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 | @dataclass
class StrikeSink(SnapShotControlSinkElement, ParallelizeSinkElement):
"""Collect background triggers and horizon distances into a StrikeObject
Feeds the ranking statistics the StrikeObject maintains per subbank,
from background triggers on one pad and horizon distance history on
the others. Online, it snapshots the resulting likelihood ratios and
zerolag PDFs for exchange with the rest of the pipeline.
Args:
ifos:
list[str], the interferometers taking part in the analysis
strike_object:
StrikeObject, the ranking statistic collection to populate
bankids_map:
dict[str, list], the subbank ids belonging to each bank
background_pad:
str, the name of the sink pad carrying background triggers
horizon_pads:
list[str], the names of the sink pads carrying horizon
distance history, one per interferometer
is_online:
bool, whether to snapshot ranking statistics for exchange as
the analysis runs
multiprocess:
bool, whether this element runs in its own process rather
than a thread
injections:
bool, whether this is an injection job, which contributes no
background
count_removal_times:
list[int] | None, the times at which SNR-chisq counts were
removed, taken from the ranking statistics when not given
"""
ifos: list[str] = None # type: ignore[assignment]
strike_object: StrikeObject = None # type: ignore[assignment]
bankids_map: dict[str, list] = None # type: ignore[assignment]
background_pad: str = None # type: ignore[assignment]
horizon_pads: list[str] = None # type: ignore[assignment]
is_online: bool = False
multiprocess: bool = False
injections: bool = False
count_removal_times: list[int] | None = None
def __post_init__(self):
self._use_threading_override = not self.multiprocess
assert isinstance(self.ifos, list)
assert isinstance(self.background_pad, str)
assert isinstance(self.horizon_pads, list)
assert self.strike_object is not None
self.sink_pad_names = (self.background_pad,) + tuple(self.horizon_pads)
self.temp = 0
SnapShotControlSinkElement.__post_init__(self)
if not self.injections:
ParallelizeSinkElement.__post_init__(self)
if self.is_online and not self.injections:
# setup bottle
self.state_dict = {
"xml": {},
"zerolagxml": {},
"count_tracker": 0,
"count_removal_times": [],
}
for bankid in self.bankids_map:
self.add_snapshot_filename(
"%s_SGNL_LIKELIHOOD_RATIO" % bankid, "xml.gz"
)
self.state_dict["xml"][bankid] = xml_string(
self.strike_object.likelihood_ratios[bankid]
)
self.state_dict["zerolagxml"][bankid] = xml_string(
self.strike_object.zerolag_rank_stat_pdfs[bankid]
)
if self.count_removal_times is None:
self.count_removal_times = (
self.strike_object.likelihood_ratios[bankid]
.terms["P_of_SNR_chisq"]
.remove_counts_times
)
self.state_dict["count_removal_times"] = self.count_removal_times
else:
assert (
self.count_removal_times
== self.strike_object.likelihood_ratios[bankid]
.terms["P_of_SNR_chisq"]
.remove_counts_times
)
self.register_snapshot()
def pull(self, pad, frame):
if frame.EOS:
self.mark_eos(pad)
if self.rsnks[pad] == self.background_pad:
# Iterate through all events in the frame
for event in frame.events:
background = event["background"]
if background is None:
continue
#
# Background triggers
#
self.strike_object.train_noise(
frame.start_ns / 1e9,
background["snrs"],
background["chisqs"],
background["single_masks"],
)
#
# Trigger rates
#
# FIXME : come up with a way to make populating the trigger rate
# object as part of train_noise
trigger_rates = event["trigger_rates"]
if trigger_rates is not None:
for ifo, trigger_rate in trigger_rates.items():
for bankid in self.bankids_map:
buf_seg, count = trigger_rate[bankid]
self.strike_object.likelihood_ratios[bankid].terms[
"P_of_tref_Dh"
].triggerrates[ifo].add_ratebin(list(buf_seg), count)
elif self.rsnks[pad] in self.horizon_pads:
if frame.start_ns - frame.end_ns == 0:
return
# Iterate through all events in the frame
for event in frame.events:
ifo = event["ifo"]
horizon = event["horizon"]
horizon_time = event["epoch"] / 1_000_000_000
# Epoch is the mid point of the most recent FFT
# interval used to obtain this PSD
if (
horizon is not None
and float(event["n_samples"] / event["navg"]) > 0.3
):
# n_samples / navg is the "stability", which is a measure of the
# fraction of the configured averaging timescale used to obtain this
# measurement.
for bankid in self.bankids_map:
self.strike_object.likelihood_ratios[bankid].terms[
"P_of_tref_Dh"
].horizon_history[ifo][horizon_time] = horizon[bankid]
else:
for bankid in self.bankids_map:
self.strike_object.likelihood_ratios[bankid].terms[
"P_of_tref_Dh"
].horizon_history[ifo][horizon_time] = 0
def process_outqueue(self, sdict):
# FIXME This reduces the ram, but is a mess
xml = sdict["xml"]
zerolagxml = sdict["zerolagxml"]
bankid = sdict["bankid"]
old_xml = self.state_dict["xml"]
old_zerolagxml = self.state_dict["zerolagxml"]
new_xml = {bid: x for bid, x in old_xml.items() if bid != bankid}
new_zerolagxml = {bid: x for bid, x in old_zerolagxml.items() if bid != bankid}
new_xml[bankid] = xml[bankid]
new_zerolagxml[bankid] = zerolagxml[bankid]
count_tracker = self.state_dict["count_tracker"]
count_removal_times = self.state_dict["count_removal_times"]
del self.state_dict
self.state_dict = {
"xml": new_xml,
"zerolagxml": new_zerolagxml,
"count_tracker": count_tracker,
"count_removal_times": count_removal_times,
}
frankenstein = sdict["frankenstein"]
likelihood_ratio_upload = sdict["likelihood_ratio_upload"]
self.strike_object.update_dynamic(
bankid, frankenstein[bankid], likelihood_ratio_upload[bankid]
)
del sdict
def get_state_from_queue(self):
try:
sdict = self.out_queue.get_nowait()
self.process_outqueue(sdict)
except Empty:
return
def internal(self):
if self.injections:
return
if self.is_online:
if self.multiprocess:
# FIXME: is this the correct logic?
ParallelizeSinkElement.internal(self)
self.get_state_from_queue()
SnapShotControlSinkElement.exchange_state(self.name, self.state_dict)
if self.state_dict["count_tracker"] != 0:
self.count_removal_callback()
if self.at_eos:
if self.is_online:
if self.terminated.is_set():
print("At EOS and subprocess is terminated")
else:
drained_outq = self.sub_process_shutdown(600)
print("after shutdown", len(drained_outq))
# Do this in the main thread
# no reason to put this in subprocess
# only write out the LR files, don't update
for bankid in self.bankids_map:
desc = "%s_SGNL_LIKELIHOOD_RATIO" % bankid
fn = self.snapshot_filenames(desc)
self.strike_object.update_array_data(bankid)
data = self.strike_object.prepare_inq_data(fn, bankid)
on_snapshot(data, shutdown=True, reset_dynamic=False)
else:
for bankid in self.bankids_map:
self.strike_object.update_array_data(bankid)
self.strike_object.likelihood_ratios[bankid].save(
self.strike_object.output_likelihood_file[bankid]
)
else:
if self.is_online:
for i, bankid in enumerate(self.bankids_map):
desc = "%s_SGNL_LIKELIHOOD_RATIO" % bankid
if self.snapshot_ready(desc):
fn = self.snapshot_filenames(desc)
self.strike_object.update_array_data(bankid)
data = self.strike_object.prepare_inq_data(fn, bankid)
self.in_queue.put(data)
if i == 0:
self.strike_object.load_rank_stat_pdf()
def count_removal_callback(self):
# FIXME : at the time of calling exchange_state() posted data is already
# added on top of existing remove counts list and converted into json
# format. where should I add the functionality?
# FIXME : how can I make this callback to be called upon posting remove
# count info?
# FIXME : how can an external program know the self.name of StrikeSink
# for each inspiral job, which is part of URL to post information to but
# not included in registry.txt?
if self.state_dict["count_tracker"] > 0:
self.count_removal_times.append(self.state_dict["count_tracker"])
elif self.state_dict["count_tracker"] < 0:
gps_time = abs(self.state_dict["count_tracker"])
if gps_time in self.count_removal_times:
self.count_removal_times.remove(gps_time)
else:
print(f"{gps_time} not in self.count_removal_times, not removing")
self.state_dict["count_removal_times"] = self.count_removal_times
for bankid, likelihood_ratio in self.strike_object.likelihood_ratios.items():
# update the internal array for the removed times
print(f"bankid: {bankid} count remove times: {self.count_removal_times}")
likelihood_ratio.terms["P_of_SNR_chisq"].remove_counts_times = (
self.count_removal_times
)
self.state_dict["count_tracker"] = 0
def worker_process(self, context: WorkerContext):
try:
data = context.input_queue.get(timeout=2)
sdict = on_snapshot(data, context.should_shutdown(), True)
if sdict is not None:
context.output_queue.put(sdict)
time.sleep(10)
except Empty:
return
|