def prepare_outdir_log(outdir='.', verbose=False, is_main=True):
"""
Prepares output directory and sets up log file.
Arguments:
outdir : Output directory
verbose : (boolean) Whether to write to screen
is_main : (boolean) If using multiple cores, some tasks only need to be performed once, not by all cores
Returns:
log : instance of Logger class
verbose : (boolean) New version of Logger can write to screen simultaneously
"""
# Create directory if it doesn't exist
if is_main:
if os.path.isdir(outdir):
sxprint("Writing to output directory: %s" % outdir)
else:
sxprint("Created output directory: %s" % outdir)
os.makedirs(
outdir) # os.mkdir() can only operate one directory deep
sp_global_def.write_command(outdir)
logname = "log_" + datetime.now().strftime("%Y%m%d_%H%M%S") + ".txt"
logname = os.path.join(outdir, logname)
#if global_def.LOGFILE:
#global_def.LOGFILE = logname
#print('LOGFILE', global_def.LOGFILE)
#exit()
# May be using old version of logger.py
try:
if verbose:
log = Logger(base_logger=BaseLogger_Files(),
base_logger2=BaseLogger_Print(),
file_name=logname)
verbose = False # logger output will be echoed to screen
else:
log = Logger(base_logger=BaseLogger_Files(), file_name=logname)
except TypeError:
if is_main: sxprint("WARNING: Using old sp_logger.py library")
log = Logger(base_logger=BaseLogger_Files()) #, file_name=logname)
logname = 'log.txt'
if is_main: sxprint("Writing log file to %s\n" % logname)
if is_main:
progbase = os.path.basename(__file__).split('.')[0].upper()
#print('progbase', progbase)
#exit()
length = len(progbase) + 4
log.add("\n" + " " * TIMESTAMP_LENGTH + "*" * length + "\n" +
" " * TIMESTAMP_LENGTH + "* " + progbase + " *\n" +
" " * TIMESTAMP_LENGTH + "*" * length)
return log, verbose
def prepare_log(outdir='.', verbose=False, main=True):
"""
Prepare log file.
Arguments:
outdir : Output directory
verbose : (boolean) Whether to write additional information to screen
main : (boolean) Whether main MPI process
Returns:
log : Instance of Logger class
verbose : (boolean) Updates flag to False if Logger class can mirror output to screen
"""
TIMESTAMP_LENGTH = 23
logname = "log_" + datetime.now().strftime("%Y%m%d_%H%M%S") + ".txt"
logname = os.path.join(outdir, logname)
# May be using old version of logger.py
try:
if verbose:
log = Logger(base_logger=BaseLogger_Files(),
base_logger2=BaseLogger_Print(),
file_name=logname)
verbose = False # logger output will be echoed to screen
else:
log = Logger(base_logger=BaseLogger_Files(), file_name=logname)
except TypeError:
print("WARNING: Using old sp_logger.py library")
log = Logger(base_logger=BaseLogger_Files()) #, file_name=logname)
logname = 'log.txt'
print("Writing log file to %s" % logname)
if main:
progbase = os.path.basename(__file__).split('.')[0].upper()
length = len(progbase) + 4
log.add("\n" + " " * TIMESTAMP_LENGTH + "*" * length + "\n" +
" " * TIMESTAMP_LENGTH + "* " + progbase + " *\n" +
" " * TIMESTAMP_LENGTH + "*" * length)
return log, verbose
def main():
def params_3D_2D_NEW(phi, theta, psi, s2x, s2y, mirror):
# the final ali2d parameters already combine shifts operation first and rotation operation second for parameters converted from 3D
if mirror:
m = 1
alpha, sx, sy, scalen = compose_transform2(0, s2x, s2y, 1.0,
540.0 - psi, 0, 0, 1.0)
else:
m = 0
alpha, sx, sy, scalen = compose_transform2(0, s2x, s2y, 1.0,
360.0 - psi, 0, 0, 1.0)
return alpha, sx, sy, m
progname = os.path.basename(sys.argv[0])
usage = progname + " prj_stack --ave2D= --var2D= --ave3D= --var3D= --img_per_grp= --fl= --aa= --sym=symmetry --CTF"
parser = OptionParser(usage, version=SPARXVERSION)
parser.add_option("--output_dir",
type="string",
default="./",
help="Output directory")
parser.add_option("--ave2D",
type="string",
default=False,
help="Write to the disk a stack of 2D averages")
parser.add_option("--var2D",
type="string",
default=False,
help="Write to the disk a stack of 2D variances")
parser.add_option("--ave3D",
type="string",
default=False,
help="Write to the disk reconstructed 3D average")
parser.add_option("--var3D",
type="string",
default=False,
help="Compute 3D variability (time consuming!)")
parser.add_option(
"--img_per_grp",
type="int",
default=100,
help="Number of neighbouring projections.(Default is 100)")
parser.add_option(
"--no_norm",
action="store_true",
default=False,
help="Do not use normalization.(Default is to apply normalization)")
#parser.add_option("--radius", type="int" , default=-1 , help="radius for 3D variability" )
parser.add_option(
"--npad",
type="int",
default=2,
help=
"Number of time to pad the original images.(Default is 2 times padding)"
)
parser.add_option("--sym",
type="string",
default="c1",
help="Symmetry. (Default is no symmetry)")
parser.add_option(
"--fl",
type="float",
default=0.0,
help=
"Low pass filter cutoff in absolute frequency (0.0 - 0.5) and is applied to decimated images. (Default - no filtration)"
)
parser.add_option(
"--aa",
type="float",
default=0.02,
help=
"Fall off of the filter. Use default value if user has no clue about falloff (Default value is 0.02)"
)
parser.add_option("--CTF",
action="store_true",
default=False,
help="Use CFT correction.(Default is no CTF correction)")
#parser.add_option("--MPI" , action="store_true", default=False, help="use MPI version")
#parser.add_option("--radiuspca", type="int" , default=-1 , help="radius for PCA" )
#parser.add_option("--iter", type="int" , default=40 , help="maximum number of iterations (stop criterion of reconstruction process)" )
#parser.add_option("--abs", type="float" , default=0.0 , help="minimum average absolute change of voxels' values (stop criterion of reconstruction process)" )
#parser.add_option("--squ", type="float" , default=0.0 , help="minimum average squared change of voxels' values (stop criterion of reconstruction process)" )
parser.add_option(
"--VAR",
action="store_true",
default=False,
help="Stack of input consists of 2D variances (Default False)")
parser.add_option(
"--decimate",
type="float",
default=0.25,
help="Image decimate rate, a number less than 1. (Default is 0.25)")
parser.add_option(
"--window",
type="int",
default=0,
help=
"Target image size relative to original image size. (Default value is zero.)"
)
#parser.add_option("--SND", action="store_true", default=False, help="compute squared normalized differences (Default False)")
#parser.add_option("--nvec", type="int" , default=0 , help="Number of eigenvectors, (Default = 0 meaning no PCA calculated)")
parser.add_option(
"--symmetrize",
action="store_true",
default=False,
help="Prepare input stack for handling symmetry (Default False)")
parser.add_option("--overhead",
type="float",
default=0.5,
help="python overhead per CPU.")
(options, args) = parser.parse_args()
#####
from mpi import mpi_comm_rank, mpi_comm_size, mpi_recv, MPI_COMM_WORLD
from mpi import mpi_barrier, mpi_reduce, mpi_bcast, mpi_send, MPI_FLOAT, MPI_SUM, MPI_INT, MPI_MAX
#from mpi import *
from sp_applications import MPI_start_end
from sp_reconstruction import recons3d_em, recons3d_em_MPI
from sp_reconstruction import recons3d_4nn_MPI, recons3d_4nn_ctf_MPI
from sp_utilities import print_begin_msg, print_end_msg, print_msg
from sp_utilities import read_text_row, get_image, get_im, wrap_mpi_send, wrap_mpi_recv
from sp_utilities import bcast_EMData_to_all, bcast_number_to_all
from sp_utilities import get_symt
# This is code for handling symmetries by the above program. To be incorporated. PAP 01/27/2015
from EMAN2db import db_open_dict
# Set up global variables related to bdb cache
if sp_global_def.CACHE_DISABLE:
from sp_utilities import disable_bdb_cache
disable_bdb_cache()
# Set up global variables related to ERROR function
sp_global_def.BATCH = True
# detect if program is running under MPI
RUNNING_UNDER_MPI = "OMPI_COMM_WORLD_SIZE" in os.environ
if RUNNING_UNDER_MPI: sp_global_def.MPI = True
if options.output_dir == "./":
current_output_dir = os.path.abspath(options.output_dir)
else:
current_output_dir = options.output_dir
if options.symmetrize:
if mpi.mpi_comm_size(MPI_COMM_WORLD) > 1:
ERROR("Cannot use more than one CPU for symmetry preparation")
if not os.path.exists(current_output_dir):
os.makedirs(current_output_dir)
sp_global_def.write_command(current_output_dir)
from sp_logger import Logger, BaseLogger_Files
if os.path.exists(os.path.join(current_output_dir, "log.txt")):
os.remove(os.path.join(current_output_dir, "log.txt"))
log_main = Logger(BaseLogger_Files())
log_main.prefix = os.path.join(current_output_dir, "./")
instack = args[0]
sym = options.sym.lower()
if (sym == "c1"):
ERROR("There is no need to symmetrize stack for C1 symmetry")
line = ""
for a in sys.argv:
line += " " + a
log_main.add(line)
if (instack[:4] != "bdb:"):
#if output_dir =="./": stack = "bdb:data"
stack = "bdb:" + current_output_dir + "/data"
delete_bdb(stack)
junk = cmdexecute("sp_cpy.py " + instack + " " + stack)
else:
stack = instack
qt = EMUtil.get_all_attributes(stack, 'xform.projection')
na = len(qt)
ts = get_symt(sym)
ks = len(ts)
angsa = [None] * na
for k in range(ks):
#Qfile = "Q%1d"%k
#if options.output_dir!="./": Qfile = os.path.join(options.output_dir,"Q%1d"%k)
Qfile = os.path.join(current_output_dir, "Q%1d" % k)
#delete_bdb("bdb:Q%1d"%k)
delete_bdb("bdb:" + Qfile)
#junk = cmdexecute("e2bdb.py "+stack+" --makevstack=bdb:Q%1d"%k)
junk = cmdexecute("e2bdb.py " + stack + " --makevstack=bdb:" +
Qfile)
#DB = db_open_dict("bdb:Q%1d"%k)
DB = db_open_dict("bdb:" + Qfile)
for i in range(na):
ut = qt[i] * ts[k]
DB.set_attr(i, "xform.projection", ut)
#bt = ut.get_params("spider")
#angsa[i] = [round(bt["phi"],3)%360.0, round(bt["theta"],3)%360.0, bt["psi"], -bt["tx"], -bt["ty"]]
#write_text_row(angsa, 'ptsma%1d.txt'%k)
#junk = cmdexecute("e2bdb.py "+stack+" --makevstack=bdb:Q%1d"%k)
#junk = cmdexecute("sxheader.py bdb:Q%1d --params=xform.projection --import=ptsma%1d.txt"%(k,k))
DB.close()
#if options.output_dir =="./": delete_bdb("bdb:sdata")
delete_bdb("bdb:" + current_output_dir + "/" + "sdata")
#junk = cmdexecute("e2bdb.py . --makevstack=bdb:sdata --filt=Q")
sdata = "bdb:" + current_output_dir + "/" + "sdata"
sxprint(sdata)
junk = cmdexecute("e2bdb.py " + current_output_dir +
" --makevstack=" + sdata + " --filt=Q")
#junk = cmdexecute("ls EMAN2DB/sdata*")
#a = get_im("bdb:sdata")
a = get_im(sdata)
a.set_attr("variabilitysymmetry", sym)
#a.write_image("bdb:sdata")
a.write_image(sdata)
else:
from sp_fundamentals import window2d
myid = mpi_comm_rank(MPI_COMM_WORLD)
number_of_proc = mpi_comm_size(MPI_COMM_WORLD)
main_node = 0
shared_comm = mpi_comm_split_type(MPI_COMM_WORLD, MPI_COMM_TYPE_SHARED,
0, MPI_INFO_NULL)
myid_on_node = mpi_comm_rank(shared_comm)
no_of_processes_per_group = mpi_comm_size(shared_comm)
masters_from_groups_vs_everything_else_comm = mpi_comm_split(
MPI_COMM_WORLD, main_node == myid_on_node, myid_on_node)
color, no_of_groups, balanced_processor_load_on_nodes = get_colors_and_subsets(main_node, MPI_COMM_WORLD, myid, \
shared_comm, myid_on_node, masters_from_groups_vs_everything_else_comm)
overhead_loading = options.overhead * number_of_proc
#memory_per_node = options.memory_per_node
#if memory_per_node == -1.: memory_per_node = 2.*no_of_processes_per_group
keepgoing = 1
current_window = options.window
current_decimate = options.decimate
if len(args) == 1: stack = args[0]
else:
sxprint("Usage: " + usage)
sxprint("Please run \'" + progname + " -h\' for detailed options")
ERROR(
"Invalid number of parameters used. Please see usage information above."
)
return
t0 = time()
# obsolete flags
options.MPI = True
#options.nvec = 0
options.radiuspca = -1
options.iter = 40
options.abs = 0.0
options.squ = 0.0
if options.fl > 0.0 and options.aa == 0.0:
ERROR("Fall off has to be given for the low-pass filter",
myid=myid)
#if options.VAR and options.SND:
# ERROR( "Only one of var and SND can be set!",myid=myid )
if options.VAR and (options.ave2D or options.ave3D or options.var2D):
ERROR(
"When VAR is set, the program cannot output ave2D, ave3D or var2D",
myid=myid)
#if options.SND and (options.ave2D or options.ave3D):
# ERROR( "When SND is set, the program cannot output ave2D or ave3D", myid=myid )
#if options.nvec > 0 :
# ERROR( "PCA option not implemented", myid=myid )
#if options.nvec > 0 and options.ave3D == None:
# ERROR( "When doing PCA analysis, one must set ave3D", myid=myid )
if current_decimate > 1.0 or current_decimate < 0.0:
ERROR("Decimate rate should be a value between 0.0 and 1.0",
myid=myid)
if current_window < 0.0:
ERROR("Target window size should be always larger than zero",
myid=myid)
if myid == main_node:
img = get_image(stack, 0)
nx = img.get_xsize()
ny = img.get_ysize()
if (min(nx, ny) < current_window): keepgoing = 0
keepgoing = bcast_number_to_all(keepgoing, main_node, MPI_COMM_WORLD)
if keepgoing == 0:
ERROR(
"The target window size cannot be larger than the size of decimated image",
myid=myid)
import string
options.sym = options.sym.lower()
# if global_def.CACHE_DISABLE:
# from utilities import disable_bdb_cache
# disable_bdb_cache()
# global_def.BATCH = True
if myid == main_node:
if not os.path.exists(current_output_dir):
os.makedirs(current_output_dir
) # Never delete output_dir in the program!
img_per_grp = options.img_per_grp
#nvec = options.nvec
radiuspca = options.radiuspca
from sp_logger import Logger, BaseLogger_Files
#if os.path.exists(os.path.join(options.output_dir, "log.txt")): os.remove(os.path.join(options.output_dir, "log.txt"))
log_main = Logger(BaseLogger_Files())
log_main.prefix = os.path.join(current_output_dir, "./")
if myid == main_node:
line = ""
for a in sys.argv:
line += " " + a
log_main.add(line)
log_main.add("-------->>>Settings given by all options<<<-------")
log_main.add("Symmetry : %s" % options.sym)
log_main.add("Input stack : %s" % stack)
log_main.add("Output_dir : %s" % current_output_dir)
if options.ave3D:
log_main.add("Ave3d : %s" % options.ave3D)
if options.var3D:
log_main.add("Var3d : %s" % options.var3D)
if options.ave2D:
log_main.add("Ave2D : %s" % options.ave2D)
if options.var2D:
log_main.add("Var2D : %s" % options.var2D)
if options.VAR: log_main.add("VAR : True")
else: log_main.add("VAR : False")
if options.CTF: log_main.add("CTF correction : True ")
else: log_main.add("CTF correction : False ")
log_main.add("Image per group : %5d" % options.img_per_grp)
log_main.add("Image decimate rate : %4.3f" % current_decimate)
log_main.add("Low pass filter : %4.3f" % options.fl)
current_fl = options.fl
if current_fl == 0.0: current_fl = 0.5
log_main.add(
"Current low pass filter is equivalent to cutoff frequency %4.3f for original image size"
% round((current_fl * current_decimate), 3))
log_main.add("Window size : %5d " % current_window)
log_main.add("sx3dvariability begins")
symbaselen = 0
if myid == main_node:
nima = EMUtil.get_image_count(stack)
img = get_image(stack)
nx = img.get_xsize()
ny = img.get_ysize()
nnxo = nx
nnyo = ny
if options.sym != "c1":
imgdata = get_im(stack)
try:
i = imgdata.get_attr("variabilitysymmetry").lower()
if (i != options.sym):
ERROR(
"The symmetry provided does not agree with the symmetry of the input stack",
myid=myid)
except:
ERROR(
"Input stack is not prepared for symmetry, please follow instructions",
myid=myid)
from sp_utilities import get_symt
i = len(get_symt(options.sym))
if ((nima / i) * i != nima):
ERROR(
"The length of the input stack is incorrect for symmetry processing",
myid=myid)
symbaselen = nima / i
else:
symbaselen = nima
else:
nima = 0
nx = 0
ny = 0
nnxo = 0
nnyo = 0
nima = bcast_number_to_all(nima)
nx = bcast_number_to_all(nx)
ny = bcast_number_to_all(ny)
nnxo = bcast_number_to_all(nnxo)
nnyo = bcast_number_to_all(nnyo)
if current_window > max(nx, ny):
ERROR("Window size is larger than the original image size")
if current_decimate == 1.:
if current_window != 0:
nx = current_window
ny = current_window
else:
if current_window == 0:
nx = int(nx * current_decimate + 0.5)
ny = int(ny * current_decimate + 0.5)
else:
nx = int(current_window * current_decimate + 0.5)
ny = nx
symbaselen = bcast_number_to_all(symbaselen)
# check FFT prime number
from sp_fundamentals import smallprime
is_fft_friendly = (nx == smallprime(nx))
if not is_fft_friendly:
if myid == main_node:
log_main.add(
"The target image size is not a product of small prime numbers"
)
log_main.add("Program adjusts the input settings!")
### two cases
if current_decimate == 1.:
nx = smallprime(nx)
ny = nx
current_window = nx # update
if myid == main_node:
log_main.add("The window size is updated to %d." %
current_window)
else:
if current_window == 0:
nx = smallprime(int(nx * current_decimate + 0.5))
current_decimate = float(nx) / nnxo
ny = nx
if (myid == main_node):
log_main.add("The decimate rate is updated to %f." %
current_decimate)
else:
nx = smallprime(
int(current_window * current_decimate + 0.5))
ny = nx
current_window = int(nx / current_decimate + 0.5)
if (myid == main_node):
log_main.add("The window size is updated to %d." %
current_window)
if myid == main_node:
log_main.add("The target image size is %d" % nx)
if radiuspca == -1: radiuspca = nx / 2 - 2
if myid == main_node:
log_main.add("%-70s: %d\n" % ("Number of projection", nima))
img_begin, img_end = MPI_start_end(nima, number_of_proc, myid)
"""
if options.SND:
from sp_projection import prep_vol, prgs
from sp_statistics import im_diff
from sp_utilities import get_im, model_circle, get_params_proj, set_params_proj
from sp_utilities import get_ctf, generate_ctf
from sp_filter import filt_ctf
imgdata = EMData.read_images(stack, range(img_begin, img_end))
if options.CTF:
vol = recons3d_4nn_ctf_MPI(myid, imgdata, 1.0, symmetry=options.sym, npad=options.npad, xysize=-1, zsize=-1)
else:
vol = recons3d_4nn_MPI(myid, imgdata, symmetry=options.sym, npad=options.npad, xysize=-1, zsize=-1)
bcast_EMData_to_all(vol, myid)
volft, kb = prep_vol(vol)
mask = model_circle(nx/2-2, nx, ny)
varList = []
for i in xrange(img_begin, img_end):
phi, theta, psi, s2x, s2y = get_params_proj(imgdata[i-img_begin])
ref_prj = prgs(volft, kb, [phi, theta, psi, -s2x, -s2y])
if options.CTF:
ctf_params = get_ctf(imgdata[i-img_begin])
ref_prj = filt_ctf(ref_prj, generate_ctf(ctf_params))
diff, A, B = im_diff(ref_prj, imgdata[i-img_begin], mask)
diff2 = diff*diff
set_params_proj(diff2, [phi, theta, psi, s2x, s2y])
varList.append(diff2)
mpi_barrier(MPI_COMM_WORLD)
"""
if options.VAR: # 2D variance images have no shifts
#varList = EMData.read_images(stack, range(img_begin, img_end))
from EMAN2 import Region
for index_of_particle in range(img_begin, img_end):
image = get_im(stack, index_of_proj)
if current_window > 0:
varList.append(
fdecimate(
window2d(image, current_window, current_window),
nx, ny))
else:
varList.append(fdecimate(image, nx, ny))
else:
from sp_utilities import bcast_number_to_all, bcast_list_to_all, send_EMData, recv_EMData
from sp_utilities import set_params_proj, get_params_proj, params_3D_2D, get_params2D, set_params2D, compose_transform2
from sp_utilities import model_blank, nearest_proj, model_circle, write_text_row, wrap_mpi_gatherv
from sp_applications import pca
from sp_statistics import avgvar, avgvar_ctf, ccc
from sp_filter import filt_tanl
from sp_morphology import threshold, square_root
from sp_projection import project, prep_vol, prgs
from sets import Set
from sp_utilities import wrap_mpi_recv, wrap_mpi_bcast, wrap_mpi_send
import numpy as np
if myid == main_node:
t1 = time()
proj_angles = []
aveList = []
tab = EMUtil.get_all_attributes(stack, 'xform.projection')
for i in range(nima):
t = tab[i].get_params('spider')
phi = t['phi']
theta = t['theta']
psi = t['psi']
x = theta
if x > 90.0: x = 180.0 - x
x = x * 10000 + psi
proj_angles.append([x, t['phi'], t['theta'], t['psi'], i])
t2 = time()
log_main.add(
"%-70s: %d\n" %
("Number of neighboring projections", img_per_grp))
log_main.add("...... Finding neighboring projections\n")
log_main.add("Number of images per group: %d" % img_per_grp)
log_main.add("Now grouping projections")
proj_angles.sort()
proj_angles_list = np.full((nima, 4), 0.0, dtype=np.float32)
for i in range(nima):
proj_angles_list[i][0] = proj_angles[i][1]
proj_angles_list[i][1] = proj_angles[i][2]
proj_angles_list[i][2] = proj_angles[i][3]
proj_angles_list[i][3] = proj_angles[i][4]
else:
proj_angles_list = 0
proj_angles_list = wrap_mpi_bcast(proj_angles_list, main_node,
MPI_COMM_WORLD)
proj_angles = []
for i in range(nima):
proj_angles.append([
proj_angles_list[i][0], proj_angles_list[i][1],
proj_angles_list[i][2],
int(proj_angles_list[i][3])
])
del proj_angles_list
proj_list, mirror_list = nearest_proj(proj_angles, img_per_grp,
range(img_begin, img_end))
all_proj = Set()
for im in proj_list:
for jm in im:
all_proj.add(proj_angles[jm][3])
all_proj = list(all_proj)
index = {}
for i in range(len(all_proj)):
index[all_proj[i]] = i
mpi_barrier(MPI_COMM_WORLD)
if myid == main_node:
log_main.add("%-70s: %.2f\n" %
("Finding neighboring projections lasted [s]",
time() - t2))
log_main.add("%-70s: %d\n" %
("Number of groups processed on the main node",
len(proj_list)))
log_main.add("Grouping projections took: %12.1f [m]" %
((time() - t2) / 60.))
log_main.add("Number of groups on main node: ", len(proj_list))
mpi_barrier(MPI_COMM_WORLD)
if myid == main_node:
log_main.add("...... Calculating the stack of 2D variances \n")
# Memory estimation. There are two memory consumption peaks
# peak 1. Compute ave, var;
# peak 2. Var volume reconstruction;
# proj_params = [0.0]*(nima*5)
aveList = []
varList = []
#if nvec > 0: eigList = [[] for i in range(nvec)]
dnumber = len(
all_proj) # all neighborhood set for assigned to myid
pnumber = len(proj_list) * 2. + img_per_grp # aveList and varList
tnumber = dnumber + pnumber
vol_size2 = nx**3 * 4. * 8 / 1.e9
vol_size1 = 2. * nnxo**3 * 4. * 8 / 1.e9
proj_size = nnxo * nnyo * len(
proj_list) * 4. * 2. / 1.e9 # both aveList and varList
orig_data_size = nnxo * nnyo * 4. * tnumber / 1.e9
reduced_data_size = nx * nx * 4. * tnumber / 1.e9
full_data = np.full((number_of_proc, 2), -1., dtype=np.float16)
full_data[myid] = orig_data_size, reduced_data_size
if myid != main_node:
wrap_mpi_send(full_data, main_node, MPI_COMM_WORLD)
if myid == main_node:
for iproc in range(number_of_proc):
if iproc != main_node:
dummy = wrap_mpi_recv(iproc, MPI_COMM_WORLD)
full_data[np.where(dummy > -1)] = dummy[np.where(
dummy > -1)]
del dummy
mpi_barrier(MPI_COMM_WORLD)
full_data = wrap_mpi_bcast(full_data, main_node, MPI_COMM_WORLD)
# find the CPU with heaviest load
minindx = np.argsort(full_data, 0)
heavy_load_myid = minindx[-1][1]
total_mem = sum(full_data)
if myid == main_node:
if current_window == 0:
log_main.add(
"Nx: current image size = %d. Decimated by %f from %d"
% (nx, current_decimate, nnxo))
else:
log_main.add(
"Nx: current image size = %d. Windowed to %d, and decimated by %f from %d"
% (nx, current_window, current_decimate, nnxo))
log_main.add("Nproj: number of particle images.")
log_main.add("Navg: number of 2D average images.")
log_main.add("Nvar: number of 2D variance images.")
log_main.add(
"Img_per_grp: user defined image per group for averaging = %d"
% img_per_grp)
log_main.add(
"Overhead: total python overhead memory consumption = %f"
% overhead_loading)
log_main.add("Total memory) = 4.0*nx^2*(nproj + navg +nvar+ img_per_grp)/1.0e9 + overhead: %12.3f [GB]"%\
(total_mem[1] + overhead_loading))
del full_data
mpi_barrier(MPI_COMM_WORLD)
if myid == heavy_load_myid:
log_main.add(
"Begin reading and preprocessing images on processor. Wait... "
)
ttt = time()
#imgdata = EMData.read_images(stack, all_proj)
imgdata = [None for im in range(len(all_proj))]
for index_of_proj in range(len(all_proj)):
#image = get_im(stack, all_proj[index_of_proj])
if (current_window > 0):
imgdata[index_of_proj] = fdecimate(
window2d(get_im(stack, all_proj[index_of_proj]),
current_window, current_window), nx, ny)
else:
imgdata[index_of_proj] = fdecimate(
get_im(stack, all_proj[index_of_proj]), nx, ny)
if (current_decimate > 0.0 and options.CTF):
ctf = imgdata[index_of_proj].get_attr("ctf")
ctf.apix = ctf.apix / current_decimate
imgdata[index_of_proj].set_attr("ctf", ctf)
if myid == heavy_load_myid and index_of_proj % 100 == 0:
log_main.add(" ...... %6.2f%% " %
(index_of_proj / float(len(all_proj)) * 100.))
mpi_barrier(MPI_COMM_WORLD)
if myid == heavy_load_myid:
log_main.add("All_proj preprocessing cost %7.2f m" %
((time() - ttt) / 60.))
log_main.add("Wait untill reading on all CPUs done...")
'''
imgdata2 = EMData.read_images(stack, range(img_begin, img_end))
if options.fl > 0.0:
for k in xrange(len(imgdata2)):
imgdata2[k] = filt_tanl(imgdata2[k], options.fl, options.aa)
if options.CTF:
vol = recons3d_4nn_ctf_MPI(myid, imgdata2, 1.0, symmetry=options.sym, npad=options.npad, xysize=-1, zsize=-1)
else:
vol = recons3d_4nn_MPI(myid, imgdata2, symmetry=options.sym, npad=options.npad, xysize=-1, zsize=-1)
if myid == main_node:
vol.write_image("vol_ctf.hdf")
print_msg("Writing to the disk volume reconstructed from averages as : %s\n"%("vol_ctf.hdf"))
del vol, imgdata2
mpi_barrier(MPI_COMM_WORLD)
'''
from sp_applications import prepare_2d_forPCA
from sp_utilities import model_blank
from EMAN2 import Transform
if not options.no_norm:
mask = model_circle(nx / 2 - 2, nx, nx)
if options.CTF:
from sp_utilities import pad
from sp_filter import filt_ctf
from sp_filter import filt_tanl
if myid == heavy_load_myid:
log_main.add("Start computing 2D aveList and varList. Wait...")
ttt = time()
inner = nx // 2 - 4
outer = inner + 2
xform_proj_for_2D = [None for i in range(len(proj_list))]
for i in range(len(proj_list)):
ki = proj_angles[proj_list[i][0]][3]
if ki >= symbaselen: continue
mi = index[ki]
dpar = Util.get_transform_params(imgdata[mi],
"xform.projection", "spider")
phiM, thetaM, psiM, s2xM, s2yM = dpar["phi"], dpar[
"theta"], dpar[
"psi"], -dpar["tx"] * current_decimate, -dpar[
"ty"] * current_decimate
grp_imgdata = []
for j in range(img_per_grp):
mj = index[proj_angles[proj_list[i][j]][3]]
cpar = Util.get_transform_params(imgdata[mj],
"xform.projection",
"spider")
alpha, sx, sy, mirror = params_3D_2D_NEW(
cpar["phi"], cpar["theta"], cpar["psi"],
-cpar["tx"] * current_decimate,
-cpar["ty"] * current_decimate, mirror_list[i][j])
if thetaM <= 90:
if mirror == 0:
alpha, sx, sy, scale = compose_transform2(
alpha, sx, sy, 1.0, phiM - cpar["phi"], 0.0,
0.0, 1.0)
else:
alpha, sx, sy, scale = compose_transform2(
alpha, sx, sy, 1.0, 180 - (phiM - cpar["phi"]),
0.0, 0.0, 1.0)
else:
if mirror == 0:
alpha, sx, sy, scale = compose_transform2(
alpha, sx, sy, 1.0, -(phiM - cpar["phi"]), 0.0,
0.0, 1.0)
else:
alpha, sx, sy, scale = compose_transform2(
alpha, sx, sy, 1.0,
-(180 - (phiM - cpar["phi"])), 0.0, 0.0, 1.0)
imgdata[mj].set_attr(
"xform.align2d",
Transform({
"type": "2D",
"alpha": alpha,
"tx": sx,
"ty": sy,
"mirror": mirror,
"scale": 1.0
}))
grp_imgdata.append(imgdata[mj])
if not options.no_norm:
for k in range(img_per_grp):
ave, std, minn, maxx = Util.infomask(
grp_imgdata[k], mask, False)
grp_imgdata[k] -= ave
grp_imgdata[k] /= std
if options.fl > 0.0:
for k in range(img_per_grp):
grp_imgdata[k] = filt_tanl(grp_imgdata[k], options.fl,
options.aa)
# Because of background issues, only linear option works.
if options.CTF:
ave, var = aves_wiener(grp_imgdata,
SNR=1.0e5,
interpolation_method="linear")
else:
ave, var = ave_var(grp_imgdata)
# Switch to std dev
# threshold is not really needed,it is just in case due to numerical accuracy something turns out negative.
var = square_root(threshold(var))
set_params_proj(ave, [phiM, thetaM, 0.0, 0.0, 0.0])
set_params_proj(var, [phiM, thetaM, 0.0, 0.0, 0.0])
aveList.append(ave)
varList.append(var)
xform_proj_for_2D[i] = [phiM, thetaM, 0.0, 0.0, 0.0]
'''
if nvec > 0:
eig = pca(input_stacks=grp_imgdata, subavg="", mask_radius=radiuspca, nvec=nvec, incore=True, shuffle=False, genbuf=True)
for k in range(nvec):
set_params_proj(eig[k], [phiM, thetaM, 0.0, 0.0, 0.0])
eigList[k].append(eig[k])
"""
if myid == 0 and i == 0:
for k in xrange(nvec):
eig[k].write_image("eig.hdf", k)
"""
'''
if (myid == heavy_load_myid) and (i % 100 == 0):
log_main.add(" ......%6.2f%% " %
(i / float(len(proj_list)) * 100.))
del imgdata, grp_imgdata, cpar, dpar, all_proj, proj_angles, index
if not options.no_norm: del mask
if myid == main_node: del tab
# At this point, all averages and variances are computed
mpi_barrier(MPI_COMM_WORLD)
if (myid == heavy_load_myid):
log_main.add("Computing aveList and varList took %12.1f [m]" %
((time() - ttt) / 60.))
xform_proj_for_2D = wrap_mpi_gatherv(xform_proj_for_2D, main_node,
MPI_COMM_WORLD)
if (myid == main_node):
write_text_row([str(entry) for entry in xform_proj_for_2D],
os.path.join(current_output_dir, "params.txt"))
del xform_proj_for_2D
mpi_barrier(MPI_COMM_WORLD)
if options.ave2D:
from sp_fundamentals import fpol
from sp_applications import header
if myid == main_node:
log_main.add("Compute ave2D ... ")
km = 0
for i in range(number_of_proc):
if i == main_node:
for im in range(len(aveList)):
aveList[im].write_image(
os.path.join(current_output_dir,
options.ave2D), km)
km += 1
else:
nl = mpi_recv(1, MPI_INT, i,
SPARX_MPI_TAG_UNIVERSAL,
MPI_COMM_WORLD)
nl = int(nl[0])
for im in range(nl):
ave = recv_EMData(i, im + i + 70000)
"""
nm = mpi_recv(1, MPI_INT, i, SPARX_MPI_TAG_UNIVERSAL, MPI_COMM_WORLD)
nm = int(nm[0])
members = mpi_recv(nm, MPI_INT, i, SPARX_MPI_TAG_UNIVERSAL, MPI_COMM_WORLD)
ave.set_attr('members', map(int, members))
members = mpi_recv(nm, MPI_FLOAT, i, SPARX_MPI_TAG_UNIVERSAL, MPI_COMM_WORLD)
ave.set_attr('pix_err', map(float, members))
members = mpi_recv(3, MPI_FLOAT, i, SPARX_MPI_TAG_UNIVERSAL, MPI_COMM_WORLD)
ave.set_attr('refprojdir', map(float, members))
"""
tmpvol = fpol(ave, nx, nx, 1)
tmpvol.write_image(
os.path.join(current_output_dir,
options.ave2D), km)
km += 1
else:
mpi_send(len(aveList), 1, MPI_INT, main_node,
SPARX_MPI_TAG_UNIVERSAL, MPI_COMM_WORLD)
for im in range(len(aveList)):
send_EMData(aveList[im], main_node, im + myid + 70000)
"""
members = aveList[im].get_attr('members')
mpi_send(len(members), 1, MPI_INT, main_node, SPARX_MPI_TAG_UNIVERSAL, MPI_COMM_WORLD)
mpi_send(members, len(members), MPI_INT, main_node, SPARX_MPI_TAG_UNIVERSAL, MPI_COMM_WORLD)
members = aveList[im].get_attr('pix_err')
mpi_send(members, len(members), MPI_FLOAT, main_node, SPARX_MPI_TAG_UNIVERSAL, MPI_COMM_WORLD)
try:
members = aveList[im].get_attr('refprojdir')
mpi_send(members, 3, MPI_FLOAT, main_node, SPARX_MPI_TAG_UNIVERSAL, MPI_COMM_WORLD)
except:
mpi_send([-999.0,-999.0,-999.0], 3, MPI_FLOAT, main_node, SPARX_MPI_TAG_UNIVERSAL, MPI_COMM_WORLD)
"""
if myid == main_node:
header(os.path.join(current_output_dir, options.ave2D),
params='xform.projection',
fimport=os.path.join(current_output_dir,
"params.txt"))
mpi_barrier(MPI_COMM_WORLD)
if options.ave3D:
from sp_fundamentals import fpol
t5 = time()
if myid == main_node: log_main.add("Reconstruct ave3D ... ")
ave3D = recons3d_4nn_MPI(myid,
aveList,
symmetry=options.sym,
npad=options.npad)
bcast_EMData_to_all(ave3D, myid)
if myid == main_node:
if current_decimate != 1.0:
ave3D = resample(ave3D, 1. / current_decimate)
ave3D = fpol(ave3D, nnxo, nnxo,
nnxo) # always to the orignal image size
set_pixel_size(ave3D, 1.0)
ave3D.write_image(
os.path.join(current_output_dir, options.ave3D))
log_main.add("Ave3D reconstruction took %12.1f [m]" %
((time() - t5) / 60.0))
log_main.add("%-70s: %s\n" %
("The reconstructed ave3D is saved as ",
options.ave3D))
mpi_barrier(MPI_COMM_WORLD)
del ave, var, proj_list, stack, alpha, sx, sy, mirror, aveList
'''
if nvec > 0:
for k in range(nvec):
if myid == main_node:log_main.add("Reconstruction eigenvolumes", k)
cont = True
ITER = 0
mask2d = model_circle(radiuspca, nx, nx)
while cont:
#print "On node %d, iteration %d"%(myid, ITER)
eig3D = recons3d_4nn_MPI(myid, eigList[k], symmetry=options.sym, npad=options.npad)
bcast_EMData_to_all(eig3D, myid, main_node)
if options.fl > 0.0:
eig3D = filt_tanl(eig3D, options.fl, options.aa)
if myid == main_node:
eig3D.write_image(os.path.join(options.outpout_dir, "eig3d_%03d.hdf"%(k, ITER)))
Util.mul_img( eig3D, model_circle(radiuspca, nx, nx, nx) )
eig3Df, kb = prep_vol(eig3D)
del eig3D
cont = False
icont = 0
for l in range(len(eigList[k])):
phi, theta, psi, s2x, s2y = get_params_proj(eigList[k][l])
proj = prgs(eig3Df, kb, [phi, theta, psi, s2x, s2y])
cl = ccc(proj, eigList[k][l], mask2d)
if cl < 0.0:
icont += 1
cont = True
eigList[k][l] *= -1.0
u = int(cont)
u = mpi_reduce([u], 1, MPI_INT, MPI_MAX, main_node, MPI_COMM_WORLD)
icont = mpi_reduce([icont], 1, MPI_INT, MPI_SUM, main_node, MPI_COMM_WORLD)
if myid == main_node:
u = int(u[0])
log_main.add(" Eigenvector: ",k," number changed ",int(icont[0]))
else: u = 0
u = bcast_number_to_all(u, main_node)
cont = bool(u)
ITER += 1
del eig3Df, kb
mpi_barrier(MPI_COMM_WORLD)
del eigList, mask2d
'''
if options.ave3D: del ave3D
if options.var2D:
from sp_fundamentals import fpol
from sp_applications import header
if myid == main_node:
log_main.add("Compute var2D...")
km = 0
for i in range(number_of_proc):
if i == main_node:
for im in range(len(varList)):
tmpvol = fpol(varList[im], nx, nx, 1)
tmpvol.write_image(
os.path.join(current_output_dir,
options.var2D), km)
km += 1
else:
nl = mpi_recv(1, MPI_INT, i,
SPARX_MPI_TAG_UNIVERSAL,
MPI_COMM_WORLD)
nl = int(nl[0])
for im in range(nl):
ave = recv_EMData(i, im + i + 70000)
tmpvol = fpol(ave, nx, nx, 1)
tmpvol.write_image(
os.path.join(current_output_dir,
options.var2D), km)
km += 1
else:
mpi_send(len(varList), 1, MPI_INT, main_node,
SPARX_MPI_TAG_UNIVERSAL, MPI_COMM_WORLD)
for im in range(len(varList)):
send_EMData(varList[im], main_node, im + myid +
70000) # What with the attributes??
mpi_barrier(MPI_COMM_WORLD)
if myid == main_node:
from sp_applications import header
header(os.path.join(current_output_dir, options.var2D),
params='xform.projection',
fimport=os.path.join(current_output_dir,
"params.txt"))
mpi_barrier(MPI_COMM_WORLD)
if options.var3D:
if myid == main_node: log_main.add("Reconstruct var3D ...")
t6 = time()
# radiusvar = options.radius
# if( radiusvar < 0 ): radiusvar = nx//2 -3
res = recons3d_4nn_MPI(myid,
varList,
symmetry=options.sym,
npad=options.npad)
#res = recons3d_em_MPI(varList, vol_stack, options.iter, radiusvar, options.abs, True, options.sym, options.squ)
if myid == main_node:
from sp_fundamentals import fpol
if current_decimate != 1.0:
res = resample(res, 1. / current_decimate)
res = fpol(res, nnxo, nnxo, nnxo)
set_pixel_size(res, 1.0)
res.write_image(os.path.join(current_output_dir,
options.var3D))
log_main.add(
"%-70s: %s\n" %
("The reconstructed var3D is saved as ", options.var3D))
log_main.add("Var3D reconstruction took %f12.1 [m]" %
((time() - t6) / 60.0))
log_main.add("Total computation time %f12.1 [m]" %
((time() - t0) / 60.0))
log_main.add("sx3dvariability finishes")
if RUNNING_UNDER_MPI:
sp_global_def.MPI = False
sp_global_def.BATCH = False
def main():
from sp_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_filter=.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("--chunk0",
type="string",
default='',
help="chunk0 for computing margin of error")
parser.add_option("--chunk1",
type="string",
default='',
help="chunk1 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:
sxprint("Usage: " + usage)
sxprint("Please run \'" + progname + " -h\' for detailed options")
ERROR(
"Invalid number of parameters used. Please see usage information above."
)
return
else:
if len(args) > 2:
mask_file = args[2]
else:
mask_file = None
orgstack = args[0]
masterdir = args[1]
sp_global_def.BATCH = True
#---initialize MPI related variables
nproc = mpi.mpi_comm_size(mpi.MPI_COMM_WORLD)
myid = mpi.mpi_comm_rank(mpi.MPI_COMM_WORLD)
mpi_comm = mpi.MPI_COMM_WORLD
main_node = 0
# import some utilities
from sp_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 sp_applications import recons3d_n_MPI, mref_ali3d_MPI, Kmref_ali3d_MPI
from sp_statistics import k_means_match_clusters_asg_new, k_means_stab_bbenum
from sp_applications import mref_ali3d_EQ_Kmeans, ali3d_mref_Kmeans_MPI
# Create the main log file
from sp_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["chunk0"] = options.chunk0
Constants["chunk1"] = options.chunk1
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 sp_utilities import sample_down_1D_curve, get_initial_ID, remove_small_groups, print_upper_triangular_matrix, print_a_line_with_timestamp
from sp_utilities import print_dict, get_resolution_mrk01, partition_to_groups, partition_independent_runs, get_outliers
from sp_utilities import merge_groups, save_alist, margin_of_error, get_margin_of_error, do_two_way_comparison, select_two_runs, get_ali3d_params
from sp_utilities import counting_projections, unload_dict, load_dict, get_stat_proj, create_random_list, get_number_of_groups, recons_mref
from sp_utilities import apply_low_pass_filter, get_groups_from_partition, get_number_of_groups, get_complementary_elements_total, update_full_dict
from sp_utilities import count_chunk_members, set_filter_parameters_from_adjusted_fsc, get_two_chunks_from_stack
####------------------------------------------------------------------
#
# Get the pixel size; if none, set to 1.0, and the original image size
from sp_utilities import get_shrink_data_huang
if (myid == main_node):
line = strftime("%Y-%m-%d_%H:%M:%S", localtime()) + " =>"
sxprint((line + "Initialization of 3-D sorting"))
a = get_im(orgstack)
nnxo = a.get_xsize()
if (Tracker["nxinit"] > nnxo):
sp_global_def.ERROR(
"Image size less than minimum permitted $d" %
Tracker["nxinit"])
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):
return
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"]):
sp_global_def.ERROR("Particle radius set too large!", myid=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 -p", masterdir)
os.system(cmd)
else:
li = 0
li = mpi.mpi_bcast(li, 1, mpi.MPI_INT, main_node,
mpi.MPI_COMM_WORLD)[0]
if li > 0:
masterdir = mpi.mpi_bcast(masterdir, li, mpi.MPI_CHAR, main_node,
mpi.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.mpi_barrier(mpi.MPI_COMM_WORLD)
from time import sleep
while not os.path.exists(masterdir):
sxprint("Node ", myid, " waiting...")
sleep(5)
mpi.mpi_barrier(mpi.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 sp_user_functions
user_func = sp_user_functions.factory[Tracker["constants"]
["user_func"]]
chunk_dict = {}
chunk_list = []
if myid == main_node:
chunk_one = read_text_file(Tracker["constants"]["chunk0"])
chunk_two = read_text_file(Tracker["constants"]["chunk1"])
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.mpi_barrier(mpi.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("sp_cpy.py", orgstack,
Tracker["constants"]["stack"])
junk = cmdexecute(cmd)
cmd = "{} {} {}".format(
"sp_header.py --params=xform.projection",
"--export=" + Tracker["constants"]["ali3d"], orgstack)
junk = cmdexecute(cmd)
cmd = "{} {} {}".format(
"sp_header.py --params=ctf",
"--export=" + Tracker["constants"]["ctf_params"], orgstack)
junk = cmdexecute(cmd)
mpi.mpi_barrier(mpi.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(
"Incorrect focused mask, after binarize all values zero"
)
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.mpi_barrier(mpi.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 range(2):
write_text_file(
chunk_list[index],
os.path.join(masterdir, "chunk%01d.txt" % index))
mpi.mpi_barrier(mpi.MPI_COMM_WORLD)
vols = []
for index in range(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.mpi_barrier(mpi.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 sp_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 range(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.mpi_barrier(mpi.MPI_COMM_WORLD)
from sp_utilities import get_input_from_string
delta = get_input_from_string(Tracker["constants"]["delta"])
delta = delta[0]
from sp_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.mpi_barrier(mpi.MPI_COMM_WORLD)
list_to_be_processed = list(range(Tracker["constants"]["total_stack"]))
Tracker["this_data_list"] = list_to_be_processed
create_random_list(Tracker)
#################################
full_dict = {}
for iptl in range(Tracker["constants"]["total_stack"]):
full_dict[iptl] = iptl
Tracker["full_ID_dict"] = full_dict
#################################
for indep_run in range(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.mpi_barrier(mpi.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.mpi_barrier(mpi.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 range(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 range(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.mpi_barrier(mpi.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 range(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.mpi_barrier(mpi.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 range(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.mpi_barrier(mpi.MPI_COMM_WORLD)
create_random_list(Tracker)
for indep_run in range(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 range(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.mpi_barrier(mpi.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.mpi_barrier(mpi.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.mpi_barrier(mpi.MPI_COMM_WORLD)
update_full_dict(Tracker["this_unaccounted_list"], Tracker)
vol_list = []
for igrp in range(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.mpi_barrier(mpi.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 range(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.mpi_barrier(mpi.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.mpi_barrier(mpi.MPI_COMM_WORLD)
return
def main():
from time import sleep
from sp_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_filter=.25 --seed=random_seed"
parser = OptionParser(usage,version=SPARXVERSION)
parser.add_option("--focus", type="string", default=None, help="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="outer radius for rotational correlation <nx-1 (set to the radius of the particle)")
parser.add_option("--maxit", type= "int", default=50, 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 1) ")
parser.add_option("--CTF", action="store_true", default=False, help="Consider CTF correction during the alignment ")
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 images per 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("--previous_run1", type="string", default='', help="two previous runs" )
parser.add_option("--previous_run2", type="string", default='', help="two previous runs" )
parser.add_option("--group_size_for_unaccounted", type="int", default=500, help="size for unaccounted particles" )
parser.add_option("--chunkdir", type="string", default='', help="chunkdir for computing margin of error")
parser.add_option("--sausage", action="store_true", default=False, help="way of filter volume")
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:
sxprint("Usage: " + usage)
sxprint("Please run \'" + progname + " -h\' for detailed options")
ERROR( "Invalid number of parameters used. Please see usage information above." )
return
else:
if len(args)>2:
mask_file = args[2]
else:
mask_file = None
orgstack =args[0]
masterdir =args[1]
sp_global_def.BATCH = True
#---initialize MPI related variables
nproc = mpi.mpi_comm_size( mpi.MPI_COMM_WORLD )
myid = mpi.mpi_comm_rank( mpi.MPI_COMM_WORLD )
mpi_comm = mpi.MPI_COMM_WORLD
main_node = 0
# Create the main log file
from sp_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["user_func"] =options.function
Constants["low_pass_filter"] =options.low_pass_filter # enforced low_pass_filter
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["previous_runs"] =options.previous_run1+" "+options.previous_run2
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
#Constants["frequency_stop_search"] =options.frequency_stop_search
#Constants["scale_of_number"] =options.scale_of_number
# -------------------------------------------------------------
#
# 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["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"] = Tracker["constants"]["low_pass_filter"]
Tracker["falloff"] = 0.1
#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 sp_utilities import sample_down_1D_curve,get_initial_ID,remove_small_groups,print_upper_triangular_matrix,print_a_line_with_timestamp
from sp_utilities import convertasi,prepare_ptp,print_dict,get_resolution_mrk01,partition_to_groups,partition_independent_runs,get_outliers
from sp_utilities import merge_groups, save_alist, margin_of_error, get_margin_of_error, do_two_way_comparison, select_two_runs, get_ali3d_params
from sp_utilities import counting_projections, unload_dict, load_dict, get_stat_proj, create_random_list, get_number_of_groups, recons_mref
from sp_utilities import apply_low_pass_filter, get_groups_from_partition, get_number_of_groups, get_complementary_elements_total, update_full_dict
from sp_utilities import count_chunk_members, set_filter_parameters_from_adjusted_fsc, get_two_chunks_from_stack
####------------------------------------------------------------------
# another part
from sp_utilities import get_class_members, remove_small_groups, get_number_of_groups, get_stable_members_from_two_runs
from sp_utilities import two_way_comparison_single, get_leftover_from_stable, get_initial_ID, Kmeans_exhaustive_run
from sp_utilities import print_a_line_with_timestamp, split_a_group
#
# Get the pixel size; if none, set to 1.0, and the original image size
from sp_utilities import get_shrink_data_huang
from time import sleep
import sp_user_functions
user_func = sp_user_functions.factory[Tracker["constants"]["user_func"]]
if(myid == main_node):
line = ''
sxprint((line+"Initialization of 3-D sorting"))
a = get_im(Tracker["orgstack"])
nnxo = a.get_xsize()
if( Tracker["nxinit"] > nnxo ):
ERROR( "Image size less than minimum permitted $d"%Tracker["nxinit"] )
nnxo = -1 # we break here, so not sure what this is supposed to accomplish
return
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 ):
return
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"]["wn"]
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!", myid=myid )
return
####-----------------------------------------------------------------------------------------
# create the master directory
if myid == main_node:
if masterdir =="":
timestring = strftime("_%d_%b_%Y_%H_%M_%S", localtime())
masterdir ="master_sort3d"+timestring
li =len(masterdir)
else:
li = 0
cmd="{} {}".format("mkdir -p", masterdir)
os.system(cmd)
sp_global_def.write_command(masterdir)
else:
li=0
li = mpi.mpi_bcast( li, 1, mpi.MPI_INT, main_node, mpi.MPI_COMM_WORLD )[0]
if li>0:
masterdir = mpi.mpi_bcast( masterdir, li,MPI_CHAR, main_node, mpi.MPI_COMM_WORLD )
masterdir = string.join(masterdir,"")
####--- masterdir done!
if myid == main_node:
print_dict(Tracker["constants"],"Permanent settings of 3-D sorting program")
from time import sleep
while not os.path.exists(masterdir): # Be sure each proc is able to access the created dir
sxprint("Node ",myid," waiting...")
sleep(5)
mpi.mpi_barrier(mpi.MPI_COMM_WORLD)
######### 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"]["partstack"] = Tracker["constants"]["ali3d"]
Tracker["constants"]["ctf_params"] = os.path.join(masterdir, "ctf_params.txt")
######
if myid == main_node:
if(Tracker["orgstack"][:4] == "bdb:"): cmd = "{} {} {}".format("e2bdb.py", Tracker["orgstack"],"--makevstack="+Tracker["constants"]["stack"])
else: cmd = "{} {} {}".format("sp_cpy.py", orgstack, Tracker["constants"]["stack"])
cmdexecute(cmd)
cmd = "{} {} {} {} ".format("sp_header.py", Tracker["constants"]["stack"],"--params=xform.projection","--export="+Tracker["constants"]["ali3d"])
cmdexecute(cmd)
cmd = "{} {} {} {} ".format("sp_header.py", Tracker["constants"]["stack"],"--params=ctf","--export="+Tracker["constants"]["ctf_params"])
cmdexecute(cmd)
#keepchecking = False
total_stack = EMUtil.get_image_count(Tracker["orgstack"])
else:
total_stack =0
total_stack = bcast_number_to_all(total_stack, source_node = main_node)
"""
if myid==main_node:
from EMAN2db import db_open_dict
OB = db_open_dict(orgstack)
DB = db_open_dict(Tracker["constants"]["stack"])
for i in xrange(total_stack):
DB[i] = OB[i]
OB.close()
DB.close()
mpi.mpi_barrier( mpi.MPI_COMM_WORLD )
if myid==main_node:
params= []
for i in xrange(total_stack):
e=get_im(orgstack,i)
phi,theta,psi,s2x,s2y = get_params_proj(e)
params.append([phi,theta,psi,s2x,s2y])
write_text_row(params,Tracker["constants"]["ali3d"])
mpi.mpi_barrier( mpi.MPI_COMM_WORLD )
"""
#Tracker["total_stack"] = total_stack
Tracker["constants"]["total_stack"] = total_stack
Tracker["shrinkage"] = float(Tracker["nxinit"])/Tracker["constants"]["nnxo"]
#####------------------------------------------------------------------------------
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"]:
get_shrink_3dmask(Tracker["nxinit"],Tracker["constants"]["mask3D"]).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" )
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
###----------------------------------------------------------------------------------
####--------------------------- Extract the previous results #####################################################
from random import shuffle
if myid ==main_node:
log_main.add(" Sphire rsort3d ")
log_main.add("extract stable groups from two previous runs")
stable_member_list = get_stable_members_from_two_runs(Tracker["constants"]["previous_runs"], Tracker["constants"]["total_stack"], log_main)
Tracker["this_unaccounted_list"], new_stable_P1 = get_leftover_from_stable(stable_member_list, Tracker["constants"]["total_stack"], Tracker["constants"]["smallest_group"])
Tracker["this_unaccounted_list"].sort()
Tracker["total_stack"] = len(Tracker["this_unaccounted_list"])
log_main.add("new stable is %d"%len(new_stable_P1))
else:
Tracker["total_stack"] = 0
Tracker["this_unaccounted_list"] = 0
stable_member_list =0
stable_member_list = wrap_mpi_bcast(stable_member_list, main_node)
Tracker["total_stack"] = bcast_number_to_all(Tracker["total_stack"], source_node = main_node)
left_one_from_old_two_runs = wrap_mpi_bcast(Tracker["this_unaccounted_list"], main_node)
if myid ==main_node:
write_text_file(left_one_from_old_two_runs, os.path.join(masterdir,"unaccounted_from_two_previous_runs.txt"))
sxprint(" Extracting results of two previous runs is done!")
#################################### Estimate resolution----------------------#############
#### make chunkdir dictionary for computing margin of error
chunk_list = []
if Tracker["constants"]["chunkdir"] !="": ##inhere previous random assignment of odd and even
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)
else: ## if traces are lost, then creating new random assignment of odd, even particles
chunks = list(range(Tracker["constants"]["total_stack"]))
shuffle(chunks)
chunk_one =chunks[0:Tracker["constants"]["total_stack"]//2]
chunk_two =chunks[Tracker["constants"]["total_stack"]//2:Tracker["constants"]["total_stack"]]
chunk_one = wrap_mpi_bcast(chunk_one, main_node)
chunk_two = wrap_mpi_bcast(chunk_two, main_node)
###### Fill chunk ID into headers when calling get_shrink_data_huang
if myid ==main_node:
sxprint(" random odd and even assignment done !")
mpi.mpi_barrier( mpi.MPI_COMM_WORLD )
#------------------------------------------------------------------------------
Tracker["chunk_dict"] = {}
for element in chunk_one: Tracker["chunk_dict"][element] = 0
for element in chunk_two: Tracker["chunk_dict"][element] = 1
Tracker["P_chunk0"] = len(chunk_one)/float(Tracker["constants"]["total_stack"])
Tracker["P_chunk1"] = len(chunk_two)/float(Tracker["constants"]["total_stack"])
### create two volumes to estimate resolution
if myid == main_node:
write_text_file(chunk_one, os.path.join(masterdir,"chunk0.txt"))
write_text_file(chunk_two, os.path.join(masterdir,"chunk1.txt"))
mpi.mpi_barrier( mpi.MPI_COMM_WORLD )
vols = []
for index in range(2):
data1,old_shifts1 = get_shrink_data_huang(Tracker,Tracker["constants"]["nxinit"], os.path.join(masterdir,"chunk%d.txt"%index), Tracker["constants"]["partstack"], myid, main_node, nproc, preshift = True)
vol1 = recons3d_4nn_ctf_MPI(myid=myid, prjlist=data1, symmetry=Tracker["constants"]["sym"], finfo=None)
if myid ==main_node:
vol1_file_name = os.path.join(masterdir, "vol%d.hdf"%index)
vol1.write_image(vol1_file_name)
vols.append(vol1)
mpi.mpi_barrier( mpi.MPI_COMM_WORLD )
if myid ==main_node:
low_pass, falloff, currentres = get_resolution_mrk01(vols, Tracker["constants"]["radius"]*Tracker["shrinkage"], 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
currentres = bcast_number_to_all(currentres,source_node = main_node)
low_pass = bcast_number_to_all(low_pass,source_node = main_node)
falloff = 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"] = low_pass*Tracker["shrinkage"]
else:
Tracker["low_pass_filter"] = Tracker["constants"]["low_pass_filter"]/Tracker["shrinkage"]
Tracker["lowpass"] = Tracker["low_pass_filter"]
Tracker["falloff"] = 0.1
Tracker["global_fsc"] = os.path.join(masterdir,"fsc.txt")
##################################################################
if myid ==main_node:
log_main.add("The command-line inputs are :")
log_main.add("**********************************************************")
for a in sys.argv:
log_main.add(a)
log_main.add("**********************************************************")
from sp_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"%(round((Tracker["constants"]["pixel_size"]/Tracker["currentres"]/Tracker["shrinkage"]),4)))
filt_tanl(get_im(os.path.join(masterdir, "vol0.hdf")), Tracker["low_pass_filter"], 0.1).write_image(os.path.join(masterdir, "volf0.hdf"))
filt_tanl(get_im(os.path.join(masterdir, "vol1.hdf")), Tracker["low_pass_filter"], 0.1).write_image(os.path.join(masterdir, "volf1.hdf"))
sxprint(" random odd and even assignment done !")
mpi.mpi_barrier( mpi.MPI_COMM_WORLD )
## ---------------------------------------------------------------------------------------------########
## Stop program and output results when the leftover from two sort3d runs is not sufficient for a new run ########
## --------------------------------------------------- --------------------------------------- ######
Tracker["number_of_groups"] = get_number_of_groups(len(left_one_from_old_two_runs), Tracker["constants"]["number_of_images_per_group"])
if Tracker["number_of_groups"] <=1 : # programs finishes
if myid == main_node:
log_main.add("the unaccounted ones are no sufficient for a simple two-group run, output results!")
log_main.add("this implies your two sort3d runs already achieved high reproducibale ratio. ")
log_main.add("Or your number_of_images_per_group is too large ")
log_main.add("the final reproducibility is %f"%((Tracker["constants"]["total_stack"]-len(Tracker["this_unaccounted_list"]))/float(Tracker["constants"]["total_stack"])))
for i in range(len(stable_member_list)): write_text_file(stable_member_list[i], os.path.join(masterdir,"P2_final_class%d.txt"%i))
mask3d = get_im(Tracker["constants"]["mask3D"])
else:
mask3d = model_blank(Tracker["constants"]["nnxo"],Tracker["constants"]["nnxo"],Tracker["constants"]["nnxo"])
bcast_EMData_to_all(mask3d, myid, main_node)
for igrp in range(len(stable_member_list)):
#name_of_class_file = os.path.join(masterdir, "P2_final_class%d.txt"%igrp)
data, old_shifts = get_shrink_data_huang(Tracker,Tracker["constants"]["nnxo"], os.path.join(masterdir, "P2_final_class%d.txt"%igrp), Tracker["constants"]["partstack"], myid, main_node, nproc,preshift = True)
if Tracker["constants"]["CTF"]:
volref, fscc = rec3D_two_chunks_MPI(data, 1.0, Tracker["constants"]["sym"], mask3d,os.path.join(masterdir,"resolution_%02d.txt"%igrp), myid, main_node, index =-1, npad=2)
else:
sxprint("Missing CTF flag!")
return
mpi.mpi_barrier( mpi.MPI_COMM_WORLD )
#nx_of_image=volref.get_xsize()
if Tracker["constants"]["PWadjustment"] : Tracker["PWadjustment"] = Tracker["PW_dict"][Tracker["constants"]["nnxo"]]
else: Tracker["PWadjustment"] = Tracker["constants"]["PWadjustment"]
if myid ==main_node:
try:
lowpass = search_lowpass(fscc)
falloff = 0.1
except:
lowpass = 0.4
falloff = 0.1
log_main.add(" lowpass and falloff from fsc are %f %f"%(lowpass, falloff))
lowpass = round(lowpass,4)
falloff = round(min(0.1,falloff),4)
Tracker["lowpass"] = lowpass
Tracker["falloff"] = falloff
refdata = [None]*4
refdata[0] = volref
refdata[1] = Tracker
refdata[2] = Tracker["constants"]["myid"]
refdata[3] = Tracker["constants"]["nproc"]
volref = user_func(refdata)
cutoff = Tracker["constants"]["pixel_size"]/lowpass
log_main.add("%d vol low pass filer %f %f cut to %f Angstrom"%(igrp,Tracker["lowpass"],Tracker["falloff"],cutoff))
volref.write_image(os.path.join(masterdir,"volf_final%d.hdf"%igrp))
mpi.mpi_barrier( mpi.MPI_COMM_WORLD )
return
else: # Continue clustering on unaccounted ones that produced by two_way comparison of two previous runs
#########################################################################################################################
#if Tracker["constants"]["number_of_images_per_group"] ==-1: # Estimate number of images per group from delta, and scale up
# or down by scale_of_number
# number_of_images_per_group = int(Tracker["constants"]["scale_of_number"]*len(n_angles))
#
#########################################################################################################################P2
if myid ==main_node:
sxprint(" Now continue clustering on accounted ones because they can make at least two groups!")
P2_partitions = []
number_of_P2_runs = 2 # Notice P2 start from two P1 runs
### input list_to_be_processed
import copy
mpi.mpi_barrier( mpi.MPI_COMM_WORLD )
for iter_P2_run in range(number_of_P2_runs): # two runs such that one can obtain reproducibility
list_to_be_processed = left_one_from_old_two_runs[:]#Tracker["this_unaccounted_list"][:]
Tracker["this_unaccounted_list"] = left_one_from_old_two_runs[:]
if myid == main_node : new_stable1 = new_stable_P1[:]
total_stack = len(list_to_be_processed) # This is the input from two P1 runs
#number_of_images_per_group = Tracker["constants"]["number_of_images_per_group"]
P2_run_dir = os.path.join(masterdir, "P2_run%d"%iter_P2_run)
Tracker["number_of_groups"] = get_number_of_groups(total_stack, Tracker["constants"]["number_of_images_per_group"])
if myid == main_node:
cmd="{} {}".format("mkdir", P2_run_dir)
os.system(cmd)
log_main.add("----------------P2 independent run %d--------------"%iter_P2_run)
log_main.add("user provided number_of_images_per_group %d"%Tracker["constants"]["number_of_images_per_group"])
sxprint("----------------P2 independent run %d--------------"%iter_P2_run)
mpi.mpi_barrier( mpi.MPI_COMM_WORLD )
#
#Tracker["number_of_groups"] = get_number_of_groups(total_stack,Tracker["constants"]["number_of_images_per_group"])
generation = 0
if myid == main_node:
log_main.add("number of groups is %d"%Tracker["number_of_groups"])
log_main.add("total stack %d"%total_stack)
while( Tracker["number_of_groups"]>=2 ):
partition_dict = {}
full_dict = {}
workdir = os.path.join(P2_run_dir,"generation%03d"%generation)
Tracker["this_dir"] = workdir
if myid ==main_node:
cmd="{} {}".format("mkdir", workdir)
os.system(cmd)
log_main.add("---- generation %5d"%generation)
log_main.add("number of images per group is set as %d"%Tracker["constants"]["number_of_images_per_group"])
log_main.add("the initial number of groups is %d "%Tracker["number_of_groups"])
log_main.add(" the number to be processed in this generation is %d"%len(list_to_be_processed))
sxprint("---- generation %5d"%generation)
#core=read_text_row(Tracker["constants"]["ali3d"],-1)
#write_text_row(core, os.path.join(workdir,"node%d.txt"%myid))
mpi.mpi_barrier( mpi.MPI_COMM_WORLD )
Tracker["this_data_list"] = list_to_be_processed # leftover of P1 runs
Tracker["total_stack"] = len(list_to_be_processed)
create_random_list(Tracker)
###------ For super computer ##############
update_full_dict(list_to_be_processed, Tracker)
###----
##### ----------------Independent runs for EQ-Kmeans ------------------------------------
for indep_run in range(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.mpi_barrier( mpi.MPI_COMM_WORLD )
#this_particle_text_file = # for get_shrink_data
if myid ==main_node:
write_text_file(list_to_be_processed, os.path.join(workdir, "independent_list_%03d.txt"%indep_run))
mref_ali3d_EQ_Kmeans(ref_vol, os.path.join(workdir, "EQ_Kmeans%03d"%indep_run), os.path.join(workdir, "independent_list_%03d.txt"%indep_run), Tracker)
partition_dict[indep_run] = Tracker["this_partition"]
del ref_vol
Tracker["partition_dict"] = partition_dict
Tracker["this_total_stack"] = Tracker["total_stack"]
do_two_way_comparison(Tracker)
##############################
if myid ==main_node: log_main.add("Now calculate stable volumes")
if myid ==main_node:
for igrp in range(len(Tracker["two_way_stable_member"])):
Tracker["this_data_list"] = Tracker["two_way_stable_member"][igrp]
write_text_file(Tracker["this_data_list"], os.path.join(workdir,"stable_class%d.txt"%igrp))
Tracker["this_data_list_file"] = -1
mpi.mpi_barrier( mpi.MPI_COMM_WORLD )
###
number_of_ref_class = []
ref_vol_list = []
for igrp in range(len(Tracker["two_way_stable_member"])):
data, old_shifts = get_shrink_data_huang(Tracker,Tracker["nxinit"], os.path.join(workdir, "stable_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)
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"])))
#ref_vol_list=apply_low_pass_filter(ref_vol_list,Tracker)
for iref in range(len(ref_vol_list)): ref_vol_list[iref].write_image(os.path.join(workdir,"vol_stable.hdf"),iref)
mpi.mpi_barrier( mpi.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_groups,res_classes,final_list = ali3d_mref_Kmeans_MPI(ref_vol_list, outdir, os.path.join(workdir,"Accounted.txt"), 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"]))
update_full_dict(Tracker["this_unaccounted_list"], Tracker)
if myid == main_node: write_text_file(Tracker["this_unaccounted_list"], Tracker["this_unaccounted_text"])
Tracker["number_of_groups"] = len(res_classes)
### Update data
mpi.mpi_barrier( mpi.MPI_COMM_WORLD )
if myid == main_node:
number_of_ref_class=[]
log_main.add(" Compute volumes of original size")
for igrp in range(Tracker["number_of_groups"]):
if os.path.exists( os.path.join( outdir,"Class%d.txt"%igrp ) ):
new_stable1.append( read_text_file( os.path.join( outdir, "Class%d.txt"%igrp ) ) )
log_main.add(" read Class file %d"%igrp)
number_of_ref_class.append(len(new_stable1))
else: number_of_ref_class = 0
number_of_ref_class = wrap_mpi_bcast(number_of_ref_class,main_node)
################################
mpi.mpi_barrier( mpi.MPI_COMM_WORLD )
if myid ==main_node:
vol_list = []
for igrp in range(Tracker["number_of_groups"]):
if myid ==main_node: log_main.add("start vol %d"%igrp)
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)
if myid == main_node:
vol_list.append(volref)
log_main.add(" vol %d is done"%igrp)
Tracker["number_of_ref_class"] = number_of_ref_class
mpi.mpi_barrier( mpi.MPI_COMM_WORLD )
generation +=1
#################################
if myid ==main_node:
for ivol in range(len(vol_list)): vol_list[ivol].write_image(os.path.join(workdir, "vol_of_Classes.hdf"),ivol)
filt_tanl(vol_list[ivol],Tracker["constants"]["low_pass_filter"],.1).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
Tracker["this_data_list"] = Tracker["this_unaccounted_list"]
Tracker["total_stack"] = len(Tracker["this_unaccounted_list"])
Tracker["this_total_stack"] = Tracker["total_stack"]
#update_full_dict(complementary)
#number_of_groups = int(float(len(Tracker["this_unaccounted_list"]))/number_of_images_per_group)
del list_to_be_processed
list_to_be_processed = copy.deepcopy(Tracker["this_unaccounted_list"])
Tracker["number_of_groups"] = get_number_of_groups(len(list_to_be_processed),Tracker["constants"]["number_of_images_per_group"])
mpi.mpi_barrier( mpi.MPI_COMM_WORLD )
#############################################################################################################################
### Reconstruct the unaccounted is only done once
if (Tracker["constants"]["unaccounted"] and (len(Tracker["this_unaccounted_list"]) != 0)):
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)
volref = filt_tanl(volref, Tracker["constants"]["low_pass_filter"],.1)
if myid ==main_node: volref.write_image(os.path.join(workdir, "volf_unaccounted.hdf"))
######## Exhaustive Kmeans #############################################
if myid ==main_node:
if len(Tracker["this_unaccounted_list"])>=Tracker["constants"]["smallest_group"]:
new_stable1.append(Tracker["this_unaccounted_list"])
unaccounted = get_complementary_elements_total(Tracker["constants"]["total_stack"], final_list)
Tracker["number_of_groups"] = len(new_stable1)
log_main.add("----------------Exhaustive Kmeans------------------")
log_main.add("number_of_groups is %d"%Tracker["number_of_groups"])
else: Tracker["number_of_groups"] = 0
### prepare references for final K-means
if myid == main_node:
final_list =[]
for alist in new_stable1:
for element in alist:final_list.append(int(element))
unaccounted = get_complementary_elements_total(Tracker["constants"]["total_stack"],final_list)
if len(unaccounted) > Tracker["constants"]["smallest_group"]: # treat unaccounted ones also as a group if it is not too small.
new_stable1.append(unaccounted)
Tracker["number_of_groups"] = len(new_stable1)
for any in unaccounted:final_list.append(any)
log_main.add("total number %d"%len(final_list))
else: final_list = 0
Tracker["number_of_groups"] = bcast_number_to_all(Tracker["number_of_groups"],source_node = main_node)
mpi.mpi_barrier( mpi.MPI_COMM_WORLD )
final_list = wrap_mpi_bcast(final_list, main_node)
workdir = os.path.join(P2_run_dir,"Exhaustive_Kmeans") # new workdir
if myid==main_node:
os.mkdir(workdir)
write_text_file(final_list, os.path.join(workdir,"final_list.txt"))
else: new_stable1 = 0
mpi.mpi_barrier( mpi.MPI_COMM_WORLD )
## Create reference volumes
if myid == main_node:
number_of_ref_class = []
for igrp in range(Tracker["number_of_groups"]):
class_file = os.path.join(workdir,"final_class%d.txt"%igrp)
write_text_file(new_stable1[igrp],class_file)
log_main.add(" group %d number of particles %d"%(igrp,len(new_stable1[igrp])))
number_of_ref_class.append(len(new_stable1[igrp]))
else: number_of_ref_class= 0
number_of_ref_class = wrap_mpi_bcast(number_of_ref_class,main_node)
mpi.mpi_barrier( mpi.MPI_COMM_WORLD )
ref_vol_list = []
for igrp in range(Tracker["number_of_groups"]):
if myid ==main_node : sxprint(" prepare reference %d"%igrp)
#Tracker["this_data_list_file"] = os.path.join(workdir,"final_class%d.txt"%igrp)
data,old_shifts = get_shrink_data_huang(Tracker, Tracker["nxinit"],os.path.join(workdir,"final_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)
#volref = filt_tanl(volref, Tracker["low_pass_filter"],.1)
#if myid == main_node:
# volref.write_image(os.path.join(masterdir,"volf_stable.hdf"),iref)
#volref = resample(volref,Tracker["shrinkage"])
bcast_EMData_to_all(volref, myid, main_node)
ref_vol_list.append(volref)
mpi.mpi_barrier( mpi.MPI_COMM_WORLD )
### -------variables used in Kmeans_exhaustive_run-----
Tracker["number_of_ref_class"] = number_of_ref_class
Tracker["this_data_list"] = final_list
Tracker["total_stack"] = len(final_list)
Tracker["this_dir"] = workdir
Tracker["this_data_list_file"] = os.path.join(workdir,"final_list.txt")
KE_group = Kmeans_exhaustive_run(ref_vol_list,Tracker) #
P2_partitions.append(KE_group[:][:])
if myid ==main_node:
log_main.add(" the number of groups after exhaustive Kmeans is %d"%len(KE_group))
for ike in range(len(KE_group)):log_main.add(" group %d number of objects %d"%(ike,len(KE_group[ike])))
del new_stable1
mpi.mpi_barrier( mpi.MPI_COMM_WORLD )
if myid == main_node: log_main.add("P2 runs are done, now start two-way comparision to exclude those that are not reproduced ")
reproduced_groups = two_way_comparison_single(P2_partitions[0],P2_partitions[1],Tracker)# Here partition IDs are original indexes.
###### ----------------Reconstruct reproduced groups------------------------#######
######
if myid == main_node:
for index_of_reproduced_groups in range(len(reproduced_groups)):
name_of_class_file = os.path.join(masterdir, "P2_final_class%d.txt"%index_of_reproduced_groups)
write_text_file(reproduced_groups[index_of_reproduced_groups],name_of_class_file)
log_main.add("-------start to reconstruct reproduced volumes individully to orignal size-----------")
mpi.mpi_barrier( mpi.MPI_COMM_WORLD )
if Tracker["constants"]["mask3D"]: mask_3d = get_shrink_3dmask(Tracker["constants"]["nnxo"],Tracker["constants"]["mask3D"])
else: mask_3d = None
for igrp in range(len(reproduced_groups)):
data,old_shifts = get_shrink_data_huang(Tracker,Tracker["constants"]["nnxo"],os.path.join(masterdir, "P2_final_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)
if Tracker["constants"]["CTF"]:
volref, fscc = rec3D_two_chunks_MPI(data,1.0,Tracker["constants"]["sym"],mask_3d, \
os.path.join(masterdir,"resolution_%02d.txt"%igrp),myid,main_node,index =-1,npad =2,finfo=None)
else:
sxprint("Missing CTF flag!")
return
mpi.mpi_barrier( mpi.MPI_COMM_WORLD )
fscc = read_text_file(os.path.join(masterdir, "resolution_%02d.txt"%igrp),-1)
nx_of_image = volref.get_xsize()
if Tracker["constants"]["PWadjustment"]: Tracker["PWadjustment"] = Tracker["PW_dict"][nx_of_image]
else: Tracker["PWadjustment"] = Tracker["constants"]["PWadjustment"]
try:
lowpass = search_lowpass(fscc)
falloff = 0.1
except:
lowpass= 0.4
falloff= 0.1
sxprint(lowpass)
lowpass=round(lowpass,4)
falloff=round(min(.1,falloff),4)
Tracker["lowpass"]= lowpass
Tracker["falloff"]= falloff
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)
cutoff = Tracker["constants"]["pixel_size"]/lowpass
log_main.add("%d vol low pass filer %f %f cut to %f Angstrom"%(igrp,Tracker["lowpass"],Tracker["falloff"],cutoff))
volref.write_image(os.path.join(masterdir,"volf_final%d.hdf"%igrp))
if myid==main_node: log_main.add(" sxsort3d_P2 finishes. ")
# Finish program
mpi.mpi_barrier( mpi.MPI_COMM_WORLD )
return