Changeset 870a2f4 in sasmodels
- Timestamp:
- Aug 8, 2017 4:18:53 PM (7 years ago)
- Branches:
- master, core_shell_microgels, costrafo411, magnetic_model, ticket-1257-vesicle-product, ticket_1156, ticket_1265_superball, ticket_822_more_unit_tests
- Children:
- 793d802
- Parents:
- c26897a
- Files:
-
- 2 deleted
- 6 edited
Legend:
- Unmodified
- Added
- Removed
-
doc/developer/calculator.rst
r2e66ef5 r870a2f4 1 1 .. currentmodule:: sasmodels 2 3 .. _Calculator_Interface: 2 4 3 5 Calculator Interface -
doc/developer/overview.rst
r2e66ef5 r870a2f4 7 7 Computational kernels 8 8 --------------------- 9 10 * :mod:`core` 11 * :mod:`modelinfo` 12 * :mod:`kernel` 13 * :mod:`product` 14 * :mod:`mixture` 9 15 10 16 At the heart of *sasmodels* are the individual computational kernels. These … … 12 18 return the expected scattering for that $q$. The instructions for writing 13 19 a kernel are documented in :ref:`Writing_a_Plugin`. The source code for 14 the kernels is stored in :mod:` sasmodels.models`.15 16 The primary interface to the models is through :mod:` sasmodels.core`, which20 the kernels is stored in :mod:`models`. 21 22 The primary interface to the models is through :mod:`core`, which 17 23 provides functions for listing available models, loading the model definition 18 and compiling the model. Use :func:`sasmodels.core.load_model` to load in 19 a model definition and compile it. 20 21 The computational kernels are difficult to call directly, but an example 22 was provided in :ref:`Scripting_Interface`. 23 24 The code for loading the kernel is in 25 :mod:`sasmodels.generate`, but a big part of it, which is interpreting the 26 definition, lives in :mod:`sasmodels.modelinfo`. This defines the properties 27 of the model in :class:`sasmodels.modelinfo.ModelInfo`, the available 28 model parameters in :class:`sasmodels.modelinfo.ParameterTable` and the 29 individual parameter attributes such as units and hard limits in 30 :class:`sasmodels.modelinfo.Parameter`. 24 and compiling the model. Use :func:`core.load_model` to load in 25 a model definition and compile it. This makes use of 26 :func:`core.load_model_info` to load the model definition and 27 :func:`core.build_model` to turn it into a computational kernel model 28 :class:`kernel.KernelModel`. 29 30 The :class:`modelinfo.ModelInfo` class defines the properties 31 of the model including the available model parameters 32 :class:`modelinfo.ParameterTable` with individual parameter attributes 33 such as units and hard limits defined in :class:`modelinfo.Parameter`. 34 35 The :class:`product.ProductModel` and :class:`mixture.MixtureModel` classes 36 are derived models, created automatically for models with names like 37 "hardsphere*sphere" and "cylinder+sphere". 38 39 Data loaders 40 ------------ 41 42 * :mod:`data` 43 44 In order to test models a minimal set of data management routines is 45 provided in :mod:`data`. In particular, it provides mock :class:`data.Data1D` 46 and :class:`data.Data2D` classes which mimic those classes in *SasView*. 47 The functions :func:`data.empty_data1D` and :func:`data.empty_data2D` 48 are handy for creating containers with a particular set of $q$, $\Delta q$ 49 points which can later be evaluated, and :func:`data.plot_theory` to show 50 the result. If *SasView* is available on the path then :func:`data.load_data` 51 can be used to load any data type defined in *SasView*. The function 52 :func:`data.plot_data` can plot that data alone without the theory value. 53 54 Kernel execution 55 ---------------- 56 57 * :mod:`resolution` 58 * :mod:`resolution2d` 59 * :mod:`sesans` 60 * :mod:`weights` 61 * :mod:`details` 62 * :mod:`direct_model` 63 * :mod:`bumps_model` 64 * :mod:`sasview_model` 65 66 To execute a computational kernel at a particular set of $q$ values, the 67 use :meth:`kernel.KernelModel.make_kernel`, which returns a callable 68 :class:`kernel.Kernel` for that $q$ vector (or a pair of $q_x$, $q_y$ 69 for 2-D datasets). 70 71 The calculated $q$ values should include the measured 72 data points as well as additional $q$ values required to properly compute the 73 $q$ resolution function. The *Resolution* subclasses in :mod:`resolution` 74 define the *q_calc* attribute for this purpose. These are 75 :class:`resolution.Perfect1D` for perfect resolution, 76 :class:`resolution.Pinhole1D` for the usual SANS pinhole aperture, 77 :class:`resolution.Slit1D` for the usual USANS slit aperture and 78 :class:`resolution2d.Pinhole2D` for 2-D pinhole data. 79 In addition, :class:`resolution2d.Slit2D` defines 1-D slit smeared data 80 for oriented samples, which require calculation at particular $q_x$ and 81 $q_y$ values instead of $|q|$ as is the case for orientationally averaged 82 USANS. The :class:`sesans.SesansTransform` class acts like a 1-D resolution, 83 having a *q_calc* attribute that defines the calculated $q$ values for 84 the SANS models that get converted to spin-echo values by the 85 :meth:`sesnas.SesansTransform.apply` method. 86 87 Polydispersity is defined by :class:`weights.Dispersion` classes, 88 :class:`weights.RectangleDispersion`, :class:`weights.ArrayDispersion`, 89 :class:`weights.LogNormalDispersion`, :class:`weights.GaussianDispersion`, 90 :class:`weights.SchulzDispersion`. The :func:`weights.get_weights` 91 function creates a dispersion object of the class matching 92 :attr:`weights.Dispersion.type`, and calls it with the current value 93 of the parameter. This returns a vector of values and weights for a 94 weighted average polydispersity. 95 96 In order to call the :class:`kernel.Kernel`, the values and weights for 97 all parameters must be composed into a :class:`details.CallDetails` object. 98 This is a compact vector representation of the entire polydispersity 99 loop that can be passed easily to the kernel. Additionally, the magnetic 100 parameters must be converted from polar to cartesian coordinates. This 101 work is done by the :func:`details.make_kernel_args` function, which returns 102 values that can be sent directly to the kernel. It uses 103 :func:`details.make_details` to set the details object and 104 :func:`details.convert_magnetism` for the coordinate transform. 105 106 In the end, making a simple theory function evaluation requires a lot of 107 setup. To make calling them a little easier, the *DirectModel* and 108 *BumpsModel* interfaces are provided. See :ref:`Scripting_Interface` 109 for an example. 110 111 The :class:`direct_model.DirectModel` interface accepts a data object 112 and a kernel model. Within the class, 113 the :meth:`direct_model.DataMixin._interpret_data` method is called to 114 query the data and set the resolution. 115 The :meth:`direct_model.DataMixin._calc_theory` takes a set of parameter 116 values, builds the kernel arguments, calls the kernel, and applies the 117 resolution function, returning the predicted value for the data $q$ values. 118 The :class:`bumps_model.Experiment` class is like the DirectModel class, 119 but it defines a Fitness class that can be handed directly to the 120 bumps optimization and uncertainty analysis program. 121 122 The :class:`sasview_model.SasviewModel` class defines a SasView 4.x 123 compatible interface to the sasmodels definitions, allowing sasmodels 124 to be used directly from SasView. Over time the SasView shim should 125 disappear as SasView access the :class:`modelinfo.ModelInfo` and 126 computational kernels directly. 127 128 Kernel execution 129 ---------------- 130 131 * :mod:`kernelcl` 132 * :mod:`kerneldll` 133 * :mod:`kernelpy` 134 * :mod:`generate` 135 31 136 32 137 The kernel functions for the most part do not define polydispersity, 33 resolution functions or magnetism. Instead, code for these is generated in 34 :mod:`sasmodels.generate` 35 36 37 38 Sasview kernels: 39 40 * :mod:` 41 ('kernel', 'Evaluator type definitions'), 42 ('kernelcl', 'OpenCL model evaluator'), 43 ('kerneldll', 'Ctypes model evaluator'), 44 ('kernelpy', 'Python model evaluator'), 45 46 Bumps fitting routines: 47 48 * :mod:`sasmodels.bumps_model` is a wrapper around 49 ('bumps_model', 'Bumps interface'), 50 ('direct_model', 'Simple interface'), 51 52 53 Utility functions: 54 55 * :mod:`sasmodels` provides :func:`sasmodels.__init__.data_files` for the installer. 56 * :mod:`sasmodels.rst2html` provides tools for converting model docs to html and viewing 57 the html. When run as a main program, it can display arbitrary rst files. 58 * :mod:`sasmodels.exception` annotates the current exception with a context string, 59 such as "while opening myfile.dat" without adjusting the traceback. 60 61 62 And so on:: 63 64 65 ('compare_many', 'Batch compare models on different compute engines'), 66 ('compare', 'Compare models on different compute engines'), 67 ('convert', 'Sasview to sasmodel converter'), 68 ('core', 'Model access'), 69 ('data', 'Data layout and plotting routines'), 70 ('details', 'Parameter packing for kernel calls'), 71 ('generate', 'Model parser'), 72 ('list_pars', 'Identify all parameters in all models'), 73 ('mixture', 'Mixture model evaluator'), 74 ('model_test', 'Unit test support'), 75 ('modelinfo', 'Parameter and model definitions'), 76 ('product', 'Product model evaluator'), 77 ('resolution', '1-D resolution functions'), 78 ('resolution2d', '2-D resolution functions'), 79 ('sasview_model', 'Sasview interface'), 80 ('sesans', 'SESANS calculation routines'), 81 ('weights', 'Distribution functions'), 138 resolution or magnetism directly. Instead sasmodels automatically 139 applies these, calling the computation kernel as needed. 140 141 The outermost loop is the resolution calculation. For the 1-D case 142 this computes a single vector of $I(q)$ values and applies the convolution 143 to the resulting set. Since the same $I(q)$ vector is used to compute the 144 convolution at each point, it can be precomputed before the convolution, 145 and so the convolution is reasonably efficient. The 2-D case is not 146 that efficient, and instead recomputes the entire shifted/scaled set 147 of $q_x$, $q_y$ values many times, or very many times depending on the 148 accuracy requested. 149 150 Polydispersity is handled as a mesh over the polydisperse parameters. 151 This is the next level of the loop. For C kernels run in a DLL or 152 using OpenCL, the polydisperisty loop is generated separately for each 153 model as C code. Inside the polydispersity loop there is a loop over 154 the magnetic cross sections for magnetic models, updating the SLD 155 parameters with the effective magnetic SLD for that particular $q$ 156 value. For OpenCL, each $q$ value loops over the 157 polydispersity mesh on a separate processor. For DLL, the outer loop 158 cycles through polydispersity, and the inner loop distributes q values 159 amongst the processors. Like the DLL, the Python kernel execution 160 cycles over the polydisperse parameters and the magnetic cross sections, 161 calling the computation kernel with a vector of $q$ values. Assuming 162 the kernel code accepts vectors, this can be fast enough (though it is 163 painfully slow if not vectorized). 164 165 Further details are provided in the next section, 166 :ref:`Calculator_Interface` 167 168 169 Testing 170 ------- 171 172 * :mod:`model_test` 173 * :mod:`compare` 174 * :mod:`compare_many` 175 * :mod:`rst2html` 176 * :mod:`list_pars` 177 178 Individual models should all have test values to make sure that the 179 evaluation is correct. This is particularly important in the context 180 of OpenCL since sasmodels doesn't control the compiler or the hardware, 181 and since GPUs are notorious for preferring speed over precision. The 182 tests can be run as a group using :mod:`model_test` as main:: 183 184 $ python -m sasmodels.model_test all 185 186 Individual models can be listed instead of *all*, which is convenient when 187 adding new models. 188 189 The :mod:`compare` module, usually invoked using *./sascomp* provides a 190 rich interface for exploring model accuracy, execution speed and parameter 191 ranges. It also allows different models to be compared. 192 The :mod:`compare_many` module does batch comparisons, keeping a list of 193 the particular random seeds which lead to large differences in output 194 between different computing platforms. 195 196 The :mod:`rst2html` module provides tools for converting model docs to 197 html and viewing the html. This is used by :mod:`compare` to display 198 the model description, such as:: 199 200 $ ./sascomp -html sphere 201 202 This makes debugging the latex much faster, though this may require 203 Qt in order for mathjax to work correctly. 204 205 When run as main, it can display arbitrary ReStructuredText files. E.g., 206 207 :: 208 209 $ python -m sasmodels.rst2html doc/developer/overview.rst 210 211 This is handy for sorting out rst and latex syntax. With some work 212 the results could be improved so that it recognizes sphinx roles 213 such as *mod*, *class* and *func*, and so that it uses the style sheets 214 from the sasmodels docs. 215 216 The :mod:`list_pars` module lists all instances of parameters across 217 all models. This helps to make sure that similar parameters have 218 similar names across the different models. With the verbose flag, 219 the particular models which use each named parameter are listed. 220 221 222 Model conversion 223 ---------------- 224 225 * :mod:`convert` 226 * :mod:`conversion_table` 227 228 Model definitions are not static. As needs change or problems are found, 229 models may be updated with new model names or may be reparameterized 230 with new parameter definitions. For example, in translating the 231 Teubner-Strey model from SasView 3.x to sasmodels, the definition 232 in terms of *drho*, *k*, *c1*, *c2*, *a2* and prefactor was replaced 233 by the defintion in terms of *volfraction_a*, *xi*, *d*, *sld_a* and 234 *sld_b*. Within :mod:`convert`, the *_hand_convert_3_1_2_to_4_1* 235 function must be called when loading a 3.x model definition to update it to 236 4.1, and then the model should be further updated to 4.2, 5.0, and so on. 237 The :func:`convert.convert_model` function does this, using the conversion 238 table in :mod:`conversion_table` (which handled the major renaming from 239 SasView 3.x to sasmodels), and using the internal *_hand_convert* function 240 for the more complicated cases. 241 242 Other 243 ----- 244 245 * :mod:`exception` 246 * :mod:`alignment` 247 248 The :func:`exception.annotate_exception` function annotates the current 249 exception with a context string, such as "while opening myfile.dat" without 250 adjusting the traceback. 251 252 The :mod:`alignment` module is unused. -
doc/guide/plugin.rst
r990d8df r870a2f4 610 610 611 611 :code:`source = ["lib/polevl.c", ...]` 612 (` link to code<https://github.com/SasView/sasmodels/tree/master/sasmodels/models/lib/polevl.c>`_)612 (`polevl.c <https://github.com/SasView/sasmodels/tree/master/sasmodels/models/lib/polevl.c>`_) 613 613 614 614 sas_gamma(x): … … 618 618 on some platforms. 619 619 620 :code:`source = ["lib/sas gamma.c", ...]`621 (` link to code<https://github.com/SasView/sasmodels/tree/master/sasmodels/models/lib/sas_gamma.c>`_)620 :code:`source = ["lib/sas_gamma.c", ...]` 621 (`sas_gamma.c <https://github.com/SasView/sasmodels/tree/master/sasmodels/models/lib/sas_gamma.c>`_) 622 622 623 623 sas_erf(x), sas_erfc(x): … … 631 631 632 632 :code:`source = ["lib/polevl.c", "lib/sas_erf.c", ...]` 633 (` link to error functions' code<https://github.com/SasView/sasmodels/tree/master/sasmodels/models/lib/sas_erf.c>`_)633 (`sas_erf.c <https://github.com/SasView/sasmodels/tree/master/sasmodels/models/lib/sas_erf.c>`_) 634 634 635 635 sas_J0(x): … … 640 640 641 641 :code:`source = ["lib/polevl.c", "lib/sas_J0.c", ...]` 642 (` link to Bessel function's code<https://github.com/SasView/sasmodels/tree/master/sasmodels/models/lib/sas_J0.c>`_)642 (`sas_J0.c <https://github.com/SasView/sasmodels/tree/master/sasmodels/models/lib/sas_J0.c>`_) 643 643 644 644 sas_J1(x): … … 649 649 650 650 :code:`source = ["lib/polevl.c", "lib/sas_J1.c", ...]` 651 (` link to Bessel function's code<https://github.com/SasView/sasmodels/tree/master/sasmodels/models/lib/sas_J1.c>`_)651 (`sas_J1.c <https://github.com/SasView/sasmodels/tree/master/sasmodels/models/lib/sas_J1.c>`_) 652 652 653 653 sas_JN(n, x): … … 660 660 661 661 :code:`source = ["lib/polevl.c", "lib/sas_J0.c", "lib/sas_J1.c", "lib/sas_JN.c", ...]` 662 (` link to Bessel function's code<https://github.com/SasView/sasmodels/tree/master/sasmodels/models/lib/sas_JN.c>`_)662 (`sas_JN.c <https://github.com/SasView/sasmodels/tree/master/sasmodels/models/lib/sas_JN.c>`_) 663 663 664 664 sas_Si(x): … … 684 684 685 685 :code:`source = ["lib/Si.c", ...]` 686 (` link to code<https://github.com/SasView/sasmodels/tree/master/sasmodels/models/lib/Si.c>`_)686 (`Si.c <https://github.com/SasView/sasmodels/tree/master/sasmodels/models/lib/Si.c>`_) 687 687 688 688 sas_3j1x_x(x): … … 695 695 696 696 :code:`source = ["lib/sas_3j1x_x.c", ...]` 697 (` link to code<https://github.com/SasView/sasmodels/tree/master/sasmodels/models/lib/sas_3j1x_x.c>`_)697 (`sas_3j1x_x.c <https://github.com/SasView/sasmodels/tree/master/sasmodels/models/lib/sas_3j1x_x.c>`_) 698 698 699 699 … … 704 704 705 705 :code:`source = ["lib/polevl.c", "lib/sas_J1.c", ...]` 706 (` link to Bessel form's code<https://github.com/SasView/sasmodels/tree/master/sasmodels/models/lib/sas_J1.c>`_)706 (`sas_J1.c <https://github.com/SasView/sasmodels/tree/master/sasmodels/models/lib/sas_J1.c>`_) 707 707 708 708 … … 715 715 716 716 :code:`source = ["lib/gauss76.c", ...]` 717 (` link to code<https://github.com/SasView/sasmodels/tree/master/sasmodels/models/lib/gauss76.c>`_)717 (`gauss76.c <https://github.com/SasView/sasmodels/tree/master/sasmodels/models/lib/gauss76.c>`_) 718 718 719 719 -
sasmodels/alignment.py
r7ae2b7f r870a2f4 42 42 view[:] = x 43 43 return view 44 -
sasmodels/models/multilayer_vesicle.py
r5d23de2 r870a2f4 71 71 sufficiently fine grained in certain cases. Please report any such occurences 72 72 to the SasView team. Generally, for the best possible experience: 73 * Start with the best possible guess 74 * Using a priori knowledge, hold as many parameters fixed as possible 75 * if N=1, tw (water thickness) must by definition be zero. Both N and tw should 73 74 - Start with the best possible guess 75 - Using a priori knowledge, hold as many parameters fixed as possible 76 - if N=1, tw (water thickness) must by definition be zero. Both N and tw should 76 77 be fixed during fitting. 77 *If N>1, use constraints to keep N > 178 *Because N only really moves in integer steps, it may get "stuck" if the78 - If N>1, use constraints to keep N > 1 79 - Because N only really moves in integer steps, it may get "stuck" if the 79 80 optimizer step size is too small so care should be taken 80 81 If you experience problems with this please contact the SasView team and let -
sasmodels/rst2html.py
r990d8df r870a2f4 176 176 from PyQt5.QtCore import QUrl 177 177 except ImportError: 178 from PyQt4.QtWeb kit import QWebView178 from PyQt4.QtWebKit import QWebView 179 179 from PyQt4.QtCore import QUrl 180 180 helpView = QWebView() … … 204 204 from PyQt5.QtCore import QUrl 205 205 except ImportError: 206 from PyQt4.QtWeb kit import QWebView206 from PyQt4.QtWebKit import QWebView 207 207 from PyQt4.QtCore import QUrl 208 208 frame = QWebView() … … 222 222 except ImportError: 223 223 try: 224 from PyQt4.QtWeb kit import QWebView224 from PyQt4.QtWebKit import QWebView 225 225 return True 226 226 except ImportError:
Note: See TracChangeset
for help on using the changeset viewer.