def opr(operation, op1, op2=()):
result = ()
if isinstance(op1, types_Number):
# prefix +/-
if op2==():
if operation=='+':
result = operator.pos(op1)
elif operation=='-':
result = operator.neg(op1)
elif isinstance(op2, types_Number):
if operation=='+':
result = operator.add(op1,op2)
elif operation=='-':
result = operator.sub(op1,op2)
# string '+'
elif operation=='+' and all([isinstance(x,types.StringTypes) for x in (op1,op2)]):
result = operator.add(op1,op2)
return result
def visit_UnaryOpExpr(self, node):
# TODO: handle TypeError 'bad operand type for ...' exceptions
operand = self.visit(node.operand)
if isinstance(node.op, nodes.UMinus):
return operator.neg(self.node_as_value(operand))
elif isinstance(node.op, nodes.UPlus):
return operator.pos(self.node_as_value(operand))
else:
raise RuntimeError() # pragma: no cover
def test_neg(self):
self.failUnlessRaises(TypeError, operator.neg)
self.failUnlessRaises(TypeError, operator.neg, None)
self.failUnless(operator.neg(5) == -5)
self.failUnless(operator.neg(-5) == 5)
self.failUnless(operator.neg(0) == 0)
self.failUnless(operator.neg(-0) == 0)
def test_neg(self):
self.assertRaises(TypeError, operator.neg)
self.assertRaises(TypeError, operator.neg, None)
self.assertTrue(operator.neg(5) == -5)
self.assertTrue(operator.neg(-5) == 5)
self.assertTrue(operator.neg(0) == 0)
self.assertTrue(operator.neg(-0) == 0)
def test_neg(self):
self.failUnlessRaises(TypeError, operator.neg)
self.failUnlessRaises(TypeError, operator.neg, None)
self.assertEqual(operator.neg(5), -5)
self.assertEqual(operator.neg(-5), 5)
self.assertEqual(operator.neg(0), 0)
self.assertEqual(operator.neg(-0), 0)
def test_neg(self):
#operator = self.module
self.assertRaises(TypeError, operator.neg)
self.assertRaises(TypeError, operator.neg, None)
self.assertEqual(operator.neg(5), -5)
self.assertEqual(operator.neg(-5), 5)
self.assertEqual(operator.neg(0), 0)
self.assertEqual(operator.neg(-0), 0)
def test_result(self):
types = np.typecodes['AllInteger'] + np.typecodes['AllFloat']
with suppress_warnings() as sup:
sup.filter(RuntimeWarning)
for dt in types:
a = np.ones((), dtype=dt)[()]
assert_equal(operator.neg(a) + a, 0)
def clonesBall( self ):
# clone ball 1 = rect( x, y, w, h ), ( fpdx, fpdy )
clone = None
try:
xbmcgui.lock()
for ball in self.balls:
if not ball.dead:
rect = ball.ballControl.getRect()
fpd = neg( ball.fpdx ), neg( ball.fpdy )
clone = rect, fpd
break
except:
print_exc()
xbmcgui.unlock()
if clone:
self.new_ball( clone )
def getlocaldata(sms,dr,flst):
for f in flst:
fullf = os.path.join(dr,f)
if os.path.isfile(fullf):
tup = (fullf, os.path.getsize(fullf))
_listOfFiles.append(tup)
tmp = sorted(_listOfFiles, key=operator.itemgetter(1))
sms[1] = tmp[operator.neg(sms[0]):]
def perturb_data(loss_func, optimizer, model, optimal_subset, numEpochs, batchSize,
constrainWeight, epsilon, invertTargets, negateLoss,
):
"""
the model parameter is pre-trained model that minimizes loss. By freezing the weights of all layers except the
embedding layer and training with either inverted target vector or negated loss, we can perturb the embedding
layer weights to the input data as a perturbed embedding matrix--the one-hot vector input type cannot be
perturbed because its values are integers whereas the embedding matrix is real. The embedding layer weight
matrix functions as a lookup-table so that each row corresponds to a one-hot index--by perturbing the
weight matrix/lookup-table we can construct perturbed embedding matrix inputs using the original one-hot vector
input types and mapping them to perturbed embedding matrix input types via the perturbed
weight matrix/lookup-table
:param model:
:param optimal_subset:
:param loss_func:
:param optimizer:
:param numEpochs:
:param constrainWeight:
:param epsilon:
:param invertTargets:
:param negateLoss:
:return:
"""
from operator import neg
def constrain_fn(embedW, embedW_orig, eps):
return K.T.nlinalg.trace(K.dot(K.transpose(embedW_orig - embedW), embedW_orig - embedW)) - eps
# freeze the weights of every layer except embedding layer
for idx, L in enumerate(model.layers):
if idx == 0:
continue
L.trainable = False
assert model.layers[0].trainable is True, "embedding layer must be trainable"
constrain_fn_args = [model.layers[0].W,
model.layers[0].W.get_value(), epsilon]
loss = constrained_loss((lambda ytru, ypred: neg(loss_func(ytru, ypred)) if negateLoss else loss_func(ytru, ypred)),
constrain_fn, constrain_fn_args,
constraint_weight=constrainWeight)
model.compile(optimizer=optimizer, loss=loss, metrics=['accuracy'])
targets = optimal_subset[1]
onehotVectors = optimal_subset[0]
if invertTargets:
targets = np.logical_not(targets)
callbacks = [Batch_EarlyStopping(verbose=1)]
model.fit(onehotVectors, targets, nb_epoch=numEpochs, batch_size=batchSize, callbacks=callbacks)
return model
def score_getter_penalize_overlap_estimated(a, b):
'''Penalizes overlap, to try to find non-overlapping segments'''
if a == None or overlap(a,b) == 0:
return len_aln(b) * (b.pctid / 100.0)
olap = overlap(a,b)
olap_penalty = b_olap_matches = olap * (b.pctid / 100.0) * neg(1)
b_hang_matches = (len_aln(b) - olap) * (b.pctid / 100.0)
return olap_penalty + b_hang_matches
def error(geoprint, radians=False):
"""Returns the error of a given geoprint.
If radians is true, return the error in radians, otherwise
degrees.
"""
e = 90.0 * pow(2, neg(len(geoprint) - 1))
if radians:
return rads(e)
return e
def candidate_killer(candidates):
scoreboard = [] # here we will sort the candidates
for i, c in enumerate(candidates): # for each candidate in the list of candidates, fill its
scoreboard.append((len(c),neg(c[-1]),i)) # length, negative of the last item, index in the list
scoreboard.sort(reverse=True) # reverse sort, longest with smallest last item first (because we use the negative)
good_candidates = [candidates[scoreboard[0][2]]] # put the first one into the list of good candidates
best_last = scoreboard[0][1] # keep the information of the smallest last item so far (we use the negative)
for s in scoreboard: # from the scoreboard
if s[1] > best_last: # if the last item is smaller than the best known (we compare its negative!)
good_candidates.append(candidates[s[2]]) # append the candidate into the list of good ones
best_last = s[1] # and update the smallest last item (negative)
return good_candidates # return the list of good candidates
def solve_pnp_distance(self, box,poses):
rows = box.shape[0]
distances = np.zeros((rows,1))
for a in range(rows):
pose = poses[:3,a]
distance = pose[2]
if distance < 0:
distance = neg(distance)
distances = np.resize(distances,(a,1))
distances = np.append(distances,distance)
return distances
def DiffInDays(self, d1, d2):
""" calculate difference between two dates rounded to days """
d1, d2 = self.__DiffCheck(d1, d2)
if d1 == 'N/A' or d2 == 'N/A':
return 'N/A'
# Calculate difference in seconds
delta = d1 - d2
if delta < 0:
delta = operator.neg(delta)
# Calculate days
days = int(round(delta / 60 / 60 / 24))
if days == 0:
days = 1
return days
def score_getter_matching_consensus_estimated(a, b):
'''Scoring Function for filtering alignments for downstream
consensus. Overlaps are sort of encouraged '''
if a == None or overlap(a,b) == 0:
return len_aln(b) * (b.pctid / 100.0)
olap = overlap(a,b)
a_olap_matches = olap * (a.pctid / 100.0) * neg(0.5)
b_olap_matches = olap * (b.pctid / 100.0) * 0.5
b_hang_matches = (len_aln(b) - olap) * (b.pctid / 100.0)
score = a_olap_matches + b_olap_matches + b_hang_matches
return score
def unary_exp(tokens):
assert len(tokens)
if tokens[0] == "-":
tokens.pop(0)
value = value_exp(tokens)
return lambda row, ctx: operator.neg(value(row, ctx))
elif tokens[0] == "not":
tokens.pop(0)
value = value_exp(tokens)
return lambda row, ctx: operator.not_(value(row, ctx))
elif tokens[0] == "+":
tokens.pop(0)
return value_exp(tokens)
def ActualYear(self, timestamp = None, name = 'ActualYear', selected = None, rangeFrom = None, rangeTo = None, plus = 0, minus = 0, withNull = 0, nullSelected = 0):
if not timestamp:
timestamp = time.time()
else:
timestamp = self.CareForTimestamp(timestamp)
if rangeFrom is None:
rangeFrom = int(time.strftime('%Y', time.localtime(timestamp)))
elif rangeFrom < 0:
# actual year - rangeFrom
rf = int(time.strftime('%Y', time.localtime(timestamp))) - operator.neg(rangeFrom)
rangeFrom = rf
if rangeTo is None:
rangeTo = rangeFrom + 1
elif rangeTo < 1000:
rangeTo = int(time.strftime('%Y', time.localtime(timestamp))) + rangeTo
return self.__GenerateActual(timestamp, name, selected, '%Y', (rangeFrom, rangeTo), plus, minus, withNull, nullSelected)
def test_operator_high_cohesive_invoke():
# f objects can be called as functions
assert add(1, 2) == 3
assert sub(7, 3) == 4
assert mul(3, 7) == 21
assert truediv(32, 4) == 8
assert neg(1) == -1
# works with arbitrary numbers of arguments
assert f(bool)() is False
assert f(max)([1]) == 1
assert f(max)(6, 1) == 6
assert f(max)(6, 1, 9) == 9
assert f(max)(*range(10)) == 9
# works with keyword arguments
assert f(int)('10', base=16) == 16
def _get_array_size(self, node):
"""Calculate the size of the array."""
if isinstance(node, c_ast.Constant):
size = int(node.value)
elif isinstance(node, c_ast.UnaryOp):
if node.op == '-':
size = operator.neg(self._get_array_size(node.expr))
else:
size = self._get_array_size(node.expr)
elif isinstance(node, c_ast.BinaryOp):
mapping = {'+': operator.add, '-': operator.sub,
'*': operator.mul, '/': operator.floordiv}
left = self._get_array_size(node.left)
right = self._get_array_size(node.right)
size = mapping[node.op](left, right)
else:
raise ParseError('This type of array not supported: %s' % node)
return size
def get_project_name_and_entry_id(self, series, feed, entry):
base_regex = "Episode [\d\.]+ - (.+)"
rep = re.compile("%s with .+" % base_regex)
if rep.match(entry.title):
title = rep.match(entry.title).groups()[0]
else:
# The episode title does not contain " with YYY"
rep = re.compile(base_regex)
if rep.match(entry.title):
title = rep.match(entry.title).groups()[0]
else:
# No episode number (probably not a project's resource...)
title = entry.title
# Strip the following bits if the name ends in it
for s in (", and more", " and more", ","):
if title.endswith(s):
title = title[: neg(len(s))]
title = title.strip()
return (title, entry.id)
def __init__(self):
self.version = None
self.options = None
self.function_stack = [] # stack of functions called, from top-level to currently executing one
# namespace includes variables and rules
self.namespace = {} # Will be hash list of input files of whatever textual content
self.namespace['report'] = OrderedDict()
self.namespace['report']['title'] = "RCQC Quality Control Report"
self.namespace['report']['tool_version'] = CODE_VERSION
self.namespace['report']['job'] = {'status': 'ok'}
self.namespace['report']['quality_control'] = {'status': 'ok'}
self.namespace['sections'] = []
self.namespace['rule_index'] = {} # rule index based on location of store(_, location) field. 1 per.
self.namespace['name_index'] = {} # index based on last (z) key of x.y.z namespace reference.
self.namespace['files'] = []
self.namespace['file_names'] = {}
self.namespace['iterator'] = {} # Provides the dictionary for each current iterator function evaluation (at call depth).
self.namespace['report_html'] = ''
self.input_file_paths = None
self.ruleset_file_path = None
self.output_json_file = None
# Really core functions below require access to RCQC class variables.
# Other functions can be added in rcqc_functions RCQCClassFnExtension and RCQCStaticFnExtension classes.
self.functions = {
'=': lambda location, value: self.storeNamespaceValue(value, location),
'store': self.storeNamespaceValue,
'store_array': self.storeNamespaceValueAsArray,
'if': self.fnIf,
'fail': self.fail,
'exit': self.exit,
'exists': lambda location: self.namespaceReadValue(location, True),
'-': lambda x: operator.neg(x),
'not': lambda x: operator.not_(x),
'function': lambda x: self.applyRules(x)
}
def diff(self, d1, d2):
# Check d1 and d2 for type; if string (no timestamp) convert
if not d1 or not d2:
return 'N/A'
if type(d1) is types.StringType:
if d1[0] == '0':
return 'N/A'
if self.isISODate(d1):
d1 = self.isoToTimestamp(d1)
if type(d2) is types.StringType:
if d2[0] == '0':
return 'N/A'
if self.isISODate(d2):
d2 = self.isoToTimestamp(d2)
# Calculate difference in seconds
delta = d1 - d2
if delta < 0:
delta = operator.neg(delta)
# Calculate days
days = int(round(delta / 60 / 60 / 24))
if days == 0:
days = 1
return days
def __neg__(self): return Quaternion([operator.neg(c) for c in self]) def __pos__(self): return Quaternion([operator.pos(c) for c in self])
def __neg__(self):
return self.__class__(operator.neg(self._magnitude), self._units)
print()
# pop
print(d.pop("dob"))
print(d)
#---------------------------------- Operators ---------------------------------
print("\n" * 3, "\t" * 2, "Operators\n\n" * 1)
import operator
print("pow(5,5)\t\t", operator.pow(5, 5))
print("abs(33)\t\t", operator.abs(33))
print("add(2,3)\t\t", operator.add(2, 3))
print("eq(2,2)\t\t", operator.eq(2, 2))
print("mod(23,8)\t\t", operator.mod(23, 8))
print("neg(66)\t\t", operator.neg(66))
print("xor(22,55)\t\t", operator.xor(22, 55))
print("sub(55,77)\t\t", operator.sub(55, 77))
print()
# != < > == >= <= and or not
# identity operator--> is / is not
print(2 is 2)
print(2 is not 2)
#---------------------------------- Operator Overloading ---------------------------------
print("\n" * 3, "\t" * 2, "Operator Overloading\n\n" * 1)
# + (Strings)
x = "3" #input()
y = "23" #input()
def run_model(self):
## initialize data structure
self.res = np.zeros([
self.duration,
len(MODEL_RUN_COLUMNS) + len(EXPORT_COLUMNS_FOR_CSV)
],
dtype=np.float32)
self.res[0, 0] = self.nf1
self.res[0, 1] = self.nf2
self.res[0, 2] = self.nf3
self.res[0, 3] = self.nm1
self.res[0, 4] = self.nm2
self.res[0, 5] = self.nm3
self.res[0, 6] = 0
self.res[0, 7] = 0
self.res[0, 8] = 0
self.res[0, 9] = self.female_promotion_probability_1
self.res[0, 10] = self.female_promotion_probability_2
self.res[0, 11] = np.float32(
sum(list([self.nf1, self.nf2, self.nf3])) / sum(
list([
self.nf1, self.nf2, self.nf3, self.nm1, self.nm2, self.nm3
])))
self.res[0, 12] = 0
self.res[0, 13] = self.res[0, 0:6].sum()
self.res[0, 14:] = 0
# I assign the state variables to temporary variables. That way I
# don't have to worry about overwriting the original state variables.
hiring_rate_female_level_1 = self.bf1
hiring_rate_female_level_2 = self.bf2
hiring_rate_female_level_3 = self.bf3
attrition_rate_female_level_1 = self.df1
attrition_rate_female_level_2 = self.df2
attrition_rate_female_level_3 = self.df3
attrition_rate_male_level_1 = self.dm1
attrition_rate_male_level_2 = self.dm2
attrition_rate_male_level_3 = self.dm3
probability_of_outside_hire_level_3 = self.phire3
probability_of_outside_hire_level_2 = self.phire2
female_promotion_probability_1_2 = self.female_promotion_probability_1
female_promotion_probability_2_3 = self.female_promotion_probability_2
department_size_upper_bound = self.upperbound
department_size_lower_bound = self.lowerbound
variation_range = self.variation_range
unfilled_vacanies = 0
change_to_level_1 = 0
change_to_level_2 = 0
change_to_level_3 = 0
for i in range(1, self.duration):
# initialize variables for this iteration
prev_number_of_females_level_1 = self.res[i - 1, 0]
prev_number_of_females_level_2 = self.res[i - 1, 1]
prev_number_of_females_level_3 = self.res[i - 1, 2]
prev_number_of_males_level_1 = self.res[i - 1, 3]
prev_number_of_males_level_2 = self.res[i - 1, 4]
prev_number_of_males_level_3 = self.res[i - 1, 5]
prev_number_of_vacancies_level_3 = self.res[i - 1, 6]
prev_number_of_vacancies_level_2 = self.res[i - 1, 7]
prev_number_of_vacancies_level_1 = self.res[i - 1, 8]
prev_promotion_rate_female_level_1 = self.female_promotion_probability_1
prev_promotion_rate_female_level_2 = self.female_promotion_probability_2
department_size = self.res[i - 1, 0:6].sum()
# Process Model
# Determine department size variation for this timestep
# first both female and males leave the department according to binomial probability.
female_attrition_level_3 = binomial(prev_number_of_females_level_3,
attrition_rate_female_level_3)
male_attrition_level_3 = binomial(prev_number_of_males_level_3,
attrition_rate_male_level_3)
# the departures create a set of vacancies. These vacancies are the basis for new hiring
total_vacancies_3 = female_attrition_level_3 + \
male_attrition_level_3 + change_to_level_3
# women are hired first and then men
hiring_female_3 = binomial(
max(0,
total_vacancies_3), probability_of_outside_hire_level_3 *
hiring_rate_female_level_3)
hiring_male_3 = binomial(
max(0, total_vacancies_3 - hiring_female_3),
probability_of_outside_hire_level_3 *
(1 - hiring_rate_female_level_3))
total_hiring_3 = hiring_female_3 + hiring_male_3
# level 3 vacancies that are not filled by new hires create opportunities
# for promotion from level 2. Again women are promoted first and men second.
# Also note the error trap that if we try to promote more professors from
# level 2 than there exist at level 2, then we will prevent this from happening.
vacancies_remaining_after_hiring_3 = total_vacancies_3 - total_hiring_3
potential_promotions_after_hiring_3 = max(
0, vacancies_remaining_after_hiring_3)
promotions_of_females_level_2_3 = binomial(
min(potential_promotions_after_hiring_3,
prev_number_of_females_level_2),
female_promotion_probability_2_3)
promotions_of_males_level_2_3 = binomial(
max(
0,
min(
vacancies_remaining_after_hiring_3 -
promotions_of_females_level_2_3,
prev_number_of_males_level_2)),
1 - female_promotion_probability_2_3)
# attrition at level 2 - either people leave from attrition or promotion
female_attrition_level_2 = binomial(
max(
0, prev_number_of_females_level_2 -
promotions_of_females_level_2_3),
attrition_rate_female_level_2)
male_attrition_level_2 = binomial(
max(
0, prev_number_of_males_level_2 -
promotions_of_males_level_2_3),
attrition_rate_male_level_2)
# the departures create a set of vacancies. These vacancies are the basis for new hiring
total_vacancies_2 = sum(
list([
female_attrition_level_2, male_attrition_level_2,
promotions_of_females_level_2_3,
promotions_of_males_level_2_3, change_to_level_2
]))
hiring_female_2 = binomial(
max(0,
total_vacancies_2), probability_of_outside_hire_level_2 *
hiring_rate_female_level_2)
hiring_male_2 = binomial(
max(0, total_vacancies_2 - hiring_female_2),
probability_of_outside_hire_level_2 *
(1 - hiring_rate_female_level_2))
total_hiring_2 = hiring_female_2 + hiring_male_2
vacancies_remaining_after_hiring_2 = total_vacancies_2 - total_hiring_2
potential_promotions_after_hiring_2 = max(
0, vacancies_remaining_after_hiring_2)
promotions_of_females_level_1_2 = binomial(
max(
0,
min(potential_promotions_after_hiring_2,
prev_number_of_females_level_1)),
female_promotion_probability_1_2)
promotions_of_males_level_1_2 = binomial(
max(
0,
min(
vacancies_remaining_after_hiring_2 -
promotions_of_females_level_1_2,
prev_number_of_females_level_1)),
probability_of_outside_hire_level_2 *
(1 - female_promotion_probability_1_2))
## Level 1
female_attrition_level_1 = binomial(
max(
0, prev_number_of_females_level_1 -
promotions_of_females_level_1_2),
attrition_rate_female_level_1)
male_attrition_level_1 = binomial(
max(
0, prev_number_of_males_level_1 -
promotions_of_males_level_1_2),
attrition_rate_male_level_1)
total_vacancies_1 = sum(
list([
female_attrition_level_1, male_attrition_level_1,
promotions_of_females_level_1_2,
promotions_of_males_level_1_2, change_to_level_1
]))
hiring_female_1 = binomial(max(0, total_vacancies_1),
hiring_rate_female_level_1)
hiring_male_1 = binomial(
max(0, total_vacancies_1 - hiring_female_1),
1 - hiring_rate_female_level_1)
# Write state variables to array and move to next iteration
self.res[i, 0] = number_of_females_level_1 = sum(
list([
prev_number_of_females_level_1,
neg(female_attrition_level_1),
neg(promotions_of_females_level_1_2), hiring_female_1
]))
assert (number_of_females_level_1 >=
0), "negative number of females 1"
self.res[i, 1] = number_of_females_level_2 = max(
0,
sum(
list([
prev_number_of_females_level_2,
neg(female_attrition_level_2),
neg(promotions_of_females_level_2_3),
promotions_of_females_level_1_2, hiring_female_2
])))
self.res[i, 2] = number_of_females_level_3 = sum(
list([
prev_number_of_females_level_3,
neg(female_attrition_level_3),
promotions_of_females_level_2_3, hiring_female_3
]))
self.res[i, 3] = number_of_males_level_1 = sum(
list([
prev_number_of_males_level_1,
neg(male_attrition_level_1),
neg(promotions_of_males_level_1_2), hiring_male_1
]))
self.res[i, 4] = number_of_males_level_2 = sum(
list([
prev_number_of_males_level_2,
neg(male_attrition_level_2),
neg(promotions_of_males_level_2_3),
promotions_of_males_level_1_2, hiring_male_2
]))
self.res[i, 5] = number_of_males_level_3 = sum(
list([
prev_number_of_males_level_3,
neg(male_attrition_level_3), promotions_of_males_level_2_3,
hiring_male_3
]))
self.res[i, 6] = sum(
list([male_attrition_level_3, female_attrition_level_3]))
self.res[i, 7] = sum(
list([
male_attrition_level_2, female_attrition_level_2,
promotions_of_females_level_2_3,
promotions_of_males_level_2_3
]))
self.res[i, 8] = sum(
list([
male_attrition_level_1, female_attrition_level_1,
promotions_of_males_level_1_2,
promotions_of_females_level_1_2
]))
self.res[i, 9] = self.female_promotion_probability_1
self.res[i, 10] = self.female_promotion_probability_2
self.res[i, 11] = np.float32(
truediv(
sum(
list([
number_of_females_level_1,
number_of_females_level_2,
number_of_females_level_3
])),
sum(
list([
number_of_females_level_1,
number_of_females_level_2,
number_of_females_level_3, number_of_males_level_1,
number_of_males_level_2, number_of_males_level_3
]))))
unfilled_vacanies = abs(department_size - self.res[i, 0:6].sum())
self.res[i, 12] = unfilled_vacanies
department_size = self.res[i, 0:6].sum()
self.res[i, 13] = department_size
self.res[i, 14] = hiring_female_3
self.res[i, 15] = hiring_male_3
self.res[i, 16] = hiring_female_2
self.res[i, 17] = hiring_male_2
self.res[i, 18] = hiring_female_1
self.res[i, 19] = hiring_male_1
self.res[i, 20] = 0
self.res[i, 21] = 0
self.res[i, 22] = promotions_of_females_level_2_3
self.res[i, 23] = promotions_of_males_level_2_3
self.res[i, 24] = promotions_of_females_level_1_2
self.res[i, 25] = promotions_of_males_level_1_2
self.res[i, 26] = hiring_rate_female_level_1
self.res[i, 27] = hiring_rate_female_level_2
self.res[i, 28] = hiring_rate_female_level_3
self.res[i, 29] = 1 - hiring_rate_female_level_1
self.res[i, 30] = 1 - hiring_rate_female_level_2
self.res[i, 31] = 1 - hiring_rate_female_level_3
self.res[i, 32] = attrition_rate_female_level_1
self.res[i, 33] = attrition_rate_female_level_2
self.res[i, 34] = attrition_rate_female_level_3
self.res[i, 35] = attrition_rate_male_level_1
self.res[i, 36] = attrition_rate_male_level_2
self.res[i, 37] = attrition_rate_male_level_3
self.res[i, 38] = 1
self.res[i, 39] = probability_of_outside_hire_level_2
self.res[i, 40] = probability_of_outside_hire_level_3
self.res[i, 41] = female_promotion_probability_1_2
self.res[i, 42] = female_promotion_probability_2_3
self.res[i, 43] = 1 - female_promotion_probability_1_2
self.res[i, 44] = 1 - female_promotion_probability_2_3
self.res[i, 45] = department_size_upper_bound
self.res[i, 46] = department_size_lower_bound
self.res[i, 47] = variation_range
self.res[i, 48] = self.duration
# this produces an array of values. Then I need to assign the
# values to levels. So if I have say a range of variation of 5. I
# will get something like [-1,0,1,-1,0] or something. I need to
# turn this into something like [2,-1,0]. That means randomly
# assigning the values in the array to levels.
flag = False
while flag == False:
changes = np.random.choice([-1, 0, 1], variation_range)
levels = np.random.choice([1, 2, 3], variation_range) #
# random level
# choice
# need to test whether the candidate changes keep the
# department size within bounds.
# print(["old dept size:", department_size,
# "new dept size:", self.res[i, 0:6].sum(),
# "candidate:", department_size +
# changes.sum(),
# " added postions: ", changes.sum(),
# "unfilled ", unfilled_vacanies])
if (department_size + changes.sum() <=
department_size_upper_bound
and department_size + changes.sum() >=
department_size_lower_bound):
change_to_level_3 = np.int(
changes[np.where(levels == 3)[0]].sum())
change_to_level_2 = np.int(
changes[np.where(levels == 2)[0]].sum())
change_to_level_1 = np.int(
changes[np.where(levels == 1)[0]].sum())
flag = True
if (department_size > department_size_upper_bound):
change_to_level_3 = 0
change_to_level_2 = 0
change_to_level_1 = 0
flag = True
if department_size < department_size_lower_bound:
changes = np.ones(variation_range)
change_to_level_3 = np.int(
changes[np.where(levels == 3)[0]].sum())
change_to_level_2 = np.int(
changes[np.where(levels == 2)[0]].sum())
change_to_level_1 = np.int(
changes[np.where(levels == 1)[0]].sum())
flag = True
df_ = pd.DataFrame(self.res)
df_.columns = MODEL_RUN_COLUMNS + EXPORT_COLUMNS_FOR_CSV
recarray_results = df_.to_records(index=True)
self.run = recarray_results
return recarray_results
def __neg__(self): return NonStandardInteger(operator.neg(self.val))
def __neg__(self):
return Coordinates(operator.neg(self.x), operator.neg(self.y),
operator.neg(self.z))
def negate_usecase(x):
return operator.neg(x)
def get_score(self):
from math import sin, cos
from operator import add, sub, mul, neg
div = lambda a, b: a / b if b != 0 else 1 # protected division operator
return div(self.mbs, neg(mul(self.ed, div(self.ls, self.mbs)))) + self.ks * Feature.KS_MAGNIFIER
def __neg__(self):
return Vec2D(operator.neg(self.x), operator.neg(self.y))
def negate_usecase(x):
return operator.neg(x)
def run_parameter_sweep(self, number_of_runs, param, llim, ulim,
num_of_steps):
'''
This function sweeps a single parameter and captures the effect of
that variation on the overall model. Any valid parameter can be chosen.
:param number_of_runs: The number of model iterations per parameter
value
:type number_of_runs: int
:param param: The name of the parameter to sweep
:type param: basestring
:param llim: lower limit of the parameter value
:type llim: float
:param ulim: upper limit of the parameter value
:type ulim: float
:param num_of_steps: total number of increments in the range between
the upper and lower limits.
:type num_of_steps: int
:return: a Dataframe containing individual model runs using the
parameter increments
:rtype: pandas Dataframe
'''
# First I will create a structured array to hold the results of the
# simulation. The size of the array should be one for each step in the
# parameter sweep. To calculate that,
parameter_sweep_increments = np.linspace(llim, ulim, num_of_steps)
parameter_sweep_results = pd.DataFrame(
np.zeros([
len(parameter_sweep_increments),
len(RESULTS_COLUMNS + FEMALE_MATRIX_COLUMNS[1:]) + 1
]))
parameter_sweep_results.columns = ['increment'] + RESULTS_COLUMNS + \
FEMALE_MATRIX_COLUMNS[1:]
parameter_sweep_results.loc[:,
'increment'] = parameter_sweep_increments
# Run simulations with parameter increments and collect into a container.
for i, val in enumerate(parameter_sweep_increments):
setattr(self, param, val)
self.run_multiple(number_of_runs)
# Sets the values the sweep data matrix to the last values in the
# multiple runs results_matrix.
parameter_sweep_results.iloc[i, 1:neg(len(FEMALE_MATRIX_COLUMNS)) -
1] = self.results_matrix.tail(1).iloc[
0, 1:-1]
# Sets the year in the sweep data matrix to the last year in the
# results_matrix.
parameter_sweep_results.iloc[i, 1] = self.results_matrix.tail(
1).iloc[0, len(RESULTS_COLUMNS)]
# Fills the sweep matrix with data from the female percentage
# matrices
parameter_sweep_results.iloc[i, len(RESULTS_COLUMNS) + 1:] = \
self.pct_female_matrix.tail(1).iloc[0, 1:]
self.parameter_sweep_results = parameter_sweep_results
# BEGIN BLOCK
# Reset the models to original settings. This is very important,
# otherwise settings from the parameter sweep will contaminate
# subsequent runs of the model.
self.load_baseline_data_mgmt()
# END BLOCK
return (0)
def __neg__(self) -> bool:
return operator.neg(self.size)
def run_model(self):
## initialize data structure
self.res = np.zeros([self.duration, 12], dtype=np.float32)
self.res[0, 0] = self.nf1
self.res[0, 1] = self.nf2
self.res[0, 2] = self.nf3
self.res[0, 3] = self.nm1
self.res[0, 4] = self.nm2
self.res[0, 5] = self.nm3
self.res[0, 6] = self.vac3
self.res[0, 7] = self.vac2
self.res[0, 8] = self.vac1
self.res[0, 9] = self.female_promotion_probability_1
self.res[0, 10] = self.female_promotion_probability_2
self.res[0, 11] = np.float32(
sum(list([self.nf1, self.nf2, self.nf3])) / sum(
list([
self.nf1, self.nf2, self.nf3, self.nm1, self.nm2, self.nm3
])))
hiring_rate_female_level_1 = self.bf1
hiring_rate_female_level_2 = self.bf2
hiring_rate_female_level_3 = self.bf3
attrition_rate_female_level_1 = self.df1
attrition_rate_female_level_2 = self.df2
attrition_rate_female_level_3 = self.df3
attrition_rate_male_level_1 = self.dm1
attrition_rate_male_level_2 = self.dm2
attrition_rate_male_level_3 = self.dm3
probability_of_outside_hire_level_3 = self.phire3
probability_of_outside_hire_level_2 = self.phire2
male_promotion_probability_1_2 = self.male_promotion_probability_1
male_promotion_probability_2_3 = self.male_promotion_probability_2
for i in range(1, self.duration):
# initialize variables for this iteration
prev_number_of_females_level_1 = self.res[i - 1, 0]
prev_number_of_females_level_2 = self.res[i - 1, 1]
prev_number_of_females_level_3 = self.res[i - 1, 2]
prev_number_of_males_level_1 = self.res[i - 1, 3]
prev_number_of_males_level_2 = self.res[i - 1, 4]
prev_number_of_males_level_3 = self.res[i - 1, 5]
prev_number_of_vacancies_level_3 = self.res[i - 1, 6]
prev_number_of_vacancies_level_2 = self.res[i - 1, 7]
prev_number_of_vacancies_level_1 = self.res[i - 1, 8]
prev_promotion_rate_female_level_1 = self.female_promotion_probability_1
prev_promotion_rate_female_level_2 = self.female_promotion_probability_2
if np.isnan(prev_promotion_rate_female_level_1):
prev_promotion_rate_female_level_1 = 0
if np.isnan(prev_promotion_rate_female_level_2):
prev_promotion_rate_female_level_2 = 0
prev_gender_proportion_of_department = np.float32(
sum(
list([
prev_number_of_females_level_1,
prev_number_of_females_level_2,
prev_number_of_females_level_3
])) / (sum(
list([
prev_number_of_females_level_1,
prev_number_of_females_level_2,
prev_number_of_females_level_3,
prev_number_of_males_level_1,
prev_number_of_males_level_2,
prev_number_of_males_level_3
]))))
# Process Model
# first both female and males leave the department according to binomial probability.
female_attrition_level_3 = binomial(prev_number_of_females_level_3,
attrition_rate_female_level_3)
male_attrition_level_3 = binomial(prev_number_of_males_level_3,
attrition_rate_male_level_3)
# the departures create a set of vacancies. These vacancies are the basis for new hiring
total_vacancies_3 = female_attrition_level_3 + male_attrition_level_3
# women are hired first and then men
hiring_female_3 = binomial(
total_vacancies_3, probability_of_outside_hire_level_3 *
hiring_rate_female_level_3)
hiring_male_3 = binomial(
max(0, total_vacancies_3 - hiring_female_3),
probability_of_outside_hire_level_3 *
(1 - hiring_rate_female_level_3))
# promotion after hiring level 3
promotions_female_after_hiring_2_3 = binomial(
max(prev_number_of_females_level_2,
total_vacancies_3 - hiring_female_3 - hiring_male_3),
prev_promotion_rate_female_level_2)
# formula should read that either the remaining vacancies or the previous number of males--whichever is smallest. But need to make sure no negative values.
promotions_of_males_level_2_3 = binomial(
min(
prev_number_of_males_level_2,
max(
0, total_vacancies_3 - hiring_female_3 -
hiring_male_3 - promotions_female_after_hiring_2_3)),
male_promotion_probability_2_3)
assert (promotions_of_males_level_2_3 >=
0), "promotions_of_males_level_2_3 is negative"
# attrition at level 2 - either people leave from attrition or promotion
female_attrition_level_2 = binomial(
max(
0, prev_number_of_females_level_2 -
promotions_female_after_hiring_2_3),
attrition_rate_female_level_2)
male_attrition_level_2 = binomial(
max(
0, prev_number_of_males_level_2 -
promotions_of_males_level_2_3),
attrition_rate_male_level_2)
# the departures create a set of vacancies. These vacancies are the basis for new hiring
total_vacancies_2 = sum(
list([
female_attrition_level_2, male_attrition_level_2,
promotions_female_after_hiring_2_3,
promotions_of_males_level_2_3
]))
assert (total_vacancies_2 >=
0), "total vacancies level 2 is less than zero"
# TODO set to hiring first
hiring_female_2 = binomial(
max(0,
total_vacancies_2), probability_of_outside_hire_level_2 *
hiring_rate_female_level_2)
hiring_male_2 = binomial(
max(0, total_vacancies_2 - hiring_female_2),
1 - probability_of_outside_hire_level_2 *
hiring_rate_female_level_2)
promotions_of_females_level_1_2 = binomial(
min(
0,
min(prev_number_of_females_level_1,
total_vacancies_2 - hiring_female_2 - hiring_male_2)),
prev_promotion_rate_female_level_1)
promotions_of_males_level_1_2 = binomial(
min(
0,
min(
prev_number_of_males_level_1,
total_vacancies_2 - hiring_female_2 - hiring_male_2 -
promotions_of_females_level_1_2)),
male_promotion_probability_1_2)
assert (promotions_of_females_level_1_2 >=
0), "promotions of females level 1-2 is negative"
assert (promotions_of_males_level_1_2 >=
0), "promotions of males level 1-2 is negative"
total_hiring_2 = hiring_female_2 + hiring_male_2
## Level 1
female_attrition_level_1 = binomial(
max(
0, prev_number_of_females_level_1 -
promotions_of_females_level_1_2),
attrition_rate_female_level_1)
male_attrition_level_1 = binomial(
max(0, prev_number_of_males_level_1),
attrition_rate_male_level_1)
total_vacancies_1 = sum(
list([
female_attrition_level_1, male_attrition_level_1,
promotions_of_females_level_1_2,
promotions_of_males_level_1_2
]))
hiring_female_1 = binomial(total_vacancies_1,
hiring_rate_female_level_1)
hiring_male_1 = binomial(
max(0, total_vacancies_1 - hiring_female_1),
1 - hiring_rate_female_level_1)
# Write state variables to array and move to next iteration
self.res[i, 0] = number_of_females_level_1 = sum(
list([
prev_number_of_females_level_1,
neg(female_attrition_level_1),
neg(promotions_of_females_level_1_2), hiring_female_1
]))
self.res[i, 1] = number_of_females_level_2 = max(
0,
sum(
list([
prev_number_of_females_level_2,
neg(female_attrition_level_2),
neg(promotions_female_after_hiring_2_3),
promotions_of_females_level_1_2, hiring_female_2
])))
self.res[i, 2] = number_of_females_level_3 = sum(
list([
prev_number_of_females_level_3,
neg(female_attrition_level_3),
promotions_female_after_hiring_2_3, hiring_female_3
]))
self.res[i, 3] = number_of_males_level_1 = sum(
list([
prev_number_of_males_level_1,
neg(male_attrition_level_1),
neg(promotions_of_males_level_1_2), hiring_male_1
]))
self.res[i, 4] = number_of_males_level_2 = sum(
list([
prev_number_of_males_level_2,
neg(male_attrition_level_2),
neg(promotions_of_males_level_2_3),
promotions_of_males_level_1_2, hiring_male_2
]))
self.res[i, 5] = number_of_males_level_3 = sum(
list([
prev_number_of_males_level_3,
neg(male_attrition_level_3), promotions_of_males_level_2_3,
hiring_male_3
]))
self.res[i, 6] = number_of_vacancies_level_3 = sum(
list([male_attrition_level_3, female_attrition_level_3]))
self.res[i, 7] = number_of_vacancies_level_2 = sum(
list([
male_attrition_level_2, female_attrition_level_2,
promotions_female_after_hiring_2_3,
promotions_of_males_level_2_3
]))
self.res[i, 8] = number_of_vacancies_level_1 = sum(
list([
male_attrition_level_1, female_attrition_level_1,
promotions_of_males_level_1_2,
promotions_of_females_level_1_2
]))
self.res[
i,
9] = promotion_rate_female_level_1 = self.female_promotion_probability_1
self.res[
i,
10] = promotion_rate_women_level_2 = self.female_promotion_probability_2
self.res[i, 11] = gender_proportion_of_department = np.float32(
truediv(
sum(
list([
number_of_females_level_1,
number_of_females_level_2,
number_of_females_level_3
])),
sum(
list([
number_of_females_level_1,
number_of_females_level_2,
number_of_females_level_3, number_of_males_level_1,
number_of_males_level_2, number_of_males_level_3
]))))
# print(self.res[i,:])
## Print Data matrix
df_ = pd.DataFrame(self.res)
df_.columns = [
'f1', 'f2', 'f3', 'm1', 'm2', 'm3', 't3', 't2', 't1', 'prom1',
'prom2', 'gendprop'
]
recarray_results = df_.to_records(index=True)
self.run = recarray_results
return recarray_results
#// 只能用于整数 #浮点数只能返回一位小数0 print operator.floordiv(9, 4) print operator.floordiv(9.04, 4) #除数取余 同a%b print operator.mod(9, 4) #精确除法 print operator.truediv(9.04, 4) #绝对值 print operator.abs(-10) #取反 相当于 -a print operator.neg(-10) #取反 相当于 ~a #~a = (-a)-1 #~10 = -11 #~(-10) = 9 print operator.inv(10) print operator.inv(-10) print operator.invert(10) print operator.invert(-10) #乘方 同a**b print operator.pow(2, 3) #向左移位 同<< 相当于乘以2的相应次方 print operator.lshift(3, 2)
def __neg__(self):
return self.__class__(operator.neg(self._magnitude), self._units)
def test_metrics_neg():
first_metric = DummyMetric(1)
final_neg = neg(first_metric)
assert isinstance(final_neg, CompositionalMetric)
assert torch.allclose(tensor(-1), final_neg.compute())
def test_neg_array(self, xp, dtype):
a = testing.shaped_arange((2, 3), xp, dtype)
return operator.neg(a)
def op_neg(x):
return _op.neg(x)
def test_neg_zerodim(self, xp, dtype):
a = xp.array(-2, dtype)
return operator.neg(a)
def __neg__(self):
return vec(operator.neg(self.x), operator.neg(self.y))
print(" bit or ", oct(10), oct(4))
print(" | : ", 10 | 4)
print(" or_(a, b) : ", op.or_(10,4))
print("x.__or__(y) : ", (10).__or__(4))
print(" bit xor ", oct(10), oct(6))
print(" ^ : ", 10 ^ 6)
print(" xor(a, b) : ", op.xor(10,6))
print("x.__xor__(y) : ", (10).__xor__(6))
print(" bit inversion ", oct(10))
print(" ~ : ", ~(10) )
print(" invert(a) : ", op.invert(10))
print("x.__invert__(y) : ", (10).__invert__())
print(" negative : ", -(10))
print(" neg(a) : ", op.neg((10)))
print("a.__neg__ : ", (10).__neg__())
print(" positive : ", -(-10), +(10))
print(" pos(a) : ", op.pos((10)))
print("a.__pos__ : ", (10).__pos__())
print(" right hand operator ")
print(" x + y ", (8).__radd__(2))
print(" x + y ", (2).__add__(8))
print(" x ** y ", (3).__rpow__(2))
print(" x ** y ", (2).__pow__(3))
x = True
y = False
print('x and y is',x and y)
def test_neg(self):
self.failUnless(operator.neg(5) == -5)
self.failUnless(operator.neg(-5) == 5)
self.failUnless(operator.neg(0) == 0)
self.failUnless(operator.neg(-0) == 0)
def __neg__(self):
return neg(self)
def neg_abs(num):
return neg(abs(num))
def operator_neg(size):
a = Array(size, 'int32')
for i in range(size):
a[i] = nb_types.int32(i)
return operator.neg(a)
def __neg__(self) -> T:
return operator.neg(object.__getattribute__(self, "_obj"))
def dom_j_up(csp, queue):
return SortedSet(queue, key=lambda t: neg(len(csp.curr_domains[t[1]])))
def __neg__ (self): return Euler([operator.neg(c) for c in self]) def __pos__ (self): return Euler([operator.pos(c) for c in self])
def run_model(self):
## initialize data structure
self.res = np.zeros([self.duration, 12], dtype=np.float32)
self.res[0, 0] = self.nf1
self.res[0, 1] = self.nf2
self.res[0, 2] = self.nf3
self.res[0, 3] = self.nm1
self.res[0, 4] = self.nm2
self.res[0, 5] = self.nm3
self.res[0, 6] = self.vac3
self.res[0, 7] = self.vac2
self.res[0, 8] = self.vac1
self.res[0, 9] = self.female_promotion_probability_1
self.res[0, 10] = self.female_promotion_probability_2
self.res[0, 11] = np.float32(
sum(list([self.nf1, self.nf2, self.nf3])) / sum(
list([
self.nf1, self.nf2, self.nf3, self.nm1, self.nm2, self.nm3
])))
hiring_rate_female_level_1 = self.bf1
hiring_rate_female_level_2 = self.bf2
hiring_rate_female_level_3 = self.bf3
attrition_rate_female_level_1 = self.df1
attrition_rate_female_level_2 = self.df2
attrition_rate_female_level_3 = self.df3
attrition_rate_male_level_1 = self.dm1
attrition_rate_male_level_2 = self.dm2
attrition_rate_male_level_3 = self.dm3
probability_of_outside_hire_level_3 = self.phire3
probability_of_outside_hire_level_2 = self.phire2
male_promotion_probability_1_2 = self.male_promotion_probability_1
male_promotion_probability_2_3 = self.male_promotion_probability_2
for i in range(1, self.duration):
# initialize variables for this iteration
prev_number_of_females_level_1 = self.res[i - 1, 0]
prev_number_of_females_level_2 = self.res[i - 1, 1]
prev_number_of_females_level_3 = self.res[i - 1, 2]
prev_number_of_males_level_1 = self.res[i - 1, 3]
prev_number_of_males_level_2 = self.res[i - 1, 4]
prev_number_of_males_level_3 = self.res[i - 1, 5]
prev_number_of_vacancies_level_3 = self.res[i - 1, 6]
prev_number_of_vacancies_level_2 = self.res[i - 1, 7]
prev_number_of_vacancies_level_1 = self.res[i - 1, 8]
prev_promotion_rate_female_level_1 = self.female_promotion_probability_1
prev_promotion_rate_female_level_2 = self.female_promotion_probability_2
if np.isnan(prev_promotion_rate_female_level_1):
prev_promotion_rate_female_level_1 = 0
if np.isnan(prev_promotion_rate_female_level_2):
prev_promotion_rate_female_level_2 = 0
prev_gender_proportion_of_department = np.float32(
sum(
list([
prev_number_of_females_level_1,
prev_number_of_females_level_2,
prev_number_of_females_level_3
])) / (sum(
list([
prev_number_of_females_level_1,
prev_number_of_females_level_2,
prev_number_of_females_level_3,
prev_number_of_males_level_1,
prev_number_of_males_level_2,
prev_number_of_males_level_3
]))))
# Process Model
# first both female and males leave the department according to binomial probability.
female_attrition_level_3 = binomial(prev_number_of_females_level_3,
attrition_rate_female_level_3)
male_attrition_level_3 = binomial(prev_number_of_males_level_3,
attrition_rate_male_level_3)
# the departures create a set of vacancies. These vacancies are the basis for new hiring
total_vacancies_3 = female_attrition_level_3 + male_attrition_level_3
# women are hired first and then men
hiring_female_3 = binomial(
total_vacancies_3, probability_of_outside_hire_level_3 *
hiring_rate_female_level_3)
hiring_male_3 = binomial(
max(0, total_vacancies_3 - hiring_female_3),
probability_of_outside_hire_level_3 *
(1 - hiring_rate_female_level_3))
total_hiring_3 = hiring_female_3 + hiring_male_3
# level 3 vacancies that are not filled by new hires create opportunities
# for promotion from level 2. Again women are promoted first and men second.
# Also note the error trap that if we try to promote more professors from
# level 2 than there exist at level 2, then we will prevent this from happening.
vacancies_remaining_after_hiring_3 = total_vacancies_3 - total_hiring_3
potential_promotions_after_hiring_3 = max(
0, vacancies_remaining_after_hiring_3)
promotions_of_females_level_2_3 = binomial(
min(potential_promotions_after_hiring_3,
prev_number_of_females_level_2),
prev_promotion_rate_female_level_2)
promotions_of_males_level_2_3 = binomial(
max(
0,
min(
potential_promotions_after_hiring_3 -
promotions_of_females_level_2_3,
prev_number_of_males_level_2)),
male_promotion_probability_2_3)
# attrition at level 2 - either people leave from attrition or promotion
female_attrition_level_2 = binomial(
max(
0, prev_number_of_females_level_2 -
promotions_of_females_level_2_3),
attrition_rate_female_level_2)
male_attrition_level_2 = binomial(
max(
0, prev_number_of_males_level_2 -
promotions_of_males_level_2_3),
attrition_rate_male_level_2)
# the departures create a set of vacancies. These vacancies are the basis for new hiring
total_vacancies_2 = sum(
list([
female_attrition_level_2, male_attrition_level_2,
promotions_of_females_level_2_3,
promotions_of_males_level_2_3
]))
hiring_female_2 = binomial(
max(0,
total_vacancies_2), probability_of_outside_hire_level_2 *
hiring_rate_female_level_2)
hiring_male_2 = binomial(
max(0, total_vacancies_2 - hiring_female_2),
probability_of_outside_hire_level_2 *
(1 - hiring_rate_female_level_2))
total_hiring_2 = hiring_female_2 + hiring_male_2
vacancies_remaining_after_hiring_2 = total_vacancies_2 - total_hiring_2
potential_promotions_after_hiring_2 = max(
0, vacancies_remaining_after_hiring_2)
promotions_of_females_level_1_2 = binomial(
max(
0,
min(potential_promotions_after_hiring_2,
prev_number_of_females_level_1)),
prev_promotion_rate_female_level_1)
promotions_of_males_level_1_2 = binomial(
max(
0,
min(
potential_promotions_after_hiring_2 -
promotions_of_females_level_1_2,
prev_number_of_males_level_1)),
male_promotion_probability_1_2)
## Level 1
female_attrition_level_1 = binomial(
max(
0, prev_number_of_females_level_1 -
promotions_of_females_level_1_2),
attrition_rate_female_level_1)
male_attrition_level_1 = binomial(
max(
0, prev_number_of_males_level_1 -
promotions_of_males_level_1_2),
attrition_rate_male_level_1)
total_vacancies_1 = sum(
list([
female_attrition_level_1, male_attrition_level_1,
promotions_of_females_level_1_2,
promotions_of_males_level_1_2
]))
hiring_female_1 = binomial(max(0, total_vacancies_1),
hiring_rate_female_level_1)
hiring_male_1 = binomial(
max(0, total_vacancies_1 - hiring_female_1),
1 - hiring_rate_female_level_1)
# Write state variables to array and move to next iteration
self.res[i, 0] = number_of_females_level_1 = sum(
list([
prev_number_of_females_level_1,
neg(female_attrition_level_1),
neg(promotions_of_females_level_1_2), hiring_female_1
]))
assert (number_of_females_level_1 >=
0), "negative number of females 1"
self.res[i, 1] = number_of_females_level_2 = max(
0,
sum(
list([
prev_number_of_females_level_2,
neg(female_attrition_level_2),
neg(promotions_of_females_level_2_3),
promotions_of_females_level_1_2, hiring_female_2
])))
self.res[i, 2] = number_of_females_level_3 = sum(
list([
prev_number_of_females_level_3,
neg(female_attrition_level_3),
promotions_of_females_level_2_3, hiring_female_3
]))
self.res[i, 3] = number_of_males_level_1 = sum(
list([
prev_number_of_males_level_1,
neg(male_attrition_level_1),
neg(promotions_of_males_level_1_2), hiring_male_1
]))
self.res[i, 4] = number_of_males_level_2 = sum(
list([
prev_number_of_males_level_2,
neg(male_attrition_level_2),
neg(promotions_of_males_level_2_3),
promotions_of_males_level_1_2, hiring_male_2
]))
self.res[i, 5] = number_of_males_level_3 = sum(
list([
prev_number_of_males_level_3,
neg(male_attrition_level_3), promotions_of_males_level_2_3,
hiring_male_3
]))
self.res[i, 6] = number_of_vacancies_level_3 = sum(
list([male_attrition_level_3, female_attrition_level_3]))
self.res[i, 7] = number_of_vacancies_level_2 = sum(
list([
male_attrition_level_2, female_attrition_level_2,
promotions_of_females_level_2_3,
promotions_of_males_level_2_3
]))
self.res[i, 8] = number_of_vacancies_level_1 = sum(
list([
male_attrition_level_1, female_attrition_level_1,
promotions_of_males_level_1_2,
promotions_of_females_level_1_2
]))
self.res[i, 9] = promotion_rate_female_level_1 = np.float32(
number_of_females_level_1 /
sum(list([number_of_females_level_1, number_of_males_level_1
])))
self.res[i, 10] = promotion_rate_women_level_2 = np.float32(
number_of_females_level_2 /
sum(list([number_of_females_level_2, number_of_males_level_2
])))
self.res[i, 11] = gender_proportion_of_department = np.float32(
truediv(
sum(
list([
number_of_females_level_1,
number_of_females_level_2,
number_of_females_level_3
])),
sum(
list([
number_of_females_level_1,
number_of_females_level_2,
number_of_females_level_3, number_of_males_level_1,
number_of_males_level_2, number_of_males_level_3
]))))
# print(self.res[i,:])
## Print Data matrix
df_ = pd.DataFrame(self.res)
df_.columns = [
'f1', 'f2', 'f3', 'm1', 'm2', 'm3', 't3', 't2', 't1', 'prom1',
'prom2', 'gendprop'
]
# print(df_)
recarray_results = df_.to_records(index=True)
self.run = recarray_results
return recarray_results
def __rsub__(self, other_obj):
self_neg = operator.neg(self)
return operator.add(self_neg, other_obj)
def __neg__(self):
return Vec2d(operator.neg(self.x), operator.neg(self.y))
class TVMScriptParser(Transformer):
"""Synr AST visitor pass which finally lowers to TIR.
Notes for Extension
-------------------
1. To support a new type of AST node, add a function transform_xxx().
2. To support new functions, add the function to the appropriate registry:
We divide allowed function calls in TVM script into 3 categories,
intrin, scope_handler and special_stmt.
1. intrin functions are low level functions like mod, load, and
constants. They correspond to a tir `IRNode`. They must have a
return value. The user can register intrin functions for the parser to
use.
2. scope_handler functions have no return value. They take two
arguments: the parser and the AST node. scope_handler functions are
used in with and for statements.
3. special_stmt functions handle cases that do not have a corresponding
tir `IRNode`. These functions take the parser and the AST node as
arguments and may return a value.
When visiting a Call node, we check the special_stmt registry first. If
no registered function is found, we then check the intrin registry.
When visiting With node, we check the with_scope registry.
When visiting For node, we check the for_scope registry.
"""
_binop_maker = {
ast.BuiltinOp.Add: tvm.tir.Add,
ast.BuiltinOp.Sub: tvm.tir.Sub,
ast.BuiltinOp.Mul: tvm.tir.Mul,
ast.BuiltinOp.Div: tvm.tir.Div,
ast.BuiltinOp.FloorDiv: tvm.tir.FloorDiv,
ast.BuiltinOp.Mod: tvm.tir.FloorMod,
ast.BuiltinOp.BitOr: lambda lhs, rhs, span: operator.or_(lhs, rhs),
ast.BuiltinOp.BitAnd: lambda lhs, rhs, span: operator.and_(lhs, rhs),
ast.BuiltinOp.BitXor: lambda lhs, rhs, span: operator.xor(lhs, rhs),
ast.BuiltinOp.GT: tvm.tir.GT,
ast.BuiltinOp.GE: tvm.tir.GE,
ast.BuiltinOp.LT: tvm.tir.LT,
ast.BuiltinOp.LE: tvm.tir.LE,
ast.BuiltinOp.Eq: tvm.tir.EQ,
ast.BuiltinOp.NotEq: tvm.tir.NE,
ast.BuiltinOp.And: tvm.tir.And,
ast.BuiltinOp.Or: tvm.tir.Or,
}
_unaryop_maker = {
ast.BuiltinOp.USub: lambda rhs, span: operator.neg(rhs),
ast.BuiltinOp.Invert: lambda rhs, span: operator.invert(rhs),
ast.BuiltinOp.Not: tvm.tir.Not,
}
def __init__(self, base_lienno, tir_namespace):
self.context = None
self.base_lineno = base_lienno
self.current_lineno = 0
self.current_col_offset = 0
self.tir_namespace = tir_namespace
self.meta = None
def init_function_parsing_env(self):
"""Initialize function parsing environment"""
self.context = ContextMaintainer(self.report_error) # scope emitter
def init_meta(self, meta_dict):
if meta_dict is not None:
self.meta = tvm.ir.load_json(json.dumps(meta_dict))
def transform(self, node):
"""Generic transformation for visiting the AST. Dispatches to
`transform_ClassName` for the appropriate ClassName."""
old_lineno, old_col_offset = self.current_lineno, self.current_col_offset
if hasattr(node, "lineno"):
self.current_lineno = self.base_lineno + node.lineno - 1
if hasattr(node, "col_offset"):
self.current_col_offset = node.col_offset
method = "transform_" + node.__class__.__name__
visitor = getattr(self, method, self.generic_visit)
transform_res = visitor(node)
self.current_lineno, self.current_col_offset = old_lineno, old_col_offset
return transform_res
def match_tir_namespace(self, identifier: str) -> bool:
"""Check if the namespace is equal to tvm.script.tir"""
return identifier in self.tir_namespace
def report_error(self, message: str, span: Union[ast.Span, tvm.ir.Span]):
"""Report an error occuring at a location.
This just dispatches to synr's DiagnosticContext.
Parameters
----------
message : str
Error message
span : Union[synr.ast.Span, tvm.ir.Span】
Location of the error
"""
if isinstance(span, tvm.ir.Span):
span = synr_span_from_tvm(span)
self.error(message, span)
def parse_body(self, parent):
"""Parse remaining statements in this scope.
Parameters
----------
parent : synr.ast.Node
Parent node of this scope. Errors will be reported here.
"""
body = []
spans = []
stmt = parent
while len(self.context.node_stack[-1]) > 0:
stmt = self.context.node_stack[-1].pop()
spans.append(stmt.span)
res = self.transform(stmt)
if res is not None:
body.append(res)
if len(body) == 0:
self.report_error(
"Expected another statement at the end of this block. Perhaps you "
"used a concise statement and forgot to include a body afterwards.",
stmt.span,
)
else:
return (tvm.tir.SeqStmt(body,
tvm_span_from_synr(ast.Span.union(spans)))
if len(body) > 1 else body[0])
def parse_arg_list(self, func, node_call):
"""Match the arguments of a function call in the AST to the required
arguments of the function. This handles positional arguments,
positional arguments specified by name, keyword arguments, and varargs.
Parameters
----------
func : Function
The function that provides the signature
node_call: ast.Call
The AST call node that calls into the function.
Returns
-------
arg_list : list
The parsed positional argument.
"""
assert isinstance(node_call, ast.Call)
# collect arguments
args = [self.transform(arg) for arg in node_call.params]
kw_args = {
self.transform(k): self.transform(v)
for k, v in node_call.keyword_params.items()
}
# get the name and parameter list of func
if isinstance(func, (Intrin, ScopeHandler, SpecialStmt)):
func_name, param_list = func.signature()
else:
self.report_error(
"Internal Error: function must be of type Intrin, ScopeHandler or SpecialStmt, "
f"but it is {type(func).__name__}",
node_call.span,
)
# check arguments and parameter list and get a list of arguments
reader = CallArgumentReader(func_name, args, kw_args, self, node_call)
pos_only, kwargs, varargs = param_list
internal_args = list()
for i, arg_name in enumerate(pos_only):
internal_args.append(reader.get_pos_only_arg(i + 1, arg_name))
for i, arg_info in enumerate(kwargs):
arg_name, default = arg_info
internal_args.append(
reader.get_kwarg(i + 1 + len(pos_only),
arg_name,
default=default))
if varargs is not None:
internal_args.extend(
reader.get_varargs(len(pos_only) + len(kwargs) + 1))
elif len(args) + len(kw_args) > len(pos_only) + len(kwargs):
self.report_error(
"Arguments mismatched. " +
f"Expected {len(pos_only) + len(kwargs)} args but got " +
f"{len(args) + len(kw_args)}",
node_call.span,
)
return internal_args
def parse_type(self, type_node, parent):
"""Parse a type annotation.
We require the parent object to the type so that we have a place to
report the error message if the type does not exist.
"""
if type_node is None:
self.report_error("A type annotation is required", parent.span)
res_type = self.transform(type_node)
return tvm.ir.TupleType(
[]) if res_type is None else res_type.evaluate()
def generic_visit(self, node):
"""Fallback visitor if node type is not handled. Reports an error."""
self.report_error(
type(node).__name__ + " AST node is not supported", node.span)
def transform_Module(self, node):
"""Module visitor
Right now, we only support two formats for TVM Script.
Example
-------
1. Generate a PrimFunc (If the code is printed, then it may also contain metadata)
.. code-block:: python
import tvm
@tvm.script
def A(...):
...
# returns a PrimFunc
func = A
2. Generate an IRModule
.. code-block:: python
import tvm
@tvm.script.ir_module
class MyMod():
@T.prim_func
def A(...):
...
@T.prim_func
def B(...):
...
__tvm_meta__ = ...
# returns an IRModule
mod = MyMod
"""
if len(node.funcs) == 1:
return self.transform(next(iter(node.funcs.values())))
elif len(node.func) == 0:
self.report_error(
"You must supply at least one class or function definition",
node.span)
else:
self.report_error(
"Only one-function, one-class or function-with-meta source code is allowed",
ast.Span.union([x.span
for x in list(node.funcs.values())[1:]]),
)
def transform_Class(self, node):
"""Class definition visitor.
A class can have multiple function definitions and a single
:code:`__tvm_meta__` statement. Each class corresponds to a single
:code:`IRModule`.
Example
-------
.. code-block:: python
@tvm.script.ir_module
class MyClass:
__tvm_meta__ = {}
def A():
T.evaluate(0)
"""
if len(node.assignments) == 1:
if not (len(node.assignments[0].lhs) == 1
and isinstance(node.assignments[0].lhs[0], ast.Var)
and node.assignments[0].lhs[0].id.name == "__tvm_meta__"):
self.report_error(
"The only top level assignments allowed are `__tvm_meta__ = ...`",
node.assignments[0].span,
)
self.init_meta(MetaUnparser().do_transform(
node.assignments[0].rhs, self._diagnostic_context))
elif len(node.assignments) > 1:
self.report_error(
"Only a single top level `__tvm_meta__` is allowed",
ast.Span.union([x.span for x in node.assignments[1:]]),
)
return IRModule({
GlobalVar(name): self.transform(func)
for name, func in node.funcs.items()
})
def transform_Function(self, node):
"""Function definition visitor.
Each function definition is translated to a single :code:`PrimFunc`.
There are a couple restrictions on TVM Script functions:
1. Function arguments must have their types specified.
2. The body of the function can contain :code:`func_attr` to specify
attributes of the function (like it's name).
3. The body of the function can also contain multiple :code:`buffer_bind`s,
which give shape and dtype information to arguments.
4. Return statements are implicit.
Example
-------
.. code-block:: python
@T.prim_func
def my_function(x: T.handle): # 1. Argument types
T.func_attr({"global_symbol": "mmult"}) # 2. Function attributes
X_1 = tir.buffer_bind(x, [1024, 1024]) # 3. Buffer binding
T.evaluate(0) # 4. This function returns 0
"""
def check_decorator(decorators: List[ast.Expr]) -> bool:
"""Check the decorator is `T.prim_func"""
if len(decorators) != 1:
return False
d: ast.Expr = decorators[0]
return (isinstance(d, ast.Attr) and isinstance(d.object, ast.Var)
and self.match_tir_namespace(d.object.id.name)
and d.field.name == "prim_func")
self.init_function_parsing_env()
self.context.enter_scope(nodes=node.body.stmts)
# add parameters of function
for arg in node.params:
arg_var = tvm.te.var(arg.name, self.parse_type(arg.ty, arg))
self.context.update_symbol(arg.name, arg_var, node)
self.context.func_params.append(arg_var)
if not check_decorator(node.decorators):
self.report_error(
"All functions should be decorated by `T.prim_func`",
node.span,
)
# fetch the body of root block
body = self.parse_body(node.body)
# return a tir.PrimFunc
dict_attr = self.context.func_dict_attr
ret_type = self.parse_type(node.ret_type,
node) if node.ret_type is not None else None
func = tvm.tir.PrimFunc(
self.context.func_params,
body,
ret_type,
buffer_map=self.context.func_buffer_map,
attrs=tvm.ir.make_node("DictAttrs", **dict_attr)
if dict_attr else None,
span=tvm_span_from_synr(node.span),
)
# New Scope : Implicit root block
# Each function contains an implicit root block in TensorIR,
# so here we need a block scope for it.
# If the PrimFunc is not a TensorIR func (e.g. TE scheduled func or low-level func),
# the root block will not be added. The logic to add root block is in `_ffi_api.Complete`
# Fix the PrimFunc
# 1. generate root block if necessary
# 2. generate surrounding loops for blocks if necessary
func = call_with_error_reporting(
self.report_error,
node.span,
_ffi_api.Complete,
func,
self.context.root_alloc_buffers,
)
self.context.exit_scope()
return func
def transform_Lambda(self, node):
"""Lambda visitor
Return an array of input parameters and the transformed lambda body.
"""
self.context.enter_scope(nodes=[node.body])
# add parameters of the lambda
arg_vars = []
for arg in node.params:
arg_var = tvm.te.var(arg.name)
arg_vars.append(arg_var)
self.context.update_symbol(arg.name, arg_var, node)
# the body of a lambda must be an expr
if not isinstance(node.body, ast.Expr):
self.report_error("The body of a lambda must be an expression",
node.span)
# transform the body of the lambda
body = self.transform(node.body)
self.context.exit_scope()
return arg_vars, body
def transform_Assign(self, node):
"""Assign visitor
AST abstract grammar:
Assign(expr* targets, expr value, string? type_comment)
By now 3 patterns of Assign is supported:
1. special stmts with return value
1.1 Buffer = T.match_buffer()/T.buffer_decl()
1.2 Var = T.var()
1.3 Var = T.env_thread()
2. (BufferStore) Buffer[PrimExpr, PrimExpr, ..., PrimExpr] = PrimExpr
3. (Store) Var[PrimExpr] = PrimExpr
4. with scope handlers with concise scoping and var def
4.1 var = T.allocate()
"""
if isinstance(node.rhs, ast.Call):
# Pattern 1 & Pattern 4
func = self.transform(node.rhs.func_name)
if isinstance(func, WithScopeHandler):
if not func.concise_scope or not func.def_symbol:
self.report_error(
"with scope handler " + func.signature()[0] +
" is not suitable here",
node.rhs.span,
)
# Pattern 4
arg_list = self.parse_arg_list(func, node.rhs)
func.enter_scope(node, self.context, arg_list,
node.rhs.func_name.span)
func.body = self.parse_body(node)
return func.exit_scope(node, self.context, arg_list,
node.rhs.func_name.span)
elif isinstance(func, SpecialStmt):
# Pattern 1
arg_list = self.parse_arg_list(func, node.rhs)
func.handle(node, self.context, arg_list,
node.rhs.func_name.span)
return self.parse_body(node)
else:
value = self.transform(node.rhs)
if len(node.lhs) == 1 and not isinstance(node.lhs[0], ast.Var):
# This is a little confusing because it only is true when
# we have taken this branch. We might need to clarify what
# exectly is allowed in Assignments in tvmscript.
self.report_error(
"Left hand side of assignment must be an unqualified variable",
node.span,
)
ast_var = node.lhs[0]
var = tvm.te.var(
ast_var.id.name,
self.parse_type(node.ty, ast_var),
span=tvm_span_from_synr(ast_var.span),
)
self.context.update_symbol(var.name, var, node)
body = self.parse_body(node)
self.context.remove_symbol(var.name)
return tvm.tir.LetStmt(var,
value,
body,
span=tvm_span_from_synr(node.span))
self.report_error("Unsupported Assign stmt", node.span)
def transform_SubscriptAssign(self, node):
"""Visitor for statements of the form :code:`x[1] = 2`."""
symbol = self.transform(node.params[0])
indexes = self.transform(node.params[1])
rhs = self.transform(node.params[2])
rhs_span = tvm_span_from_synr(node.params[2].span)
if isinstance(symbol, tvm.tir.Buffer):
# BufferStore
return tvm.tir.BufferStore(
symbol,
tvm.runtime.convert(rhs, span=rhs_span),
indexes,
span=tvm_span_from_synr(node.span),
)
else:
if len(indexes) != 1:
self.report_error(
f"Store is only allowed with one index, but {len(indexes)} were provided.",
node.params[1].span,
)
# Store
return tvm.tir.Store(
symbol,
tvm.runtime.convert(rhs, span=rhs_span),
indexes[0],
tvm.runtime.convert(True, span=tvm_span_from_synr(node.span)),
span=tvm_span_from_synr(node.span),
)
def transform_Assert(self, node):
"""Assert visitor
Pattern corresponds to concise mode of :code:`with T.Assert()`.
"""
condition = self.transform(node.condition)
if node.msg is None:
self.report_error("Assert statements must have an error message.",
node.span)
message = self.transform(node.msg)
body = self.parse_body(node)
return tvm.tir.AssertStmt(condition,
tvm.runtime.convert(message),
body,
span=tvm_span_from_synr(node.span))
def transform_For(self, node):
"""For visitor
AST abstract grammar:
For(expr target, expr iter, stmt* body, stmt* orelse, string? type_comment)
By now 1 pattern of For is supported:
1. for scope handler
for name in T.serial()/T.parallel()/T.vectorized()/T.unroll()/range()/
T.grid()/T.thread_binding()
"""
if not isinstance(node.rhs, ast.Call):
self.report_error("The loop iterator should be a function call.",
node.rhs.span)
func = self.transform(node.rhs.func_name)
if not isinstance(func, ForScopeHandler):
self.report_error(
"Only For scope handlers can be used in a for statement.",
node.rhs.func_name.span)
# prepare for new for scope
old_lineno, old_col_offset = self.current_lineno, self.current_col_offset
self.current_lineno = node.span.start_line
self.current_col_offset = node.span.start_column
self.context.enter_scope(nodes=node.body.stmts)
# for scope handler process the scope
arg_list = self.parse_arg_list(func, node.rhs)
func.enter_scope(node, self.context, arg_list, node.rhs.func_name.span)
func.body = self.parse_body(node)
res = func.exit_scope(node, self.context, arg_list,
node.rhs.func_name.span)
# exit the scope
self.context.exit_scope()
self.current_lineno, self.current_col_offset = old_lineno, old_col_offset
return res
def transform_While(self, node):
"""While visitor
AST abstract grammar:
While(expr condition, stmt* body)
"""
condition = self.transform(node.condition)
# body
self.context.enter_scope(nodes=node.body.stmts)
body = self.parse_body(node)
self.context.exit_scope()
return tvm.tir.While(condition,
body,
span=tvm_span_from_synr(node.span))
def transform_With(self, node):
"""With visitor
AST abstract grammar:
With(withitem* items, stmt* body, string? type_comment)
withitem = (expr context_expr, expr? optional_vars)
By now 2 patterns of With is supported:
1. with scope handler with symbol def
with T.block(*axes)/T.allocate() as targets:
2. with scope handler without symbol def
with T.let()/T.Assert()/T.attr()/T.realize()
"""
if not isinstance(node.rhs, ast.Call):
self.report_error(
"The context expression of a `with` statement should be a function call.",
node.rhs.span,
)
func = self.transform(node.rhs.func_name)
if not isinstance(func, WithScopeHandler):
self.report_error(
f"Function {func} cannot be used in a `with` statement.",
node.rhs.func_name.span)
# prepare for new block scope
old_lineno, old_col_offset = self.current_lineno, self.current_col_offset
self.current_lineno = node.body.span.start_line
self.current_col_offset = node.body.span.start_column
self.context.enter_block_scope(nodes=node.body.stmts)
# with scope handler process the scope
arg_list = self.parse_arg_list(func, node.rhs)
func.enter_scope(node, self.context, arg_list, node.rhs.func_name.span)
func.body = self.parse_body(node)
res = func.exit_scope(node, self.context, arg_list,
node.rhs.func_name.span)
# exit the scope
self.context.exit_block_scope()
self.current_lineno, self.current_col_offset = old_lineno, old_col_offset
return res
def transform_If(self, node):
"""If visitor
AST abstract grammar:
If(expr test, stmt* body, stmt* orelse)
"""
condition = self.transform(node.condition)
# then body
self.context.enter_scope(nodes=node.true.stmts)
then_body = self.parse_body(node)
self.context.exit_scope()
# else body
if len(node.false.stmts) > 0:
self.context.enter_scope(nodes=node.false.stmts)
else_body = self.parse_body(node)
self.context.exit_scope()
else:
else_body = None
return tvm.tir.IfThenElse(condition,
then_body,
else_body,
span=tvm_span_from_synr(node.span))
def transform_Call(self, node):
"""Call visitor
3 different Call patterns are allowed:
1. Intrin representing a PrimExpr/IterVar
1.1 tir.int/uint/float8/16/32/64/floormod/floordiv/load/cast/ramp/broadcast/max
1.2 tir.range/reduce_axis/scan_axis/opaque_axis
2. tir.Op(dtype, ...)
3. other callable functions
"""
if isinstance(node.func_name, ast.Op):
if node.func_name.name == ast.BuiltinOp.Subscript:
return self.transform_Subscript(node)
if node.func_name.name in self._binop_maker:
lhs = self.transform(node.params[0])
rhs = self.transform(node.params[1])
return self._binop_maker[node.func_name.name](
lhs, rhs, span=tvm_span_from_synr(node.span))
if node.func_name.name in self._unaryop_maker:
rhs = self.transform(node.params[0])
return self._unaryop_maker[node.func_name.name](
rhs, span=tvm_span_from_synr(node.span))
self.report_error(f"Unsupported operator {node.func_name.name}.",
node.func_name.span)
else:
func = self.transform(node.func_name)
if isinstance(func, Intrin) and not func.stmt:
# pattern 1
arg_list = self.parse_arg_list(func, node)
return call_with_error_reporting(
self.report_error,
node.func_name.span,
func.handle,
arg_list,
node.func_name.span,
)
else:
args = [self.transform(arg) for arg in node.params]
kw_args = {
self.transform(k): self.transform(v)
for k, v in node.keyword_params.items()
}
if isinstance(func, tvm.tir.op.Op):
# pattern 2
return tvm.tir.Call(kw_args["dtype"],
func,
args,
span=tvm_span_from_synr(node.span))
elif callable(func):
# pattern 3
return func(*args, **kw_args)
else:
self.report_error(
f"Function is neither callable nor a tvm.tir.op.Op (it is a {type(func)}).",
node.func_name.span,
)
def transform_UnassignedCall(self, node):
"""Visitor for statements that are function calls.
This handles function calls that appear on thier own line like `tir.realize`.
Examples
--------
.. code-block:: python
@T.prim_func
def f():
A = T.buffer_decl([10, 10])
T.realize(A[1:2, 1:2], "") # This is an UnassignedCall
A[1, 1] = 2 # This is also an UnassignedCall
"""
# Only allowed builtin operator that can be a statement is x[1] = 3 i.e. subscript assign.
if isinstance(node.call.func_name, ast.Op):
if node.call.func_name.name != ast.BuiltinOp.SubscriptAssign:
self.report_error(
"Binary and unary operators are not allowed as a statement",
node.span)
else:
return self.transform_SubscriptAssign(node.call)
# handle a regular function call
func = self.transform(node.call.func_name)
arg_list = self.parse_arg_list(func, node.call)
if isinstance(func, tir.scope_handler.AssertHandler):
self.report_error(
"A standalone `T.Assert` is not allowed. Use `assert condition, message` "
"instead.",
node.call.func_name.span,
)
if isinstance(func, Intrin):
if func.stmt:
return call_with_error_reporting(
self.report_error,
node.call.func_name.span,
func.handle,
arg_list,
node.call.func_name.span,
)
else:
self.report_error(
f"This intrinsic cannot be used as a statement.",
node.call.span)
elif isinstance(func, WithScopeHandler
) and func.concise_scope and not func.def_symbol:
func.enter_scope(node, self.context, arg_list,
node.call.func_name.span)
func.body = self.parse_body(node)
return func.exit_scope(node, self.context, arg_list,
node.call.func_name.span)
elif isinstance(func, SpecialStmt) and not func.def_symbol:
func.handle(node, self.context, arg_list, node.call.func_name.span)
return
self.report_error(
"Unexpected statement. Expected an assert, an intrinsic, a with statement, or a "
f"special statement, but got {type(func).__name__}.",
node.call.func_name.span,
)
def transform_Slice(self, node):
start = self.transform(node.start)
end = self.transform(node.end)
if not (isinstance(node.step, ast.Constant) and node.step.value == 1):
self.report_error("Only step size 1 is supported for slices.",
node.step.span)
return Slice(start, end)
def transform_Subscript(self, node):
"""Array access visitor.
By now only 3 types of Subscript are supported:
1. Buffer[index, index, ...], Buffer element access(BufferLoad & BufferStore)
Var[index] Buffer element access()
2. Buffer[start: stop, start: stop, ...], BufferRealize(realize(buffer[...]))
3. Array[index], Buffer element access
"""
symbol = self.transform(node.params[0])
if symbol is None:
self.report_error(
f"Variable {node.params[0].id.name} is not defined.",
node.params[0].span)
indexes = [self.transform(x) for x in node.params[1].values]
if isinstance(symbol, tvm.tir.expr.Var):
for index in indexes:
if not isinstance(index, (tvm.tir.PrimExpr, int)):
self.report_error(
"Buffer load indexes should be int or PrimExpr, but they are "
+ type(index),
node.span,
)
return tvm.tir.Load("float32",
symbol,
indexes,
True,
span=tvm_span_from_synr(node.span))
elif isinstance(symbol, tvm.tir.Buffer):
return BufferSlice(symbol,
indexes,
self.report_error,
span=tvm_span_from_synr(node.span))
elif isinstance(symbol, tvm.container.Array):
if len(indexes) > 1:
self.report_error(
"Array access should be one-dimension access, but the indices are "
+ str(indexes),
node.span,
)
index = indexes[0]
if not isinstance(index, (int, tvm.tir.expr.IntImm)):
self.report_error(
"Array access index expected int or IntImm, but got " +
type(index),
node.span,
)
if int(index) >= len(symbol):
self.report_error(
f"Array access out of bound, size: {len(symbol)}, got index {index}.",
node.span,
)
return symbol[int(index)]
else:
self.report_error(
f"Cannot subscript from a {type(symbol).__name__}. Only variables and "
"buffers are supported.",
node.params[0].span,
)
def transform_Attr(self, node):
"""Visitor for field access of the form `x.y`.
This visitor is used to lookup function and symbol names. We have two
cases to handle here:
1. If we have a statement of the form `tir.something`, then we lookup
`tir.something` in the `Registry`. If the function is not in the
registry, then we try to find a `tvm.ir.op.Op` with the same name.
2. All other names `tvm.something` are lookup up in this current python
namespace.
"""
def get_full_attr_name(node: ast.Attr) -> str:
reverse_field_names = [node.field.name]
while isinstance(node.object, ast.Attr):
node = node.object
reverse_field_names.append(node.field.name)
if isinstance(node.object, ast.Var):
reverse_field_names.append(node.object.id.name)
return ".".join(reversed(reverse_field_names))
if isinstance(node.object, (ast.Var, ast.Attr)):
full_attr_name = get_full_attr_name(node)
attr_object, fields = full_attr_name.split(".", maxsplit=1)
if self.match_tir_namespace(attr_object):
func_name = "tir." + fields
res = Registry.lookup(func_name)
if res is not None:
return res
try:
return tvm.ir.op.Op.get(func_name)
except TVMError as e:
# Check if we got an attribute error
if e.args[0].find("AttributeError"):
self.report_error(
f"Unregistered function `tir.{fields}`.",
node.span)
else:
raise e
symbol = self.transform(node.object)
if symbol is None:
self.report_error("Unsupported Attribute expression.",
node.object.span)
if not hasattr(symbol, node.field.name):
self.report_error(
f"Type {type(symbol)} does not have a field called `{node.field.name}`.",
node.span)
res = getattr(symbol, node.field.name)
return res
def transform_TypeAttr(self, node):
"""Visitor for field access of the form `x.y` for types.
We have two cases here:
1. If the type is of the form `T.something`, we look up the type in
the `tir` namespace in this module.
2. If the type is of the form `tvm.x.something` then we look up
`tvm.x.something` in this modules namespace.
"""
if isinstance(node.object, ast.TypeVar):
if self.match_tir_namespace(node.object.id.name):
if not hasattr(tir, node.field.name):
self.report_error(
f"Invalid type annotation `tir.{node.field.name}`.",
node.span)
return getattr(tir, node.field.name)
symbol = self.transform(node.object)
if symbol is None:
self.report_error("Unsupported Attribute expression",
node.object.span)
if not hasattr(symbol, node.field):
self.report_error(
f"Type {type(symbol)} does not have a field called `{node.field}`.",
node.span)
res = getattr(symbol, node.field)
return res
def transform_DictLiteral(self, node):
"""Dictionary literal visitor.
Handles dictionary literals of the form `{x:y, z:2}`.
"""
keys = [self.transform(key) for key in node.keys]
values = [self.transform(value) for value in node.values]
return dict(zip(keys, values))
def transform_Tuple(self, node):
"""Tuple visitor.
Handles tuples of the form `(x, y, 2)`.
"""
return tuple(self.transform(element) for element in node.values)
def transform_ArrayLiteral(self, node):
"""List literal visitor.
Handles lists of the form `[x, 2, 3]`.
"""
return [self.transform(element) for element in node.values]
def transform_Var(self, node):
"""Variable visitor
Handles variables like `x` in `x = 2`.
"""
name = node.id.name
if name == "meta":
return self.meta
symbol = Registry.lookup(name)
if symbol is not None:
return symbol
symbol = self.context.lookup_symbol(name)
if symbol is not None:
return symbol
self.report_error(f"Unknown identifier {name}.", node.span)
def transform_TypeVar(self, node):
"""Type variable visitor.
Equivalent to `transform_Var` but for types.
"""
name = node.id.name
symbol = Registry.lookup(name) or self.context.lookup_symbol(name)
if symbol is not None:
return symbol
self.report_error(f"Unknown identifier {name}.", node.span)
def transform_Constant(self, node):
"""Constant value visitor.
Constant values include `None`, `"strings"`, `2` (integers), `4.2`
(floats), and `true` (booleans).
"""
return tvm.runtime.convert(node.value,
span=tvm_span_from_synr(node.span))
def transform_TypeConstant(self, node):
"""Constant value visitor for types.
See `transform_Constant`.
"""
return node.value
def transform_Return(self, node):
self.report_error(
"TVM script does not support return statements. Instead the last statement in any "
"block is implicitly returned.",
node.span,
)
def notBitStream(bs1):
return opBitStream(lambda x: operator.neg(x)-1, bs1)
def __neg__(self):
return vec2d(operator.neg(self.x), operator.neg(self.y))
from operator import abs, neg, pos, add, floordiv, mod, mul, pow, sub, and_, truediv, invert, lshift, or_, rshift, xor
a = -1
b = 5.0
c = 2
d = 6
print('a = ', a)
print('b = ', b)
print('c = ', c)
print('d = ', d)
print('\nPositive/Negative:')
print(f'abs({a}):', abs(a))
print(f'neg({a}):', neg(a))
print(f'neg({b}):', neg(b))
print(f'pos({a}):', pos(a))
print(f'pos({b}):', pos(b))
print('\nArithmetic:')
print(f'add({a}, {b}):', add(a, b))
print(f'floordiv({a}, {b}):', floordiv(a, b))
print(f'mod({a},{b}): ', mod(a, b))
print(f'mul({a},{b}): ', mul(a, b))
print(f'pow({c},{d}):', pow(c, d))
print(f'sub({b},{a}):', sub(b, a))
print(f'truediv({a},{b}):', truediv(a, b))
print(f'truediv({d},{c}):', truediv(d, c))
print('\nBitwise:')
print(f'and_({c}, {d}) :', and_(c, d))
def p_expression_uminus(p):
"""expression : MINUS expression %prec UMINUS"""
p[0] = operator.neg(p[2])
def __sub__(self, other):
assert type(other) == type(self)
# basic check: negating twice should be identity
assert op.neg(op.neg(other)) == other
return self + (-other)
