR is use a generator to feed data from your file into the child processes as they spawn. It's the easiest way. Read on for harder ways.)

  1. An h5py file handle can't be pickled and therefore can't be passed as an argument using pool.map()
  2. If you set the handle as a global and access it from the child processes you run the risk of racing which leads to corrupted reads. My personal runin was that my code sometimes ran fine but sometimes would complain that there are NaNs or Infinity in the data. This wasted some time tracking down. Other people have had this kind of problem [1].
  3. Same problem if you pass the filename and have the different processes open individual instances of the file separately.
  4. The hard way to solve this problem is to switch your workflow over to MPI and use mpi4py or somesuch. The details are here. My problem is embarrassingly parallel and the computations result in heavy data reduction, so I ended up simply doing ...
  5. The easy way to solve the problem is to have a generator in the parent thread that reads in the data from the file and passeses it to the child process in a JIT manner.
Pseudocode:

fin = h5py.File(fname,'r')
X = (read_chunk(fin, [n,n+Nstep]) for n in range(0,N,Nstep))
pool = mp.Pool()
result = pool.map(child_process, X)
fin.close()
Labels:

Comments

Post a Comment

Popular posts from this blog

A note on Python's __exit__() and errors

Python's context managers are a very neat way of handling code that needs a teardown once you are done. Python objects have do have a destructor method ( __del__ ) called right before the last instance of the object is about to be destroyed. You can do a teardown there. However there is a lot of fine print to the __del__ method. A cleaner way of doing tear-downs is through Python's context manager , manifested as the with keyword. class CrushMe: def __init__(self): self.f = open('test.txt', 'w') def foo(self, a, b): self.f.write(str(a - b)) def __enter__(self): return self def __exit__(self, exc_type, exc_val, exc_tb): self.f.close() return True with CrushMe() as c: c.foo(2, 3) One thing that is important, and that got me just now, is error handling. I made the mistake of ignoring all those 'junk' arguments ( exc_type, exc_val, exc_tb ). I just skimmed the docs and what popped out is that you need to return True or...
Post a Comment
Read more

Store numpy arrays in sqlite

Use numpy.getbuffer (or sqlite3.Binary ) in combination with numpy.frombuffer to lug numpy data in and out of the sqlite3 database: import sqlite3, numpy r1d = numpy.random.randn(10) con = sqlite3.connect(':memory:') con.execute("CREATE TABLE eye(id INTEGER PRIMARY KEY, desc TEXT, data BLOB)") con.execute("INSERT INTO eye(desc,data) VALUES(?,?)", ("1d", sqlite3.Binary(r1d))) con.execute("INSERT INTO eye(desc,data) VALUES(?,?)", ("1d", numpy.getbuffer(r1d))) res = con.execute("SELECT * FROM eye").fetchall() con.close() #res -> #[(1, u'1d', <read-write buffer ptr 0x10371b220, size 80 at 0x10371b1e0>), # (2, u'1d', <read-write buffer ptr 0x10371b190, size 80 at 0x10371b150>)] print r1d - numpy.frombuffer(res[0][2]) #->[ 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.] print r1d - numpy.frombuffer(res[1][2]) #->[ 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.] Note that for work where data ty...
Post a Comment
Read more