Python Using multiprocessing.pool with mpi4py generates Cannot allocate memory error

Question

I am working on a cluster of computers, and recently started using parallel programming with python.

My current understanding is: mpi4py helps to manage work between different nodes, and multiprocessing manages work in a node's cores.

I have divided a big for loop into parts based on the number of procs running,

comm.Get_size() = nprocs

and then tried to generate a pool of processes by the cpu_count of the node, and then give them work. The code is as follows:

       if rank == proc:
            global output_dictionary
            output_dictionary = {}
            p=Pool(processes=cpu_count())                                                                                                     
            print "rank", rank, "started backpropagating. Async mapping..."
            results = []
            for key in serialize_and_divide(n_muscles,n_mn,rank,nprocs):
                r = p.apply_async(calc_neuronij_grad,key,callback=append_grads_list)
                results.append(r)
            for r in results:                                                                                                                     
                r.wait()                                                                                                                          
            p.close()
            p.join()

My problem is that the moment the code reaches p = Pool(processes=cpu_count()), it generates the error "Cannot allocate memory". The Traceback is the same as Python cannot allocate memory using multiprocessing.pool , but the solution doesn't help.

Any hints/help/explanations are appreciated