def sim_all(Nshot_max=None, jid=0, n_jobs=1): odir = "/global/project/projectdirs/lcls/dermen/bigsim" prefix = odir + "/run62_%06d" spec_file = h5py.File(full_path("simMe_data_run62.h5"), "r") spec_data = spec_file["hist_spec"] Umat_data = spec_file["Umats"] en_chans = spec_file["energy_bins"][()] wave_chans = ENERGY_CONV / en_chans sfall = sim_spectra.load_spectra(full_path("test_sfall.h5")) # his jobs id #jid = int(sys.argv[1]) #global device_Id device_Id = jid Nshot = spec_data.shape[0] if Nshot_max is not None: Nshot = min(Nshot_max, Nshot) print("Job %d: Simulating %d shots" % (jid, Nshot)) # total number of available jobs #n_jobs = int(sys.argv[2]) idx_range = np.array_split(np.arange(Nshot), n_jobs) Nshot_per_job = len(idx_range[jid]) for idx in idx_range[jid]: print("<><><><><><>") print("Job %d; Image %d / %d" % (jid, idx + 1, Nshot_per_job)) print("<><><><><><>") if (jid == 0): print("GPU status") os.system("nvidia-smi") spec = spec_data[idx] C_ori = sqr(Umat_data[idx]) wavelength_A = np.mean(wave_chans) C = microcrystal(Deff_A=Deff_A, length_um=length_um, beam_diameter_um=beam_size_mm * 1000, verbose=False) iterator = iter([(wave_chans, spec, wavelength_A)]) run_sim2smv( prefix=prefix % idx, # "run62_%06d" % idx, crystal=C, spectra=iterator, sfall_main=sfall, rotation=C_ori, quick=False, rank=jid, save_bragg=False)
def tst_one(quick=False, prefix="step5"): spec_file = h5py.File(full_path("test_data.h5"), "r") idx = 2 en_chans = spec_file["energy_bins"][()] wave_chans = ENERGY_CONV / en_chans spec = spec_file["hist_spec"][idx] C_ori = sqr(spec_file["Umats"][idx]) wavelength_A = np.mean(wave_chans) sfall = sim_spectra.load_spectra(full_path("test_sfall.h5")) C = microcrystal(Deff_A=Deff_A, length_um=length_um, beam_diameter_um=beam_size_mm * 1000) iterator = iter([(wave_chans, spec, wavelength_A)]) run_sim2smv(prefix="tst_one_test_data_%06d" % idx, crystal=C, spectra=iterator, sfall_main=sfall, rotation=C_ori, quick=False, rank=0, save_bragg=False) #, actually_use_smv=False)
parser.add_argument("-o", dest='o', help='outputfile prefix (prefix.npz) ', type=str, default=None) parser.add_argument("-plot", dest='plot', action='store_true', help='plot Ymodel vs Yobs') parser.add_argument("-anom", dest='anom', action='store_true', help='group h by anom') args = parser.parse_args() SFall = sim_spectra.load_spectra("test_sfall.h5") # need to load the structure factors at the channel A and B wavelengths spec_f = h5py.File("simMe_data_run62.h5", "r") en_bins = spec_f["energy_bins"][()] ilow = abs(en_bins - ENERGY_LOW).argmin() ihigh = abs(en_bins - ENERGY_HIGH).argmin() SA = SFall[ilow].amplitudes() # energy A (lower energy) SB = SFall[ihigh].amplitudes() # energy B (higher energy) # load the integrated data df = pandas.read_pickle(args.i) Kfact = 1e20 if args.anom: hkl = tuple(map(tuple, df[['hAnom', 'kAnom', 'lAnom']].values.astype(int)))
def run_sim2smv(Nshot_max, odir, prefix, rank, n_jobs, save_bragg=False, save_smv=True, save_h5 =False, return_pixels=False): from six.moves import range, StringIO from six.moves import cPickle as pickle from cxid9114.sim import sim_utils import os import h5py import math import sys import numpy as np from IPython import embed from cxid9114.bigsim.bigsim_geom import DET,BEAM import scitbx from scitbx.array_family import flex from scitbx.matrix import sqr,col from simtbx.nanoBragg import shapetype from simtbx.nanoBragg import nanoBragg import libtbx.load_env # possibly implicit from libtbx.development.timers import Profiler from cctbx import crystal,crystal_orientation from LS49.sim.step4_pad import microcrystal from cxid9114 import utils from cxid9114.parameters import ENERGY_CONV, ENERGY_HIGH, ENERGY_LOW from cxid9114.bigsim import sim_spectra odir_j = os.path.join( odir, "job%d" % rank) if not os.path.exists(odir_j): os.makedirs(odir_j) add_noise = False add_background = False overwrite = True #$False sample_thick_mm = 0.005 # 50 micron GDVN nozzle makes a ~5ish micron jet air_thick_mm =0 # mostly vacuum, maybe helium layer of 1 micron flux_ave=2e11 add_spots_algorithm="cuda" big_data = "." # directory location for reference files detpixels_slowfast = (1800,1800) pixsize_mm=0.11 distance_mm = 125 offset_adu=30 mos_spread_deg=0.015 mos_doms=1000 beam_size_mm=0.001 exposure_s=1 use_microcrystal=True #False Ncells_abc=(120,120,120) Deff_A = 2200 length_um = 2.2 timelog = False background = utils.open_flex("background") crystal = microcrystal(Deff_A = Deff_A, length_um = length_um, beam_diameter_um = beam_size_mm*1000, verbose=False) spec_file = h5py.File("simMe_data_run62.h5", "r") spec_data = spec_file["hist_spec"] Umat_data = spec_file["Umats"] en_chans = spec_file["energy_bins"][()] ilow = np.abs(en_chans - ENERGY_LOW).argmin() ihigh = np.abs(en_chans - ENERGY_HIGH).argmin() wave_chans = ENERGY_CONV/en_chans sfall_main = sim_spectra.load_spectra("test_sfall.h5") Nshot = spec_data.shape[0] idx_range = np.array_split(np.arange(Nshot), n_jobs) Nshot_per_job = len(idx_range[rank]) if Nshot_max > 0 : Nshot_per_job = min( Nshot_max, Nshot_per_job) print ("Job %d: Simulating %d shots" % (rank, Nshot_per_job)) istart = idx_range[rank][0] istop = istart + Nshot_per_job smi_stride = 10 for idx in range( istart, istop): print ("<><><><><><><><><><><><><><>") print ("Job %d; Image %d (%d - %d)" % (rank, idx+1, istart, istop)) print ("<><><><><><><><><><><><><><>") smv_fileout = os.path.join( odir_j, prefix % idx + ".img") h5_fileout = smv_fileout + ".h5" if os.path.exists(smv_fileout) and not overwrite and save_smv: print("Shot %s exists: moving on" % smv_fileout) continue if os.path.exists(h5_fileout) and not overwrite and save_h5: print("Shot %s exists: moving on" % h5_fileout) continue if (rank==0 and idx % smi_stride==0): print("GPU status") os.system("nvidia-smi") print("\n\n") print("CPU memory usage") mem_usg= """ps -U dermen --no-headers -o rss | awk '{ sum+=$1} END {print int(sum/1024) "MB consumed by CPU user"}'""" os.system(mem_usg) spec = spec_data[2] rotation = sqr(Umat_data[2]) wavelength_A = np.mean(wave_chans) spectra = iter([(wave_chans, spec, wavelength_A)]) direct_algo_res_limit = 1.7 wavlen, flux, wavelength_A = next(spectra) # list of lambdas, list of fluxes, average wavelength assert wavelength_A > 0 assert (len(wavlen)==len(flux)==len(sfall_main)) N = crystal.number_of_cells(sfall_main[0].unit_cell()) if use_mcrocrystal: Ncells_abc = (N,N,N) flux *= 0 flux[ilow] = 1e12 flux[ihigh]=1e12 UMAT_nm = flex.mat3_double() mersenne_twister = flex.mersenne_twister(seed=0) scitbx.random.set_random_seed(1234) rand_norm = scitbx.random.normal_distribution(mean=0, sigma=mos_spread_deg * math.pi/180.) g = scitbx.random.variate(rand_norm) mosaic_rotation = g(mos_doms) for m in mosaic_rotation: site = col(mersenne_twister.random_double_point_on_sphere()) UMAT_nm.append( site.axis_and_angle_as_r3_rotation_matrix(m,deg=False) ) Amatrix_rot = (rotation * sqr(sfall_main[0].unit_cell().orthogonalization_matrix())).transpose() #SIM = nanoBragg(detpixels_slowfast=detpixels_slowfast, # pixel_size_mm=pixsize_mm,Ncells_abc=Ncells_abc, # wavelength_A=wavelength_A,verbose=verbose) SIM = nanoBragg(detector=DET, beam=BEAM, panel_id=0,verbose=verbose) SIM.adc_offset_adu = offset_adu # Do not offset by 40 SIM.mosaic_spread_deg = mos_spread_deg # interpreted by UMAT_nm as a half-width stddev SIM.mosaic_domains = mos_doms SIM.distance_mm=distance_mm SIM.set_mosaic_blocks(UMAT_nm) SIM.seed = 1 SIM.oversample=1 SIM.wavelength_A = wavelength_A SIM.polarization=1 SIM.default_F=0 SIM.Fhkl=sfall_main[0].amplitudes() SIM.progress_meter=False SIM.flux=flux_ave SIM.exposure_s=exposure_s SIM.beamsize_mm=beam_size_mm SIM.Ncells_abc=Ncells_abc SIM.xtal_shape=shapetype.Gauss idxpath = "try3_idx2/job0/dump_0_data.pkl" Amat = sim_utils.Amatrix_dials2nanoBragg(utils.open_flex(idxpath)["crystalAB"]) #SIM.Umatrix = rotation.elems #SIM.unit_cell_Adeg = sfall_main[0].unit_cell() #SIM2 = nanoBragg(detector=DET, beam=BEAM, panel_id=0,verbose=verbose) #SIM2.adc_offset_adu = offset_adu # Do not offset by 40 #SIM2.mosaic_spread_deg = mos_spread_deg # interpreted by UMAT_nm as a half-width stddev #SIM2.mosaic_domains = mos_doms #SIM2.distance_mm=distance_mm #SIM2.set_mosaic_blocks(UMAT_nm) #SIM2.seed = 1 #SIM2.oversample=1 #SIM2.wavelength_A = wavelength_A #SIM2.polarization=1 #SIM2.default_F=0 #SIM2.Fhkl=sfall_main[0].amplitudes() #SIM2.progress_meter=False #SIM2.flux=flux_ave #SIM2.exposure_s=exposure_s #SIM2.beamsize_mm=beam_size_mm #SIM2.Ncells_abc=Ncells_abc #SIM2.xtal_shape=shapetype.Gauss #SIM2.Umatrix = rotation.elems #SIM2.unit_cell_Adeg = sfall_main[0].unit_cell() #SIM.Amatrix_RUB = Amatrix_rot #SIM.Amatrix = Amatrix_rot.inverse() raw_pixel_sum = flex.double(len(SIM.raw_pixels)) Nflux = len(flux) print("Beginning the loop") for x in range(Nflux): if x % 10==0: print("+++++++++++++++++++++++++++++++++++++++ Wavelength %d / %d" % (x+1, Nflux) , end="\r") if flux[x] ==0: continue #SIM.Amatrix = Amatrix_rot.inverse() print (SIM.Ncells_abc) SIM.wavelength_A=wavlen[x] SIM.flux=flux[x] SIM.Fhkl=sfall_main[x].amplitudes() SIM.Amatrix = Amat#Amatrix_rot.inverse() #sim_utils.compare_sims(SIM, SIM2) #SIM.Ncells_abc=Ncells_abc #SIM.adc_offset_adu = offset_adu #SIM.mosaic_spread_deg = mos_spread_deg # interpreted by UMAT_nm as a half-width stddev #SIM.mosaic_domains = mos_doms # #SIM.distance_mm=distance_mm #SIM.set_mosaic_blocks(UMAT_nm) #SIM.seed = 1 #SIM.polarization=1 #SIM.default_F=0 #SIM.xtal_shape=shapetype.Gauss #SIM.progress_meter=False #SIM.exposure_s = exposure_s #SIM.beamsize_mm=beam_size_mm SIM.timelog=timelog SIM.device_Id=rank SIM.raw_pixels *= 0 # just in case! SIM.add_nanoBragg_spots_cuda() if use_microcrystal: raw_pixel_sum += SIM.raw_pixels * crystal.domains_per_crystal print() SIM.raw_pixels = raw_pixel_sum if add_background: SIM.raw_pixels = SIM.raw_pixels + background SIM.detector_psf_kernel_radius_pixels=5; #SIM.detector_psf_fwhm_mm=0.08; #SIM.detector_psf_type=shapetype.Fiber # rayonix=Fiber, CSPAD=None (or small Gaussian) SIM.detector_psf_type=shapetype.Unknown # for CSPAD SIM.detector_psf_fwhm_mm=0 SIM.quantum_gain = 28. #SIM.apply_psf() if add_noise: SIM.add_noise() #converts phtons to ADU. extra = "PREFIX=%s;\nRANK=%d;\n"%(prefix,rank) out = SIM.raw_pixels.as_numpy_array() if save_smv: SIM.to_smv_format_py(fileout=smv_fileout,intfile_scale=1,rotmat=True,extra=extra,gz=True) elif save_h5: f = h5py.File(h5_fileout, "w") f.create_dataset("bigsim_d9114", data=SIM.raw_pixels.as_numpy_array().astype(np.uint16).reshape(detpixels_slowfast), compression="lzf") f.close() if npout is not None: np.save(npout, SIM.raw_pixels.as_numpy_array()) SIM.free_all()
def main(rank): device_Id = rank % ngpu worker_Id = node_id * ngpu + rank import os import sys from copy import deepcopy import glob from itertools import izip from scipy.spatial import distance import h5py import scipy.ndimage from IPython import embed import numpy as np import pandas from scipy.spatial import cKDTree from simtbx.nanoBragg import shapetype, nanoBragg from libtbx.phil import parse from scitbx.matrix import sqr import dxtbx from dxtbx.model.experiment_list import ExperimentListFactory from dxtbx.model.crystal import CrystalFactory from dials.algorithms.indexing.compare_orientation_matrices \ import rotation_matrix_differences from dials.array_family import flex from dials.command_line.find_spots import phil_scope as find_spots_phil_scope from cxid9114.refine import metrics from cxid9114 import utils from cxid9114.geom import geom_utils from cxid9114.spots import integrate, spot_utils from cxid9114 import parameters from cxid9114.sim import sim_utils from cctbx import miller, sgtbx from cxid9114 import utils from cxid9114.bigsim import sim_spectra from cxid9114.refine.jitter_refine import make_param_list spot_par = find_spots_phil_scope.fetch(source=parse("")).extract() spot_par.spotfinder.threshold.dispersion.global_threshold = 40 spot_par.spotfinder.threshold.dispersion.gain = 28 spot_par.spotfinder.threshold.dispersion.kernel_size = [2, 2] spot_par.spotfinder.threshold.dispersion.sigma_strong = 1 spot_par.spotfinder.threshold.dispersion.sigma_background = 6 spot_par.spotfinder.filter.min_spot_size = 3 spot_par.spotfinder.force_2d = True odir = args.odir odirj = os.path.join(odir, "job%d" % worker_Id) #all_pkl_files = [s for sl in \ # [ files for _,_, files in os.walk(odir)]\ # for s in sl if s.endswith("pkl")] #print "Found %d pkl files already in %s!" \ # % (len(all_pkl_files), odir) if not os.path.exists(odirj): os.makedirs(odirj) hkl_tol = .15 run = 61 shot_idx = 0 ENERGIES = [parameters.ENERGY_LOW, parameters.ENERGY_HIGH] # colors of the beams FF = [10000, None] cryst_descr = { '__id__': 'crystal', 'real_space_a': (79, 0, 0), 'real_space_b': (0, 79, 0), 'real_space_c': (0, 0, 38), 'space_group_hall_symbol': '-P 4 2' } crystalAB = CrystalFactory.from_dict(cryst_descr) sfall_main = sim_spectra.load_spectra("../bigsim/test_sfall.h5") FFdat = [sfall_main[19], sfall_main[110]] FLUX = [1e11, 1e11] # fluxes of the beams chanA_flux = 1e11 chanB_flux = 1e11 FLUXdat = [chanA_flux, chanB_flux] GAIN = 1 waveA = parameters.ENERGY_CONV / ENERGIES[0] waveB = parameters.ENERGY_CONV / ENERGIES[1] from cxid9114.bigsim.bigsim_geom import DET, BEAM detector = DET print("Rank %d Begin" % worker_Id) for i_data in range(args.num_trials): pklname = "%s_rank%d_data%d.pkl" % (ofile, worker_Id, i_data) pklname = os.path.join(odirj, pklname) print("<><><><><><><") print("Job %d: trial %d / %d" % (worker_Id, i_data + 1, args.num_trials)) print("<><><><><><><") if (worker_Id == 0 and i_data % smi_stride == 0 and cuda): print("GPU status") os.system("nvidia-smi") print("\n\n") print("CPU memory usage") mem_usg = """ps -U dermen --no-headers -o rss | awk '{ sum+=$1} END {print int(sum/1024) "MB consumed by CPU user"}'""" os.system(mem_usg) beamA = deepcopy(BEAM) beamB = deepcopy(BEAM) beamA.set_wavelength(waveA) beamB.set_wavelength(waveB) np.random.seed(args.seed) crystalAB = CrystalFactory.from_dict(cryst_descr) randnums = np.random.random(3) Rrand = random_rotation(1, randnums) crystalAB.set_U(Rrand.ravel()) #pert = np.random.uniform(0.0001/2/np.pi, 0.0003 / 2. /np.pi) #print("PERT %f" % pert) #Rsmall = random_rotation(0.00001, randnums ) #pert) params_lst = make_param_list(crystalAB, DET, BEAM, 1, rot=0.08, cell=.0000001, eq=(1, 1, 0), min_Ncell=23, max_Ncell=24, min_mos_spread=0.02, max_mos_spread=0.08) Ctruth = params_lst[0]['crystal'] print Ctruth.get_unit_cell().parameters() print crystalAB.get_unit_cell().parameters() init_comp = rotation_matrix_differences((Ctruth, crystalAB)) init_rot = float(init_comp.split("\n")[-2].split()[2]) if use_data_spec: print "NOT IMPLEMENTED, Using a phony 2col spectrum to simulate the data" data_fluxes = FLUXdat data_energies = [parameters.ENERGY_LOW, parameters.ENERGY_HIGH] data_ff = FFdat else: print "Using a phony two color spectrum to simulate the data" data_fluxes = FLUXdat data_energies = [parameters.ENERGY_LOW, parameters.ENERGY_HIGH] data_ff = FFdat print("Truth crystal Misorientation deviation: %f deg" % init_rot) if args.truth_cryst: print "Using truth crystal" dataCryst = Ctruth else: print "Not using truth crystal" dataCryst = crystalAB if not make_background: print "SIMULATING Flat-Fhkl IMAGES" simsAB = sim_utils.sim_twocolors2( crystalAB, detector, BEAM, FF, [parameters.ENERGY_LOW, parameters.ENERGY_HIGH], FLUX, pids=None, Gauss=Gauss, cuda=cuda, oversample=oversample, Ncells_abc=Ncells_abc, mos_dom=mos_doms, mos_spread=mos_spread, exposure_s=exposure_s, beamsize_mm=beamsize_mm, device_Id=device_Id, boost=boost) if make_background: print("MAKING BACKGROUND") spec_file = h5py.File("../bigsim/simMe_data_run62.h5", "r") ave_spec = np.mean(spec_file["hist_spec"][()], axis=0) data_fluxes = [ave_spec[19], ave_spec[110]] data_energies = spec_file["energy_bins"][()][[19, 110]] data_ff = [1, 1] #*len(data_energies) only_water = True else: only_water = False print "SIULATING DATA IMAGE" print data_fluxes simsDataSum = sim_utils.sim_twocolors2(dataCryst, detector, BEAM, data_ff, data_energies, data_fluxes, pids=None, Gauss=Gauss, cuda=cuda, oversample=oversample, Ncells_abc=Ncells_abc, accumulate=True, mos_dom=mos_doms, mos_spread=mos_spread, boost=boost, exposure_s=exposure_s, beamsize_mm=beamsize_mm, only_water=only_water, device_Id=device_Id) simsDataSum = np.array(simsDataSum) if make_background: bg_out = h5py.File(bg_name, "w") bg_out.create_dataset("bigsim_d9114", data=simsDataSum[0]) print "Background made! Saved to file %s" % bg_name sys.exit() if add_background: print("ADDING BG") background = h5py.File(bg_name, "r")['bigsim_d9114'][()] bg_scale = np.sum([39152412349.12075, 32315440627.406036]) bg_scale = np.sum(data_fluxes) / bg_scale print "%.3e backgorund scale" % bg_scale print "BG shape", background.shape simsDataSum[0] += background * bg_scale if add_noise: print("ADDING NOISE") for pidx in range(1): SIM = nanoBragg(detector=DET, beam=BEAM, panel_id=pidx) SIM.exposure_s = exposure_s SIM.beamsize_mm = beamsize_mm SIM.flux = np.sum(data_fluxes) SIM.detector_psf_kernel_radius_pixels = 5 SIM.detector_psf_type = shapetype.Unknown # for CSPAD SIM.detector_psf_fwhm_mm = 0 SIM.quantum_gain = 28 SIM.raw_pixels = flex.double(simsDataSum[pidx].ravel()) SIM.add_noise() simsDataSum[pidx] = SIM.raw_pixels.as_numpy_array()\ .reshape(simsDataSum[0].shape) SIM.free_all() del SIM if args.write_img: print "SAVING DATAFILE" h5name = "%s_rank%d_data%d.h5" % (ofile, worker_Id, i_data) h5name = os.path.join(odirj, h5name) fout = h5py.File(h5name, "w") fout.create_dataset("bigsim_d9114", data=simsDataSum[0]) fout.create_dataset("crystalAB", data=crystalAB.get_A()) fout.create_dataset("dataCryst", data=dataCryst.get_A()) fout.close() if args.write_sim_img: print "SAVING DATAFILE" for i_sim in simsAB: sim_h5name = "%s_rank%d_sim%d_%d.h5" % (ofile, worker_Id, i_data, i_sim) sim_h5name = os.path.join(odirj, sim_h5name) from IPython import embed embed() fout = h5py.File(sim_h5name, "w") fout.create_dataset("bigsim_d9114", data=simsAB[i_sim][0]) fout.create_dataset("crystalAB", data=crystalAB.get_A()) fout.create_dataset("dataCryst", data=dataCryst.get_A()) fout.close() print "RELFS FROM SIMS" refl_simA = spot_utils.refls_from_sims(simsAB[0], detector, beamA, thresh=thresh) refl_simB = spot_utils.refls_from_sims(simsAB[1], detector, beamB, thresh=thresh) if use_dials_spotter: print("DIALS SPOTTING") El = utils.explist_from_numpyarrays(simsDataSum, DET, beamA) refl_data = flex.reflection_table.from_observations(El, spot_par) print("Found %d refls using DIALS spot finder" % len(refl_data)) else: refl_data = spot_utils.refls_from_sims(simsDataSum, detector, beamA,\ thresh=thresh) print("Found %d refls using threshold" % len(refl_data)) if len(refl_data) == 0: print "Rank %d: No reflections found! " % (worker_Id) continue residA = metrics.check_indexable2(refl_data, refl_simA, detector, beamA, crystalAB, hkl_tol) residB = metrics.check_indexable2(refl_data, refl_simB, detector, beamB, crystalAB, hkl_tol) sg96 = sgtbx.space_group(" P 4nw 2abw") FA = sfall_main[19] # utils.open_flex('SA.pkl') # ground truth values FB = sfall_main[110] #utils.open_flex('SB.pkl') # ground truth values HA = tuple([hkl for hkl in FA.indices()]) HB = tuple([hkl for hkl in FB.indices()]) HA_val_map = {h: data for h, data in izip(FA.indices(), FA.data())} HB_val_map = {h: data for h, data in izip(FB.indices(), FB.data())} def get_val_at_hkl(hkl, val_map): poss_equivs = [ i.h() for i in miller.sym_equiv_indices(sg96, hkl).indices() ] in_map = False for hkl2 in poss_equivs: if hkl2 in val_map: # fast lookup in_map = True break if in_map: return hkl2, val_map[hkl2] else: return (None, None, None), -1 filt = 1 #True #`False #True if filt: _, all_HiA, _ = spot_utils.refls_to_hkl(refl_simA, detector, beamA, crystal=crystalAB, returnQ=True) all_treeA = cKDTree(all_HiA) nnA = all_treeA.query_ball_point(all_HiA, r=1e-7) _, all_HiB, _ = spot_utils.refls_to_hkl(refl_simB, detector, beamB, crystal=crystalAB, returnQ=True) all_treeB = cKDTree(all_HiB) nnB = all_treeB.query_ball_point(all_HiB, r=1e-7) NreflA = len(refl_simA) NreflB = len(refl_simB) drop_meA = [] for i, vals in enumerate(nnA): if i in drop_meA: continue if len(vals) > 1: pids = [refl_simA[v]['panel'] for v in vals] if len(set(pids)) == 1: refl_vals = refl_simA.select( flex.bool( [i_v in vals for i_v in np.arange(NreflA)])) x, y, z = spot_utils.xyz_from_refl(refl_vals) allI = [r['intensity.sum.value'] for r in refl_vals] allI = sum(allI) xm = np.mean(x) ym = np.mean(y) zm = np.mean(z) drop_meA.extend(vals[1:]) x1b, x2b, y1b, y2b, z1b, z2b = zip( *[r['bbox'] for r in refl_vals]) keep_me = vals[0] # indexing order is important to modify as reference refl_simA['intensity.sum.value'][keep_me] = allI refl_simA['xyzobs.px.value'][keep_me] = (xm, ym, zm) refl_simA['bbox'][keep_me] = (min(x1b), max(x2b),\ min(y1b), max(y2b), min(z1b), max(z2b)) else: drop_meA.append(vals) print vals if drop_meA: keep_meA = np.array([i not in drop_meA for i in range(NreflA)]) refl_simA = refl_simA.select(flex.bool(keep_meA)) NreflA = len(refl_simA) drop_meB = [] for i, vals in enumerate(nnB): if i in drop_meB: continue if len(vals) > 1: pids = [refl_simB[v]['panel'] for v in vals] if len(set(pids)) == 1: print vals # merge_spots(vals) refl_vals = refl_simB.select( flex.bool( [i_v in vals for i_v in np.arange(NreflB)])) x, y, z = spot_utils.xyz_from_refl(refl_vals) allI = [r['intensity.sum.value'] for r in refl_vals] allI = sum(allI) xm = np.mean(x) ym = np.mean(y) zm = np.mean(z) drop_meB.extend(vals[1:]) x1b, x2b, y1b, y2b, z1b, z2b = zip( *[r['bbox'] for r in refl_vals]) keep_me = vals[0] refl_simB['intensity.sum.value'][keep_me] = allI refl_simB['xyzobs.px.value'][keep_me] = (xm, ym, zm) refl_simB['bbox'][keep_me] = (min(x1b), max(x2b), min(y1b),\ max(y2b), min(z1b), max(z2b)) else: drop_meB.append(vals) print vals if drop_meB: keep_meB = [i not in drop_meB for i in range(NreflB)] refl_simB = refl_simB.select(flex.bool(keep_meB)) NreflB = len(refl_simB) ## remake the trees given the drops _, all_HiA = spot_utils.refls_to_hkl(refl_simA, detector, beamA, crystal=crystalAB, returnQ=False) all_treeA = cKDTree(all_HiA) _, all_HiB = spot_utils.refls_to_hkl(refl_simB, detector, beamB, crystal=crystalAB, returnQ=False) #all_treeB = cKDTree(all_HiB) ## CHECK if same HKL, indexed by both colors # exists on multiple panels, and if so, delete... nnAB = all_treeA.query_ball_point(all_HiB, r=1e-7) drop_meA = [] drop_meB = [] for iB, iA_vals in enumerate(nnAB): if len(iA_vals) > 0: assert (len(iA_vals) == 1) iA = iA_vals[0] pidA = refl_simA[iA]['panel'] pidB = refl_simB[iB]['panel'] if pidA != pidB: drop_meA.append(iA) drop_meB.append(iB) if drop_meA: keep_meA = [i not in drop_meA for i in range(NreflA)] refl_simA = refl_simA.select(flex.bool(keep_meA)) if drop_meB: keep_meB = [i not in drop_meB for i in range(NreflB)] refl_simB = refl_simB.select(flex.bool(keep_meB)) # ---- Done with edge case filters# # reflections per panel rpp = spot_utils.refls_by_panelname(refl_data) rppA = spot_utils.refls_by_panelname(refl_simA) rppB = spot_utils.refls_by_panelname(refl_simB) DATA = { "D": [], "IA": [], "IB": [], "h2": [], "k2": [], "l2": [], "h": [], "k": [], "l": [], "PA": [], "PB": [], "FA": [], "FB": [], "iA": [], "iB": [], "Nstrong": [], "pid": [], "delta_pix": [], "deltaX": [], "deltaY": [] } all_int_me = [] # now set up boundboxes and integrate if tilt_plane_integration: mask = np.ones(simsDataSum.shape).astype(np.bool) print "Using tilt plane integration!" else: print "Not using tilt plane integration, just basic spot thresh integration " for pid in rpp: if tilt_plane_integration: Is, Ibk, noise, pix_per = \ integrate.integrate3( rpp[pid], mask[pid], simsDataSum[pid], gain=28) #nom_gain) R = rpp[pid] if pid in rppA: # are there A-channel reflections on this panel inA = True RA = rppA[pid] xA, yA, _ = spot_utils.xyz_from_refl(RA) pointsA = np.array(zip(xA, yA)) HA, HiA, QA = spot_utils.refls_to_hkl(RA, detector, beamA, crystal=crystalAB, returnQ=True) else: inA = False if pid in rppB: # are there B channel reflections on this channel inB = True RB = rppB[pid] xB, yB, _ = spot_utils.xyz_from_refl(RB) pointsB = np.array(zip(xB, yB)) HB, HiB, QB = spot_utils.refls_to_hkl(RB, detector, beamB, crystal=crystalAB, returnQ=True) else: inB = False x, y, _ = spot_utils.xyz_from_refl(R) x = np.array(x) y = np.array(y) panX, panY = detector[pid].get_image_size() mergesA = [] mergesB = [] if inA and inB: # are there both A and B channel reflections ? If so, lets find out which ones have same hkl # make tree structure for merging the spots treeA = cKDTree(pointsA) treeB = cKDTree(pointsB) QA = geom_utils.res_on_panel(detector[pid], beamA) QAmag = np.linalg.norm(QA, axis=2) * 2 * np.pi detdist = detector[pid].get_distance() pixsize = detector[pid].get_pixel_size()[0] merge_me = [] for p in pointsA: iix, iiy = int(p[0]), int(p[1]) q = QAmag[iiy, iix] radA = detdist * np.tan( 2 * np.arcsin(q * waveA / 4 / np.pi)) / pixsize radB = detdist * np.tan( 2 * np.arcsin(q * waveB / 4 / np.pi)) / pixsize rmax = np.abs(radA - radB) split_spot_pairs = treeB.query_ball_point(x=p, r=rmax + sz) merge_me.append(split_spot_pairs) #rmax = geom_utils.twocolor_deltapix(detector[pid], beamA, beamB) #merge_me = treeA.query_ball_tree(treeB, r=rmax + sz) for iA, iB in enumerate(merge_me): if not iB: continue iB = iB[0] # check that the miller indices are the same if not all([i == j for i, j in zip(HiA[iA], HiB[iB])]): continue x1A, x2A, y1A, y2A, _, _ = RA[iA]['bbox'] # shoebox'].bbox x1B, x2B, y1B, y2B, _, _ = RB[iB]['bbox'] # shoebox'].bbox xlow = max([0, min((x1A, x1B)) - sz]) xhigh = min([panX, max((x2A, x2B)) + sz]) ylow = max([0, min((y1A, y1B)) - sz]) yhigh = min([panY, max((y2A, y2B)) + sz]) #if iA==79: # embed() # integrate me if I am in the bounding box! int_me = np.where((xlow < x) & (x < xhigh) & (ylow < y) & (y < yhigh))[0] if not int_me.size: continue mergesA.append(iA) mergesB.append(iB) # integrate the spot, this will change depending on data or simulation totalI = 0 totalCOM = 0 for ref_idx in int_me: if tilt_plane_integration: totalI += Is[ref_idx] else: totalI += rpp[pid][ref_idx]["intensity.sum.value"] totalCOM += np.array( rpp[pid][ref_idx]["xyzobs.px.value"]) totalCOM /= len(int_me) PA = RA[iA]['intensity.sum.value'] PB = RB[iB]['intensity.sum.value'] # get the hkl structure factor, and the sym equiv hkl (h, k, l) = HiA[iA] # NOTE: same for A and B channels (h2, k2, l2), FA = get_val_at_hkl((h, k, l), HA_val_map) _, FB = get_val_at_hkl( (h, k, l), HB_val_map) # NOTE: no need to return h2,k2,l2 twice #if FB==-1 or FA==-1: # continue DATA['h'].append(h) DATA['k'].append(k) DATA['l'].append(l) DATA['h2'].append(h2) DATA['k2'].append(k2) DATA['l2'].append(l2) DATA['D'].append(totalI) DATA['PA'].append(PA) DATA['PB'].append(PB) DATA['FA'].append(FA) DATA['FB'].append(FB) DATA['IA'].append(abs(FA)**2) DATA['IB'].append(abs(FB)**2) DATA['pid'].append(pid) DATA["Nstrong"].append(int_me.size) DATA["iA"].append(iA) DATA["iB"].append(iB) all_int_me.append(int_me) # NOTE: stash the sim-data distance (COM to COM) posA = RA[iA]['xyzobs.px.value'] posB = RB[iB]['xyzobs.px.value'] simCOM = np.mean([posA, posB], axis=0) DATA["delta_pix"].append( distance.euclidean(totalCOM[:2], simCOM[:2])) DATA["deltaX"].append(totalCOM[0] - simCOM[0]) DATA["deltaY"].append(totalCOM[1] - simCOM[1]) if inA: for iA, ref in enumerate(RA): if iA in mergesA: continue x1A, x2A, y1A, y2A, _, _ = RA[iA][ 'bbox'] # ['shoebox'].bbox xlow = max((0, x1A - sz)) xhigh = min((panX, x2A + sz)) ylow = max((0, y1A - sz)) yhigh = min((panY, y2A + sz)) int_me = np.where((xlow < x) & (x < xhigh) & (ylow < y) & (y < yhigh))[0] if not int_me.size: continue totalI = 0 totalCOM = 0 for ref_idx in int_me: if tilt_plane_integration: totalI += Is[ref_idx] else: totalI += rpp[pid][ref_idx]["intensity.sum.value"] totalCOM += np.array( rpp[pid][ref_idx]["xyzobs.px.value"]) totalCOM /= len(int_me) PA = RA[iA]['intensity.sum.value'] PB = 0 # crucial ;) # get the hkl structure factor, and the sym equiv hkl (h, k, l) = HiA[iA] # NOTE: same for A and B channels (h2, k2, l2), FA = get_val_at_hkl((h, k, l), HA_val_map) _, FB = get_val_at_hkl( (h, k, l), HB_val_map) # NOTE: no need to return h2,k2,l2 twice #if FA==-1 or FB==-1: # continue DATA['h'].append(h) DATA['k'].append(k) DATA['l'].append(l) DATA['h2'].append(h2) DATA['k2'].append(k2) DATA['l2'].append(l2) DATA['D'].append(totalI) DATA['PA'].append(PA) DATA['PB'].append(PB) DATA['FA'].append(FA) DATA['FB'].append(FB) DATA['IA'].append(abs(FA)**2) DATA['IB'].append(abs(FB)**2) DATA['pid'].append(pid) DATA["Nstrong"].append(int_me.size) DATA["iA"].append(iA) DATA["iB"].append(np.nan) all_int_me.append(int_me) # NOTE: stash the sim-data distance (COM to COM) simCOM = np.array(RA[iA]['xyzobs.px.value']) DATA["delta_pix"].append( distance.euclidean(totalCOM[:2], simCOM[:2])) DATA["deltaX"].append(totalCOM[0] - simCOM[0]) DATA["deltaY"].append(totalCOM[1] - simCOM[1]) if inB: for iB, ref in enumerate(RB): if iB in mergesB: continue x1B, x2B, y1B, y2B, _, _ = RB[iB]['bbox'] # shoebox'].bbox xlow = max((0, x1B - sz)) xhigh = min((panX, x2B + sz)) ylow = max((0, y1B - sz)) yhigh = min((panY, y2B + sz)) # subimg = simsDataSum[pid][ylow:yhigh, xlow:xhigh] # bg = 0 int_me = np.where((xlow < x) & (x < xhigh) & (ylow < y) & (y < yhigh))[0] if not int_me.size: continue totalI = 0 totalCOM = 0 for ref_idx in int_me: if tilt_plane_integration: totalI += Is[ref_idx] else: totalI += rpp[pid][ref_idx]["intensity.sum.value"] totalCOM += np.array( rpp[pid][ref_idx]["xyzobs.px.value"]) totalCOM /= len(int_me) PA = 0 # crucial ;) PB = RB[iB]['intensity.sum.value'] # get the hkl structure factor, and the sym equiv hkl (h, k, l) = HiB[iB] # NOTE: same for A and B channels (h2, k2, l2), FB = get_val_at_hkl((h, k, l), HB_val_map) _, FA = get_val_at_hkl( (h, k, l), HA_val_map) # NOTE: no need to return h2,k2,l2 twice #if FA==-1 or FB==-1: # continue DATA['h'].append(h) DATA['k'].append(k) DATA['l'].append(l) DATA['h2'].append(h2) DATA['k2'].append(k2) DATA['l2'].append(l2) DATA['D'].append(totalI) DATA['PA'].append(PA) DATA['PB'].append(PB) DATA['FA'].append(FA) DATA['FB'].append(FB) DATA['IA'].append(abs(FA)**2) DATA['IB'].append(abs(FB)**2) DATA['pid'].append(pid) DATA["Nstrong"].append(int_me.size) DATA["iA"].append(np.nan) DATA["iB"].append(iB) all_int_me.append(int_me) # NOTE: stash the sim-data distance (COM to COM) simCOM = np.array(RB[iB]['xyzobs.px.value']) DATA["delta_pix"].append( distance.euclidean(totalCOM[:2], simCOM[:2])) DATA["deltaX"].append(totalCOM[0] - simCOM[0]) DATA["deltaY"].append(totalCOM[1] - simCOM[1]) df = pandas.DataFrame(DATA) df["run"] = run df["shot_idx"] = shot_idx df['gain'] = GAIN if use_data_spec: print "Setting LA, LB as sums over flux regions A,B" df['LA'] = data_fluxes[:75].sum() df['LB'] = data_fluxes[75:].sum() else: print "Setting LA LB as data_fluxes" df['LA'] = data_fluxes[0] df["LB"] = data_fluxes[1] df['K'] = FF[0]**2 * FLUX[0] df["rhs"] = df.gain * (df.IA * df.LA * (df.PA / df.K) + df.IB * df.LB * (df.PB / df.K)) df["lhs"] = df.D #df['data_name'] = data_name df['init_rot'] = init_rot df.to_pickle(pklname) print("PLOT") if args.plot: import pylab as plt plt.plot(df.lhs, df.rhs, '.') plt.show() print("DonDonee")
parser.add_argument('--filt', action='store_true', dest='filt') parser.add_argument('--gauss', action='store_true') args = parser.parse_args() import pandas from joblib import Parallel, delayed import numpy as np from itertools import izip import sys #from cxid9114 import utils from cxid9114.bigsim import sim_spectra #from cxid9114.parameters import ENERGY_HIGH, ENERGY_LOW import h5py from cxid9114.solvers.filter_outliers import filter_outliers spec = sim_spectra.load_spectra("bigsim_rocketships/test_sfall.h5") spec_f = h5py.File("bigsim_rocketships/test_data.h5", "r") en_bins = spec_f["energy_bins"][()] ilow = 19 # abs(en_bins - ENERGY_LOW).argmin() ihigh = 110 #abs(en_bins - ENERGY_HIGH).argmin() SA = spec[ilow].amplitudes() # energy A (lower energy) SB = spec[ihigh].amplitudes() # energy B (higher energy) # load the integrated data df = pandas.read_pickle(args.i) df.reset_index(inplace=True) if args.hanom: hkey = ['hAnom', 'kAnom', 'lAnom'] else:
def run_paramList(Ntrials, odir, tag, rank, n_jobs, pkl_file): import os import sys from copy import deepcopy import numpy as np import h5py from IPython import embed from scipy.ndimage.morphology import binary_dilation import scitbx from scitbx.array_family import flex from scitbx.matrix import sqr,col from simtbx.nanoBragg import shapetype from simtbx.nanoBragg import nanoBragg import libtbx.load_env # possibly implicit from libtbx.development.timers import Profiler from dxtbx.model.crystal import CrystalFactory from cctbx import crystal,crystal_orientation from cxid9114 import utils from cxid9114.sim import sim_utils from cxid9114.spots import spot_utils from cxid9114.bigsim.bigsim_geom import DET,BEAM from cxid9114.parameters import ENERGY_CONV, ENERGY_HIGH, ENERGY_LOW from cxid9114.refine.jitter_refine import make_param_list from cxid9114.bigsim import sim_spectra from LS49.sim.step4_pad import microcrystal data_pack = utils.open_flex(pkl_file) CRYST = data_pack['crystalAB'] mos_spread_deg=0.015 mos_doms=1000 beam_size_mm=0.001 exposure_s=1 use_microcrystal=True Deff_A = 2200 length_um = 2.2 timelog = False crystal = microcrystal(Deff_A = Deff_A, length_um = length_um, beam_diameter_um = beam_size_mm*1000, verbose=False) spec_file = h5py.File("simMe_data_run62.h5", "r") spec_data = spec_file["hist_spec"] Umat_data = spec_file["Umats"] en_chans = spec_file["energy_bins"][()] ilow = np.abs(en_chans - ENERGY_LOW).argmin() ihigh = np.abs(en_chans - ENERGY_HIGH).argmin() wave_chans = ENERGY_CONV/en_chans sfall_main = sim_spectra.load_spectra("test_sfall.h5") refls_strong = data_pack['refls_strong'] strong_mask_img = spot_utils.strong_spot_mask( refls_strong, (1800,1800) ) # edge detection in the ground truth strong mask image reference_img = (binary_dilation(strong_mask_img, iterations=1).astype(int) - strong_mask_img.astype(int) ).astype(bool) param_fileout = os.path.join( odir, "rank%d_%s.pkl" % (rank, tag)) param_list = make_param_list( CRYST, DET, BEAM, Ntrials, rot=0.09, cell=0.1, eq=(1,1,0), min_Ncell=20, max_Ncell=40, min_mos_spread=0.005, max_mos_spread=0.02) for p in param_list: print(p['crystal'].get_unit_cell().parameters()) shot_idx = int(data_pack['img_f'].split("_")[-1].split(".")[0]) Fluxes = spec_data[2] #shot_idx] Pmax = param_list[0] F1max = 0 for i_trial in range(Ntrials): print ("<><><><><><><><><><><><><><>") print ("Job %d; Trial %d / %d" % (rank, i_trial+1, Ntrials)) print ("<><><><><><><><><><><><><><>") if (rank==0 and i_trial % smi_stride==0): print("GPU status") os.system("nvidia-smi") print("\n\n") print("CPU memory usage") mem_usg= """ps -U dermen --no-headers -o rss | awk '{ sum+=$1} END {print int(sum/1024) "MB consumed by CPU user"}'""" os.system(mem_usg) assert (len(wave_chans)==len(Fluxes)==len(sfall_main)) if np.sum(Fluxes)==0: print ("Cannot simulate with an all-zeros spectrum!") sys.exit() N = crystal.number_of_cells(sfall_main[0].unit_cell()) Ncells_abc = (N,N,N) if force_twocolor: Fluxes *= 0 Fluxes[ilow] = 1e12 Fluxes[ihigh]=1e12 P = param_list[i_trial] simsAB = sim_utils.sim_twocolors2( P['crystal'], DET, BEAM, sfall_main, en_chans, Fluxes, pids = None, profile="gauss", oversample=0, Ncells_abc = Ncells_abc, #P['Ncells_abc'], mos_dom=mos_doms, verbose=verbose, mos_spread=mos_spread_deg, #@mos_spread=P['mos_spread'], cuda=True, device_Id=rank, beamsize_mm=beamsize_mm, exposure_s=exposure_s, boost=crystal.domains_per_crystal) out = np.sum( [ simsAB[i][0] for i in simsAB.keys() if simsAB[i]], axis=0) if out.shape==(): print("This simsAB output has an empty shape, something is wrong!") sys.exit() trial_refls = spot_utils.refls_from_sims([out], DET, BEAM, thresh=thresh) trial_spotmask = spot_utils.strong_spot_mask( trial_refls, (1800,1800) ) trial_img = (binary_dilation(trial_spotmask, iterations=1).astype(int) - trial_spotmask.astype(int) ).astype(bool) comp_img = trial_img.astype(int) + reference_img.astype(int)*2 Nfalse_neg = (comp_img==2).sum() # reference has signal but trial doesnt Nfalse_pos = (comp_img==1).sum() # trial has signal but reference doesnt Ntrue_pos = (comp_img==3).sum() # Precision = float(Ntrue_pos) / (Ntrue_pos + Nfalse_pos) Recall = float(Ntrue_pos) / (Ntrue_pos + Nfalse_neg) F1 = 2.*(Precision*Recall) / (Precision+Recall) if F1 > F1max: Pmax = {'crystal': P['crystal'], 'mos_spread': P['mos_spread'], 'Ncells_abc': P['Ncells_abc'], "F1": F1} F1max = F1 print("Rank %d, Trial %d: F1score = %.5f" % (rank, i_trial, F1)) utils.save_flex(Pmax, param_fileout) return F1max
def run_sim2smv(Nshot_max, odir, tag, rank, n_jobs, save_bragg=False, save_smv=True, save_h5 =False, return_pixels=False): import os import h5py import math import sys import numpy as np from IPython import embed from cxid9114.bigsim.bigsim_geom import DET,BEAM import scitbx from scitbx.array_family import flex from scitbx.matrix import sqr,col from simtbx.nanoBragg import shapetype from simtbx.nanoBragg import nanoBragg import libtbx.load_env # possibly implicit from libtbx.development.timers import Profiler from cctbx import crystal,crystal_orientation from LS49.sim.step4_pad import microcrystal from cxid9114 import utils from cxid9114.sim import sim_utils from cxid9114.parameters import ENERGY_CONV, ENERGY_HIGH, ENERGY_LOW from cxid9114.bigsim import sim_spectra from dxtbx.model.crystal import CrystalFactory odir_j = os.path.join( odir, "job%d" % rank) if not os.path.exists(odir_j): os.makedirs(odir_j) cryst_descr = {'__id__': 'crystal', 'real_space_a': (79, 0, 0), 'real_space_b': (0, 79, 0), 'real_space_c': (0, 0, 38), 'space_group_hall_symbol': '-P 4 2'} Crystal = CrystalFactory.from_dict(cryst_descr) mos_spread_deg=0.015 mos_doms=1000 beam_size_mm=0.001 exposure_s=1 use_microcrystal=True Deff_A = 2200 length_um = Deff_A/1000. timelog = False if add_background: background = utils.open_flex("background").as_numpy_array() crystal = microcrystal(Deff_A = Deff_A, length_um = length_um, beam_diameter_um = beam_size_mm*1000, verbose=False) spec_file = h5py.File("simMe_data_run62.h5", "r") spec_data = spec_file["hist_spec"] Umat_data = spec_file["Umats"] en_chans = spec_file["energy_bins"][()] ilow = np.abs(en_chans - ENERGY_LOW).argmin() ihigh = np.abs(en_chans - ENERGY_HIGH).argmin() wave_chans = ENERGY_CONV/en_chans sfall_main = sim_spectra.load_spectra("test_sfall.h5") Nshot = spec_data.shape[0] idx_range = np.array_split(np.arange(Nshot), n_jobs) Nshot_per_job = len(idx_range[rank]) if Nshot_max > 0 : Nshot_per_job = min( Nshot_max, Nshot_per_job) print ("Job %d: Simulating %d shots" % (rank, Nshot_per_job)) istart = idx_range[rank][0] istop = istart + Nshot_per_job smi_stride = 10 for idx in range( istart, istop): print ("<><><><><><><><><><><><><><>") print ("Job %d; Image %d (%d - %d)" % (rank, idx+1, istart, istop)) print ("<><><><><><><><><><><><><><>") smv_fileout = os.path.join( odir_j, "%s_%d.img" % (tag,idx)) h5_fileout = smv_fileout + ".h5" if os.path.exists(smv_fileout) and not overwrite and save_smv: print("Shot %s exists: moving on" % smv_fileout) continue if os.path.exists(h5_fileout) and not overwrite and save_h5: print("Shot %s exists: moving on" % h5_fileout) continue if (rank==0 and idx % smi_stride==0): print("GPU status") os.system("nvidia-smi") print("\n\n") print("CPU memory usage") mem_usg= """ps -U dermen --no-headers -o rss | awk '{ sum+=$1} END {print int(sum/1024) "MB consumed by CPU user"}'""" os.system(mem_usg) if force_index is not None: spec = spec_data[force_index] rotation = sqr(Umat_data[force_index]) else: spec = spec_data[idx] rotation = sqr(Umat_data[idx]) wavelength_A = np.mean(wave_chans) spectra = iter([(wave_chans, spec, wavelength_A)]) wavlen, flux, wavelength_A = next(spectra) # list of lambdas, list of fluxes, average wavelength assert wavelength_A > 0 assert (len(wavlen)==len(flux)==len(sfall_main)) if np.sum(flux)==0: continue N = crystal.number_of_cells(sfall_main[0].unit_cell()) Ncells_abc = (N,N,N) if not on_axis: Crystal.set_U(rotation) if idx_path is not None: assert( model_file is None) Crystal = utils.open_flex(idx_path)['crystalAB'] if model_file is not None: assert( idx_path is None) P = utils.open_flex(model_file) Crystal = P['crystal'] #mos_spread_deg = P['mos_spread'] #Ncells_abc = P['Ncells_abc'] if force_twocolor: flux *= 0 flux[ilow] = 1e11 flux[ihigh]=1e11 simsAB = sim_utils.sim_twocolors2( Crystal, DET, BEAM, sfall_main, en_chans, flux, pids = None, profile="gauss", oversample=1, Ncells_abc = Ncells_abc, mos_dom=mos_doms, verbose=verbose, mos_spread=mos_spread_deg, cuda=True, device_Id =rank, beamsize_mm=beamsize_mm, exposure_s=exposure_s, accumulate=True, boost=crystal.domains_per_crystal) if not simsAB: continue out = simsAB[0] if add_background: out = out + background if add_noise: SIM = nanoBragg(detector=DET, beam=BEAM) SIM.exposure_s = exposure_s SIM.flux = np.sum(flux) SIM.raw_pixels = flex.double(out.ravel()) SIM.detector_psf_kernel_radius_pixels=5; SIM.detector_psf_type=shapetype.Unknown # for CSPAD SIM.detector_psf_fwhm_mm=0 SIM.quantum_gain = 28 SIM.add_noise() out = SIM.raw_pixels.as_numpy_array().reshape(out.shape) f = h5py.File(h5_fileout, "w") f.create_dataset("bigsim_d9114", data=out, compression="lzf") ua,ub,uc = Crystal.get_real_space_vectors() f.create_dataset("real_space_a", data=ua) f.create_dataset("real_space_b", data=ub) f.create_dataset("real_space_c", data=uc) f.create_dataset("space_group_hall_symbol", data=Crystal.get_space_group().info().type().hall_symbol()) f.create_dataset("Umatrix", data=Crystal.get_U()) f.create_dataset("fluxes",data=flux) f.create_dataset("energies", data=en_chans) f.close() if npout is not None: np.save(npout, out) # SIM.raw_pixels.as_numpy_array())
def main(rank): device_Id = rank % ngpu worker_Id = node_id*ngpu + rank import os import sys from copy import deepcopy import glob from itertools import izip from scipy.spatial import distance import h5py import scipy.ndimage from IPython import embed import numpy as np import pandas from scipy.spatial import cKDTree from simtbx.nanoBragg import shapetype, nanoBragg from libtbx.phil import parse from scitbx.matrix import sqr import dxtbx from dxtbx.model.experiment_list import ExperimentListFactory from dxtbx.model.crystal import CrystalFactory from dials.algorithms.indexing.compare_orientation_matrices \ import rotation_matrix_differences from dials.array_family import flex from dials.command_line.find_spots import phil_scope as find_spots_phil_scope from cxid9114.refine import metrics from cxid9114 import utils from cxid9114.geom import geom_utils from cxid9114.spots import integrate, spot_utils from cxid9114 import parameters from cxid9114.sim import sim_utils from cctbx import miller, sgtbx from cxid9114 import utils from cxid9114.bigsim import sim_spectra from cxid9114.refine.jitter_refine import make_param_list spot_par = find_spots_phil_scope.fetch(source=parse("")).extract() spot_par.spotfinder.threshold.dispersion.global_threshold = 40 spot_par.spotfinder.threshold.dispersion.gain = GAIN spot_par.spotfinder.threshold.dispersion.kernel_size = [2,2] spot_par.spotfinder.threshold.dispersion.sigma_strong = 1 spot_par.spotfinder.threshold.dispersion.sigma_background = 6 spot_par.spotfinder.filter.min_spot_size = 3 spot_par.spotfinder.force_2d = True odir = args.odir odirj = os.path.join(odir, "job%d" % worker_Id) #all_pkl_files = [s for sl in \ # [ files for _,_, files in os.walk(odir)]\ # for s in sl if s.endswith("pkl")] #print "Found %d pkl files already in %s!" \ # % (len(all_pkl_files), odir) if not os.path.exists(odirj): os.makedirs(odirj) hkl_tol = .15 run = 61 shot_idx = 0 ENERGIES = [parameters.ENERGY_LOW, parameters.ENERGY_HIGH] # colors of the beams FF = [10000, None] cryst_descr = {'__id__': 'crystal', 'real_space_a': (79, 0, 0), 'real_space_b': (0, 79, 0), 'real_space_c': (0, 0, 38), 'space_group_hall_symbol': '-P 4 2'} crystalAB = CrystalFactory.from_dict(cryst_descr) sfall_main = sim_spectra.load_spectra("../bigsim/test_sfall.h5") FFdat = [sfall_main[19], sfall_main[110]] FLUX = [1e11, 1e11] # fluxes of the beams chanA_flux = np.random.uniform(1e11,1e12) chanB_flux = np.random.uniform(1e11,1e12) FLUXdat = [chanA_flux, chanB_flux] waveA = parameters.ENERGY_CONV / ENERGIES[0] waveB = parameters.ENERGY_CONV / ENERGIES[1] from cxid9114.bigsim.bigsim_geom import DET,BEAM detector = DET print("Rank %d Begin" % worker_Id) for i_data in range( args.num_trials): pklname = "%s_rank%d_data%d.pkl" % (ofile, worker_Id, i_data) pklname = os.path.join( odirj, pklname) print("<><><><><><><") print("Job %d: trial %d / %d" % ( worker_Id, i_data+1, args.num_trials )) print("<><><><><><><") if (worker_Id==0 and i_data % smi_stride==0 and cuda): print("GPU status") os.system("nvidia-smi") print("\n\n") print("CPU memory usage") mem_usg= """ps -U dermen --no-headers -o rss | awk '{ sum+=$1} END {print int(sum/1024) "MB consumed by CPU user"}'""" os.system(mem_usg) beamA = deepcopy(BEAM) beamB = deepcopy(BEAM) beamA.set_wavelength(waveA) beamB.set_wavelength(waveB) SCALE = np.random.uniform(0.1,10) np.random.seed(args.seed) crystalAB = CrystalFactory.from_dict(cryst_descr) randnums = np.random.random(3) Rrand = random_rotation(1, randnums) crystalAB.set_U(Rrand.ravel()) #pert = np.random.uniform(0.0001/2/np.pi, 0.0003 / 2. /np.pi) #print("PERT %f" % pert) #Rsmall = random_rotation(0.00001, randnums ) #pert) params_lst = make_param_list(crystalAB, DET, BEAM, 1, rot=0.08, cell=.0000001, eq=(1,1,0), min_Ncell=23, max_Ncell=24, min_mos_spread=0.02, max_mos_spread=0.08) Ctruth = params_lst[0]['crystal'] print Ctruth.get_unit_cell().parameters() print crystalAB.get_unit_cell().parameters() init_comp = rotation_matrix_differences((Ctruth, crystalAB)) init_rot = float(init_comp.split("\n")[-2].split()[2]) if use_data_spec: print "NOT IMPLEMENTED, Using a phony 2col spectrum to simulate the data" data_fluxes = FLUXdat data_energies = [parameters.ENERGY_LOW, parameters.ENERGY_HIGH] data_ff = FFdat else: print "Using a phony two color spectrum to simulate the data" data_fluxes = FLUXdat data_energies = [parameters.ENERGY_LOW, parameters.ENERGY_HIGH] data_ff = FFdat print ("Truth crystal Misorientation deviation: %f deg" % init_rot ) if args.truth_cryst: print "Using truth crystal" dataCryst = Ctruth else: print "Not using truth crystal" dataCryst = crystalAB if not make_background: print "SIMULATING Flat-Fhkl IMAGES" simsAB = sim_utils.sim_twocolors2( crystalAB, detector, BEAM, FF, [parameters.ENERGY_LOW, parameters.ENERGY_HIGH], FLUX, pids=None, Gauss=Gauss, cuda=cuda, oversample=oversample, Ncells_abc=Ncells_abc, mos_dom=mos_doms, mos_spread=mos_spread, exposure_s=exposure_s, beamsize_mm=beamsize_mm, device_Id=device_Id, boost=boost) if make_background: print("MAKING BACKGROUND") spec_file = h5py.File("../bigsim/simMe_data_run62.h5", "r") ave_spec = np.mean( spec_file["hist_spec"][()], axis=0) data_fluxes=[ave_spec[19], ave_spec[110] ] data_energies = spec_file["energy_bins"][()][[19,110]] data_ff = [1,1] #*len(data_energies) only_water=True else: only_water=False print "SIULATING DATA IMAGE" print data_fluxes simsDataSum = sim_utils.sim_twocolors2( dataCryst, detector, BEAM, data_ff, data_energies, data_fluxes, pids=None, Gauss=Gauss, cuda=cuda,oversample=oversample, Ncells_abc=Ncells_abc, accumulate=True, mos_dom=mos_doms, mos_spread=mos_spread, boost=boost, exposure_s=exposure_s, beamsize_mm=beamsize_mm, only_water=only_water, device_Id=device_Id) simsDataSum = SCALE*np.array(simsDataSum) if make_background: bg_out = h5py.File(bg_name, "w") bg_out.create_dataset("bigsim_d9114",data=simsDataSum[0]) print "Background made! Saved to file %s" % bg_name sys.exit() if add_background: print("ADDING BG") background = h5py.File(bg_name, "r")['bigsim_d9114'][()] bg_scale = np.sum([39152412349.12075, 32315440627.406036] ) bg_scale = np.sum(data_fluxes) / bg_scale print "%.3e backgorund scale" % bg_scale print "BG shape", background.shape simsDataSum[0] += background * bg_scale if add_noise: print("ADDING NOISE") for pidx in range(1): SIM = nanoBragg(detector=DET, beam=BEAM, panel_id=pidx) SIM.exposure_s = exposure_s SIM.beamsize_mm = beamsize_mm SIM.flux = np.sum(data_fluxes) SIM.detector_psf_kernel_radius_pixels=5; #SIM.detector_psf_type=shapetype.Gauss SIM.detector_psf_type=shapetype.Unknown # for CSPAD SIM.detector_psf_fwhm_mm=0 SIM.quantum_gain = GAIN SIM.raw_pixels = flex.double(simsDataSum[pidx].ravel()) SIM.add_noise() simsDataSum[pidx] = SIM.raw_pixels.as_numpy_array()\ .reshape(simsDataSum[0].shape) SIM.free_all() del SIM if args.write_img: print "SAVING DATAFILE" h5name = "%s_rank%d_data%d.h5" % (ofile, worker_Id, i_data) h5name = os.path.join( odirj, h5name) fout = h5py.File(h5name,"w" ) fout.create_dataset("bigsim_d9114", data=simsDataSum[0]) fout.create_dataset("crystalAB", data=crystalAB.get_A() ) fout.create_dataset("dataCryst", data=dataCryst.get_A() ) fout.close() if args.write_sim_img: print "SAVING DATAFILE" for i_sim in simsAB: sim_h5name = "%s_rank%d_sim%d_%d.h5" % (ofile, worker_Id, i_data, i_sim) sim_h5name = os.path.join( odirj, sim_h5name) fout = h5py.File(sim_h5name,"w" ) fout.create_dataset("bigsim_d9114", data=simsAB[i_sim][0]) fout.create_dataset("crystalAB", data=crystalAB.get_A() ) fout.create_dataset("dataCryst", data=dataCryst.get_A() ) fout.close() print "RELFS FROM SIMS" refl_simA = spot_utils.refls_from_sims(simsAB[0], detector, beamA, thresh=thresh) refl_simB = spot_utils.refls_from_sims(simsAB[1], detector, beamB, thresh=thresh) if use_dials_spotter: print("DIALS SPOTTING") El = utils.explist_from_numpyarrays(simsDataSum,DET,beamA) refl_data = flex.reflection_table.from_observations(El, spot_par) print("Found %d refls using DIALS spot finder" % len(refl_data)) else: refl_data = spot_utils.refls_from_sims(simsDataSum, detector, beamA,\ thresh=thresh) print ("Found %d refls using threshold" % len(refl_data)) if len(refl_data)==0: print "Rank %d: No reflections found! " % (worker_Id) continue residA = metrics.check_indexable2( refl_data, refl_simA, detector, beamA, crystalAB, hkl_tol) residB = metrics.check_indexable2( refl_data, refl_simB, detector, beamB, crystalAB, hkl_tol) FA = sfall_main[19] # utils.open_flex('SA.pkl') # ground truth values FB = sfall_main[110] #utils.open_flex('SB.pkl') # ground truth values HA = tuple([hkl for hkl in FA.indices()]) HB = tuple([hkl for hkl in FB.indices()]) HA_val_map = { h:data for h,data in izip(FA.indices(), FA.data())} HB_val_map = { h:data for h,data in izip(FB.indices(), FB.data())} Hmaps = [HA_val_map, HB_val_map] def get_val_at_hkl(hkl, val_map): sg96 = sgtbx.space_group(" P 4nw 2abw") poss_equivs = [i.h() for i in miller.sym_equiv_indices(sg96, hkl).indices()] in_map=False for hkl2 in poss_equivs: if hkl2 in val_map: # fast lookup in_map=True break if in_map: return hkl2, val_map[hkl2] else: return (None,None,None),-1 if use_data_spec: print "Setting LA, LB as sums over flux regions A,B" LA = data_fluxes[:75].sum() LB = data_fluxes[75:].sum() else: print "Setting LA LB as data_fluxes" LA = data_fluxes[0] LB = data_fluxes[1] K=FF[0] ** 2 * FLUX[0] L_at_color = [LA,LB] out = spot_utils.integrate_boxes( refl_data, simsDataSum[0], [refl_simA, refl_simB], DET, [beamA,beamB], crystalAB, delta_q=args.deltaq, gain=GAIN ) Nh = len(out[0]) rhs = [] lhs = [] all_H2 = [] all_PA = [] all_PB = [] all_FA = [] all_FB = [] for i in range(Nh): HKL = out[0][i] yobs = out[1][i] Pvals = out[2][i] ycalc = 0 for i_P, P in enumerate(Pvals): L = L_at_color[i_P] H2, F = get_val_at_hkl(HKL, Hmaps[i_P]) if i_P==0: all_FA.append(F) else: all_FB.append(F) ycalc += SCALE*L*P*abs(F)**2/K all_PA.append( Pvals[0]) all_PB.append( Pvals[1]) all_H2.append(H2) rhs.append(ycalc) lhs.append(yobs) df = pandas.DataFrame({"rhs":rhs, "lhs": lhs, "PA":all_PA, "PB":all_PB, "FA": all_FA, "FB": all_FB}) df["run"] = run df["shot_idx"] = shot_idx df['gain'] = SCALE df['LA'] = LA df['LB'] = LB df['K'] = K df['init_rot'] = init_rot h,k,l = zip(*all_H2) df['h2'] = h df['k2'] = k df['l2'] = l df.to_pickle(pklname) if args.plot: print("PLOT") import pylab as plt plt.plot( df.lhs, df.rhs, '.') plt.show() print("DonDonee")