def get_image_size( imgdata, myid ):
from mpi import mpi_gather, mpi_bcast, MPI_COMM_WORLD, MPI_INT
nimg = len(imgdata)
nimgs = mpi_gather( nimg, 1, MPI_INT, 1, MPI_INT, 0, MPI_COMM_WORLD )
if myid==0:
src = -1
for i in xrange( len(nimgs) ):
if int(nimgs[i]) > 0 :
src = i
break
if src==-1:
return 0
else:
src = -1
size_src = mpi_bcast( src, 1, MPI_INT, 0, MPI_COMM_WORLD )
if myid==int(size_src[0]):
assert nimg > 0
size = imgdata[0].get_xsize()
else:
size = -1
nx = mpi_bcast( size, 1, MPI_INT, size_src[0], MPI_COMM_WORLD )
return int(nx[0])
def identify_outliers(myid, main_node, rviper_iter, no_of_viper_runs_analyzed_together,
no_of_viper_runs_analyzed_together_from_user_options, masterdir, bdb_stack_location, outlier_percentile,
criterion_name, outlier_index_threshold_method, angle_threshold):
no_of_viper_runs_analyzed_together_must_be_incremented = 0
do_calculation = 1
if (myid == main_node):
mainoutputdir = masterdir + DIR_DELIM + NAME_OF_MAIN_DIR + ("%03d" + DIR_DELIM) % (rviper_iter)
if(os.path.exists(mainoutputdir + DIR_DELIM + "list_of_viper_runs_included_in_outlier_elimination.json")):
# list_of_independent_viper_run_indices_used_for_outlier_elimination = map(int, read_text_file(mainoutputdir + DIR_DELIM + "list_of_viper_runs_included_in_outlier_elimination.txt"))
import json; f = open(mainoutputdir + "list_of_viper_runs_included_in_outlier_elimination.json", 'r')
list_of_independent_viper_run_indices_used_for_outlier_elimination = json.load(f); f.close()
do_calculation = 0
do_calculation = mpi_bcast(do_calculation, 1, MPI_INT, 0, MPI_COMM_WORLD)[0]
else:
do_calculation = mpi_bcast(do_calculation, 1, MPI_INT, 0, MPI_COMM_WORLD)[0]
if do_calculation:
list_of_independent_viper_run_indices_used_for_outlier_elimination = calculate_list_of_independent_viper_run_indices_used_for_outlier_elimination(no_of_viper_runs_analyzed_together,
no_of_viper_runs_analyzed_together_from_user_options, masterdir, rviper_iter, criterion_name)
# only master has the actual list: list_of_independent_viper_run_indices_used_for_outlier_elimination
# only master has the actual list: list_of_independent_viper_run_indices_used_for_outlier_elimination
# only master has the actual list: list_of_independent_viper_run_indices_used_for_outlier_elimination
error_status = 0
if (myid == main_node):
# if len(list_of_independent_viper_run_indices_used_for_outlier_elimination) == 0:
if list_of_independent_viper_run_indices_used_for_outlier_elimination[0] == EMPTY_VIPER_RUN_INDICES_LIST:
if no_of_viper_runs_analyzed_together > MAXIMUM_NO_OF_VIPER_RUNS_ANALYZED_TOGETHER:
error_status = 1
print "RVIPER reached maximum number of VIPER runs analyzed together without finding a core set of stable projections for the current RVIPER iteration (%d)! Finishing."%rviper_iter
cmd = "{} {}".format("mkdir ", masterdir + "MAXIMUM_NO_OF_VIPER_RUNS_ANALYZED_TOGETHER__Reached"); cmdexecute(cmd)
else:
# No set of solutions has been found to make a selection for outlier elimination.
# A new independent viper run will be performed
no_of_viper_runs_analyzed_together_must_be_incremented = 1
cmd = "{} {}".format("rm ", mainoutputdir + "list_of_viper_runs_included_in_outlier_elimination.json"); cmdexecute(cmd)
else:
# Outliers are eliminated based on the viper runs contained in "list_of_independent_viper_run_indices_used_for_outlier_elimination"
if list_of_independent_viper_run_indices_used_for_outlier_elimination[0] == MUST_END_PROGRAM_THIS_ITERATION:
no_of_viper_runs_analyzed_together_must_be_incremented = MUST_END_PROGRAM_THIS_ITERATION
found_outliers(list_of_independent_viper_run_indices_used_for_outlier_elimination[1:], outlier_percentile,
rviper_iter, masterdir, bdb_stack_location, "use all images", angle_threshold)
else:
# still need to eliminate DUMMY_INDEX_USED_AS_BUFFER
found_outliers(list_of_independent_viper_run_indices_used_for_outlier_elimination[1:], outlier_percentile,
rviper_iter, masterdir, bdb_stack_location, outlier_index_threshold_method, angle_threshold)
if_error_then_all_processes_exit_program(error_status)
no_of_viper_runs_analyzed_together_must_be_incremented = mpi_bcast(no_of_viper_runs_analyzed_together_must_be_incremented, 1, MPI_INT, 0, MPI_COMM_WORLD)[0]
return no_of_viper_runs_analyzed_together_must_be_incremented
def mpi_bcast_recv(src):
"""Unlike the C routine, in this python module, mpi_bcast is split into a send and a receive method. Send must be
called on exactly one core, and receive called on all of the others. This routine also coordinates transfer of
variable length objects. src is the rank of the sender"""
from mpi import mpi_bcast, MPI_CHAR, MPI_COMM_WORLD
l = mpi_bcast(None, 4, MPI_CHAR, src, MPI_COMM_WORLD)
l=unpack("I",l)[0]
data = mpi_bcast(None, l, MPI_CHAR, src, MPI_COMM_WORLD)
return loads(str(data.data))
def mpi_bcast_send(data):
"""Unlike the C routine, in this python module, mpi_bcast is split into a send and a receive method. Send must be
called on exactly one core, and receive called on all of the others. This routine also coordinates transfer of
variable length objects."""
from mpi import mpi_comm_rank, mpi_bcast, MPI_CHAR, MPI_COMM_WORLD
data=dumps(data,-1)
l=pack("I",len(data))
rank = mpi_comm_rank(MPI_COMM_WORLD)
mpi_bcast(l, len(l), MPI_CHAR, rank, MPI_COMM_WORLD)
mpi_bcast(data, len(data), MPI_CHAR, rank, MPI_COMM_WORLD)
def if_error_then_all_processes_exit_program(error_status):
import sys, os
from utilities import print_msg
if "OMPI_COMM_WORLD_SIZE" not in os.environ:
def mpi_comm_rank(n): return 0
def mpi_bcast(*largs):
return [largs[0]]
def mpi_finalize():
return None
MPI_INT, MPI_COMM_WORLD = 0, 0
else:
from mpi import mpi_comm_rank, mpi_bcast, mpi_finalize, MPI_INT, MPI_COMM_WORLD
myid = mpi_comm_rank(MPI_COMM_WORLD)
if error_status != None and error_status != 0:
error_status_info = error_status
error_status = 1
else:
error_status = 0
error_status = mpi_bcast(error_status, 1, MPI_INT, 0, MPI_COMM_WORLD)
error_status = int(error_status[0])
if error_status > 0:
if myid == 0:
if type(error_status_info) == type((1,1)):
if len(error_status_info) == 2:
frameinfo = error_status_info[1]
print_msg("***********************************\n")
print_msg("** Error: %s\n"%error_status_info[0])
print_msg("***********************************\n")
print_msg("** Location: %s\n"%(frameinfo.filename + ":" + str(frameinfo.lineno)))
print_msg("***********************************\n")
sys.stdout.flush()
mpi_finalize()
sys.exit(1)
def main():
from logger import Logger, BaseLogger_Files
arglist = []
i = 0
while( i < len(sys.argv) ):
if sys.argv[i]=='-p4pg':
i = i+2
elif sys.argv[i]=='-p4wd':
i = i+2
else:
arglist.append( sys.argv[i] )
i = i+1
progname = os.path.basename(arglist[0])
usage = progname + " stack outdir <mask> --focus=3Dmask --radius=outer_radius --delta=angular_step" +\
"--an=angular_neighborhood --maxit=max_iter --CTF --sym=c1 --function=user_function --independent=indenpendent_runs --number_of_images_per_group=number_of_images_per_group --low_pass_frequency=.25 --seed=random_seed"
parser = OptionParser(usage,version=SPARXVERSION)
parser.add_option("--focus", type ="string", default ='', help="bineary 3D mask for focused clustering ")
parser.add_option("--ir", type = "int", default =1, help="inner radius for rotational correlation > 0 (set to 1)")
parser.add_option("--radius", type = "int", default =-1, help="particle radius in pixel for rotational correlation <nx-1 (set to the radius of the particle)")
parser.add_option("--maxit", type = "int", default =25, help="maximum number of iteration")
parser.add_option("--rs", type = "int", default =1, help="step between rings in rotational correlation >0 (set to 1)" )
parser.add_option("--xr", type ="string", default ='1', help="range for translation search in x direction, search is +/-xr ")
parser.add_option("--yr", type ="string", default ='-1', help="range for translation search in y direction, search is +/-yr (default = same as xr)")
parser.add_option("--ts", type ="string", default ='0.25', help="step size of the translation search in both directions direction, search is -xr, -xr+ts, 0, xr-ts, xr ")
parser.add_option("--delta", type ="string", default ='2', help="angular step of reference projections")
parser.add_option("--an", type ="string", default ='-1', help="angular neighborhood for local searches")
parser.add_option("--center", type ="int", default =0, help="0 - if you do not want the volume to be centered, 1 - center the volume using cog (default=0)")
parser.add_option("--nassign", type ="int", default =1, help="number of reassignment iterations performed for each angular step (set to 3) ")
parser.add_option("--nrefine", type ="int", default =0, help="number of alignment iterations performed for each angular step (set to 0)")
parser.add_option("--CTF", action ="store_true", default =False, help="do CTF correction during clustring")
parser.add_option("--stoprnct", type ="float", default =3.0, help="Minimum percentage of assignment change to stop the program")
parser.add_option("--sym", type ="string", default ='c1', help="symmetry of the structure ")
parser.add_option("--function", type ="string", default ='do_volume_mrk05', help="name of the reference preparation function")
parser.add_option("--independent", type ="int", default = 3, help="number of independent run")
parser.add_option("--number_of_images_per_group", type ="int", default =1000, help="number of groups")
parser.add_option("--low_pass_filter", type ="float", default =-1.0, help="absolute frequency of low-pass filter for 3d sorting on the original image size" )
parser.add_option("--nxinit", type ="int", default =64, help="initial image size for sorting" )
parser.add_option("--unaccounted", action ="store_true", default =False, help="reconstruct the unaccounted images")
parser.add_option("--seed", type ="int", default =-1, help="random seed for create initial random assignment for EQ Kmeans")
parser.add_option("--smallest_group", type ="int", default =500, help="minimum members for identified group")
parser.add_option("--sausage", action ="store_true", default =False, help="way of filter volume")
parser.add_option("--chunkdir", type ="string", default ='', help="chunkdir for computing margin of error")
parser.add_option("--PWadjustment", type ="string", default ='', help="1-D power spectrum of PDB file used for EM volume power spectrum correction")
parser.add_option("--protein_shape", type ="string", default ='g', help="protein shape. It defines protein preferred orientation angles. Currently it has g and f two types ")
parser.add_option("--upscale", type ="float", default =0.5, help=" scaling parameter to adjust the power spectrum of EM volumes")
parser.add_option("--wn", type ="int", default =0, help="optimal window size for data processing")
parser.add_option("--interpolation", type ="string", default ="4nn", help="3-d reconstruction interpolation method, two options trl and 4nn")
(options, args) = parser.parse_args(arglist[1:])
if len(args) < 1 or len(args) > 4:
print "usage: " + usage
print "Please run '" + progname + " -h' for detailed options"
else:
if len(args)>2:
mask_file = args[2]
else:
mask_file = None
orgstack =args[0]
masterdir =args[1]
global_def.BATCH = True
#---initialize MPI related variables
from mpi import mpi_init, mpi_comm_size, MPI_COMM_WORLD, mpi_comm_rank,mpi_barrier,mpi_bcast, mpi_bcast, MPI_INT,MPI_CHAR
sys.argv = mpi_init(len(sys.argv),sys.argv)
nproc = mpi_comm_size(MPI_COMM_WORLD)
myid = mpi_comm_rank(MPI_COMM_WORLD)
mpi_comm = MPI_COMM_WORLD
main_node= 0
# import some utilities
from utilities import get_im,bcast_number_to_all,cmdexecute,write_text_file,read_text_file,wrap_mpi_bcast, get_params_proj, write_text_row
from applications import recons3d_n_MPI, mref_ali3d_MPI, Kmref_ali3d_MPI
from statistics import k_means_match_clusters_asg_new,k_means_stab_bbenum
from applications import mref_ali3d_EQ_Kmeans, ali3d_mref_Kmeans_MPI
# Create the main log file
from logger import Logger,BaseLogger_Files
if myid ==main_node:
log_main=Logger(BaseLogger_Files())
log_main.prefix = masterdir+"/"
else:
log_main =None
#--- fill input parameters into dictionary named after Constants
Constants ={}
Constants["stack"] = args[0]
Constants["masterdir"] = masterdir
Constants["mask3D"] = mask_file
Constants["focus3Dmask"] = options.focus
Constants["indep_runs"] = options.independent
Constants["stoprnct"] = options.stoprnct
Constants["number_of_images_per_group"] = options.number_of_images_per_group
Constants["CTF"] = options.CTF
Constants["maxit"] = options.maxit
Constants["ir"] = options.ir
Constants["radius"] = options.radius
Constants["nassign"] = options.nassign
Constants["rs"] = options.rs
Constants["xr"] = options.xr
Constants["yr"] = options.yr
Constants["ts"] = options.ts
Constants["delta"] = options.delta
Constants["an"] = options.an
Constants["sym"] = options.sym
Constants["center"] = options.center
Constants["nrefine"] = options.nrefine
#Constants["fourvar"] = options.fourvar
Constants["user_func"] = options.function
Constants["low_pass_filter"] = options.low_pass_filter # enforced low_pass_filter
#Constants["debug"] = options.debug
Constants["main_log_prefix"] = args[1]
#Constants["importali3d"] = options.importali3d
Constants["myid"] = myid
Constants["main_node"] = main_node
Constants["nproc"] = nproc
Constants["log_main"] = log_main
Constants["nxinit"] = options.nxinit
Constants["unaccounted"] = options.unaccounted
Constants["seed"] = options.seed
Constants["smallest_group"] = options.smallest_group
Constants["sausage"] = options.sausage
Constants["chunkdir"] = options.chunkdir
Constants["PWadjustment"] = options.PWadjustment
Constants["upscale"] = options.upscale
Constants["wn"] = options.wn
Constants["3d-interpolation"] = options.interpolation
Constants["protein_shape"] = options.protein_shape
# -----------------------------------------------------
#
# Create and initialize Tracker dictionary with input options
Tracker = {}
Tracker["constants"] = Constants
Tracker["maxit"] = Tracker["constants"]["maxit"]
Tracker["radius"] = Tracker["constants"]["radius"]
#Tracker["xr"] = ""
#Tracker["yr"] = "-1" # Do not change!
#Tracker["ts"] = 1
#Tracker["an"] = "-1"
#Tracker["delta"] = "2.0"
#Tracker["zoom"] = True
#Tracker["nsoft"] = 0
#Tracker["local"] = False
#Tracker["PWadjustment"] = Tracker["constants"]["PWadjustment"]
Tracker["upscale"] = Tracker["constants"]["upscale"]
#Tracker["upscale"] = 0.5
Tracker["applyctf"] = False # Should the data be premultiplied by the CTF. Set to False for local continuous.
#Tracker["refvol"] = None
Tracker["nxinit"] = Tracker["constants"]["nxinit"]
#Tracker["nxstep"] = 32
Tracker["icurrentres"] = -1
#Tracker["ireachedres"] = -1
#Tracker["lowpass"] = 0.4
#Tracker["falloff"] = 0.2
#Tracker["inires"] = options.inires # Now in A, convert to absolute before using
Tracker["fuse_freq"] = 50 # Now in A, convert to absolute before using
#Tracker["delpreviousmax"] = False
#Tracker["anger"] = -1.0
#Tracker["shifter"] = -1.0
#Tracker["saturatecrit"] = 0.95
#Tracker["pixercutoff"] = 2.0
#Tracker["directory"] = ""
#Tracker["previousoutputdir"] = ""
#Tracker["eliminated-outliers"] = False
#Tracker["mainiteration"] = 0
#Tracker["movedback"] = False
#Tracker["state"] = Tracker["constants"]["states"][0]
#Tracker["global_resolution"] =0.0
Tracker["orgstack"] = orgstack
#--------------------------------------------------------------------
# import from utilities
from utilities import sample_down_1D_curve,get_initial_ID,remove_small_groups,print_upper_triangular_matrix,print_a_line_with_timestamp
from utilities import print_dict,get_resolution_mrk01,partition_to_groups,partition_independent_runs,get_outliers
from utilities import merge_groups, save_alist, margin_of_error, get_margin_of_error, do_two_way_comparison, select_two_runs, get_ali3d_params
from utilities import counting_projections, unload_dict, load_dict, get_stat_proj, create_random_list, get_number_of_groups, recons_mref
from utilities import apply_low_pass_filter, get_groups_from_partition, get_number_of_groups, get_complementary_elements_total, update_full_dict
from utilities import count_chunk_members, set_filter_parameters_from_adjusted_fsc, adjust_fsc_down, get_two_chunks_from_stack
####------------------------------------------------------------------
#
# Get the pixel size; if none, set to 1.0, and the original image size
from utilities import get_shrink_data_huang
if(myid == main_node):
line = strftime("%Y-%m-%d_%H:%M:%S", localtime()) + " =>"
print(line+"Initialization of 3-D sorting")
a = get_im(orgstack)
nnxo = a.get_xsize()
if( Tracker["nxinit"] > nnxo ):
ERROR("Image size less than minimum permitted $d"%Tracker["nxinit"],"sxsort3d.py",1)
nnxo = -1
else:
if Tracker["constants"]["CTF"]:
i = a.get_attr('ctf')
pixel_size = i.apix
fq = pixel_size/Tracker["fuse_freq"]
else:
pixel_size = 1.0
# No pixel size, fusing computed as 5 Fourier pixels
fq = 5.0/nnxo
del a
else:
nnxo = 0
fq = 0.0
pixel_size = 1.0
nnxo = bcast_number_to_all(nnxo, source_node = main_node)
if( nnxo < 0 ):
mpi_finalize()
exit()
pixel_size = bcast_number_to_all(pixel_size, source_node = main_node)
fq = bcast_number_to_all(fq, source_node = main_node)
if Tracker["constants"]["wn"]==0:
Tracker["constants"]["nnxo"] = nnxo
else:
Tracker["constants"]["nnxo"] = Tracker["constants"]["wn"]
nnxo = Tracker["constants"]["nnxo"]
Tracker["constants"]["pixel_size"] = pixel_size
Tracker["fuse_freq"] = fq
del fq, nnxo, pixel_size
if(Tracker["constants"]["radius"] < 1):
Tracker["constants"]["radius"] = Tracker["constants"]["nnxo"]//2-2
elif((2*Tracker["constants"]["radius"] +2) > Tracker["constants"]["nnxo"]):
ERROR("Particle radius set too large!","sxsort3d.py",1,myid)
####-----------------------------------------------------------------------------------------
# Master directory
if myid == main_node:
if masterdir =="":
timestring = strftime("_%d_%b_%Y_%H_%M_%S", localtime())
masterdir ="master_sort3d"+timestring
li =len(masterdir)
cmd="{} {}".format("mkdir", masterdir)
os.system(cmd)
else:
li=0
li = mpi_bcast(li,1,MPI_INT,main_node,MPI_COMM_WORLD)[0]
if li>0:
masterdir = mpi_bcast(masterdir,li,MPI_CHAR,main_node,MPI_COMM_WORLD)
import string
masterdir = string.join(masterdir,"")
if myid ==main_node:
print_dict(Tracker["constants"],"Permanent settings of 3-D sorting program")
######### create a vstack from input stack to the local stack in masterdir
# stack name set to default
Tracker["constants"]["stack"] = "bdb:"+masterdir+"/rdata"
Tracker["constants"]["ali3d"] = os.path.join(masterdir, "ali3d_init.txt")
Tracker["constants"]["ctf_params"] = os.path.join(masterdir, "ctf_params.txt")
Tracker["constants"]["partstack"] = Tracker["constants"]["ali3d"] # also serves for refinement
if myid == main_node:
total_stack = EMUtil.get_image_count(Tracker["orgstack"])
else:
total_stack = 0
total_stack = bcast_number_to_all(total_stack, source_node = main_node)
mpi_barrier(MPI_COMM_WORLD)
from time import sleep
while not os.path.exists(masterdir):
print "Node ",myid," waiting..."
sleep(5)
mpi_barrier(MPI_COMM_WORLD)
if myid == main_node:
log_main.add("Sphire sort3d ")
log_main.add("the sort3d master directory is "+masterdir)
#####
###----------------------------------------------------------------------------------
# Initial data analysis and handle two chunk files
from random import shuffle
# Compute the resolution
#### make chunkdir dictionary for computing margin of error
import user_functions
user_func = user_functions.factory[Tracker["constants"]["user_func"]]
chunk_dict = {}
chunk_list = []
if myid == main_node:
chunk_one = read_text_file(os.path.join(Tracker["constants"]["chunkdir"],"chunk0.txt"))
chunk_two = read_text_file(os.path.join(Tracker["constants"]["chunkdir"],"chunk1.txt"))
else:
chunk_one = 0
chunk_two = 0
chunk_one = wrap_mpi_bcast(chunk_one, main_node)
chunk_two = wrap_mpi_bcast(chunk_two, main_node)
mpi_barrier(MPI_COMM_WORLD)
######################## Read/write bdb: data on main node ############################
if myid==main_node:
if(orgstack[:4] == "bdb:"): cmd = "{} {} {}".format("e2bdb.py", orgstack,"--makevstack="+Tracker["constants"]["stack"])
else: cmd = "{} {} {}".format("sxcpy.py", orgstack, Tracker["constants"]["stack"])
cmdexecute(cmd)
cmd = "{} {} {}".format("sxheader.py --params=xform.projection", "--export="+Tracker["constants"]["ali3d"],orgstack)
cmdexecute(cmd)
cmd = "{} {} {}".format("sxheader.py --params=ctf", "--export="+Tracker["constants"]["ctf_params"],orgstack)
cmdexecute(cmd)
mpi_barrier(MPI_COMM_WORLD)
########-----------------------------------------------------------------------------
Tracker["total_stack"] = total_stack
Tracker["constants"]["total_stack"] = total_stack
Tracker["shrinkage"] = float(Tracker["nxinit"])/Tracker["constants"]["nnxo"]
Tracker["radius"] = Tracker["constants"]["radius"]*Tracker["shrinkage"]
if Tracker["constants"]["mask3D"]:
Tracker["mask3D"] = os.path.join(masterdir,"smask.hdf")
else:
Tracker["mask3D"] = None
if Tracker["constants"]["focus3Dmask"]:
Tracker["focus3D"] = os.path.join(masterdir,"sfocus.hdf")
else:
Tracker["focus3D"] = None
if myid == main_node:
if Tracker["constants"]["mask3D"]:
mask_3D = get_shrink_3dmask(Tracker["nxinit"],Tracker["constants"]["mask3D"])
mask_3D.write_image(Tracker["mask3D"])
if Tracker["constants"]["focus3Dmask"]:
mask_3D = get_shrink_3dmask(Tracker["nxinit"],Tracker["constants"]["focus3Dmask"])
st = Util.infomask(mask_3D, None, True)
if( st[0] == 0.0 ): ERROR("sxrsort3d","incorrect focused mask, after binarize all values zero",1)
mask_3D.write_image(Tracker["focus3D"])
del mask_3D
if Tracker["constants"]["PWadjustment"] !='':
PW_dict = {}
nxinit_pwsp = sample_down_1D_curve(Tracker["constants"]["nxinit"],Tracker["constants"]["nnxo"],Tracker["constants"]["PWadjustment"])
Tracker["nxinit_PW"] = os.path.join(masterdir,"spwp.txt")
if myid == main_node: write_text_file(nxinit_pwsp,Tracker["nxinit_PW"])
PW_dict[Tracker["constants"]["nnxo"]] = Tracker["constants"]["PWadjustment"]
PW_dict[Tracker["constants"]["nxinit"]] = Tracker["nxinit_PW"]
Tracker["PW_dict"] = PW_dict
mpi_barrier(MPI_COMM_WORLD)
#-----------------------From two chunks to FSC, and low pass filter-----------------------------------------###
for element in chunk_one: chunk_dict[element] = 0
for element in chunk_two: chunk_dict[element] = 1
chunk_list =[chunk_one, chunk_two]
Tracker["chunk_dict"] = chunk_dict
Tracker["P_chunk0"] = len(chunk_one)/float(total_stack)
Tracker["P_chunk1"] = len(chunk_two)/float(total_stack)
### create two volumes to estimate resolution
if myid == main_node:
for index in xrange(2): write_text_file(chunk_list[index],os.path.join(masterdir,"chunk%01d.txt"%index))
mpi_barrier(MPI_COMM_WORLD)
vols = []
for index in xrange(2):
data,old_shifts = get_shrink_data_huang(Tracker,Tracker["constants"]["nxinit"], os.path.join(masterdir,"chunk%01d.txt"%index), Tracker["constants"]["partstack"],myid,main_node,nproc,preshift=True)
vol = recons3d_4nn_ctf_MPI(myid=myid, prjlist=data,symmetry=Tracker["constants"]["sym"], finfo=None)
if myid == main_node:
vol.write_image(os.path.join(masterdir, "vol%d.hdf"%index))
vols.append(vol)
mpi_barrier(MPI_COMM_WORLD)
if myid ==main_node:
low_pass, falloff,currentres = get_resolution_mrk01(vols,Tracker["constants"]["radius"],Tracker["constants"]["nxinit"],masterdir,Tracker["mask3D"])
if low_pass >Tracker["constants"]["low_pass_filter"]: low_pass= Tracker["constants"]["low_pass_filter"]
else:
low_pass =0.0
falloff =0.0
currentres =0.0
bcast_number_to_all(currentres,source_node = main_node)
bcast_number_to_all(low_pass,source_node = main_node)
bcast_number_to_all(falloff,source_node = main_node)
Tracker["currentres"] = currentres
Tracker["falloff"] = falloff
if Tracker["constants"]["low_pass_filter"] ==-1.0:
Tracker["low_pass_filter"] = min(.45,low_pass/Tracker["shrinkage"]) # no better than .45
else:
Tracker["low_pass_filter"] = min(.45,Tracker["constants"]["low_pass_filter"]/Tracker["shrinkage"])
Tracker["lowpass"] = Tracker["low_pass_filter"]
Tracker["falloff"] =.1
Tracker["global_fsc"] = os.path.join(masterdir, "fsc.txt")
############################################################################################
if myid == main_node:
log_main.add("The command-line inputs are as following:")
log_main.add("**********************************************************")
for a in sys.argv:
if myid == main_node:log_main.add(a)
if myid == main_node:
log_main.add("number of cpus used in this run is %d"%Tracker["constants"]["nproc"])
log_main.add("**********************************************************")
from filter import filt_tanl
### START 3-D sorting
if myid ==main_node:
log_main.add("----------3-D sorting program------- ")
log_main.add("current resolution %6.3f for images of original size in terms of absolute frequency"%Tracker["currentres"])
log_main.add("equivalent to %f Angstrom resolution"%(Tracker["constants"]["pixel_size"]/Tracker["currentres"]/Tracker["shrinkage"]))
log_main.add("the user provided enforced low_pass_filter is %f"%Tracker["constants"]["low_pass_filter"])
#log_main.add("equivalent to %f Angstrom resolution"%(Tracker["constants"]["pixel_size"]/Tracker["constants"]["low_pass_filter"]))
for index in xrange(2):
filt_tanl(get_im(os.path.join(masterdir,"vol%01d.hdf"%index)), Tracker["low_pass_filter"],Tracker["falloff"]).write_image(os.path.join(masterdir, "volf%01d.hdf"%index))
mpi_barrier(MPI_COMM_WORLD)
from utilities import get_input_from_string
delta = get_input_from_string(Tracker["constants"]["delta"])
delta = delta[0]
from utilities import even_angles
n_angles = even_angles(delta, 0, 180)
this_ali3d = Tracker["constants"]["ali3d"]
sampled = get_stat_proj(Tracker,delta,this_ali3d)
if myid ==main_node:
nc = 0
for a in sampled:
if len(sampled[a])>0:
nc += 1
log_main.add("total sampled direction %10d at angle step %6.3f"%(len(n_angles), delta))
log_main.add("captured sampled directions %10d percentage covered by data %6.3f"%(nc,float(nc)/len(n_angles)*100))
number_of_images_per_group = Tracker["constants"]["number_of_images_per_group"]
if myid ==main_node: log_main.add("user provided number_of_images_per_group %d"%number_of_images_per_group)
Tracker["number_of_images_per_group"] = number_of_images_per_group
number_of_groups = get_number_of_groups(total_stack,number_of_images_per_group)
Tracker["number_of_groups"] = number_of_groups
generation =0
partition_dict ={}
full_dict ={}
workdir =os.path.join(masterdir,"generation%03d"%generation)
Tracker["this_dir"] = workdir
if myid ==main_node:
log_main.add("---- generation %5d"%generation)
log_main.add("number of images per group is set as %d"%number_of_images_per_group)
log_main.add("the initial number of groups is %10d "%number_of_groups)
cmd="{} {}".format("mkdir",workdir)
os.system(cmd)
mpi_barrier(MPI_COMM_WORLD)
list_to_be_processed = range(Tracker["constants"]["total_stack"])
Tracker["this_data_list"] = list_to_be_processed
create_random_list(Tracker)
#################################
full_dict ={}
for iptl in xrange(Tracker["constants"]["total_stack"]):
full_dict[iptl] = iptl
Tracker["full_ID_dict"] = full_dict
#################################
for indep_run in xrange(Tracker["constants"]["indep_runs"]):
Tracker["this_particle_list"] = Tracker["this_indep_list"][indep_run]
ref_vol = recons_mref(Tracker)
if myid == main_node: log_main.add("independent run %10d"%indep_run)
mpi_barrier(MPI_COMM_WORLD)
Tracker["this_data_list"] = list_to_be_processed
Tracker["total_stack"] = len(Tracker["this_data_list"])
Tracker["this_particle_text_file"] = os.path.join(workdir,"independent_list_%03d.txt"%indep_run) # for get_shrink_data
if myid == main_node: write_text_file(Tracker["this_data_list"], Tracker["this_particle_text_file"])
mpi_barrier(MPI_COMM_WORLD)
outdir = os.path.join(workdir, "EQ_Kmeans%03d"%indep_run)
ref_vol = apply_low_pass_filter(ref_vol,Tracker)
mref_ali3d_EQ_Kmeans(ref_vol, outdir, Tracker["this_particle_text_file"], Tracker)
partition_dict[indep_run]=Tracker["this_partition"]
Tracker["partition_dict"] = partition_dict
Tracker["total_stack"] = len(Tracker["this_data_list"])
Tracker["this_total_stack"] = Tracker["total_stack"]
###############################
do_two_way_comparison(Tracker)
###############################
ref_vol_list = []
from time import sleep
number_of_ref_class = []
for igrp in xrange(len(Tracker["two_way_stable_member"])):
Tracker["this_data_list"] = Tracker["two_way_stable_member"][igrp]
Tracker["this_data_list_file"] = os.path.join(workdir,"stable_class%d.txt"%igrp)
if myid == main_node:
write_text_file(Tracker["this_data_list"], Tracker["this_data_list_file"])
data,old_shifts = get_shrink_data_huang(Tracker,Tracker["nxinit"], Tracker["this_data_list_file"], Tracker["constants"]["partstack"], myid, main_node, nproc, preshift = True)
volref = recons3d_4nn_ctf_MPI(myid=myid, prjlist = data, symmetry=Tracker["constants"]["sym"], finfo = None)
ref_vol_list.append(volref)
number_of_ref_class.append(len(Tracker["this_data_list"]))
if myid == main_node:
log_main.add("group %d members %d "%(igrp,len(Tracker["this_data_list"])))
Tracker["number_of_ref_class"] = number_of_ref_class
nx_of_image = ref_vol_list[0].get_xsize()
if Tracker["constants"]["PWadjustment"]:
Tracker["PWadjustment"] = Tracker["PW_dict"][nx_of_image]
else:
Tracker["PWadjustment"] = Tracker["constants"]["PWadjustment"] # no PW adjustment
if myid == main_node:
for iref in xrange(len(ref_vol_list)):
refdata = [None]*4
refdata[0] = ref_vol_list[iref]
refdata[1] = Tracker
refdata[2] = Tracker["constants"]["myid"]
refdata[3] = Tracker["constants"]["nproc"]
volref = user_func(refdata)
volref.write_image(os.path.join(workdir,"volf_stable.hdf"),iref)
mpi_barrier(MPI_COMM_WORLD)
Tracker["this_data_list"] = Tracker["this_accounted_list"]
outdir = os.path.join(workdir,"Kmref")
empty_group, res_groups, final_list = ali3d_mref_Kmeans_MPI(ref_vol_list,outdir,Tracker["this_accounted_text"],Tracker)
Tracker["this_unaccounted_list"] = get_complementary_elements(list_to_be_processed,final_list)
if myid == main_node:
log_main.add("the number of particles not processed is %d"%len(Tracker["this_unaccounted_list"]))
write_text_file(Tracker["this_unaccounted_list"],Tracker["this_unaccounted_text"])
update_full_dict(Tracker["this_unaccounted_list"], Tracker)
#######################################
number_of_groups = len(res_groups)
vol_list = []
number_of_ref_class = []
for igrp in xrange(number_of_groups):
data,old_shifts = get_shrink_data_huang(Tracker, Tracker["constants"]["nnxo"], os.path.join(outdir,"Class%d.txt"%igrp), Tracker["constants"]["partstack"],myid,main_node,nproc,preshift = True)
volref = recons3d_4nn_ctf_MPI(myid=myid, prjlist = data, symmetry=Tracker["constants"]["sym"], finfo=None)
vol_list.append(volref)
if( myid == main_node ): npergroup = len(read_text_file(os.path.join(outdir,"Class%d.txt"%igrp)))
else: npergroup = 0
npergroup = bcast_number_to_all(npergroup, main_node )
number_of_ref_class.append(npergroup)
Tracker["number_of_ref_class"] = number_of_ref_class
mpi_barrier(MPI_COMM_WORLD)
nx_of_image = vol_list[0].get_xsize()
if Tracker["constants"]["PWadjustment"]:
Tracker["PWadjustment"]=Tracker["PW_dict"][nx_of_image]
else:
Tracker["PWadjustment"]=Tracker["constants"]["PWadjustment"]
if myid == main_node:
for ivol in xrange(len(vol_list)):
refdata =[None]*4
refdata[0] = vol_list[ivol]
refdata[1] = Tracker
refdata[2] = Tracker["constants"]["myid"]
refdata[3] = Tracker["constants"]["nproc"]
volref = user_func(refdata)
volref.write_image(os.path.join(workdir,"volf_of_Classes.hdf"),ivol)
log_main.add("number of unaccounted particles %10d"%len(Tracker["this_unaccounted_list"]))
log_main.add("number of accounted particles %10d"%len(Tracker["this_accounted_list"]))
Tracker["this_data_list"] = Tracker["this_unaccounted_list"] # reset parameters for the next round calculation
Tracker["total_stack"] = len(Tracker["this_unaccounted_list"])
Tracker["this_total_stack"] = Tracker["total_stack"]
number_of_groups = get_number_of_groups(len(Tracker["this_unaccounted_list"]),number_of_images_per_group)
Tracker["number_of_groups"] = number_of_groups
while number_of_groups >= 2 :
generation +=1
partition_dict ={}
workdir =os.path.join(masterdir,"generation%03d"%generation)
Tracker["this_dir"] = workdir
if myid ==main_node:
log_main.add("*********************************************")
log_main.add("----- generation %5d "%generation)
log_main.add("number of images per group is set as %10d "%number_of_images_per_group)
log_main.add("the number of groups is %10d "%number_of_groups)
log_main.add(" number of particles for clustering is %10d"%Tracker["total_stack"])
cmd ="{} {}".format("mkdir",workdir)
os.system(cmd)
mpi_barrier(MPI_COMM_WORLD)
create_random_list(Tracker)
for indep_run in xrange(Tracker["constants"]["indep_runs"]):
Tracker["this_particle_list"] = Tracker["this_indep_list"][indep_run]
ref_vol = recons_mref(Tracker)
if myid == main_node:
log_main.add("independent run %10d"%indep_run)
outdir = os.path.join(workdir, "EQ_Kmeans%03d"%indep_run)
Tracker["this_data_list"] = Tracker["this_unaccounted_list"]
#ref_vol=apply_low_pass_filter(ref_vol,Tracker)
mref_ali3d_EQ_Kmeans(ref_vol,outdir,Tracker["this_unaccounted_text"],Tracker)
partition_dict[indep_run] = Tracker["this_partition"]
Tracker["this_data_list"] = Tracker["this_unaccounted_list"]
Tracker["total_stack"] = len(Tracker["this_unaccounted_list"])
Tracker["partition_dict"] = partition_dict
Tracker["this_total_stack"] = Tracker["total_stack"]
total_list_of_this_run = Tracker["this_unaccounted_list"]
###############################
do_two_way_comparison(Tracker)
###############################
ref_vol_list = []
number_of_ref_class = []
for igrp in xrange(len(Tracker["two_way_stable_member"])):
Tracker["this_data_list"] = Tracker["two_way_stable_member"][igrp]
Tracker["this_data_list_file"] = os.path.join(workdir,"stable_class%d.txt"%igrp)
if myid == main_node: write_text_file(Tracker["this_data_list"], Tracker["this_data_list_file"])
mpi_barrier(MPI_COMM_WORLD)
data,old_shifts = get_shrink_data_huang(Tracker,Tracker["constants"]["nxinit"],Tracker["this_data_list_file"],Tracker["constants"]["partstack"],myid,main_node,nproc,preshift = True)
volref = recons3d_4nn_ctf_MPI(myid=myid, prjlist = data, symmetry=Tracker["constants"]["sym"],finfo= None)
#volref = filt_tanl(volref, Tracker["constants"]["low_pass_filter"],.1)
if myid == main_node:volref.write_image(os.path.join(workdir,"vol_stable.hdf"),iref)
#volref = resample(volref,Tracker["shrinkage"])
ref_vol_list.append(volref)
number_of_ref_class.append(len(Tracker["this_data_list"]))
mpi_barrier(MPI_COMM_WORLD)
Tracker["number_of_ref_class"] = number_of_ref_class
Tracker["this_data_list"] = Tracker["this_accounted_list"]
outdir = os.path.join(workdir,"Kmref")
empty_group, res_groups, final_list = ali3d_mref_Kmeans_MPI(ref_vol_list,outdir,Tracker["this_accounted_text"],Tracker)
# calculate the 3-D structure of original image size for each group
number_of_groups = len(res_groups)
Tracker["this_unaccounted_list"] = get_complementary_elements(total_list_of_this_run,final_list)
if myid == main_node:
log_main.add("the number of particles not processed is %d"%len(Tracker["this_unaccounted_list"]))
write_text_file(Tracker["this_unaccounted_list"],Tracker["this_unaccounted_text"])
mpi_barrier(MPI_COMM_WORLD)
update_full_dict(Tracker["this_unaccounted_list"],Tracker)
vol_list = []
for igrp in xrange(number_of_groups):
data,old_shifts = get_shrink_data_huang(Tracker,Tracker["constants"]["nnxo"], os.path.join(outdir,"Class%d.txt"%igrp), Tracker["constants"]["partstack"], myid, main_node, nproc,preshift = True)
volref = recons3d_4nn_ctf_MPI(myid=myid, prjlist = data, symmetry=Tracker["constants"]["sym"],finfo= None)
vol_list.append(volref)
mpi_barrier(MPI_COMM_WORLD)
nx_of_image=ref_vol_list[0].get_xsize()
if Tracker["constants"]["PWadjustment"]:
Tracker["PWadjustment"] = Tracker["PW_dict"][nx_of_image]
else:
Tracker["PWadjustment"] = Tracker["constants"]["PWadjustment"]
if myid == main_node:
for ivol in xrange(len(vol_list)):
refdata = [None]*4
refdata[0] = vol_list[ivol]
refdata[1] = Tracker
refdata[2] = Tracker["constants"]["myid"]
refdata[3] = Tracker["constants"]["nproc"]
volref = user_func(refdata)
volref.write_image(os.path.join(workdir, "volf_of_Classes.hdf"),ivol)
log_main.add("number of unaccounted particles %10d"%len(Tracker["this_unaccounted_list"]))
log_main.add("number of accounted particles %10d"%len(Tracker["this_accounted_list"]))
del vol_list
mpi_barrier(MPI_COMM_WORLD)
number_of_groups = get_number_of_groups(len(Tracker["this_unaccounted_list"]),number_of_images_per_group)
Tracker["number_of_groups"] = number_of_groups
Tracker["this_data_list"] = Tracker["this_unaccounted_list"]
Tracker["total_stack"] = len(Tracker["this_unaccounted_list"])
if Tracker["constants"]["unaccounted"]:
data,old_shifts = get_shrink_data_huang(Tracker,Tracker["constants"]["nnxo"],Tracker["this_unaccounted_text"],Tracker["constants"]["partstack"],myid,main_node,nproc,preshift = True)
volref = recons3d_4nn_ctf_MPI(myid=myid, prjlist = data, symmetry=Tracker["constants"]["sym"],finfo= None)
nx_of_image = volref.get_xsize()
if Tracker["constants"]["PWadjustment"]:
Tracker["PWadjustment"]=Tracker["PW_dict"][nx_of_image]
else:
Tracker["PWadjustment"]=Tracker["constants"]["PWadjustment"]
if( myid == main_node ):
refdata = [None]*4
refdata[0] = volref
refdata[1] = Tracker
refdata[2] = Tracker["constants"]["myid"]
refdata[3] = Tracker["constants"]["nproc"]
volref = user_func(refdata)
#volref = filt_tanl(volref, Tracker["constants"]["low_pass_filter"],.1)
volref.write_image(os.path.join(workdir,"volf_unaccounted.hdf"))
# Finish program
if myid ==main_node: log_main.add("sxsort3d finishes")
mpi_barrier(MPI_COMM_WORLD)
from mpi import mpi_finalize
mpi_finalize()
exit()
def main():
from logger import Logger, BaseLogger_Files
import user_functions
from optparse import OptionParser, SUPPRESS_HELP
from global_def import SPARXVERSION
from EMAN2 import EMData
main_node = 0
mpi_init(0, [])
mpi_comm = MPI_COMM_WORLD
myid = mpi_comm_rank(MPI_COMM_WORLD)
mpi_size = mpi_comm_size(MPI_COMM_WORLD) # Total number of processes, passed by --np option.
# mpi_barrier(mpi_comm)
# from mpi import mpi_finalize
# mpi_finalize()
# print "mpi finalize"
# from sys import exit
# exit()
progname = os.path.basename(sys.argv[0])
usage = progname + " stack [output_directory] --ir=inner_radius --radius=outer_radius --rs=ring_step --xr=x_range --yr=y_range --ts=translational_search_step --delta=angular_step --an=angular_neighborhood --center=center_type --maxit1=max_iter1 --maxit2=max_iter2 --L2threshold=0.1 --fl --aa --ref_a=S --sym=c1"
usage += """
stack 2D images in a stack file: (default required string)
output_directory: directory name into which the output files will be written. If it does not exist, the directory will be created. If it does exist, the program will continue executing from where it stopped (if it did not already reach the end). The "--use_latest_master_directory" option can be used to choose the most recent directory that starts with "master".
"""
parser = OptionParser(usage,version=SPARXVERSION)
parser.add_option("--radius", type="int", help="radius of the particle: has to be less than < int(nx/2)-1 (default required int)")
parser.add_option("--ir", type="int", default=1, help="inner radius for rotational search: > 0 (default 1)")
parser.add_option("--rs", type="int", default=1, help="step between rings in rotational search: >0 (default 1)")
parser.add_option("--xr", type="string", default='0', help="range for translation search in x direction: search is +/xr in pixels (default '0')")
parser.add_option("--yr", type="string", default='0', help="range for translation search in y direction: if omitted will be set to xr, search is +/yr in pixels (default '0')")
parser.add_option("--ts", type="string", default='1.0', help="step size of the translation search in x-y directions: search is -xr, -xr+ts, 0, xr-ts, xr, can be fractional (default '1.0')")
parser.add_option("--delta", type="string", default='2.0', help="angular step of reference projections: (default '2.0')")
#parser.add_option("--an", type="string", default= "-1", help="angular neighborhood for local searches (phi and theta)")
parser.add_option("--center", type="float", default=-1.0, help="centering of 3D template: average shift method; 0: no centering; 1: center of gravity (default -1.0)")
parser.add_option("--maxit1", type="int", default=400, help="maximum number of iterations performed for the GA part: (default 400)")
parser.add_option("--maxit2", type="int", default=50, help="maximum number of iterations performed for the finishing up part: (default 50)")
parser.add_option("--L2threshold", type="float", default=0.03, help="stopping criterion of GA: given as a maximum relative dispersion of volumes' L2 norms: (default 0.03)")
parser.add_option("--doga", type="float", default=0.1, help="do GA when fraction of orientation changes less than 1.0 degrees is at least doga: (default 0.1)")
parser.add_option("--n_shc_runs", type="int", default=4, help="number of quasi-independent shc runs (same as '--nruns' parameter from sxviper.py): (default 4)")
parser.add_option("--n_rv_runs", type="int", default=10, help="number of rviper iterations: (default 10)")
parser.add_option("--n_v_runs", type="int", default=3, help="number of viper runs for each r_viper cycle: (default 3)")
parser.add_option("--outlier_percentile", type="float", default=95.0, help="percentile above which outliers are removed every rviper iteration: (default 95.0)")
parser.add_option("--iteration_start", type="int", default=0, help="starting iteration for rviper: 0 means go to the most recent one (default 0)")
#parser.add_option("--CTF", action="store_true", default=False, help="NOT IMPLEMENTED Consider CTF correction during the alignment ")
#parser.add_option("--snr", type="float", default= 1.0, help="Signal-to-Noise Ratio of the data (default 1.0)")
parser.add_option("--ref_a", type="string", default='S', help="method for generating the quasi-uniformly distributed projection directions: (default S)")
parser.add_option("--sym", type="string", default='c1', help="point-group symmetry of the structure: (default c1)")
# parser.add_option("--function", type="string", default="ref_ali3d", help="name of the reference preparation function (ref_ali3d by default)")
##### XXXXXXXXXXXXXXXXXXXXXX option does not exist in docs XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
parser.add_option("--function", type="string", default="ref_ali3d", help=SUPPRESS_HELP)
parser.add_option("--npad", type="int", default=2, help="padding size for 3D reconstruction: (default 2)")
# parser.add_option("--npad", type="int", default= 2, help="padding size for 3D reconstruction (default 2)")
#options introduced for the do_volume function
parser.add_option("--fl", type="float", default=0.25, help="cut-off frequency applied to the template volume: using a hyperbolic tangent low-pass filter (default 0.25)")
parser.add_option("--aa", type="float", default=0.1, help="fall-off of hyperbolic tangent low-pass filter: (default 0.1)")
parser.add_option("--pwreference", type="string", default='', help="text file with a reference power spectrum: (default none)")
parser.add_option("--mask3D", type="string", default=None, help="3D mask file: (default sphere)")
parser.add_option("--moon_elimination", type="string", default='', help="elimination of disconnected pieces: two arguments: mass in KDa and pixel size in px/A separated by comma, no space (default none)")
# used for debugging, help is supressed with SUPPRESS_HELP
##### XXXXXXXXXXXXXXXXXXXXXX option does not exist in docs XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
parser.add_option("--my_random_seed", type="int", default=123, help = SUPPRESS_HELP)
##### XXXXXXXXXXXXXXXXXXXXXX option does not exist in docs XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
parser.add_option("--run_get_already_processed_viper_runs", action="store_true", dest="run_get_already_processed_viper_runs", default=False, help = SUPPRESS_HELP)
##### XXXXXXXXXXXXXXXXXXXXXX option does not exist in docs XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
parser.add_option("--use_latest_master_directory", action="store_true", dest="use_latest_master_directory", default=False, help = SUPPRESS_HELP)
parser.add_option("--criterion_name", type="string", default='80th percentile',help="criterion deciding if volumes have a core set of stable projections: '80th percentile', other options:'fastest increase in the last quartile' (default '80th percentile')")
parser.add_option("--outlier_index_threshold_method",type="string", default='discontinuity_in_derivative',help="method that decides which images to keep: discontinuity_in_derivative, other options:percentile, angle_measure (default discontinuity_in_derivative)")
parser.add_option("--angle_threshold", type="int", default=30, help="angle threshold for projection removal if using 'angle_measure': (default 30)")
required_option_list = ['radius']
(options, args) = parser.parse_args(sys.argv[1:])
options.CTF = False
options.snr = 1.0
options.an = -1
if options.moon_elimination == "":
options.moon_elimination = []
else:
options.moon_elimination = map(float, options.moon_elimination.split(","))
# Making sure all required options appeared.
for required_option in required_option_list:
if not options.__dict__[required_option]:
print "\n ==%s== mandatory option is missing.\n"%required_option
print "Please run '" + progname + " -h' for detailed options"
return 1
mpi_barrier(MPI_COMM_WORLD)
if(myid == main_node):
print "****************************************************************"
Util.version()
print "****************************************************************"
sys.stdout.flush()
mpi_barrier(MPI_COMM_WORLD)
# this is just for benefiting from a user friendly parameter name
options.ou = options.radius
my_random_seed = options.my_random_seed
criterion_name = options.criterion_name
outlier_index_threshold_method = options.outlier_index_threshold_method
use_latest_master_directory = options.use_latest_master_directory
iteration_start_default = options.iteration_start
number_of_rrr_viper_runs = options.n_rv_runs
no_of_viper_runs_analyzed_together_from_user_options = options.n_v_runs
no_of_shc_runs_analyzed_together = options.n_shc_runs
outlier_percentile = options.outlier_percentile
angle_threshold = options.angle_threshold
run_get_already_processed_viper_runs = options.run_get_already_processed_viper_runs
get_already_processed_viper_runs(run_get_already_processed_viper_runs)
import random
random.seed(my_random_seed)
if len(args) < 1 or len(args) > 3:
print "usage: " + usage
print "Please run '" + progname + " -h' for detailed options"
return 1
# if len(args) > 2:
# ref_vol = get_im(args[2])
# else:
ref_vol = None
# error_status = None
# if myid == 0:
# number_of_images = EMUtil.get_image_count(args[0])
# if mpi_size > number_of_images:
# error_status = ('Number of processes supplied by --np in mpirun needs to be less than or equal to %d (total number of images) ' % number_of_images, getframeinfo(currentframe()))
# if_error_then_all_processes_exit_program(error_status)
bdb_stack_location = ""
masterdir = ""
if len(args) == 2:
masterdir = args[1]
if masterdir[-1] != DIR_DELIM:
masterdir += DIR_DELIM
elif len(args) == 1:
if use_latest_master_directory:
all_dirs = [d for d in os.listdir(".") if os.path.isdir(d)]
import re; r = re.compile("^master.*$")
all_dirs = filter(r.match, all_dirs)
if len(all_dirs)>0:
# all_dirs = max(all_dirs, key=os.path.getctime)
masterdir = max(all_dirs, key=os.path.getmtime)
masterdir += DIR_DELIM
log = Logger(BaseLogger_Files())
error_status = 0
if mpi_size % no_of_shc_runs_analyzed_together != 0:
ERROR('Number of processes needs to be a multiple of the number of quasi-independent runs (shc) within each viper run. '
'Total quasi-independent runs by default are 3, you can change it by specifying '
'--n_shc_runs option (in sxviper this option is called --nruns). Also, to improve communication time it is recommended that '
'the number of processes divided by the number of quasi-independent runs is a power '
'of 2 (e.g. 2, 4, 8 or 16 depending on how many physical cores each node has).', 'sxviper', 1)
error_status = 1
if_error_then_all_processes_exit_program(error_status)
#Create folder for all results or check if there is one created already
if(myid == main_node):
#cmd = "{}".format("Rmycounter ccc")
#cmdexecute(cmd)
if( masterdir == ""):
timestring = strftime("%Y_%m_%d__%H_%M_%S" + DIR_DELIM, localtime())
masterdir = "master"+timestring
if not os.path.exists(masterdir):
cmd = "{} {}".format("mkdir", masterdir)
cmdexecute(cmd)
if ':' in args[0]:
bdb_stack_location = args[0].split(":")[0] + ":" + masterdir + args[0].split(":")[1]
org_stack_location = args[0]
if(not os.path.exists(os.path.join(masterdir,"EMAN2DB" + DIR_DELIM))):
# cmd = "{} {}".format("cp -rp EMAN2DB", masterdir, "EMAN2DB" DIR_DELIM)
# cmdexecute(cmd)
cmd = "{} {} {}".format("e2bdb.py", org_stack_location,"--makevstack=" + bdb_stack_location + "_000")
cmdexecute(cmd)
from applications import header
try:
header(bdb_stack_location + "_000", params='original_image_index', fprint=True)
print "Images were already indexed!"
except KeyError:
print "Indexing images"
header(bdb_stack_location + "_000", params='original_image_index', consecutive=True)
else:
filename = os.path.basename(args[0])
bdb_stack_location = "bdb:" + masterdir + os.path.splitext(filename)[0]
if(not os.path.exists(os.path.join(masterdir,"EMAN2DB" + DIR_DELIM))):
cmd = "{} {} {}".format("sxcpy.py ", args[0], bdb_stack_location + "_000")
cmdexecute(cmd)
from applications import header
try:
header(bdb_stack_location + "_000", params='original_image_index', fprint=True)
print "Images were already indexed!"
except KeyError:
print "Indexing images"
header(bdb_stack_location + "_000", params='original_image_index', consecutive=True)
# send masterdir to all processes
dir_len = len(masterdir)*int(myid == main_node)
dir_len = mpi_bcast(dir_len,1,MPI_INT,0,MPI_COMM_WORLD)[0]
masterdir = mpi_bcast(masterdir,dir_len,MPI_CHAR,main_node,MPI_COMM_WORLD)
masterdir = string.join(masterdir,"")
if masterdir[-1] != DIR_DELIM:
masterdir += DIR_DELIM
global_def.LOGFILE = os.path.join(masterdir, global_def.LOGFILE)
print_program_start_information()
# mpi_barrier(mpi_comm)
# from mpi import mpi_finalize
# mpi_finalize()
# print "mpi finalize"
# from sys import exit
# exit()
# send bdb_stack_location to all processes
dir_len = len(bdb_stack_location)*int(myid == main_node)
dir_len = mpi_bcast(dir_len,1,MPI_INT,0,MPI_COMM_WORLD)[0]
bdb_stack_location = mpi_bcast(bdb_stack_location,dir_len,MPI_CHAR,main_node,MPI_COMM_WORLD)
bdb_stack_location = string.join(bdb_stack_location,"")
iteration_start = get_latest_directory_increment_value(masterdir, "main")
if (myid == main_node):
if (iteration_start < iteration_start_default):
ERROR('Starting iteration provided is greater than last iteration performed. Quiting program', 'sxviper', 1)
error_status = 1
if iteration_start_default!=0:
iteration_start = iteration_start_default
if (myid == main_node):
if (number_of_rrr_viper_runs < iteration_start):
ERROR('Please provide number of rviper runs (--n_rv_runs) greater than number of iterations already performed.', 'sxviper', 1)
error_status = 1
if_error_then_all_processes_exit_program(error_status)
for rviper_iter in range(iteration_start, number_of_rrr_viper_runs + 1):
if(myid == main_node):
all_projs = EMData.read_images(bdb_stack_location + "_%03d"%(rviper_iter - 1))
print "XXXXXXXXXXXXXXXXX"
print "Number of projections (in loop): " + str(len(all_projs))
print "XXXXXXXXXXXXXXXXX"
subset = range(len(all_projs))
else:
all_projs = None
subset = None
runs_iter = get_latest_directory_increment_value(masterdir + NAME_OF_MAIN_DIR + "%03d"%rviper_iter, DIR_DELIM + NAME_OF_RUN_DIR, start_value=0) - 1
no_of_viper_runs_analyzed_together = max(runs_iter + 2, no_of_viper_runs_analyzed_together_from_user_options)
first_time_entering_the_loop_need_to_do_full_check_up = True
while True:
runs_iter += 1
if not first_time_entering_the_loop_need_to_do_full_check_up:
if runs_iter >= no_of_viper_runs_analyzed_together:
break
first_time_entering_the_loop_need_to_do_full_check_up = False
this_run_is_NOT_complete = 0
if (myid == main_node):
independent_run_dir = masterdir + DIR_DELIM + NAME_OF_MAIN_DIR + ('%03d' + DIR_DELIM + NAME_OF_RUN_DIR + "%03d" + DIR_DELIM)%(rviper_iter, runs_iter)
if run_get_already_processed_viper_runs:
cmd = "{} {}".format("mkdir -p", masterdir + DIR_DELIM + NAME_OF_MAIN_DIR + ('%03d' + DIR_DELIM)%(rviper_iter)); cmdexecute(cmd)
cmd = "{} {}".format("rm -rf", independent_run_dir); cmdexecute(cmd)
cmd = "{} {}".format("cp -r", get_already_processed_viper_runs() + " " + independent_run_dir); cmdexecute(cmd)
if os.path.exists(independent_run_dir + "log.txt") and (string_found_in_file("Finish VIPER2", independent_run_dir + "log.txt")):
this_run_is_NOT_complete = 0
else:
this_run_is_NOT_complete = 1
cmd = "{} {}".format("rm -rf", independent_run_dir); cmdexecute(cmd)
cmd = "{} {}".format("mkdir -p", independent_run_dir); cmdexecute(cmd)
this_run_is_NOT_complete = mpi_bcast(this_run_is_NOT_complete,1,MPI_INT,main_node,MPI_COMM_WORLD)[0]
dir_len = len(independent_run_dir)
dir_len = mpi_bcast(dir_len,1,MPI_INT,main_node,MPI_COMM_WORLD)[0]
independent_run_dir = mpi_bcast(independent_run_dir,dir_len,MPI_CHAR,main_node,MPI_COMM_WORLD)
independent_run_dir = string.join(independent_run_dir,"")
else:
this_run_is_NOT_complete = mpi_bcast(this_run_is_NOT_complete,1,MPI_INT,main_node,MPI_COMM_WORLD)[0]
dir_len = 0
independent_run_dir = ""
dir_len = mpi_bcast(dir_len,1,MPI_INT,main_node,MPI_COMM_WORLD)[0]
independent_run_dir = mpi_bcast(independent_run_dir,dir_len,MPI_CHAR,main_node,MPI_COMM_WORLD)
independent_run_dir = string.join(independent_run_dir,"")
if this_run_is_NOT_complete:
mpi_barrier(MPI_COMM_WORLD)
if independent_run_dir[-1] != DIR_DELIM:
independent_run_dir += DIR_DELIM
log.prefix = independent_run_dir
options.user_func = user_functions.factory[options.function]
# for debugging purposes
#if (myid == main_node):
#cmd = "{} {}".format("cp ~/log.txt ", independent_run_dir)
#cmdexecute(cmd)
#cmd = "{} {}{}".format("cp ~/paramdir/params$(mycounter ccc).txt ", independent_run_dir, "param%03d.txt"%runs_iter)
#cmd = "{} {}{}".format("cp ~/paramdir/params$(mycounter ccc).txt ", independent_run_dir, "params.txt")
#cmdexecute(cmd)
if (myid == main_node):
store_value_of_simple_vars_in_json_file(masterdir + 'program_state_stack.json', locals(), exclude_list_of_vars=["usage"],
vars_that_will_show_only_size = ["subset"])
store_value_of_simple_vars_in_json_file(masterdir + 'program_state_stack.json', options.__dict__, write_or_append='a')
# mpi_barrier(mpi_comm)
# from mpi import mpi_finalize
# mpi_finalize()
# print "mpi finalize"
# from sys import exit
# exit()
out_params, out_vol, out_peaks = multi_shc(all_projs, subset, no_of_shc_runs_analyzed_together, options,
mpi_comm=mpi_comm, log=log, ref_vol=ref_vol)
# end of: if this_run_is_NOT_complete:
if runs_iter >= (no_of_viper_runs_analyzed_together_from_user_options - 1):
increment_for_current_iteration = identify_outliers(myid, main_node, rviper_iter,
no_of_viper_runs_analyzed_together, no_of_viper_runs_analyzed_together_from_user_options, masterdir,
bdb_stack_location, outlier_percentile, criterion_name, outlier_index_threshold_method, angle_threshold)
if increment_for_current_iteration == MUST_END_PROGRAM_THIS_ITERATION:
break
no_of_viper_runs_analyzed_together += increment_for_current_iteration
# end of independent viper loop
calculate_volumes_after_rotation_and_save_them(options, rviper_iter, masterdir, bdb_stack_location, myid,
mpi_size, no_of_viper_runs_analyzed_together, no_of_viper_runs_analyzed_together_from_user_options)
if increment_for_current_iteration == MUST_END_PROGRAM_THIS_ITERATION:
if (myid == main_node):
print "RVIPER found a core set of stable projections for the current RVIPER iteration (%d), the maximum angle difference between corresponding projections from different VIPER volumes is less than %.2f. Finishing."%(rviper_iter, ANGLE_ERROR_THRESHOLD)
break
else:
if (myid == main_node):
print "After running the last iteration (%d), RVIPER did not find a set of projections with the maximum angle difference between corresponding projections from different VIPER volumes less than %.2f Finishing."%(rviper_iter, ANGLE_ERROR_THRESHOLD)
# end of RVIPER loop
#mpi_finalize()
#sys.exit()
mpi_barrier(MPI_COMM_WORLD)
mpi_finalize()
def cml_find_structure2(Prj, Ori, Rot, outdir, outname, maxit, first_zero, flag_weights, myid, main_node, number_of_proc):
from projection import cml_export_progress, cml_disc, cml_export_txtagls
import time, sys
from random import shuffle,random
from mpi import MPI_FLOAT, MPI_INT, MPI_SUM, MPI_COMM_WORLD
from mpi import mpi_reduce, mpi_bcast, mpi_barrier
# global vars
global g_i_prj, g_n_prj, g_n_anglst, g_anglst, g_d_psi, g_debug, g_n_lines, g_seq
# list of free orientation
ocp = [-1] * g_n_anglst
if first_zero:
listprj = range(1, g_n_prj)
ocp[0] = 0
else: listprj = range(g_n_prj)
# to stop when the solution oscillates
period_disc = [0, 0, 0]
period_ct = 0
period_th = 2
#if not flag_weights: weights = [1.0] * g_n_lines
# iteration loop
for ite in xrange(maxit):
#print ">>>>>>>>>>>>>>>>>>>>>>>>>>>>>> ite = ", ite, " myid = ", myid
t_start = time.time()
# loop over i prj
change = False
tlistprj = listprj[:]
shuffle(tlistprj)
nnn = len(tlistprj)
tlistprj = mpi_bcast(tlistprj, nnn, MPI_INT, main_node, MPI_COMM_WORLD)
tlistprj = map(int, tlistprj)
"""
if(ite>1 and ite%5 == 0 and ite<140):
if(myid == main_node):
for i in xrange(0,len(tlistprj),5):
ind = 4*i
Ori[ind] = 360.*random()
Ori[ind+1] = 180.*random()
Ori[ind+2] = 360.*random()
Ori[ind+3] = -1
for i in xrange(len(tlistprj)):
ind = 4*i
Ori[ind+3] = float(Ori[ind+3])
nnn = len(Ori)
Ori = mpi_bcast(Ori, nnn, MPI_FLOAT, main_node, MPI_COMM_WORLD)
Ori = map(float, Ori)
for i in xrange(len(tlistprj)):
ind = 4*i
Ori[ind+3] = int(Ori[ind+3])
"""
for iprj in tlistprj:
#print "********************************** iprj = ", iprj, g_n_anglst
# Store current the current orientation
ind = 4*iprj
store_phi = Ori[ind]
store_theta = Ori[ind+1]
store_psi = Ori[ind+2]
cur_agl = Ori[ind+3]
if cur_agl != -1: ocp[cur_agl] = -1
# prepare active index of cml for weighting in order to earn time later
iw = [0] * (g_n_prj - 1)
c = 0
ct = 0
for i in xrange(g_n_prj):
for j in xrange(i+1, g_n_prj):
if i == iprj or j == iprj:
iw[ct] = c
ct += 1
c += 1
# loop over all angles
best_disc_list = [0]*g_n_anglst
best_psi_list = [0]*g_n_anglst
for iagl in xrange(myid, g_n_anglst, number_of_proc):
# if orientation is free
if ocp[iagl] == -1:
# assign new orientation
Ori[ind] = g_anglst[iagl][0]
Ori[ind+1] = g_anglst[iagl][1]
Rot = Util.cml_update_rot(Rot, iprj, Ori[ind], Ori[ind+1], 0.0)
# weights
if flag_weights:
cml = Util.cml_line_in3d(Ori, g_seq, g_n_prj, g_n_lines)
weights = Util.cml_weights(cml)
mw = max(weights)
for i in xrange(g_n_lines): weights[i] = mw - weights[i]
sw = sum(weights)
if sw == 0:
weights = [6.28 / float(g_n_lines)] * g_n_lines
else:
for i in xrange(g_n_lines):
weights[i] /= sw
weights[i] *= weights[i]
# spin all psi
com = Util.cml_line_insino(Rot, iprj, g_n_prj)
if flag_weights:
res = Util.cml_spin_psi(Prj, com, weights, iprj, iw, g_n_psi, g_d_psi, g_n_prj)
else:
res = Util.cml_spin_psi_now(Prj, com, iprj, iw, g_n_psi, g_d_psi, g_n_prj)
# select the best
best_disc_list[iagl] = res[0]
best_psi_list[iagl] = res[1]
if g_debug: cml_export_progress(outdir, ite, iprj, iagl, res[1], res[0], 'progress')
else:
if g_debug: cml_export_progress(outdir, ite, iprj, iagl, -1, -1, 'progress')
best_disc_list = mpi_reduce(best_disc_list, g_n_anglst, MPI_FLOAT, MPI_SUM, main_node, MPI_COMM_WORLD)
best_psi_list = mpi_reduce(best_psi_list, g_n_anglst, MPI_FLOAT, MPI_SUM, main_node, MPI_COMM_WORLD)
best_psi = -1
best_iagl = -1
if myid == main_node:
best_disc = 1.0e20
for iagl in xrange(g_n_anglst):
if best_disc_list[iagl] > 0.0 and best_disc_list[iagl] < best_disc:
best_disc = best_disc_list[iagl]
best_psi = best_psi_list[iagl]
best_iagl = iagl
best_psi = mpi_bcast(best_psi, 1, MPI_FLOAT, main_node, MPI_COMM_WORLD)
best_iagl = mpi_bcast(best_iagl, 1, MPI_INT, main_node, MPI_COMM_WORLD)
best_psi = float(best_psi[0])
best_iagl = int(best_iagl[0])
#print "xxxxx myid = ", myid, " best_psi = ", best_psi, " best_ialg = ", best_iagl
# if change, assign
if best_iagl != cur_agl:
ocp[best_iagl] = iprj
Ori[ind] = g_anglst[best_iagl][0] # phi
Ori[ind+1] = g_anglst[best_iagl][1] # theta
Ori[ind+2] = best_psi * g_d_psi # psi
Ori[ind+3] = best_iagl # index
change = True
else:
if cur_agl != -1: ocp[cur_agl] = iprj
Ori[ind] = store_phi
Ori[ind+1] = store_theta
Ori[ind+2] = store_psi
Ori[ind+3] = cur_agl
Rot = Util.cml_update_rot(Rot, iprj, Ori[ind], Ori[ind+1], Ori[ind+2])
if g_debug: cml_export_progress(outdir, ite, iprj, best_iagl, best_psi * g_d_psi, best_disc, 'choose')
# if one change, compute new full disc
disc = cml_disc(Prj, Ori, Rot, flag_weights)
# display in the progress file
if myid == main_node:
cml_export_txtagls(outdir, outname, Ori, disc, 'Ite: %03i' % (ite + 1))
if not change: break
# to stop when the solution oscillates
period_disc.pop(0)
period_disc.append(disc)
if period_disc[0] == period_disc[2]:
period_ct += 1
if period_ct >= period_th and min(period_disc) == disc and myid == main_node:
angfile = open(outdir + '/' + outname, 'a')
angfile.write('\nSTOP SOLUTION UNSTABLE\n')
angfile.write('Discrepancy period: %s\n' % period_disc)
angfile.close()
break
else:
period_ct = 0
mpi_barrier(MPI_COMM_WORLD)
return Ori, disc, ite
def compare(compare_ref_free, outfile_repro,ref_free_output,yrng, xrng, rstep,nx,apix,ref_free_cutoff, nproc, myid, main_node):
from alignment import Numrinit, ringwe, Applyws
from random import seed, randint
from utilities import get_params2D, set_params2D, model_circle, inverse_transform2, combine_params2
from fundamentals import rot_shift2D
from mpi import MPI_COMM_WORLD, mpi_barrier, mpi_bcast, MPI_INT
from statistics import fsc_mask
from filter import fit_tanh
from numpy import array
fout = "%s.hdf" % ref_free_output
frc_out = "%s_frc" % ref_free_output
res_out = "%s_res" % ref_free_output
nima = EMUtil.get_image_count(compare_ref_free)
image_start, image_end = MPI_start_end(nima, nproc, myid)
ima = EMData()
ima.read_image(compare_ref_free, image_start)
last_ring = nx/2-2
first_ring = 1
mask = model_circle(last_ring, nx, nx)
refi = []
numref = EMUtil.get_image_count(outfile_repro)
cnx = nx/2 +1
cny = cnx
mode = "F"
numr = Numrinit(first_ring, last_ring, rstep, mode)
wr = ringwe(numr, mode)
ima.to_zero()
for j in xrange(numref):
temp = EMData()
temp.read_image(outfile_repro, j)
# even, odd, numer of even, number of images. After frc, totav
refi.append(temp)
# for each node read its share of data
data = EMData.read_images(compare_ref_free, range(image_start, image_end))
for im in xrange(image_start, image_end):
data[im-image_start].set_attr('ID', im)
set_params2D(data[im-image_start],[0,0,0,0,1])
ringref = []
for j in xrange(numref):
refi[j].process_inplace("normalize.mask", {"mask":mask, "no_sigma":1}) # normalize reference images to N(0,1)
cimage = Util.Polar2Dm(refi[j], cnx, cny, numr, mode)
Util.Frngs(cimage, numr)
Applyws(cimage, numr, wr)
ringref.append(cimage)
if myid == main_node: seed(1000)
data_shift = []
frc = []
res = []
for im in xrange(image_start, image_end):
alpha, sx, sy, mirror, scale = get_params2D(data[im-image_start])
alphai, sxi, syi, scalei = inverse_transform2(alpha, sx, sy, 1.0)
# normalize
data[im-image_start].process_inplace("normalize.mask", {"mask":mask, "no_sigma":1}) # subtract average under the mask
# align current image to the reference
[angt, sxst, syst, mirrort, xiref, peakt] = Util.multiref_polar_ali_2d(data[im-image_start], ringref, xrng, yrng, 1, mode, numr, cnx+sxi, cny+syi)
iref = int(xiref)
[alphan, sxn, syn, mn] = combine_params2(0.0, -sxi, -syi, 0, angt, sxst, syst, (int)(mirrort))
set_params2D(data[im-image_start], [alphan, sxn, syn, int(mn), scale])
temp = rot_shift2D(data[im-image_start], alphan, sxn, syn, mn)
temp.set_attr('assign',iref)
tfrc = fsc_mask(temp,refi[iref],mask = mask)
temp.set_attr('frc',tfrc[1])
res = fit_tanh(tfrc)
temp.set_attr('res',res)
data_shift.append(temp)
for node in xrange(nproc):
if myid == node:
for image in data_shift:
image.write_image(fout,-1)
refindex = image.get_attr('assign')
refi[refindex].write_image(fout,-1)
mpi_barrier(MPI_COMM_WORLD)
rejects = []
if myid == main_node:
a = EMData()
index = 0
frc = []
res = []
temp = []
classes = []
for im in xrange(nima):
a.read_image(fout, index)
frc.append(a.get_attr("frc"))
if ref_free_cutoff != -1: classes.append(a.get_attr("class_ptcl_idxs"))
tmp = a.get_attr("res")
temp.append(tmp[0])
res.append("%12f" %(apix/tmp[0]))
res.append("\n")
index = index + 2
res_num = array(temp)
mean_score = res_num.mean(axis=0)
std_score = res_num.std(axis=0)
std = std_score / 2
if ref_free_cutoff !=-1:
cutoff = mean_score - std * ref_free_cutoff
reject = res_num < cutoff
index = 0
for i in reject:
if i: rejects.extend(classes[index])
index = index + 1
rejects.sort()
length = mpi_bcast(len(rejects),1,MPI_INT,main_node, MPI_COMM_WORLD)
rejects = mpi_bcast(rejects,length , MPI_INT, main_node, MPI_COMM_WORLD)
del a
fout_frc = open(frc_out,'w')
fout_res = open(res_out,'w')
fout_res.write("".join(res))
temp = zip(*frc)
datstrings = []
for i in temp:
for j in i:
datstrings.append(" %12f" % (j))
datstrings.append("\n")
fout_frc.write("".join(datstrings))
fout_frc.close()
del refi
del ringref
return rejects
def ali3d_MPI(stack, ref_vol, outdir, maskfile = None, ir = 1, ou = -1, rs = 1,
xr = "4 2 2 1", yr = "-1", ts = "1 1 0.5 0.25", delta = "10 6 4 4", an = "-1",
center = 0, maxit = 5, term = 95, CTF = False, fourvar = False, snr = 1.0, ref_a = "S", sym = "c1",
sort=True, cutoff=999.99, pix_cutoff="0", two_tail=False, model_jump="1 1 1 1 1", restart=False, save_half=False,
protos=None, oplane=None, lmask=-1, ilmask=-1, findseam=False, vertstep=None, hpars="-1", hsearch="73.0 170.0",
full_output = False, compare_repro = False, compare_ref_free = "-1", ref_free_cutoff= "-1 -1 -1 -1",
wcmask = None, debug = False, recon_pad = 4):
from alignment import Numrinit, prepare_refrings
from utilities import model_circle, get_image, drop_image, get_input_from_string
from utilities import bcast_list_to_all, bcast_number_to_all, reduce_EMData_to_root, bcast_EMData_to_all
from utilities import send_attr_dict
from utilities import get_params_proj, file_type
from fundamentals import rot_avg_image
import os
import types
from utilities import print_begin_msg, print_end_msg, print_msg
from mpi import mpi_bcast, mpi_comm_size, mpi_comm_rank, MPI_FLOAT, MPI_COMM_WORLD, mpi_barrier, mpi_reduce
from mpi import mpi_reduce, MPI_INT, MPI_SUM, mpi_finalize
from filter import filt_ctf
from projection import prep_vol, prgs
from statistics import hist_list, varf3d_MPI, fsc_mask
from numpy import array, bincount, array2string, ones
number_of_proc = mpi_comm_size(MPI_COMM_WORLD)
myid = mpi_comm_rank(MPI_COMM_WORLD)
main_node = 0
if myid == main_node:
if os.path.exists(outdir): ERROR('Output directory exists, please change the name and restart the program', "ali3d_MPI", 1)
os.mkdir(outdir)
mpi_barrier(MPI_COMM_WORLD)
if debug:
from time import sleep
while not os.path.exists(outdir):
print "Node ",myid," waiting..."
sleep(5)
info_file = os.path.join(outdir, "progress%04d"%myid)
finfo = open(info_file, 'w')
else:
finfo = None
mjump = get_input_from_string(model_jump)
xrng = get_input_from_string(xr)
if yr == "-1": yrng = xrng
else : yrng = get_input_from_string(yr)
step = get_input_from_string(ts)
delta = get_input_from_string(delta)
ref_free_cutoff = get_input_from_string(ref_free_cutoff)
pix_cutoff = get_input_from_string(pix_cutoff)
lstp = min(len(xrng), len(yrng), len(step), len(delta))
if an == "-1":
an = [-1] * lstp
else:
an = get_input_from_string(an)
# make sure pix_cutoff is set for all iterations
if len(pix_cutoff)<lstp:
for i in xrange(len(pix_cutoff),lstp):
pix_cutoff.append(pix_cutoff[-1])
# don't waste time on sub-pixel alignment for low-resolution ang incr
for i in range(len(step)):
if (delta[i] > 4 or delta[i] == -1) and step[i] < 1:
step[i] = 1
first_ring = int(ir)
rstep = int(rs)
last_ring = int(ou)
max_iter = int(maxit)
center = int(center)
nrefs = EMUtil.get_image_count( ref_vol )
nmasks = 0
if maskfile:
# read number of masks within each maskfile (mc)
nmasks = EMUtil.get_image_count( maskfile )
# open masks within maskfile (mc)
maskF = EMData.read_images(maskfile, xrange(nmasks))
vol = EMData.read_images(ref_vol, xrange(nrefs))
nx = vol[0].get_xsize()
## make sure box sizes are the same
if myid == main_node:
im=EMData.read_images(stack,[0])
bx = im[0].get_xsize()
if bx!=nx:
print_msg("Error: Stack box size (%i) differs from initial model (%i)\n"%(bx,nx))
sys.exit()
del im,bx
# for helical processing:
helicalrecon = False
if protos is not None or hpars != "-1" or findseam is True:
helicalrecon = True
# if no out-of-plane param set, use 5 degrees
if oplane is None:
oplane=5.0
if protos is not None:
proto = get_input_from_string(protos)
if len(proto) != nrefs:
print_msg("Error: insufficient protofilament numbers supplied")
sys.exit()
if hpars != "-1":
hpars = get_input_from_string(hpars)
if len(hpars) != 2*nrefs:
print_msg("Error: insufficient helical parameters supplied")
sys.exit()
## create helical parameter file for helical reconstruction
if helicalrecon is True and myid == main_node:
from hfunctions import createHpar
# create initial helical parameter files
dp=[0]*nrefs
dphi=[0]*nrefs
vdp=[0]*nrefs
vdphi=[0]*nrefs
for iref in xrange(nrefs):
hpar = os.path.join(outdir,"hpar%02d.spi"%(iref))
params = False
if hpars != "-1":
# if helical parameters explicitly given, set twist & rise
params = [float(hpars[iref*2]),float(hpars[(iref*2)+1])]
dp[iref],dphi[iref],vdp[iref],vdphi[iref] = createHpar(hpar,proto[iref],params,vertstep)
# get values for helical search parameters
hsearch = get_input_from_string(hsearch)
if len(hsearch) != 2:
print_msg("Error: specify outer and inner radii for helical search")
sys.exit()
if last_ring < 0 or last_ring > int(nx/2)-2 : last_ring = int(nx/2) - 2
if myid == main_node:
# import user_functions
# user_func = user_functions.factory[user_func_name]
print_begin_msg("ali3d_MPI")
print_msg("Input stack : %s\n"%(stack))
print_msg("Reference volume : %s\n"%(ref_vol))
print_msg("Output directory : %s\n"%(outdir))
if nmasks > 0:
print_msg("Maskfile (number of masks) : %s (%i)\n"%(maskfile,nmasks))
print_msg("Inner radius : %i\n"%(first_ring))
print_msg("Outer radius : %i\n"%(last_ring))
print_msg("Ring step : %i\n"%(rstep))
print_msg("X search range : %s\n"%(xrng))
print_msg("Y search range : %s\n"%(yrng))
print_msg("Translational step : %s\n"%(step))
print_msg("Angular step : %s\n"%(delta))
print_msg("Angular search range : %s\n"%(an))
print_msg("Maximum iteration : %i\n"%(max_iter))
print_msg("Center type : %i\n"%(center))
print_msg("CTF correction : %s\n"%(CTF))
print_msg("Signal-to-Noise Ratio : %f\n"%(snr))
print_msg("Reference projection method : %s\n"%(ref_a))
print_msg("Symmetry group : %s\n"%(sym))
print_msg("Fourier padding for 3D : %i\n"%(recon_pad))
print_msg("Number of reference models : %i\n"%(nrefs))
print_msg("Sort images between models : %s\n"%(sort))
print_msg("Allow images to jump : %s\n"%(mjump))
print_msg("CC cutoff standard dev : %f\n"%(cutoff))
print_msg("Two tail cutoff : %s\n"%(two_tail))
print_msg("Termination pix error : %f\n"%(term))
print_msg("Pixel error cutoff : %s\n"%(pix_cutoff))
print_msg("Restart : %s\n"%(restart))
print_msg("Full output : %s\n"%(full_output))
print_msg("Compare reprojections : %s\n"%(compare_repro))
print_msg("Compare ref free class avgs : %s\n"%(compare_ref_free))
print_msg("Use cutoff from ref free : %s\n"%(ref_free_cutoff))
if protos:
print_msg("Protofilament numbers : %s\n"%(proto))
print_msg("Using helical search range : %s\n"%hsearch)
if findseam is True:
print_msg("Using seam-based reconstruction\n")
if hpars != "-1":
print_msg("Using hpars : %s\n"%hpars)
if vertstep != None:
print_msg("Using vertical step : %.2f\n"%vertstep)
if save_half is True:
print_msg("Saving even/odd halves\n")
for i in xrange(100) : print_msg("*")
print_msg("\n\n")
if maskfile:
if type(maskfile) is types.StringType: mask3D = get_image(maskfile)
else: mask3D = maskfile
else: mask3D = model_circle(last_ring, nx, nx, nx)
numr = Numrinit(first_ring, last_ring, rstep, "F")
mask2D = model_circle(last_ring,nx,nx) - model_circle(first_ring,nx,nx)
fscmask = model_circle(last_ring,nx,nx,nx)
if CTF:
from filter import filt_ctf
from reconstruction_rjh import rec3D_MPI_noCTF
if myid == main_node:
active = EMUtil.get_all_attributes(stack, 'active')
list_of_particles = []
for im in xrange(len(active)):
if active[im]: list_of_particles.append(im)
del active
nima = len(list_of_particles)
else:
nima = 0
total_nima = bcast_number_to_all(nima, source_node = main_node)
if myid != main_node:
list_of_particles = [-1]*total_nima
list_of_particles = bcast_list_to_all(list_of_particles, source_node = main_node)
image_start, image_end = MPI_start_end(total_nima, number_of_proc, myid)
# create a list of images for each node
list_of_particles = list_of_particles[image_start: image_end]
nima = len(list_of_particles)
if debug:
finfo.write("image_start, image_end: %d %d\n" %(image_start, image_end))
finfo.flush()
data = EMData.read_images(stack, list_of_particles)
t_zero = Transform({"type":"spider","phi":0,"theta":0,"psi":0,"tx":0,"ty":0})
transmulti = [[t_zero for i in xrange(nrefs)] for j in xrange(nima)]
for iref,im in ((iref,im) for iref in xrange(nrefs) for im in xrange(nima)):
if nrefs == 1:
transmulti[im][iref] = data[im].get_attr("xform.projection")
else:
# if multi models, keep track of eulers for all models
try:
transmulti[im][iref] = data[im].get_attr("eulers_txty.%i"%iref)
except:
data[im].set_attr("eulers_txty.%i"%iref,t_zero)
scoremulti = [[0.0 for i in xrange(nrefs)] for j in xrange(nima)]
pixelmulti = [[0.0 for i in xrange(nrefs)] for j in xrange(nima)]
ref_res = [0.0 for x in xrange(nrefs)]
apix = data[0].get_attr('apix_x')
# for oplane parameter, create cylindrical mask
if oplane is not None and myid == main_node:
from hfunctions import createCylMask
cmaskf=os.path.join(outdir, "mask3D_cyl.mrc")
mask3D = createCylMask(data,ou,lmask,ilmask,cmaskf)
# if finding seam of helix, create wedge masks
if findseam is True:
wedgemask=[]
for pf in xrange(nrefs):
wedgemask.append(EMData())
# wedgemask option
if wcmask is not None:
wcmask = get_input_from_string(wcmask)
if len(wcmask) != 3:
print_msg("Error: wcmask option requires 3 values: x y radius")
sys.exit()
# determine if particles have helix info:
try:
data[0].get_attr('h_angle')
original_data = []
boxmask = True
from hfunctions import createBoxMask
except:
boxmask = False
# prepare particles
for im in xrange(nima):
data[im].set_attr('ID', list_of_particles[im])
data[im].set_attr('pix_score', int(0))
if CTF:
# only phaseflip particles, not full CTF correction
ctf_params = data[im].get_attr("ctf")
st = Util.infomask(data[im], mask2D, False)
data[im] -= st[0]
data[im] = filt_ctf(data[im], ctf_params, sign = -1, binary=1)
data[im].set_attr('ctf_applied', 1)
# for window mask:
if boxmask is True:
h_angle = data[im].get_attr("h_angle")
original_data.append(data[im].copy())
bmask = createBoxMask(nx,apix,ou,lmask,h_angle)
data[im]*=bmask
del bmask
if debug:
finfo.write( '%d loaded \n' % nima )
finfo.flush()
if myid == main_node:
# initialize data for the reference preparation function
ref_data = [ mask3D, max(center,0), None, None, None, None ]
# for method -1, switch off centering in user function
from time import time
# this is needed for gathering of pixel errors
disps = []
recvcount = []
disps_score = []
recvcount_score = []
for im in xrange(number_of_proc):
if( im == main_node ):
disps.append(0)
disps_score.append(0)
else:
disps.append(disps[im-1] + recvcount[im-1])
disps_score.append(disps_score[im-1] + recvcount_score[im-1])
ib, ie = MPI_start_end(total_nima, number_of_proc, im)
recvcount.append( ie - ib )
recvcount_score.append((ie-ib)*nrefs)
pixer = [0.0]*nima
cs = [0.0]*3
total_iter = 0
volodd = EMData.read_images(ref_vol, xrange(nrefs))
voleve = EMData.read_images(ref_vol, xrange(nrefs))
if restart:
# recreate initial volumes from alignments stored in header
itout = "000_00"
for iref in xrange(nrefs):
if(nrefs == 1):
modout = ""
else:
modout = "_model_%02d"%(iref)
if(sort):
group = iref
for im in xrange(nima):
imgroup = data[im].get_attr('group')
if imgroup == iref:
data[im].set_attr('xform.projection',transmulti[im][iref])
else:
group = int(999)
for im in xrange(nima):
data[im].set_attr('xform.projection',transmulti[im][iref])
fscfile = os.path.join(outdir, "fsc_%s%s"%(itout,modout))
vol[iref], fscc, volodd[iref], voleve[iref] = rec3D_MPI_noCTF(data, sym, fscmask, fscfile, myid, main_node, index = group, npad = recon_pad)
if myid == main_node:
if helicalrecon:
from hfunctions import processHelicalVol
vstep=None
if vertstep is not None:
vstep=(vdp[iref],vdphi[iref])
print_msg("Old rise and twist for model %i : %8.3f, %8.3f\n"%(iref,dp[iref],dphi[iref]))
hvals=processHelicalVol(vol[iref],voleve[iref],volodd[iref],iref,outdir,itout,
dp[iref],dphi[iref],apix,hsearch,findseam,vstep,wcmask)
(vol[iref],voleve[iref],volodd[iref],dp[iref],dphi[iref],vdp[iref],vdphi[iref])=hvals
print_msg("New rise and twist for model %i : %8.3f, %8.3f\n"%(iref,dp[iref],dphi[iref]))
# get new FSC from symmetrized half volumes
fscc = fsc_mask( volodd[iref], voleve[iref], mask3D, rstep, fscfile)
else:
vol[iref].write_image(os.path.join(outdir, "vol_%s.hdf"%itout),-1)
if save_half is True:
volodd[iref].write_image(os.path.join(outdir, "volodd_%s.hdf"%itout),-1)
voleve[iref].write_image(os.path.join(outdir, "voleve_%s.hdf"%itout),-1)
if nmasks > 1:
# Read mask for multiplying
ref_data[0] = maskF[iref]
ref_data[2] = vol[iref]
ref_data[3] = fscc
# call user-supplied function to prepare reference image, i.e., center and filter it
vol[iref], cs,fl = ref_ali3d(ref_data)
vol[iref].write_image(os.path.join(outdir, "volf_%s.hdf"%(itout)),-1)
if (apix == 1):
res_msg = "Models filtered at spatial frequency of:\t"
res = fl
else:
res_msg = "Models filtered at resolution of: \t"
res = apix / fl
ares = array2string(array(res), precision = 2)
print_msg("%s%s\n\n"%(res_msg,ares))
bcast_EMData_to_all(vol[iref], myid, main_node)
# write out headers, under MPI writing has to be done sequentially
mpi_barrier(MPI_COMM_WORLD)
# projection matching
for N_step in xrange(lstp):
terminate = 0
Iter = -1
while(Iter < max_iter-1 and terminate == 0):
Iter += 1
total_iter += 1
itout = "%03g_%02d" %(delta[N_step], Iter)
if myid == main_node:
print_msg("ITERATION #%3d, inner iteration #%3d\nDelta = %4.1f, an = %5.2f, xrange = %5.2f, yrange = %5.2f, step = %5.2f\n\n"%(N_step, Iter, delta[N_step], an[N_step], xrng[N_step],yrng[N_step],step[N_step]))
for iref in xrange(nrefs):
if myid == main_node: start_time = time()
volft,kb = prep_vol( vol[iref] )
## constrain projections to out of plane parameter
theta1 = None
theta2 = None
if oplane is not None:
theta1 = 90-oplane
theta2 = 90+oplane
refrings = prepare_refrings( volft, kb, nx, delta[N_step], ref_a, sym, numr, MPI=True, phiEqpsi = "Minus", initial_theta=theta1, delta_theta=theta2)
del volft,kb
if myid== main_node:
print_msg( "Time to prepare projections for model %i: %s\n" % (iref, legibleTime(time()-start_time)) )
start_time = time()
for im in xrange( nima ):
data[im].set_attr("xform.projection", transmulti[im][iref])
if an[N_step] == -1:
t1, peak, pixer[im] = proj_ali_incore(data[im],refrings,numr,xrng[N_step],yrng[N_step],step[N_step],finfo)
else:
t1, peak, pixer[im] = proj_ali_incore_local(data[im],refrings,numr,xrng[N_step],yrng[N_step],step[N_step],an[N_step],finfo)
#data[im].set_attr("xform.projection"%iref, t1)
if nrefs > 1: data[im].set_attr("eulers_txty.%i"%iref,t1)
scoremulti[im][iref] = peak
from pixel_error import max_3D_pixel_error
# t1 is the current param, t2 is old
t2 = transmulti[im][iref]
pixelmulti[im][iref] = max_3D_pixel_error(t1,t2,numr[-3])
transmulti[im][iref] = t1
if myid == main_node:
print_msg("Time of alignment for model %i: %s\n"%(iref, legibleTime(time()-start_time)))
start_time = time()
# gather scoring data from all processors
from mpi import mpi_gatherv
scoremultisend = sum(scoremulti,[])
pixelmultisend = sum(pixelmulti,[])
tmp = mpi_gatherv(scoremultisend,len(scoremultisend),MPI_FLOAT, recvcount_score, disps_score, MPI_FLOAT, main_node,MPI_COMM_WORLD)
tmp1 = mpi_gatherv(pixelmultisend,len(pixelmultisend),MPI_FLOAT, recvcount_score, disps_score, MPI_FLOAT, main_node,MPI_COMM_WORLD)
tmp = mpi_bcast(tmp,(total_nima * nrefs), MPI_FLOAT,0, MPI_COMM_WORLD)
tmp1 = mpi_bcast(tmp1,(total_nima * nrefs), MPI_FLOAT,0, MPI_COMM_WORLD)
tmp = map(float,tmp)
tmp1 = map(float,tmp1)
score = array(tmp).reshape(-1,nrefs)
pixelerror = array(tmp1).reshape(-1,nrefs)
score_local = array(scoremulti)
mean_score = score.mean(axis=0)
std_score = score.std(axis=0)
cut = mean_score - (cutoff * std_score)
cut2 = mean_score + (cutoff * std_score)
res_max = score_local.argmax(axis=1)
minus_cc = [0.0 for x in xrange(nrefs)]
minus_pix = [0.0 for x in xrange(nrefs)]
minus_ref = [0.0 for x in xrange(nrefs)]
#output pixel errors
if(myid == main_node):
from statistics import hist_list
lhist = 20
pixmin = pixelerror.min(axis=1)
region, histo = hist_list(pixmin, lhist)
if(region[0] < 0.0): region[0] = 0.0
print_msg("Histogram of pixel errors\n ERROR number of particles\n")
for lhx in xrange(lhist):
print_msg(" %10.3f %7d\n"%(region[lhx], histo[lhx]))
# Terminate if 95% within 1 pixel error
im = 0
for lhx in xrange(lhist):
if(region[lhx] > 1.0): break
im += histo[lhx]
print_msg( "Percent of particles with pixel error < 1: %f\n\n"% (im/float(total_nima)*100))
term_cond = float(term)/100
if(im/float(total_nima) > term_cond):
terminate = 1
print_msg("Terminating internal loop\n")
del region, histo
terminate = mpi_bcast(terminate, 1, MPI_INT, 0, MPI_COMM_WORLD)
terminate = int(terminate[0])
for im in xrange(nima):
if(sort==False):
data[im].set_attr('group',999)
elif (mjump[N_step]==1):
data[im].set_attr('group',int(res_max[im]))
pix_run = data[im].get_attr('pix_score')
if (pix_cutoff[N_step]==1 and (terminate==1 or Iter == max_iter-1)):
if (pixelmulti[im][int(res_max[im])] > 1):
data[im].set_attr('pix_score',int(777))
if (score_local[im][int(res_max[im])]<cut[int(res_max[im])]) or (two_tail and score_local[im][int(res_max[im])]>cut2[int(res_max[im])]):
data[im].set_attr('group',int(888))
minus_cc[int(res_max[im])] = minus_cc[int(res_max[im])] + 1
if(pix_run == 777):
data[im].set_attr('group',int(777))
minus_pix[int(res_max[im])] = minus_pix[int(res_max[im])] + 1
if (compare_ref_free != "-1") and (ref_free_cutoff[N_step] != -1) and (total_iter > 1):
id = data[im].get_attr('ID')
if id in rejects:
data[im].set_attr('group',int(666))
minus_ref[int(res_max[im])] = minus_ref[int(res_max[im])] + 1
minus_cc_tot = mpi_reduce(minus_cc,nrefs,MPI_FLOAT,MPI_SUM,0,MPI_COMM_WORLD)
minus_pix_tot = mpi_reduce(minus_pix,nrefs,MPI_FLOAT,MPI_SUM,0,MPI_COMM_WORLD)
minus_ref_tot = mpi_reduce(minus_ref,nrefs,MPI_FLOAT,MPI_SUM,0,MPI_COMM_WORLD)
if (myid == main_node):
if(sort):
tot_max = score.argmax(axis=1)
res = bincount(tot_max)
else:
res = ones(nrefs) * total_nima
print_msg("Particle distribution: \t\t%s\n"%(res*1.0))
afcut1 = res - minus_cc_tot
afcut2 = afcut1 - minus_pix_tot
afcut3 = afcut2 - minus_ref_tot
print_msg("Particle distribution after cc cutoff:\t\t%s\n"%(afcut1))
print_msg("Particle distribution after pix cutoff:\t\t%s\n"%(afcut2))
print_msg("Particle distribution after ref cutoff:\t\t%s\n\n"%(afcut3))
res = [0.0 for i in xrange(nrefs)]
for iref in xrange(nrefs):
if(center == -1):
from utilities import estimate_3D_center_MPI, rotate_3D_shift
dummy=EMData()
cs[0], cs[1], cs[2], dummy, dummy = estimate_3D_center_MPI(data, total_nima, myid, number_of_proc, main_node)
cs = mpi_bcast(cs, 3, MPI_FLOAT, main_node, MPI_COMM_WORLD)
cs = [-float(cs[0]), -float(cs[1]), -float(cs[2])]
rotate_3D_shift(data, cs)
if(sort):
group = iref
for im in xrange(nima):
imgroup = data[im].get_attr('group')
if imgroup == iref:
data[im].set_attr('xform.projection',transmulti[im][iref])
else:
group = int(999)
for im in xrange(nima):
data[im].set_attr('xform.projection',transmulti[im][iref])
if(nrefs == 1):
modout = ""
else:
modout = "_model_%02d"%(iref)
fscfile = os.path.join(outdir, "fsc_%s%s"%(itout,modout))
vol[iref], fscc, volodd[iref], voleve[iref] = rec3D_MPI_noCTF(data, sym, fscmask, fscfile, myid, main_node, index=group, npad=recon_pad)
if myid == main_node:
print_msg("3D reconstruction time for model %i: %s\n"%(iref, legibleTime(time()-start_time)))
start_time = time()
# Compute Fourier variance
if fourvar:
outvar = os.path.join(outdir, "volVar_%s.hdf"%(itout))
ssnr_file = os.path.join(outdir, "ssnr_%s"%(itout))
varf = varf3d_MPI(data, ssnr_text_file=ssnr_file, mask2D=None, reference_structure=vol[iref], ou=last_ring, rw=1.0, npad=1, CTF=None, sign=1, sym=sym, myid=myid)
if myid == main_node:
print_msg("Time to calculate 3D Fourier variance for model %i: %s\n"%(iref, legibleTime(time()-start_time)))
start_time = time()
varf = 1.0/varf
varf.write_image(outvar,-1)
else: varf = None
if myid == main_node:
if helicalrecon:
from hfunctions import processHelicalVol
vstep=None
if vertstep is not None:
vstep=(vdp[iref],vdphi[iref])
print_msg("Old rise and twist for model %i : %8.3f, %8.3f\n"%(iref,dp[iref],dphi[iref]))
hvals=processHelicalVol(vol[iref],voleve[iref],volodd[iref],iref,outdir,itout,
dp[iref],dphi[iref],apix,hsearch,findseam,vstep,wcmask)
(vol[iref],voleve[iref],volodd[iref],dp[iref],dphi[iref],vdp[iref],vdphi[iref])=hvals
print_msg("New rise and twist for model %i : %8.3f, %8.3f\n"%(iref,dp[iref],dphi[iref]))
# get new FSC from symmetrized half volumes
fscc = fsc_mask( volodd[iref], voleve[iref], mask3D, rstep, fscfile)
print_msg("Time to search and apply helical symmetry for model %i: %s\n\n"%(iref, legibleTime(time()-start_time)))
start_time = time()
else:
vol[iref].write_image(os.path.join(outdir, "vol_%s.hdf"%(itout)),-1)
if save_half is True:
volodd[iref].write_image(os.path.join(outdir, "volodd_%s.hdf"%(itout)),-1)
voleve[iref].write_image(os.path.join(outdir, "voleve_%s.hdf"%(itout)),-1)
if nmasks > 1:
# Read mask for multiplying
ref_data[0] = maskF[iref]
ref_data[2] = vol[iref]
ref_data[3] = fscc
ref_data[4] = varf
# call user-supplied function to prepare reference image, i.e., center and filter it
vol[iref], cs,fl = ref_ali3d(ref_data)
vol[iref].write_image(os.path.join(outdir, "volf_%s.hdf"%(itout)),-1)
if (apix == 1):
res_msg = "Models filtered at spatial frequency of:\t"
res[iref] = fl
else:
res_msg = "Models filtered at resolution of: \t"
res[iref] = apix / fl
del varf
bcast_EMData_to_all(vol[iref], myid, main_node)
if compare_ref_free != "-1": compare_repro = True
if compare_repro:
outfile_repro = comp_rep(refrings, data, itout, modout, vol[iref], group, nima, nx, myid, main_node, outdir)
mpi_barrier(MPI_COMM_WORLD)
if compare_ref_free != "-1":
ref_free_output = os.path.join(outdir,"ref_free_%s%s"%(itout,modout))
rejects = compare(compare_ref_free, outfile_repro,ref_free_output,yrng[N_step], xrng[N_step], rstep,nx,apix,ref_free_cutoff[N_step], number_of_proc, myid, main_node)
# retrieve alignment params from all processors
par_str = ['xform.projection','ID','group']
if nrefs > 1:
for iref in xrange(nrefs):
par_str.append('eulers_txty.%i'%iref)
if myid == main_node:
from utilities import recv_attr_dict
recv_attr_dict(main_node, stack, data, par_str, image_start, image_end, number_of_proc)
else: send_attr_dict(main_node, data, par_str, image_start, image_end)
if myid == main_node:
ares = array2string(array(res), precision = 2)
print_msg("%s%s\n\n"%(res_msg,ares))
dummy = EMData()
if full_output:
nimat = EMUtil.get_image_count(stack)
output_file = os.path.join(outdir, "paramout_%s"%itout)
foutput = open(output_file, 'w')
for im in xrange(nimat):
# save the parameters for each of the models
outstring = ""
dummy.read_image(stack,im,True)
param3d = dummy.get_attr('xform.projection')
g = dummy.get_attr("group")
# retrieve alignments in EMAN-format
pE = param3d.get_params('eman')
outstring += "%f\t%f\t%f\t%f\t%f\t%i\n" %(pE["az"], pE["alt"], pE["phi"], pE["tx"], pE["ty"],g)
foutput.write(outstring)
foutput.close()
del dummy
mpi_barrier(MPI_COMM_WORLD)
# mpi_finalize()
if myid == main_node: print_end_msg("ali3d_MPI")
def program_state_stack(full_current_state, frame_info, file_name_of_saved_state=None, last_call="", force_starting_execution = False):
"""
When used it needs: from inspect import currentframe, getframeinfo
This function is used for restarting time consuming data processing programs/steps from the last saved point.
This static variable must be defined before the first call:
program_state_stack.PROGRAM_STATE_VARIABLES = {"local_var_i", "local_var_j", "local_var_h", "local_var_g"}
It contains local variables at any level of the stack that define uniquely the state(flow/logic) of the program.
It is assumed that the processed data is saved at each step and it is independent from the variables that uniquely define
the state(flow/logic) of the program. All the variables that are used in more than one step must be calculated before
the "if program_state_stack(locals(), getframeinfo(currentframe())):" call. It is assumed that they are not time consuming.
Passing processed data from one step to the next is done only through files.
First call needs to contain "file_name_of_saved_state".
Then, the next calls are "if program_state_stack(locals(), getframeinfo(currentframe())):" to demarcate the blocks of
processing steps that take a long time (hours/days).
Example of initialization:
program_state_stack.PROGRAM_STATE_VARIABLES = {"local_var_i", "local_var_j", "local_var_h", "local_var_g"}
program_state_stack(locals(), getframeinfo(currentframe()), "my_state.json")
Then regular usage in the data analysis program:
if program_state_stack(locals(), getframeinfo(currentframe())):
data_analysis_1()
if program_state_stack(locals(), getframeinfo(currentframe())):
data_analysis_2()
"""
import os
from traceback import extract_stack
from mpi import mpi_comm_rank, mpi_bcast, MPI_COMM_WORLD, MPI_INT
from inspect import currentframe, getframeinfo
def get_current_stack_info():
return [[x[0], x[2]] for x in extract_stack()[:-2]]
START_EXECUTING_FALSE = 0
START_EXECUTING_TRUE = 1
START_EXECUTING_ONLY_ONE_TIME_THEN_REVERT = 2
current_state = dict()
for var in program_state_stack.PROGRAM_STATE_VARIABLES & set(full_current_state) :
current_state[var] = full_current_state[var]
if "restart_location_title" in program_state_stack.__dict__:
location_in_program = frame_info.filename + "___" + program_state_stack.restart_location_title
del program_state_stack.restart_location_title
else:
location_in_program = frame_info.filename + "___" + str(frame_info.lineno) + "_" + last_call
current_state["location_in_program"] = location_in_program
current_stack = get_current_stack_info()
error_status = 0
# not a real while, an if with the possibility of jumping with break
while mpi_comm_rank(MPI_COMM_WORLD) == 0:
if "file_name_of_saved_state" not in program_state_stack.__dict__:
if type(file_name_of_saved_state) != type(""):
error_status = ("Must provide the file name of saved state as a string in the first call of the function!", getframeinfo(currentframe()))
break
program_state_stack.file_name_of_saved_state = os.getcwd() + os.sep + file_name_of_saved_state
program_state_stack.counter = 0
program_state_stack.track_stack = get_current_stack_info()
program_state_stack.track_state = [dict() for i in xrange(len(program_state_stack.track_stack))]
program_state_stack.track_state[-1] = current_state
file_name_of_saved_state_contains_information = False
if (os.path.exists(file_name_of_saved_state)):
statinfo = os.stat(file_name_of_saved_state)
file_name_of_saved_state_contains_information = statinfo.st_size > 0
if file_name_of_saved_state_contains_information:
program_state_stack.saved_stack, \
program_state_stack.saved_state = restore_program_stack_and_state(file_name_of_saved_state)
program_state_stack.start_executing = START_EXECUTING_FALSE
else:
# check to see if file can be created
f = open(file_name_of_saved_state, "w"); f.close()
program_state_stack.start_executing = START_EXECUTING_TRUE
else:
program_state_stack.counter += 1
# print "counter: ", program_state_stack.counter
if program_state_stack.counter == program_state_stack.CCC:
# error_status = ("Reached %d calls!"%program_state_stack.CCC, getframeinfo(currentframe()))
error_status = 1
break
if program_state_stack.start_executing == START_EXECUTING_ONLY_ONE_TIME_THEN_REVERT:
program_state_stack.start_executing = START_EXECUTING_FALSE
# correct track_state to reflect track_stack
for i in xrange(len(current_stack)):
if i < len(program_state_stack.track_state):
if program_state_stack.track_stack[i] != current_stack[i]:
program_state_stack.track_state[i] = dict()
else:
program_state_stack.track_state.append(dict())
program_state_stack.track_state[i] = current_state
# correct track_stack to reflect current_stack
program_state_stack.track_stack = current_stack
# delete additional elements in track_state so that size of track_state is the same as current_stack
program_state_stack.track_state[len(current_stack):len(program_state_stack.track_state)] = []
if program_state_stack.start_executing == START_EXECUTING_TRUE or last_call != "" or force_starting_execution:
store_program_state(program_state_stack.file_name_of_saved_state, program_state_stack.track_state, current_stack)
program_state_stack.start_executing = START_EXECUTING_TRUE
else:
if len(program_state_stack.saved_state) >= len(current_stack):
for i in range(len(program_state_stack.saved_state)):
if i < len(current_stack):
if program_state_stack.track_stack[i] == current_stack[i]:
if program_state_stack.track_state[i] == program_state_stack.saved_state[i]:
continue
break
else:
program_state_stack.start_executing = START_EXECUTING_ONLY_ONE_TIME_THEN_REVERT
# print "////////////////////////////"
# print "Entering function: ", location_in_program
# print "////////////////////////////"
break
else:
program_state_stack.start_executing = START_EXECUTING_TRUE
# print "////////////////////////////"
# print "Start executing: ", location_in_program
# print "////////////////////////////"
break
else:
## needs to be initialized for all processes except master
program_state_stack.start_executing = START_EXECUTING_FALSE
if_error_then_all_processes_exit_program(error_status)
program_state_stack.start_executing = mpi_bcast(program_state_stack.start_executing, 1, MPI_INT, 0, MPI_COMM_WORLD)
program_state_stack.start_executing = int(program_state_stack.start_executing[0])
return program_state_stack.start_executing
def resample( prjfile, outdir, bufprefix, nbufvol, nvol, seedbase,\
delta, d, snr, CTF, npad,\
MPI, myid, ncpu, verbose = 0 ):
from utilities import even_angles
from random import seed, jumpahead, shuffle
import os
from sys import exit
nprj = EMUtil.get_image_count( prjfile )
if MPI:
from mpi import mpi_barrier, MPI_COMM_WORLD
if myid == 0:
if os.path.exists(outdir): nx = 1
else: nx = 0
else: nx = 0
ny = bcast_number_to_all(nx, source_node = 0)
if ny == 1: ERROR('Output directory exists, please change the name and restart the program', "resample", 1,myid)
mpi_barrier(MPI_COMM_WORLD)
if myid == 0:
os.mkdir(outdir)
mpi_barrier(MPI_COMM_WORLD)
else:
if os.path.exists(outdir):
ERROR('Output directory exists, please change the name and restart the program', "resample", 1,0)
os.mkdir(outdir)
if(verbose == 1): finfo=open( os.path.join(outdir, "progress%04d.txt" % myid), "w" )
else: finfo = None
#print " before evenangles",myid
from utilities import getvec
from numpy import array, reshape
refa = even_angles(delta)
nrefa = len(refa)
refnormal = zeros((nrefa,3),'float32')
tetref = [0.0]*nrefa
for i in xrange(nrefa):
tr = getvec( refa[i][0], refa[i][1] )
for j in xrange(3): refnormal[i][j] = tr[j]
tetref[i] = refa[i][1]
del refa
vct = array([0.0]*(3*nprj),'float32')
if myid == 0:
print " will read ",myid
tr = EMUtil.get_all_attributes(prjfile,'xform.projection')
tetprj = [0.0]*nprj
for i in xrange(nprj):
temp = tr[i].get_params("spider")
tetprj[i] = temp["theta"]
if(tetprj[i] > 90.0): tetprj[i] = 180.0 - tetprj[i]
vct[3*i+0] = tr[i].at(2,0)
vct[3*i+1] = tr[i].at(2,1)
vct[3*i+2] = tr[i].at(2,2)
del tr
else:
tetprj = [0.0]*nprj
#print " READ ",myid
if MPI:
#print " will bcast",myid
from mpi import mpi_bcast, MPI_FLOAT, MPI_COMM_WORLD
vct = mpi_bcast(vct,len(vct),MPI_FLOAT,0,MPI_COMM_WORLD)
from utilities import bcast_list_to_all
tetprj = bcast_list_to_all(tetprj, myid, 0)
#print " reshape ",myid
vct = reshape(vct,(nprj,3))
assignments = [[] for i in xrange(nrefa)]
dspn = 1.25*delta
for k in xrange(nprj):
best_s = -1.0
best_i = -1
for i in xrange( nrefa ):
if(abs(tetprj[k] - tetref[i]) <= dspn):
s = abs(refnormal[i][0]*vct[k][0] + refnormal[i][1]*vct[k][1] + refnormal[i][2]*vct[k][2])
if s > best_s:
best_s = s
best_i = i
assignments[best_i].append(k)
am = len(assignments[0])
mufur = 1.0/am
for i in xrange(1,len(assignments)):
ti = len(assignments[i])
am = min(am, ti)
if(ti>0): mufur += 1.0/ti
del tetprj,tetref
dp = 1.0 - d # keep that many in each direction
keep = int(am*dp +0.5)
mufur = keep*nrefa/(1.0 - mufur*keep/float(nrefa))
if myid == 0:
print " Number of projections ",nprj,". Number of reference directions ",nrefa,", multiplicative factor for the variance ",mufur
print " Minimum number of assignments ",am," Number of projections used per stratum ", keep," Number of projections in resampled structure ",int(am*dp +0.5)*nrefa
if am <2 or am == keep:
print "incorrect settings"
exit() # FIX
if(seedbase < 1):
seed()
jumpahead(17*myid+123)
else:
seed(seedbase)
jumpahead(17*myid+123)
volfile = os.path.join(outdir, "bsvol%04d.hdf" % myid)
from random import randint
niter = nvol/ncpu/nbufvol
for kiter in xrange(niter):
if(verbose == 1):
finfo.write( "Iteration %d: \n" % kiter )
finfo.flush()
iter_start = time()
# the following has to be converted to resample mults=1 means take given projection., mults=0 means omit
mults = [ [0]*nprj for i in xrange(nbufvol) ]
for i in xrange(nbufvol):
for l in xrange(nrefa):
mass = assignments[l][:]
shuffle(mass)
mass = mass[:keep]
mass.sort()
#print l, " * ",mass
for k in xrange(keep):
mults[i][mass[k]] = 1
'''
lout = []
for l in xrange(len(mults[i])):
if mults[i][l] == 1: lout.append(l)
write_text_file(lout, os.path.join(outdir, "list%04d_%03d.txt" %(i, myid)))
del lout
'''
del mass
rectors, fftvols, wgtvols = resample_prepare( prjfile, nbufvol, snr, CTF, npad )
resample_insert( bufprefix, fftvols, wgtvols, mults, CTF, npad, finfo )
del mults
resample_finish( rectors, fftvols, wgtvols, volfile, kiter, nprj, finfo )
rectors = None
fftvols = None
wgtvols = None
if(verbose == 1):
finfo.write( "time for iteration: %10.3f\n" % (time() - iter_start) )
finfo.flush()
from Numeric import *
import mpi
import sys
import math
#print "before",len(sys.argv),sys.argv
sys.argv = mpi.mpi_init(len(sys.argv), sys.argv)
#print "after ",len(sys.argv),sys.argv
myid = mpi.mpi_comm_rank(mpi.MPI_COMM_WORLD)
numnodes = mpi.mpi_comm_size(mpi.MPI_COMM_WORLD)
print "hello from ", myid, " of ", numnodes
color = myid % 2
new_comm = mpi.mpi_comm_split(mpi.MPI_COMM_WORLD, color, myid)
new_id = mpi.mpi_comm_rank(new_comm)
new_nodes = mpi.mpi_comm_size(new_comm)
zero_one = -1
if new_id == 0:
zero_one = color
zero_one = mpi.mpi_bcast(zero_one, 1, mpi.MPI_INT, 0, new_comm)
if zero_one == 0:
print myid, " part of even processor communicator ", new_id
if zero_one == 1:
print myid, " part of odd processor communicator ", new_id
print "old_id=", myid, "new_id=", new_id
mpi.mpi_finalize()
##### start up remote tasks ####
toRun = getcwd() + "/worker.py"
print mpi.mpi_get_processor_name(), "starting", toRun
newcom1 = mpi.mpi_comm_spawn(toRun, "from_P_", copies, mpi.MPI_INFO_NULL, 0,
mpi.MPI_COMM_WORLD)
errors = mpi.mpi_array_of_errcodes()
print "errors=", errors
newcom1Size = mpi.mpi_comm_size(newcom1)
print "newcom1Size", newcom1Size, " yes it is strange but it should be 1"
##### bcast ####
x = array(([1, 2, 3, 4]), "i")
count = 4
print "head starting bcast", x
junk = mpi.mpi_bcast(x, count, mpi.MPI_INT, mpi.MPI_ROOT, newcom1)
print "head did bcast"
##### scatter ####
scat = array([10, 20, 30], "i")
junk = mpi.mpi_scatter(scat, 1, mpi.MPI_INT, 1, mpi.MPI_INT, mpi.MPI_ROOT,
newcom1)
##### send/recv ####
for i in range(0, copies):
k = (i + 1) * 100
mpi.mpi_send(k, 1, mpi.MPI_INT, i, 1234, newcom1)
back = mpi.mpi_recv(1, mpi.MPI_INT, i, 5678, newcom1)
print "from ", i, back
##### reduce ####
sys.argv = mpi.mpi_init(len(sys.argv), sys.argv) myid = mpi.mpi_comm_rank(mpi.MPI_COMM_WORLD) numprocs = mpi.mpi_comm_size(mpi.MPI_COMM_WORLD) parent = mpi.mpi_comm_get_parent() parentSize = mpi.mpi_comm_size(parent) print "parentSize", parentSize tod = stamp() s = sys.argv[1] + "%2.2d" % myid print "hello from python worker", myid, " writing to ", s x = array([5, 3, 4, 2], 'i') print "starting bcast" buffer = mpi.mpi_bcast(x, 4, mpi.MPI_INT, 0, parent) out = open(s, "w") out.write(str(buffer)) out.write(tod + "\n") out.close() print myid, " got ", buffer junk = mpi.mpi_scatter(x, 1, mpi.MPI_INT, 1, mpi.MPI_INT, 0, parent) print myid, " got scatter ", junk back = mpi.mpi_recv(1, mpi.MPI_INT, 0, 1234, parent) back[0] = back[0] + 1 mpi.mpi_send(back, 1, mpi.MPI_INT, 0, 5678, parent) dummy = myid final = mpi.mpi_reduce(dummy, 1, mpi.MPI_INT, mpi.MPI_SUM, 0, parent)
