-
Notifications
You must be signed in to change notification settings - Fork 0
/
oct.py
563 lines (486 loc) · 21.4 KB
/
oct.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
"""Tools and Wrappers for OCT"""
import os
import re
import uuid
import logging
import subprocess as sp
from glob import glob
from shutil import rmtree, copytree
from scipy.optimize import minimize
from QDYN.analytical_pulse import AnalyticalPulse
from math import atanh, tanh
import numpy as np
import QDYN
from QDYN.pulse import Pulse
from QDYN.analytical_pulse import AnalyticalPulse
from QDYN.prop_gate import get_prop_gate_of_t
from model import transmon_model
HADAMARD = np.array([[1, 1], [1, -1]], dtype=np.complex128) / np.sqrt(2)
PHASE = np.array([[1, 0], [0, np.exp(-0.25j*np.pi)]], dtype=np.complex128)
# gate targets
GATE = {
'H_left' : QDYN.gate2q.Gate2Q(np.kron(HADAMARD, np.eye(2)), name='H_L'),
'H_right' : QDYN.gate2q.Gate2Q(np.kron(np.eye(2), HADAMARD), name='H_R'),
'Ph_left' : QDYN.gate2q.Gate2Q(np.kron(PHASE, np.eye(2)), name='S_L'),
'Ph_right': QDYN.gate2q.Gate2Q(np.kron(np.eye(2), PHASE), name='S_R'),
'BGATE' : QDYN.gate2q.BGATE,
}
MAX_TRIALS = 200
w1 = 6000.0 # MHz
w2 = 5900.0 # MHz
wc = 6200.0 # MHz
alpha1 = -290.0 # MHz
alpha2 = -310.0 # MHz
g = 70.0 # MHz
def reset_pulse(pulse_dat, iteration):
"""Reset pulse_dat at the given iteration to the last available snapshot,
assuming that snapshots are available using the same name as pulse_dat,
with the `iteration` append to the file name (e.g. pulse.dat.100 for
pulse.dat). Snapshots must be at least 10 iterations older than the
current pulse"""
snapshot_list = glob("%s.*" % pulse_dat)
snapshots = {}
logger = logging.getLogger(__name__)
logger.debug("resetting in iter %d", iteration)
logger.debug("available snapshots: %s", str(snapshot_list))
for snapshot in snapshot_list:
try:
snapshot_iter = int(os.path.splitext(snapshot)[1][1:])
snapshots[snapshot_iter] = snapshot
except ValueError:
pass # ignore pulse.dat.prev
snapshot_iters = sorted(snapshots.keys())
os.unlink(pulse_dat)
while len(snapshot_iters) > 0:
snapshot_iter = snapshot_iters.pop()
if (iteration == 0) or (snapshot_iter + 10 < iteration):
logger.debug("accepted snapshot: %s (iter %d)",
snapshots[snapshot_iter], snapshot_iter)
QDYN.shutil.copy(snapshots[snapshot_iter], pulse_dat)
return
else:
logger.debug("rejected snapshot: %s (iter %d)",
snapshots[snapshot_iter], snapshot_iter)
logger.debug("no accepted snapshot")
def get_temp_runfolder(runfolder, scratch_root=None):
"""Return the path for an appropriate temporary runfolder (inside
$SCRATCH_ROOT) for the given "real" runfolder. The runfolder is guaranteed
to exist"""
if scratch_root is None:
assert 'SCRATCH_ROOT' in os.environ, \
"SCRATCH_ROOT environment variable must be defined"
scratch_root = os.environ['SCRATCH_ROOT']
temp_runfolder = str(uuid.uuid4())
if 'SLURM_JOB_ID' in os.environ:
temp_runfolder = "%s_%s" % (os.environ['SLURM_JOB_ID'], temp_runfolder)
temp_runfolder = os.path.join(scratch_root, runfolder, temp_runfolder)
QDYN.shutil.mkdir(temp_runfolder)
return temp_runfolder
def run_oct(
runfolder, continue_oct=False, g_a_int_min_initial=1.0e-5,
g_a_int_max=1.0e-1, g_a_int_converged=1.0e-7, use_threads=True,
scratch_root=None, print_stdout=True, monotonic=True, backtrack=True):
"""Run optimal control on the given runfolder. Adjust lambda_a if
necessary.
Assumes that the runfolder contains the files config and pulse1.dat, and
optionally, pulse.dat, and oct_iters.dat.
Creates (overwrites) the files pulse.dat and oct_iters.dat.
Also, a file config.oct is created that contains the last update to
lambda_a. The original config file will remain unchanged.
"""
logger = logging.getLogger(__name__)
config_data = QDYN.config.read_config_file(
os.path.join(runfolder, 'config'))
assert len(config_data['pulse'])
# prepare clean temp_runfolder
pulse_guess_dat = config_data['pulse'][0]['filename']
pulse_opt_dat = config_data['pulse'][0]['oct_outfile']
rf_pulse_guess_dat = os.path.join(runfolder, pulse_guess_dat)
rf_pulse_opt_dat = os.path.join(runfolder, pulse_opt_dat)
temp_runfolder = get_temp_runfolder(runfolder, scratch_root)
temp_pulse_opt_dat = os.path.join(temp_runfolder, pulse_opt_dat)
temp_config = os.path.join(temp_runfolder, 'config')
assert 'basis' in config_data['user_strings']
assert 'J_T' in config_data['user_strings']
assert 'gate' in config_data['user_strings']
QDYN.config.write_config(config_data, temp_config)
if pulse_is_converged(rf_pulse_opt_dat):
logger.warning("pulse %s is already converged. Skipping.",
rf_pulse_opt_dat)
return
ham_files = [line.get('filename', None) for line in config_data['ham']]
psi_files = [line.get('filename', None) for line in config_data['psi']]
pulse_files = [line.get('filename', None) for line in config_data['pulse']]
pulse_files.extend([line.get('oct_spectral_filter', None)
for line in config_data['pulse']])
user_files = [config_data['user_strings'].get(key, None)
for key in ['rwa_vector', 'gate']]
required_files = [fn for fn in (
ham_files + psi_files + pulse_files + user_files)
if fn is not None]
files_to_copy = list(required_files)
if continue_oct:
files_to_copy.extend([pulse_opt_dat, 'oct_iters.dat'])
for file in files_to_copy:
if os.path.isfile(os.path.join(runfolder, file)):
QDYN.shutil.copy(os.path.join(runfolder, file), temp_runfolder)
logger.debug("%s to temp_runfolder %s", file, temp_runfolder)
else:
if file in required_files:
raise IOError("%s does not exist in %s" % (file, runfolder))
logger.info("Starting optimization of %s (in %s)", runfolder,
temp_runfolder)
# run while monitoring convergence
with open(os.path.join(runfolder, 'oct.log'), 'wb', 0) as stdout:
# we assume that the value for lambda_a is badly chosen and iterate
# over optimizations until we find a good value
bad_lambda = True
pulse_explosion = False
trial = 0
given_up = False
while bad_lambda:
trial += 1
if trial > MAX_TRIALS:
bad_lambda = False
given_up = True
break # give up
env = os.environ.copy()
env['OMP_NUM_THREADS'] = '1'
if use_threads:
env['OMP_NUM_THREADS'] = '4'
if int(use_threads) > 1:
env['OMP_NUM_THREADS'] = '%d' % int(use_threads)
oct_proc = sp.Popen(
['qdyn_optimize', '--internal-units=GHz_units.txt', '.'],
cwd=temp_runfolder, env=env, stdout=sp.PIPE,
universal_newlines=True)
iteration = 0
g_a_int = 0.0
while True: # monitor STDOUT from oct
line = oct_proc.stdout.readline()
if print_stdout:
print(line, end='')
if line != '':
stdout.write(line.encode('ascii'))
m = re.search(r'^\s*(\d+) \| [\d.E+-]+ \| ([\d.E+-]+) \|',
line)
if m:
iteration = int(m.group(1))
try:
g_a_int = float(m.group(2))
except ValueError:
# account for Fortran dropping the 'E' in negative
# 3-digit exponents
g_a_int = float(m.group(2).replace('-', 'E-'))
# Every 50 iterations, we take a snapshot of the current
# pulse, so that "bad lambda" restarts continue from there
if (iteration > 0) and (iteration % 50 == 0):
QDYN.shutil.copy(temp_pulse_opt_dat,
temp_pulse_opt_dat+'.'+str(iteration))
# if the pulse changes in first iteration are too small, we
# lower lambda_a, unless lambda_a was previously adjusted
# to avoid exploding pulse values
if ((iteration == 1) and
(g_a_int < g_a_int_min_initial) and
(not pulse_explosion)):
logger.debug("pulse update too small: %g < %g",
g_a_int, g_a_int_min_initial)
logger.debug("Kill %d", oct_proc.pid)
if backtrack:
oct_proc.kill()
scale_lambda_a(temp_config, 0.5)
reset_pulse(temp_pulse_opt_dat, iteration)
break # next bad_lambda loop
# if the pulse update explodes, we increase lambda_a (and
# prevent it from decreasing again)
need_to_increase_lambda = False
if 'amplitude exceeds maximum value' in line:
need_to_increase_lambda = True
if monotonic:
if 'Loss of monotonic convergence' in line:
need_to_increase_lambda = True
if g_a_int > g_a_int_max:
need_to_increase_lambda = True
if need_to_increase_lambda:
pulse_explosion = True
if "Loss of monotonic convergence" in line:
logger.debug("loss of monotonic conversion")
else:
if g_a_int > g_a_int_max:
logger.debug("g_a_int = %g > %g",
g_a_int, g_a_int_max)
logger.debug("pulse explosion")
if backtrack:
logger.debug("Kill %d", oct_proc.pid)
oct_proc.kill()
scale_lambda_a(temp_config, 1.25)
reset_pulse(temp_pulse_opt_dat, iteration)
break # next bad_lambda loop
# if there are no significant pulse changes anymore, we
# stop the optimization prematurely
if iteration > 10 and g_a_int < g_a_int_converged:
logger.debug(
("pulse update insignificant "
"(converged): g_a_int = %g < %g"),
g_a_int, g_a_int_converged)
logger.debug("Kill %d", oct_proc.pid)
oct_proc.kill()
bad_lambda = False
# add a comment to pulse.dat to mark it converged
mark_pulse_converged(temp_pulse_opt_dat)
break # effectively break from bad_lambda loop
else: # line == ''
# OCT finished
bad_lambda = False
break # effectively break from bad_lambda loop
for file in [pulse_opt_dat, 'oct_iters.dat']:
if os.path.isfile(os.path.join(temp_runfolder, file)):
QDYN.shutil.copy(os.path.join(temp_runfolder, file), runfolder)
if os.path.isfile(temp_config):
QDYN.shutil.copy(temp_config, os.path.join(runfolder, 'config.oct'))
QDYN.shutil.rmtree(temp_runfolder)
logger.debug("Removed temp_runfolder %s", temp_runfolder)
if given_up:
# Giving up is permanent, so we can mark the guess pulse as final
# by storing it as the optimized pulse. That should prevent pointlessly
# re-runing OCT
if not os.path.isfile(rf_pulse_opt_dat):
QDYN.shutil.copy(rf_pulse_guess_dat, rf_pulse_opt_dat)
mark_pulse_converged(rf_pulse_opt_dat)
logger.info("Finished optimization (given up after too many "
"attempts): %s", runfolder)
else:
logger.info("Finished optimization: %s", runfolder)
def scale_lambda_a(config_file, factor):
"""Scale lambda_a in the given config file with the given factor"""
config = QDYN.config.read_config_file(config_file)
assert len(config['pulse']) == 1
config['pulse'][0]['oct_lambda_a'] = float(
"%.2e" % (factor * config['pulse'][0]['oct_lambda_a']))
QDYN.config.write_config(config, config_file)
def mark_pulse_converged(pulse_file):
"""Mark pulse file a converged by writing a comment to the header"""
if not pulse_is_converged(pulse_file):
p = Pulse.read(pulse_file)
p.preamble.append("# converged")
p.write(pulse_file)
def pulse_is_converged(pulse_file):
"""Check if pulse file has 'converged' mark in header"""
if not os.path.isfile(pulse_file):
return False
p = Pulse.read(pulse_file)
for line in p.preamble:
if 'converged' in line:
return True
return False
def blackman100ns(tgrid, E0):
from QDYN.pulse import blackman
assert (tgrid[-1] + tgrid[0] - 100) < 1e-10
T = 100
return E0 * blackman(tgrid, 0, T)
AnalyticalPulse.register_formula('blackman100ns', blackman100ns)
def get_U(pulse, wd, gate=None, J_T=None, dissipation=True,
keep_runfolder=None):
"""Propagate pulse in the given rotating frame, using the non-Hermitian
Schrödinger equation, and return the resulting (non-unitary, due to
population loss) gate U"""
assert 5000 < wd < 7000
assert isinstance(pulse, QDYN.pulse.Pulse)
rf = get_temp_runfolder('evaluate_universal_hs')
n_qubit = 5
n_cavity = 6
kappa = list(np.arange(n_cavity) * 0.05)[1:-1] + [10000.0, ] # MHz
gamma = [0.012, 0.024, 0.033, 10000.0] # MHz
if not dissipation:
kappa = list(np.arange(n_cavity) * 0.0)[1:-1] + [0.0, ] # MHz
gamma = [0.0, 0.0, 0.0, 0.0] # MHz
if gate is None:
gate = GATE['BGATE']
assert isinstance(gate, QDYN.gate2q.Gate2Q)
if J_T is None:
J_T = 'sm'
model = transmon_model(
n_qubit, n_cavity, w1, w2, wc, wd, alpha1, alpha2, g, gamma, kappa,
lambda_a=1.0, pulse=pulse, dissipation_model='non-Hermitian',
gate=gate, J_T=J_T, iter_stop=1)
# write to runfolder
model.write_to_runfolder(rf)
np.savetxt(
os.path.join(rf, 'rwa_vector.dat'),
model.rwa_vector, header='rwa vector [MHz]')
gate.write(os.path.join(rf, 'target_gate.dat'), format='array')
# propagate
env = os.environ.copy()
env['OMP_NUM_THREADS'] = '4'
try:
stdout = sp.check_output(
['qdyn_prop_gate', '--internal-units=GHz_units.txt', rf], env=env,
universal_newlines=True)
except sp.CalledProcessError as exc_info:
from IPython.core.debugger import Tracer
Tracer()()
print(exc_info)
# evaluate error
for U_t in get_prop_gate_of_t(os.path.join(rf, 'U_over_t.dat')):
U = U_t
if keep_runfolder is not None:
if os.path.isdir(keep_runfolder):
rmtree(keep_runfolder)
copytree(rf, keep_runfolder)
rmtree(rf)
return U
def evaluate_pulse_rho(pulse, gate, wd, n_qubit=5, n_cavity=6, silent=False):
"""Propagate pulse in Liouville space"""
n_qubit = n_qubit
n_cavity = n_cavity
kappa = 0.05 # MHz
gamma = 0.012 # MHz
rf = get_temp_runfolder('evaluate_universal_rho')
if isinstance(gate, str):
gate = GATE[gate]
assert isinstance(gate, QDYN.gate2q.Gate2Q)
if not silent:
print("preprocessing in %s" % rf)
model = transmon_model(
n_qubit, n_cavity, w1, w2, wc, wd, alpha1, alpha2, g, gamma, kappa,
lambda_a=1.0, pulse=pulse, dissipation_model='dissipator',
gate=gate)
# write to runfolder
model.write_to_runfolder(rf)
np.savetxt(
os.path.join(rf, 'rwa_vector.dat'),
model.rwa_vector, header='rwa vector [MHz]')
gate.write(os.path.join(rf, 'target_gate.dat'), format='array')
# propagate
if not silent:
print("starting propagation in %s" % rf)
env = os.environ.copy()
env['OMP_NUM_THREADS'] = '16'
try:
stdout = sp.check_output(
['qdyn_prop_gate', '--rho', '--internal-units=GHz_units.txt', rf],
env=env, universal_newlines=True)
except sp.CalledProcessError as exc_info:
from IPython.core.debugger import Tracer
Tracer()()
print(exc_info)
err = float(re.search(r'1-F_avg\(U, O\)\s*=\s*([Ee.0-9+-]*)',
stdout).group(1))
if not silent:
print("err_avg = %.4e" % err)
return err
def evaluate_pulse(pulse, gate, wd, dissipation=True):
"""Evaluate figure of merit for how well the pulse implements the given
gate (for simplex). For local gates, the figure of merit is 1-Favg, for
BGATE it is J_T_LI + population loss"""
# calculate model
if isinstance(gate, QDYN.gate2q.Gate2Q):
O = gate
gate = 'O'
else:
O = GATE[gate]
J_T = 'sm'
if gate == 'BGATE':
J_T = 'LI'
U = get_U(pulse, wd, gate=O, J_T=J_T, dissipation=dissipation)
err = 1-U.F_avg(O)
if gate == 'BGATE':
err = QDYN.weyl.J_T_LI(O, U) + U.pop_loss()
return err
def krotov_from_pulse(
gate, wd, pulse, iter_stop=100, dissipation=True,
ens_pulse_scale=None, freq_window=200, lambda_a=1.0,
g_a_int_converged=1.0e-7):
"""Run a Krotov optimization from the given guess pulse"""
n_qubit = 5
n_cavity = 6
kappa = list(np.arange(n_cavity) * 0.05)[1:-1] + [10000.0, ] # MHz
gamma = [0.012, 0.024, 0.033, 10000.0] # MHz
if not dissipation:
kappa = list(np.arange(n_cavity) * 0.0)[1:-1] + [10000.0, ] # MHz
gamma = [0.0, 0.0, 0.0, 10000.0] # MHz
assert 5000 < wd < 7000
assert isinstance(pulse, QDYN.pulse.Pulse)
pulse.config_attribs['is_complex'] = True
if freq_window is not None:
pulse.config_attribs['oct_spectral_filter'] = 'filter.dat'
if isinstance(gate, QDYN.gate2q.Gate2Q):
rf = get_temp_runfolder('krotov_O')
O = gate
gate = 'O'
else:
rf = get_temp_runfolder('krotov_%s' % gate)
O = GATE[gate]
J_T = 'sm'
if gate == 'BGATE':
J_T = 'LI'
use_threads = True
if ens_pulse_scale is not None:
use_threads = (len(ens_pulse_scale) + 1) * 4
model = transmon_model(
n_qubit, n_cavity, w1, w2, wc, wd, alpha1, alpha2, g, gamma, kappa,
lambda_a=lambda_a, pulse=pulse, dissipation_model='non-Hermitian',
gate=O, iter_stop=iter_stop, J_T=J_T, ens_pulse_scale=ens_pulse_scale)
model.write_to_runfolder(rf)
np.savetxt(
os.path.join(rf, 'rwa_vector.dat'),
model.rwa_vector, header='rwa vector [MHz]')
O.write(os.path.join(rf, 'target_gate.dat'), format='array')
def filter(freq):
"""Filter to ± `freq_window` MHz window."""
return np.abs(freq) < freq_window
if freq_window is not None:
pulse.write_oct_spectral_filter(
os.path.join(rf, 'filter.dat'), filter_func=filter,
freq_unit='MHz')
print("Runfolder: %s" % rf)
run_oct(rf, scratch_root=rf, monotonic=False, use_threads=use_threads,
g_a_int_converged=g_a_int_converged)
print("Runfolder: %s" % rf)
opt_pulse = Pulse.read(os.path.join(rf, "pulse.oct.dat"))
err = evaluate_pulse(opt_pulse, O, wd, dissipation=dissipation)
print("1-F_avg = %.5e" % err)
return opt_pulse
def u_tanh(v, v_min, v_max):
return atanh((2 * v - (v_max + v_min)) / (v_max - v_min))
def v_tanh(u, v_min, v_max):
return 0.5 * (v_max - v_min) * tanh(u) + 0.5 * (v_max + v_min)
def simplex_wd_E0(gate, x0, T=100):
def fun(x):
u_wd, u_E0 = x
wd = v_tanh(u_wd, 5830, 6035)
E0 = v_tanh(u_E0, 10, 300)
pulse = AnalyticalPulse(
"blackman100ns", T=T, nt=2000, time_unit='ns',
ampl_unit='MHz', parameters={'E0': E0}).to_num_pulse()
f = evaluate_pulse(pulse, gate, wd)
print("E0 = %.2f\tw_d = %.2f:\t%.4e" % (E0, wd, f))
return f
wd_0, E0_0 = x0
u_wd_0 = u_tanh(wd_0, 5830, 6035)
u_E0_0 = u_tanh(E0_0, 10, 300)
res = minimize(fun, (u_wd_0, u_E0_0), method='Nelder-Mead')
return res.x
def simplex_E0(gate, wd, E0_0, T=100):
def fun(x):
u_E0 = x[0]
E0 = v_tanh(u_E0, 10, 300)
pulse = AnalyticalPulse(
"blackman100ns", T=T, nt=2000, time_unit='ns',
ampl_unit='MHz', parameters={'E0': E0}).to_num_pulse()
f = evaluate_pulse(pulse, gate, wd)
print("E0 = %.2f\tw_d = %.2f:\t%.4e" % (E0, wd, f))
return f
u_E0_0 = u_tanh(E0_0, 10, 300)
res = minimize(fun, (u_E0_0, ), method='Nelder-Mead')
return res.x
def krotov_from_blackman(gate, wd, E0, T=100, iter_stop=100, dissipation=True,
ens_pulse_scale=None, freq_window=200):
from oct import krotov_from_pulse
pulse = AnalyticalPulse(
"blackman100ns", T=T, nt=2000, time_unit='ns',
ampl_unit='MHz', parameters={'E0': E0}).to_num_pulse()
return krotov_from_pulse(gate, wd, pulse, iter_stop=iter_stop,
dissipation=dissipation,
ens_pulse_scale=ens_pulse_scale,
freq_window=freq_window)