def print_output(data, ostr=None, sort=None, renamings=Ellipsis):
assert sort is None or sort == '-' or sort == '+'
scriptname = os.path.basename(__file__)
timestamp = format_rfc5322(localtime=True)
print("% -*- coding:utf-8; mode:latex; -*-", file=ostr)
print("", file=ostr)
print("%% THIS IS A GENERATED FILE; PLEASE DO NOT EDIT IT MANUALLY!",
file=ostr)
print("%% Generated by {!r} on {!s}".format(scriptname, timestamp),
file=ostr)
print("", file=ostr)
sortkey = lambda kv: (kv[1][sort][0]
if sort in kv[1] else math.nan, kv[0].name)
for (prop, val) in sorted(data.items(), key=sortkey):
enumname = renamings[prop]
assert re.match(r'^[A-Z][0-9A-Z_]*$', enumname)
texname = enumname.replace('_', '\\_')
mx = format(100.0 *
val['-'][0], '.2f') if '-' in val else str(math.nan)
sx = format(100.0 *
val['-'][1], '.2f') if '-' in val else str(math.nan)
my = format(100.0 *
val['+'][0], '.2f') if '+' in val else str(math.nan)
sy = format(100.0 *
val['+'][1], '.2f') if '+' in val else str(math.nan)
argvec = ''.join("{" + value + "}" for value in [mx, sx, my, sy])
print(r"\PunctureResult[\enum{" + texname + r"}]" + argvec, file=ostr)
def write_output(ostr, results, individuals, renamings, base_runs,
base_success):
scriptname = os.path.basename(__file__)
timestamp = format_rfc5322(localtime=True)
print("% -*- coding:utf-8; mode:latex; -*-", file=ostr)
print("", file=ostr)
print("%% THIS IS A GENERATED FILE; PLEASE DO NOT EDIT IT MANUALLY!",
file=ostr)
print("%% Generated by {!s} on {!s}".format(scriptname, timestamp),
file=ostr)
print("", file=ostr)
for (test, (mean, stdev)) in sorted(results.items(),
key=(lambda kv: kv[1][0]),
reverse=True):
enumname = renamings[test]
assert re.match(r'^[A-Z][0-9A-Z_]*$', enumname)
texname = enumname.replace('_', '\\_')
if individuals is not None and BASELINE in individuals:
differences = [
individuals[BASELINE][i] - individuals[test][i]
for i in range(base_runs)
]
try:
diff_mean = statistics.mean(differences)
diff_stdev = statistics.stdev(differences, xbar=diff_mean)
except statistics.StatisticsError:
(diff_mean, diff_stdev) = (math.nan, math.nan)
else:
diff_mean = base_success[0] - results[test][0]
diff_stdev = math.sqrt(base_success[1]**2 + results[test][1]**2)
argtokens = ''.join(r"{" + format(100.0 * x, '.2f') + r"}"
for x in [mean, stdev, diff_mean, diff_stdev])
print(r"\CompetingMetricResult[\enum{" + texname + r"}]" + argtokens,
file=ostr)
def update_progress_report(filename: str,
t0: float,
t: float,
progress: float = None,
remaining: float = None,
create: bool = False):
if filename is None:
return
if not filename or filename == '-':
_warning("Progress report can only be written to a regular file")
return
mode = 'w' if create else 'a'
__ = lambda x: x if x is not None else math.nan
with open(filename, mode) as ostr:
if create:
print("#! /bin/false", file=ostr)
print("#! -*- coding:utf-8; mode:config-space; -*-", file=ostr)
print("", file=ostr)
print("## Job started on {!s}".format(
format_rfc5322(t0, localtime=True)),
file=ostr)
print("## {:>21s}{:>24s}{:>24s}".format("elapsed / s",
"progress / 1",
"remaining / s"),
file=ostr)
print("", file=ostr)
print("{:24.10E}{:24.10E}{:24.10E}".format(t - t0, __(progress),
t + __(remaining) - t0),
file=ostr)
def main():
ap = argparse.ArgumentParser(
description="Format Huang parameters as TeX code.")
ap.add_argument('src', metavar='FILE', help="read from JSON file FILE")
ap.add_argument('-o',
'--output',
metavar='FILE',
type=argparse.FileType('w'),
help="write to FILE")
ap.add_argument('-r',
'--rename',
metavar='FILE',
type=argparse.FileType('r'),
help="rename enums according to mapping read from FILE")
ns = ap.parse_args()
try:
with open(ns.src, 'r') if ns.src != '-' else contextlib.nullcontext(
sys.stdin) as istr:
info = json.load(istr)
except FileNotFoundError:
check_lazy_okay(ns.src)
info = {m: (math.nan, math.nan) for m in const.HUANG_METRICS}
else:
unpack = lambda d: (d['mean'], d['stdev'])
info = {
m: unpack(info[const.enum_to_json(m)])
for m in const.HUANG_METRICS
}
renamings = {m: m.name for m in const.HUANG_METRICS}
if ns.rename is not None:
load_rename_table(ns.rename, renamings)
scriptname = os.path.basename(__file__)
timestamp = format_rfc5322(localtime=True)
print("% -*- coding:utf-8; mode:latex; -*-", file=ns.output)
print("", file=ns.output)
print("%% THIS IS A GENERATED FILE; PLEASE DO NOT EDIT IT MANUALLY!",
file=ns.output)
print("%% Generated by {!r} on {!s}".format(scriptname, timestamp),
file=ns.output)
print("", file=ns.output)
for (metric, params) in sorted(info.items(),
key=(lambda kv: abs(kv[1][0])),
reverse=True):
enumname = renamings[metric]
assert re.match(r'^[A-Z][0-9A-Z_]*$', enumname)
texname = enumname.replace('_', '\\_')
arg1 = format(params[0], '+.4f')
arg2 = format(params[1], '.4f')
print(r"\HuangWeightResult[" + texname + r"]{" + arg1 + r"}{" + arg2 +
r"}",
file=ns.output)
def try_download(url, checksums, tempname, traceinfo):
hashes = {algo: hashlib.new(algo) for (algo, chksum) in checksums}
def update_hashers(chunk):
for hasher in hashes.values():
hasher.update(chunk)
with open(tempname, 'wb') as ostr:
info("Downloading {!r} ...".format(url))
try:
t0 = time.time()
traceinfo['date'] = format_rfc5322(t0, localtime=True)
with urllib.request.urlopen(url) as istr:
size = copy_iostreams(istr, ostr, callback=update_hashers)
t1 = time.time()
info(
"Download complete; fetched {size:,d} bytes in {time:,.3f} seconds ({rate[0]:.3f} {rate[1]:s})"
.format(size=size,
time=(t1 - t0),
rate=get_human_rate(size, t1 - t0)))
traceinfo['size'] = size
traceinfo['time'] = t1 - t0
except urllib.error.HTTPError as e:
what = traceinfo['error'] = str(e)
info(what)
return False
except OSError as e:
info(e.strerror)
return False
hashes = {algo: hasher.digest() for (algo, hasher) in hashes.items()}
for (algo, expected) in checksums:
actual = hashes[algo]
traceinfo['digest-' + algo] = hexhash(actual)
if expected == actual:
info("Verifying {:s} checksum".format(algo.upper()), "OK")
else:
info("Verifying {:s} checksum".format(algo.upper()), "FAILED")
info("Expected: " + hexhash(expected))
info("Actual: " + hexhash(actual))
traceinfo['error'] = "{:s} checksum does not match".format(
algo.upper())
return False
return True
def main():
ap = argparse.ArgumentParser(
description="Format neural network information as TeX code.")
ap.add_argument('-o',
'--output',
metavar='FILE',
type=argparse.FileType('w'),
help="write to FILE")
ap.add_argument(
'--shared-info',
metavar='FILE',
type=argparse.FileType('r'),
required=True,
help=
"read information about the shared model from FILE (Keras debug output)"
)
ap.add_argument(
'--total-info',
metavar='FILE',
type=argparse.FileType('r'),
required=True,
help=
"read information about the total model from FILE (Keras debug output)"
)
ap.add_argument('--corpus',
metavar='FILE',
type=argparse.FileType('r'),
required=True,
help="read data corpus summary from JSON FILE")
ns = ap.parse_args()
(shared_layers, shared_info) = Parser(ns.shared_info)()
(total_layers, total_info) = Parser(ns.total_info)()
corpus = json.load(ns.corpus)
scriptname = os.path.basename(__file__)
timestamp = format_rfc5322(localtime=True)
print("% -*- coding:utf-8; mode:latex; -*-", file=ns.output)
print("", file=ns.output)
print("%% THIS IS A GENERATED FILE; PLEASE DO NOT EDIT IT MANUALLY!",
file=ns.output)
print("%% Generated by {!r} on {!s}".format(scriptname, timestamp),
file=ns.output)
print("", file=ns.output)
fmtint = lambda n: format(n, ',d').replace(',', '\\,')
print(get_layer_dim_def('NNSharedInputDims', shared_layers, 'in',
'InputLayer'),
file=ns.output)
print(get_layer_dim_def('NNSharedHiddenDims', shared_layers, 'l1',
'Dense'),
file=ns.output)
print(get_layer_dim_def('NNSharedOutputDims', shared_layers, 'l2',
'Dense'),
file=ns.output)
print(get_layer_dim_def('NNTotalAuxInputDims', total_layers, 'auxin',
'InputLayer'),
file=ns.output)
print(get_layer_dim_def('NNTotalAuxHiddenDims', total_layers, 'aux',
'Dense'),
file=ns.output)
print(get_layer_dim_def('NNTotalCatDims', total_layers, 'cat',
'Concatenate'),
file=ns.output)
print(get_layer_dim_def('NNTotalOutputDims', total_layers, 'out', 'Dense'),
file=ns.output)
print(r"\def\NNTotalTrainableParams{" +
fmtint(total_info['trainable-params']) + r"}",
file=ns.output)
print(r"\def\NNCorpusSize{" + fmtint(corpus['size']) + r"}",
file=ns.output)
print(r"\def\NNCorpusSizeApprox{" +
fmtint(1000 * round(corpus['size'] / 1000)) + r"}",
file=ns.output)
def run_toolchain(
maindep,
workdir=None,
jobname=None,
columns=None,
tex=None,
bib=None,
idx=None,
oneshot=False,
epoch=None,
):
if jobname is None:
(root, ext) = os.path.splitext(maindep)
jobname = root
if epoch is not None:
logging.info("SOURCE_DATE_EPOCH={!s} # {:s}".format(
epoch, format_rfc5322(epoch, localtime=True)))
auxpattern = os.path.join(workdir,
'*.aux') if workdir is not None else '*.aux'
if tex is not None:
tex_command = [
TOOLS[tex], *TOOLFLAGS['TEX'], '-interaction', 'batchmode',
'-jobname', jobname, maindep
]
if bib is not None:
bib_command = [TOOLS[bib], *TOOLFLAGS['BIB'], jobname]
if idx is not None:
idx_command = [TOOLS[idx], *TOOLFLAGS['IDX'], jobname]
oldchksum = None if oneshot else get_aux_checksum(auxpattern)
if oldchksum is not None:
logging.debug(
"Initial combined checksum of auxiliary files is {!s}".format(
oldchksum))
for i in range(10):
if tex is not None:
logging.info("Running {:s} ({:d}) ...".format(tex, i + 1))
run_command(tex_command,
workdir=workdir,
what=tex,
columns=columns,
epoch=epoch)
newchksum = None if oneshot else get_aux_checksum(auxpattern)
if newchksum is not None:
logging.debug(
"Combined checksum of auxiliary files is {!s}".format(
newchksum))
if i == 0 and (bib or idx):
if bib is not None:
logging.info("Running {:s} ...".format(bib))
run_command(
bib_command,
workdir=workdir,
what=bib,
columns=columns,
epoch=epoch,
stdout='{:s}-stdout.log'.format(bib.lower()),
stderr='{:s}-stderr.log'.format(bib.lower()),
)
if idx is not None:
logging.info("Running {:s} ...".format(idx))
run_command(
idx_command,
workdir=workdir,
what=idx,
columns=columns,
epoch=epoch,
stdout='{:s}-stdout.log'.format(idx.lower()),
stderr='{:s}-stderr.log'.format(idx.lower()),
)
elif oldchksum == newchksum:
break
oldchksum = newchksum
else:
raise Error("Combined checksum of auxiliary files did not converge")
def main():
ap = argparse.ArgumentParser(
description="Formats a JSON confusion matrix as TeX code.")
ap.add_argument('src', metavar='FILE', help="read from FILE")
ap.add_argument('-t',
'--test',
metavar='TEST',
type=_test,
help="only consider TEST (may not be repeated)")
ap.add_argument('-o',
'--output',
metavar='FILE',
type=argparse.FileType('w'),
help="write to FILE")
ns = ap.parse_args()
try:
with open(ns.src, 'r') if ns.src != '-' else contextlib.nullcontext(
sys.stdin) as istr:
infos = json.load(istr)
except FileNotFoundError:
check_lazy_okay(ns.src)
infos = None
if ns.test is not None:
key0 = const.enum_to_json(ns.test)
_get = lambda *args, **kwargs: get(infos, key0, *args, **kwargs)
else:
_get = lambda *args, **kwargs: get(infos, *args, **kwargs)
number_d = lambda *keys: format(round(_get(*keys, na=0)), 'd').replace(
',', '\\,')
number_f2 = lambda *keys: format(_get(*keys, na=math.nan), '.2f')
percentage = lambda *keys: format(100.0 * _get(*keys, na=math.nan), '.2f')
scriptname = os.path.basename(__file__)
timestamp = format_rfc5322(localtime=True)
with contextlib.redirect_stdout(ns.output):
print("% -*- coding:utf-8; mode:latex; -*-")
print("")
print("%% THIS IS A GENERATED FILE; PLEASE DO NOT EDIT IT MANUALLY!")
print("%% Generated by {!r} on {!s}".format(scriptname, timestamp))
print("")
print(r"\def\XValTestRuns{" + number_d('test-runs') + r"}")
print(r"\def\XValCountApprox{" + number_d('test-count', 'mean') + r"}")
print(r"\def\XValCountMean{" + number_f2('test-count', 'mean') + r"}")
print(r"\def\XValCountStdev{" + number_f2('test-count', 'stdev') +
r"}")
#
print(r"\def\XValSuccessMean{" + percentage('success', 'mean') + r"}")
print(r"\def\XValSuccessStdev{" + percentage('success', 'stdev') +
r"}")
print(r"\def\XValFailureMean{" + percentage('failure', 'mean') + r"}")
print(r"\def\XValFailureStdev{" + percentage('failure', 'stdev') +
r"}")
#
print(r"\def\XValCondPosMean{" +
percentage('condition-positive', 'mean') + r"}")
print(r"\def\XValCondPosStdev{" +
percentage('condition-positive', 'stdev') + r"}")
print(r"\def\XValCondNegMean{" +
percentage('condition-negative', 'mean') + r"}")
print(r"\def\XValCondNegStdev{" +
percentage('condition-negative', 'stdev') + r"}")
#
print(r"\def\XValPredPosMean{" +
percentage('prediction-positive', 'mean') + r"}")
print(r"\def\XValPredPosStdev{" +
percentage('prediction-positive', 'stdev') + r"}")
print(r"\def\XValPredNegMean{" +
percentage('prediction-negative', 'mean') + r"}")
print(r"\def\XValPredNegStdev{" +
percentage('prediction-negative', 'stdev') + r"}")
#
print(r"\def\XValTrueNegMean{" + percentage('true-negative', 'mean') +
r"}")
print(r"\def\XValTrueNegStdev{" +
percentage('true-negative', 'stdev') + r"}")
print(r"\def\XValFalseNegMean{" +
percentage('false-negative', 'mean') + r"}")
print(r"\def\XValFalseNegStdev{" +
percentage('false-negative', 'stdev') + r"}")
print(r"\def\XValTruePosMean{" + percentage('true-positive', 'mean') +
r"}")
print(r"\def\XValTruePosStdev{" +
percentage('true-positive', 'stdev') + r"}")
print(r"\def\XValFalsePosMean{" +
percentage('false-positive', 'mean') + r"}")
print(r"\def\XValFalsePosStdev{" +
percentage('false-positive', 'stdev') + r"}")
def main():
ap = argparse.ArgumentParser(
description="Formats a discrimination results file as TeX code.")
ap.add_argument('src', metavar='FILE', help="read from FILE")
ap.add_argument('-o',
'--output',
metavar='FILE',
type=argparse.FileType('w'),
help="write to FILE")
ap.add_argument(
'-n',
'--name',
metavar='MODEL=NAME',
dest='dmnames',
action='append',
type=parse_dmnamespec,
help="use TeX name NAME for discriminator model MODEL (can be repeated)"
)
ap.add_argument('--print-names',
action='store_true',
help="print discriminator model names and exit")
ns = ap.parse_args()
dmnames = {
dm: ''.join(filter(str.isalnum, dm.name.lower()))
for dm in const.Tests
}
if ns.dmnames is not None:
dmnames.update(ns.dmnames)
if ns.print_names:
for dm in sorted(dmnames.keys()):
print(dm.name, dmnames[dm])
raise SystemExit()
try:
with open(ns.src, 'r') if ns.src != '-' else contextlib.nullcontext(
sys.stdin) as istr:
disco = list(parse_disco_file(istr))
except FileNotFoundError:
check_lazy_okay(ns.src)
disco = None
shortid = lambda id: str(id)[:8]
dmname = lambda dm: dmnames[dm]
scriptname = os.path.basename(__file__)
timestamp = format_rfc5322(localtime=True)
with contextlib.redirect_stdout(
ns.output) if ns.output is not None else contextlib.nullcontext():
print("% -*- coding:utf-8; mode:latex; -*-")
print("")
print("%% THIS IS A GENERATED FILE; PLEASE DO NOT EDIT IT MANUALLY!")
print("%% Generated by {!r} on {!s}".format(scriptname, timestamp))
print("")
if disco is None:
print("% This file contains no records (note: {!s}={!r}).".format(
LAZYOK_ENVVAR, 1))
return
print("% This file contains {:,d} records.".format(len(disco)))
for record in disco:
print("")
print("% lhsid: {!s}".format(record.lhs))
print("% rhsid: {!s}".format(record.rhs))
print("% value: {:+.10f}".format(record.value))
print("% model: {:s}".format(record.model.name))
print(
r"\expandafter\def\csname msc@{lhs:s}-{rhs:s}@{dm:s}\endcsname{{{p:.3f}}}"
.format(lhs=shortid(record.lhs),
rhs=shortid(record.rhs),
p=(100.0 * record.value),
dm=dmname(record.model)))
def fmttabs(timestamp):
return format_rfc5322(timestamp, localtime=True)