c3/c3objs.py
"""Basic custom objects."""
import hjson
from typing import List
import numpy as np
import tensorflow as tf
from c3.utils.utils import num3str
from tensorflow.python.framework import ops
import copy
class QuantityOOBException(ValueError):
pass
class C3obj:
"""
Represents an abstract object with parameters. To be inherited from.
Parameters
----------
name: str
short name that will be used as identifier
desc: str
longer description of the component
comment: str
additional information about the component
params: dict
Parameters in this dict can be accessed and optimized
"""
def __init__(self, name, desc="", comment="", params=None):
self.name = name
self.desc = desc
self.comment = comment
self.params = {}
if params:
for pname, par in params.items():
# TODO params here should be the dict representation only
if isinstance(par, dict):
self.params[pname] = Quantity(**par)
else:
self.params[pname] = par
def __str__(self) -> str:
return hjson.dumps(self.asdict(), default=hjson_encode)
def asdict(self) -> dict:
params = {}
for key, item in self.params.items():
params[key] = item.asdict()
return {"c3type": self.__class__.__name__, "params": params, "name": self.name}
class Quantity:
"""
Represents any physical quantity used in the model or the pulse
specification. For arithmetic operations just the numeric value is used. The
value itself is stored in an optimizer friendly way as a float between -1
and 1. The conversion is given by
scale (value + 1) / 2 + offset
Parameters
----------
value: np.array(np.float64) or np.float64
value of the quantity
min_val: np.array(np.float64) or np.float64
minimum this quantity is allowed to take
max_val: np.array(np.float64) or np.float64
maximum this quantity is allowed to take
unit: str
physical unit
symbol: str
latex representation
"""
def __init__(
self, value, unit="undefined", min_val=None, max_val=None, symbol=r"\alpha"
):
pref = 1.0
value = np.array(value)
if hasattr(value, "shape"):
self.shape = value.shape
self.length = int(np.prod(value.shape))
else:
self.shape = (1,)
self.length = 1
if "pi" in unit:
pref = np.pi
if "2pi" in unit:
pref = 2 * np.pi
self.pref = np.array(pref)
if min_val is None and max_val is None:
if value.any():
minmax = [0.9 * value, 1.1 * value]
min_val = np.min(minmax)
max_val = np.max(minmax)
else:
min_val = np.array(-1)
max_val = np.array(1)
self._set_limits(min_val, max_val)
self.unit = unit
self.symbol = symbol
self.value = None
self.set_value(np.array(value))
def asdict(self) -> dict:
"""
Return a config-compatible dictionary representation.
"""
pref = self.pref
return {
"value": self.numpy().tolist(),
"min_val": (self.offset / pref).tolist(),
"max_val": (self.scale / pref + self.offset / pref).tolist(),
"unit": self.unit,
"symbol": self.symbol,
}
def tolist(self) -> List:
if self.length > 1:
tolist = self.get_value().numpy().tolist()
else:
tolist = [self.get_value().numpy().tolist()]
return tolist
def __add__(self, other):
out_val = copy.deepcopy(self)
out_val._set_value_extend(self.get_value() + other)
return out_val
def __radd__(self, other):
out_val = copy.deepcopy(self)
out_val._set_value_extend(self.get_value() + other)
return out_val
def __sub__(self, other):
out_val = copy.deepcopy(self)
out_val._set_value_extend(self.get_value() - other)
return out_val
def __rsub__(self, other):
out_val = copy.deepcopy(self)
out_val._set_value_extend(other - self.get_value())
return out_val
def __mul__(self, other):
out_val = copy.deepcopy(self)
out_val._set_value_extend(self.get_value() * other)
return out_val
def __rmul__(self, other):
out_val = copy.deepcopy(self)
out_val._set_value_extend(self.get_value() * other)
return out_val
def __pow__(self, other):
out_val = copy.deepcopy(self)
out_val._set_value_extend(self.get_value() ** other)
return out_val
def __rpow__(self, other):
out_val = copy.deepcopy(self)
out_val._set_value_extend(other ** self.get_value())
return out_val
def __truediv__(self, other):
out_val = copy.deepcopy(self)
out_val._set_value_extend(self.get_value() / other)
return out_val
def __rtruediv__(self, other):
out_val = copy.deepcopy(self)
out_val._set_value_extend(other / self.get_value())
return out_val
def __mod__(self, other):
out_val = copy.deepcopy(self)
out_val._set_value_extend(self.get_value() % other)
return out_val
def __lt__(self, other):
return self.get_value() < other
def __le__(self, other):
return self.get_value() <= other
def __eq__(self, other):
return self.get_value() == other
def __ne__(self, other):
return self.get_value() != other
def __ge__(self, other):
return self.get_value() >= other
def __gt__(self, other):
return self.get_value() > other
def __array__(self):
return np.array(self.numpy())
def __len__(self):
return self.length
def __getitem__(self, key):
if self.length == 1 and key == 0:
return self.numpy()
return self.numpy().__getitem__(key)
def __float__(self):
if self.length > 1:
raise NotImplementedError
return float(self.numpy())
def __repr__(self):
return self.__str__()[:-1]
def __str__(self):
val = self.numpy()
ret = ""
for entry in np.nditer(val):
if self.unit == "pi":
ret += f"{entry} {self.unit} "
elif self.unit != "undefined":
ret += num3str(entry) + self.unit + " "
else:
ret += num3str(entry, use_prefix=False) + " "
return ret
def subtract(self, val):
self.set_value(self.get_value() - val)
def add(self, val):
self.set_value(self.get_value() + val)
def numpy(self) -> np.ndarray:
"""
Return the value of this quantity as numpy.
"""
# TODO should be removed to be consistent with get_value
return self.get_value().numpy() / self.pref
def get_value(self) -> tf.Tensor:
"""
Return the value of this quantity as tensorflow.
Parameters
----------
val : tf.float64
dtype: tf.dtypes
"""
return self.scale * (self.value + 1) / 2 + self.offset
def get_other_value(self, val) -> tf.Tensor:
"""
Return an arbitrary value of the same scale as this quantity as tensorflow.
Parameters
----------
val : tf.float64
dtype: tf.dtypes
"""
return (self.scale * (val + 1) / 2 + self.offset) / self.pref
def set_value(self, val, extend_bounds=False):
if extend_bounds:
self._set_value_extend(np.reshape(val, self.shape))
else:
self._set_value(np.reshape(val, self.shape))
def _set_value(self, val) -> None:
"""Set the value of this quantity as tensorflow. Value needs to be
within specified min and max."""
# setting can be numpyish
val = tf.constant(val, tf.float64)
tmp = (
2 * (tf.reshape(val, self.shape) * self.pref - self.offset) / self.scale - 1
)
const_1 = tf.constant(1.0, tf.float64)
if np.any(
tf.math.logical_and(
tf.math.abs(tmp) > const_1,
tf.math.logical_not(
tf.experimental.numpy.isclose(tf.math.abs(tmp), const_1)
),
)
):
raise QuantityOOBException(
f"Value {num3str(val.numpy())}{self.unit} out of bounds for quantity with "
f"min_val: {num3str(self.get_limits()[0])}{self.unit} and "
f"max_val: {num3str(self.get_limits()[1])}{self.unit}",
)
self.value = tf.cast(tmp, tf.float64)
def _set_value_extend(self, val) -> None:
"""Set the value of this quantity as tensorflow. If needed, limits will be extended."""
min_val, max_val = self.get_limits()
# Extra bounds included to not be directly at border due to differentiability
# val can be matrix valued
minmax = [
tf.math.reduce_min(val * 0.9),
tf.math.reduce_max(val * 0.9),
tf.math.reduce_min(val * 1.1),
tf.math.reduce_max(val * 1.1),
min_val,
max_val,
]
min_val = tf.math.reduce_min(minmax)
max_val = tf.math.reduce_max(minmax)
self._set_limits(min_val, max_val)
self._set_value(val)
def get_opt_value(self) -> tf.Tensor:
"""Get an optimizer friendly representation of the value."""
return tf.reshape(self.value, (-1,))
def set_opt_value(self, val: float) -> None:
"""Set value optimizer friendly.
Parameters
----------
val : tf.float64
Tensorflow number that will be mapped to a value between -1 and 1.
"""
bound_val = tf.cos((tf.reshape(val, self.shape) + 1) * np.pi / 2)
self.value = tf.acos(bound_val) / np.pi * 2 - 1
def get_limits(self):
min_val = self.offset / self.pref
max_val = (self.scale + self.offset) / self.pref
return min_val, max_val
def _set_limits(self, min_val, max_val):
"""Sets the allowed minimum and maximum of this quantity. WARNING: Calling this
manually leads to inconstistencies with the previously stored value.
Parameters
----------
min_val : float
max_val : float
"""
self.offset = np.array(min_val) * self.pref
self.scale = np.abs(np.array(max_val) - np.array(min_val)) * self.pref
def jsonify_list(data, transform_arrays=True):
# try:
if data is None:
return
if isinstance(data, dict):
return {str(k): jsonify_list(v) for k, v in data.items()}
elif isinstance(data, list):
return [jsonify_list(v) for v in data]
elif isinstance(data, tuple):
return tuple(jsonify_list(v) for v in data)
elif isinstance(data, np.ndarray) and transform_arrays:
return data.tolist()
elif isinstance(data, ops.EagerTensor) and transform_arrays:
return data.numpy().tolist()
elif isinstance(data, C3obj) or isinstance(data, Quantity):
return data.asdict()
elif (
isinstance(data, str)
or isinstance(data, bool)
or isinstance(data, float)
or isinstance(data, int)
):
return data
else:
return data
def hjson_encode(z):
if isinstance(z, complex):
return {"__complex__": str(z)}
elif isinstance(z, np.ndarray):
return {"__array__": (z.tolist())}
elif isinstance(z, tf.Tensor) or isinstance(z, ops.EagerTensor):
return {"__array__": (z.numpy().tolist())}
elif isinstance(z, Quantity):
return {"__quantity__": z.asdict()}
elif isinstance(z, C3obj):
return z.asdict()
elif isinstance(z, dict) and np.any([not isinstance(k, str) for k in z.keys()]):
return {str(k): v for k, v in z.items()}
return z
def hjson_decode(z):
if len(z) == 1:
if z[0][0] == "__complex__":
return complex(z[0][1])
elif z[0][0] == "__array__":
return np.array(z[0][1])
elif z[0][0] == "__quantity__":
return Quantity(**z[0][1])
return dict(z)