paramrf 0.25.2__tar.gz → 0.26.0__tar.gz

{paramrf-0.25.2/paramrf.egg-info → paramrf-0.26.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: paramrf
-Version: 0.25.2
+Version: 0.26.0
 Summary: Parametric radio frequency modeling
 Author-email: Gary Allen <gvcallen@gmail.com>
 Project-URL: homepage, https://github.com/gvcallen/paramrf

{paramrf-0.25.2 → paramrf-0.26.0}/docs/api/index.rst

@@ -76,4 +76,4 @@ Utilities
    pmrf.unwrap
    pmrf.unwrap_self
    pmrf.as_fixed
-   pmrf.as_variable
+   pmrf.as_free

{paramrf-0.25.2 → paramrf-0.26.0}/docs/examples/custom_parametric_models.rst

@@ -58,12 +58,12 @@ The code below demonstrates this by extending the previous class, while constrai
    from pmrf.constraints import Positive
    
    class Capacitor(Capacitor):
-       C: prf.Param = prf.param(constraint=Positive(), as_variable=True, scale=1e-12)
+       C: prf.Param = prf.param(constraint=Positive(), as_free=True, scale=1e-12)
    
        # def a(self, freq: prf.Frequency) -> jnp.ndarray:
            # <same as before>
 
-By passing ``as_variable=True``, ParamRF will enforce that the incoming value is a tunable parameter even if a float is passed. Similarly, ``as_fixed=True`` can be used to fix any incoming parameters. However, these converters are entirely optionaly, and by default the parameter's "tunability" is left unchanged, which is the most common use-case (simply registering the value in the parameter hierarchy).
+By passing ``as_free=True``, ParamRF will enforce that the incoming value is a tunable parameter even if a float is passed. Similarly, ``as_fixed=True`` can be used to fix any incoming parameters. However, these converters are entirely optionaly, and by default the parameter's "tunability" is left unchanged, which is the most common use-case (simply registering the value in the parameter hierarchy).
 
 Note that constraints will also always be enforced (even for unconstrained optimizers!), and will also automatically be intersected with any new constraints provided by the caller.
 

{paramrf-0.25.2 → paramrf-0.26.0}/docs/tutorials/1_cable_fitting.ipynb

@@ -108,12 +108,12 @@
    "source": [
     "ParamRF provides a few transmission line models, such as ``PhysicalLine``, ``CoaxialLine`` and ``DatasheetLine``. Looking at the datasheet, we see no mention of physical parameters like loss tangent, conductor loss or detailed geometry dimensions. However, there is a velocity factor, as well as loss factors ``k1`` and ``k2``. This is just what ``DatasheetLine`` was meant for.\n",
     "\n",
-    "Let's try fitting the model with its expected length and default parameters as a quick check. Since parameters are fixed by default, we need to set them as variable using ``pmrf.as_variable``. We can do this by either manually iterating over the model's named parameters or simply mapping them. In this tutorial, we demonstrate both."
+    "Let's try fitting the model with its expected length and default parameters as a quick check. Note that if we construct a default model or pass in float values, all parameters in the model are seen as fixed (non-tunable). For demonstration purposes, if we want to optimize the default model but don't have an initial guess, we can set all the parameters to be variable (tunable) using ``pmrf.as_variable``."
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 18,
+   "execution_count": null,
    "id": "40843bab",
    "metadata": {},
    "outputs": [
@@ -155,10 +155,10 @@
     "\n",
     "# Free all parameters\n",
     "for name in default_model.named_params().keys():\n",
-    "    default_model = default_model.at(name).apply(prf.as_variable)\n",
+    "    default_model = default_model.at(name).apply(prf.as_free)\n",
     "\n",
-    "# We could also have applied a map over the parameters\n",
-    "# default_model = default_model.map(lambda m: prf.as_variable(m), predicate=prf.is_param)\n",
+    "# We could also have used a map\n",
+    "# default_model = default_model.map(prf.as_free, predicate=prf.is_param)\n",
     "\n",
     "default_results = prf.fitting.fit(default_model, measured)\n",
     "\n",

{paramrf-0.25.2 → paramrf-0.26.0/paramrf.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: paramrf
-Version: 0.25.2
+Version: 0.26.0
 Summary: Parametric radio frequency modeling
 Author-email: Gary Allen <gvcallen@gmail.com>
 Project-URL: homepage, https://github.com/gvcallen/paramrf

{paramrf-0.25.2 → paramrf-0.26.0}/pmrf/__init__.py

@@ -41,7 +41,7 @@ Param: TypeAlias = parax.AbstractVariable | ArrayLike
 
 from pmrf.parameters import (
     param as param,
-    as_variable as as_variable,
+    as_free as as_free,
     as_fixed as as_fixed,
     Unconstrained as Unconstrained,
     Fixed as Fixed,
@@ -103,7 +103,7 @@ __all__ = [
 
     "Param",
     "param",
-    "as_variable",
+    "as_free",
     "as_fixed",
     "Unconstrained",
     "Fixed",

{paramrf-0.25.2 → paramrf-0.26.0}/pmrf/models/composite/nodal.py

@@ -345,13 +345,27 @@ class CoupledTwoPorts(Model):
         n = len(self.coupled)
         
         z_branch_list = []
+        y_shunt1_list = []
+        y_shunt2_list = []
+        
         for m in self.coupled:
             y_i = m.y(freq)
-            z_series = -1.0 / y_i[..., 0, 1]
+            
+            # Decompose the Pi-Network
+            # Series branch
+            y_series = -y_i[..., 0, 1]
+            z_series = 1.0 / y_series
             z_branch_list.append(z_series)
             
+            # Shunt branches to ground
+            y_shunt1_list.append(y_i[..., 0, 0] + y_i[..., 0, 1])
+            y_shunt2_list.append(y_i[..., 1, 1] + y_i[..., 1, 0])
+            
         z_branch = jnp.stack(z_branch_list, axis=-1)
+        y_p1 = jnp.stack(y_shunt1_list, axis=-1)
+        y_p2 = jnp.stack(y_shunt2_list, axis=-1)
         
+        # Couple the series branches exactly as before
         x_branch = jnp.imag(z_branch)
         x_outer = x_branch[..., :, jnp.newaxis] * x_branch[..., jnp.newaxis, :]
         
@@ -362,14 +376,20 @@ class CoupledTwoPorts(Model):
         i = jnp.arange(n)
         z_b_matrix = z_b_matrix.at[..., i, i].set(z_branch)
         
-        # Invert to get branch admittances
         y_b_matrix = jnp.linalg.inv(z_b_matrix)
         
-        # Nodal incidence matrix translation
         A = jnp.zeros((2 * n, n), dtype=jnp.float64)
         A = A.at[0::2, :].set(jnp.eye(n))
         A = A.at[1::2, :].set(-jnp.eye(n))
         
         y_nodal = jnp.einsum('pi,...ij,qj->...pq', A, y_b_matrix, A)
         
+        # Glue the substrate/shunt parasitics back onto the final nodes
+        # Model 'm' has Port 1 at index 2*m, and Port 2 at index 2*m + 1
+        even_indices = jnp.arange(0, 2 * n, 2)
+        odd_indices = jnp.arange(1, 2 * n, 2)
+        
+        y_nodal = y_nodal.at[..., even_indices, even_indices].add(y_p1)
+        y_nodal = y_nodal.at[..., odd_indices, odd_indices].add(y_p2)
+        
         return y_nodal

{paramrf-0.25.2 → paramrf-0.26.0}/pmrf/parameters.py

@@ -22,7 +22,7 @@ from pmrf.types import Param
 from pmrf.utils import unfreeze
 
 def apply_wrappers(value: Any, scale: float, fixed: bool, name: str | None) -> Param:
-    value = prx.as_variable(value)
+    value = prx.as_free(value)
     if scale != 1.0:
         scale_val = jnp.asarray(scale, dtype=float)
         try:
@@ -73,7 +73,7 @@ def extract_name(value: Any) -> str | None:
     return named_variable[0].metadata['name']
     
 
-def as_variable(
+def as_free(
     value: Any = None,
     *,
     distribution: Optional[AbstractDistribution] = None,
@@ -82,7 +82,7 @@ def as_variable(
     name: Optional[str] = None,
 ) -> Param:
     """
-    Coerces a value into a variable parameter.
+    Coerces a value into a free parameter.
 
     The incoming value can be an existing parameter or any
     parameter-like object (float, array etc.).
@@ -183,7 +183,7 @@ def as_fixed(
     pmrf.Param
         A fixed parameter.
     """
-    variable = as_variable(
+    variable = as_free(
         value,
         distribution=distribution,
         constraint=constraint,
@@ -197,7 +197,7 @@ def as_fixed(
 def param(
     *,
     default: Any = dataclasses.MISSING,
-    as_variable: bool = False,
+    as_free: bool = False,
     as_fixed: bool = False,
     distribution: Optional[AbstractDistribution] = None,
     constraint: Optional[AbstractConstraint] = None,
@@ -241,7 +241,7 @@ def param(
     ----------
     default : Any, optional
         The default value of the parameter.
-    as_variable : bool, optional
+    as_free : bool, optional
         Whether to enforce that the value is a variable parameter.
         If False, incoming values will keep the variability (e.g. constants will remain constants).
         If True, all values will be co-erced into variable parameters.
@@ -263,22 +263,22 @@ def param(
     Any
         An equinox field with a built-in converter for parameter rules.
     """
-    if as_fixed and as_variable:
-        raise ValueError("Cannot pass both `as_fixed=True` and `as_variable=True` to `pmrf.param`")
+    if as_fixed and as_free:
+        raise ValueError("Cannot pass both `as_fixed=True` and `as_free=True` to `pmrf.param`")
     
-    as_variable_func = globals()['as_variable']
+    as_free_func = globals()['as_free']
     as_fixed_func = globals()['as_fixed']
 
     def converter(x):
-        if as_variable:
-            return as_variable_func(value=x, distribution=distribution, constraint=constraint, scale=scale)
+        if as_free:
+            return as_free_func(value=x, distribution=distribution, constraint=constraint, scale=scale)
         if as_fixed:
             return as_fixed_func(value=x, distribution=distribution, constraint=constraint, scale=scale)
         
         if x is None:
             return None
         elif (prx.is_variable(x) or isinstance(x, jax.Array)) and not prx.is_constant(x):
-            return as_variable_func(value=x, distribution=distribution, constraint=constraint, scale=scale)
+            return as_free_func(value=x, distribution=distribution, constraint=constraint, scale=scale)
         return as_fixed_func(value=x, distribution=distribution, constraint=constraint, scale=scale)
 
     return eqx.field(default=default, converter=converter, **kwargs)
@@ -322,7 +322,7 @@ def Unconstrained(
     name: Optional[str] = None,
 ) -> Param:
     """
-    Create an unconstrained variable parameter.
+    Create an unconstrained free parameter.
 
     Parameters
     ----------
@@ -338,7 +338,7 @@ def Unconstrained(
     pmrf.Param
         An unconstrained parameter.
     """
-    return as_variable(value, scale=scale, name=name)
+    return as_free(value, scale=scale, name=name)
 
 
 def Constrained(
@@ -349,7 +349,7 @@ def Constrained(
     name: Optional[str] = None,
 ) -> Param:
     """
-    Create a variable parameter constrained to a specific domain.
+    Create a free parameter constrained to a specific domain.
 
     See :mod:`pmrf.constraints` for built-in constraints.
 
@@ -369,7 +369,7 @@ def Constrained(
     pmrf.Param
         The constrained parameter.
     """
-    return as_variable(value, constraint=constraint, scale=scale, name=name)
+    return as_free(value, constraint=constraint, scale=scale, name=name)
 
 
 def Bounded(
@@ -381,7 +381,7 @@ def Bounded(
     name: Optional[str] = None,
 ) -> Param:
     """
-    Create a variable parameter constrained within a specific interval.
+    Create a free parameter constrained within a specific interval.
 
     Used as the main factory to define parameters for bounded optimization.
 
@@ -405,7 +405,7 @@ def Bounded(
     pmrf.Param
         The bounded parameter.
     """
-    return as_variable(value, constraint=Interval(lower, upper), scale=scale, name=name)
+    return as_free(value, constraint=Interval(lower, upper), scale=scale, name=name)
 
 
 def Random(
@@ -417,7 +417,7 @@ def Random(
     name: Optional[str] = None,
 ) -> Param:
     """
-    Create a variable parameter with an associated probability distribution.
+    Create a free parameter with an associated probability distribution.
 
     Used as the main factory to define parameters for Bayesian inference.
     Can also be used for bounded optimization, in which case the random
@@ -451,11 +451,12 @@ def Random(
     ValueError
         If `value` is None and the distribution does not implement `mean()`.
     """
-    return as_variable(value, distribution=distribution, constraint=constraint, scale=scale, name=name)
+    return as_free(value, distribution=distribution, constraint=constraint, scale=scale, name=name)
 
 
 __all__ = [
-    "as_variable",
+    "as_free",
+    "as_fixed",
     "Unconstrained",
     "Fixed",
     "Bounded",

{paramrf-0.25.2 → paramrf-0.26.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "paramrf"
-version = "0.25.2"
+version = "0.26.0"
 description = "Parametric radio frequency modeling"
 readme = "README.rst"
 authors = [

{paramrf-0.25.2 → paramrf-0.26.0}/tests/test_models/test_adapters.py

@@ -74,7 +74,7 @@ def test_single_property_routing(fine_freq):
 
 class DummyHostModel(Host):
     """A dummy host model representing an external simulator."""
-    val: Param = param(default=10.0, as_variable=True)
+    val: Param = param(default=10.0, as_free=True)
     
     @property
     def number_of_ports(self): 

{paramrf-0.25.2 → paramrf-0.26.0}/tests/test_optimize_minimize.py

@@ -15,7 +15,7 @@ from pmrf.optimize.solvers.scipy import ScipyMinimize
 
 class DummyOptModel(Model):
     """A simple 1-port model with one free parameter for optimization."""
-    val: prf.Param = prf.param(default=1.0, as_variable=True)
+    val: prf.Param = prf.param(default=1.0, as_free=True)
 
     def s(self, freq: Frequency) -> jnp.ndarray:
         nf = freq.npoints