1 | #!/usr/bin/env python |
---|
2 | # -*- coding: utf-8 -*- |
---|
3 | """ |
---|
4 | Program to compare models using different compute engines. |
---|
5 | |
---|
6 | This program lets you compare results between OpenCL and DLL versions |
---|
7 | of the code and between precision (half, fast, single, double, quad), |
---|
8 | where fast precision is single precision using native functions for |
---|
9 | trig, etc., and may not be completely IEEE 754 compliant. This lets |
---|
10 | make sure that the model calculations are stable, or if you need to |
---|
11 | tag the model as double precision only. |
---|
12 | |
---|
13 | Run using ./compare.sh (Linux, Mac) or compare.bat (Windows) in the |
---|
14 | sasmodels root to see the command line options. |
---|
15 | |
---|
16 | Note that there is no way within sasmodels to select between an |
---|
17 | OpenCL CPU device and a GPU device, but you can do so by setting the |
---|
18 | PYOPENCL_CTX environment variable ahead of time. Start a python |
---|
19 | interpreter and enter:: |
---|
20 | |
---|
21 | import pyopencl as cl |
---|
22 | cl.create_some_context() |
---|
23 | |
---|
24 | This will prompt you to select from the available OpenCL devices |
---|
25 | and tell you which string to use for the PYOPENCL_CTX variable. |
---|
26 | On Windows you will need to remove the quotes. |
---|
27 | """ |
---|
28 | |
---|
29 | from __future__ import print_function |
---|
30 | |
---|
31 | import sys |
---|
32 | import os |
---|
33 | import math |
---|
34 | import datetime |
---|
35 | import traceback |
---|
36 | import re |
---|
37 | |
---|
38 | import numpy as np # type: ignore |
---|
39 | |
---|
40 | from . import core |
---|
41 | from . import kerneldll |
---|
42 | from . import exception |
---|
43 | from .data import plot_theory, empty_data1D, empty_data2D, load_data |
---|
44 | from .direct_model import DirectModel, get_mesh |
---|
45 | from .convert import revert_name, revert_pars, constrain_new_to_old |
---|
46 | from .generate import FLOAT_RE |
---|
47 | from .weights import plot_weights |
---|
48 | |
---|
49 | try: |
---|
50 | from typing import Optional, Dict, Any, Callable, Tuple |
---|
51 | except Exception: |
---|
52 | pass |
---|
53 | else: |
---|
54 | from .modelinfo import ModelInfo, Parameter, ParameterSet |
---|
55 | from .data import Data |
---|
56 | Calculator = Callable[[float], np.ndarray] |
---|
57 | |
---|
58 | USAGE = """ |
---|
59 | usage: sascomp model [options...] [key=val] |
---|
60 | |
---|
61 | Generate and compare SAS models. If a single model is specified it shows |
---|
62 | a plot of that model. Different models can be compared, or the same model |
---|
63 | with different parameters. The same model with the same parameters can |
---|
64 | be compared with different calculation engines to see the effects of precision |
---|
65 | on the resultant values. |
---|
66 | |
---|
67 | model or model1,model2 are the names of the models to compare (see below). |
---|
68 | |
---|
69 | Options (* for default): |
---|
70 | |
---|
71 | === data generation === |
---|
72 | -data="path" uses q, dq from the data file |
---|
73 | -noise=0 sets the measurement error dI/I |
---|
74 | -res=0 sets the resolution width dQ/Q if calculating with resolution |
---|
75 | -lowq*/-midq/-highq/-exq use q values up to 0.05, 0.2, 1.0, 10.0 |
---|
76 | -q=min:max alternative specification of qrange |
---|
77 | -nq=128 sets the number of Q points in the data set |
---|
78 | -1d*/-2d computes 1d or 2d data |
---|
79 | -zero indicates that q=0 should be included |
---|
80 | |
---|
81 | === model parameters === |
---|
82 | -preset*/-random[=seed] preset or random parameters |
---|
83 | -sets=n generates n random datasets with the seed given by -random=seed |
---|
84 | -pars/-nopars* prints the parameter set or not |
---|
85 | -default/-demo* use demo vs default parameters |
---|
86 | -sphere[=150] set up spherical integration over theta/phi using n points |
---|
87 | |
---|
88 | === calculation options === |
---|
89 | -mono*/-poly force monodisperse or allow polydisperse random parameters |
---|
90 | -cutoff=1e-5* cutoff value for including a point in polydispersity |
---|
91 | -magnetic/-nonmagnetic* suppress magnetism |
---|
92 | -accuracy=Low accuracy of the resolution calculation Low, Mid, High, Xhigh |
---|
93 | -neval=1 sets the number of evals for more accurate timing |
---|
94 | |
---|
95 | === precision options === |
---|
96 | -engine=default uses the default calcution precision |
---|
97 | -single/-double/-half/-fast sets an OpenCL calculation engine |
---|
98 | -single!/-double!/-quad! sets an OpenMP calculation engine |
---|
99 | -sasview sets the sasview calculation engine |
---|
100 | |
---|
101 | === plotting === |
---|
102 | -plot*/-noplot plots or suppress the plot of the model |
---|
103 | -linear/-log*/-q4 intensity scaling on plots |
---|
104 | -hist/-nohist* plot histogram of relative error |
---|
105 | -abs/-rel* plot relative or absolute error |
---|
106 | -title="note" adds note to the plot title, after the model name |
---|
107 | -weights shows weights plots for the polydisperse parameters |
---|
108 | |
---|
109 | === output options === |
---|
110 | -edit starts the parameter explorer |
---|
111 | -help/-html shows the model docs instead of running the model |
---|
112 | |
---|
113 | The interpretation of quad precision depends on architecture, and may |
---|
114 | vary from 64-bit to 128-bit, with 80-bit floats being common (1e-19 precision). |
---|
115 | On unix and mac you may need single quotes around the DLL computation |
---|
116 | engines, such as -engine='single!,double!' since !, is treated as a history |
---|
117 | expansion request in the shell. |
---|
118 | |
---|
119 | Key=value pairs allow you to set specific values for the model parameters. |
---|
120 | Key=value1,value2 to compare different values of the same parameter. The |
---|
121 | value can be an expression including other parameters. |
---|
122 | |
---|
123 | Items later on the command line override those that appear earlier. |
---|
124 | |
---|
125 | Examples: |
---|
126 | |
---|
127 | # compare single and double precision calculation for a barbell |
---|
128 | sascomp barbell -engine=single,double |
---|
129 | |
---|
130 | # generate 10 random lorentz models, with seed=27 |
---|
131 | sascomp lorentz -sets=10 -seed=27 |
---|
132 | |
---|
133 | # compare ellipsoid with R = R_polar = R_equatorial to sphere of radius R |
---|
134 | sascomp sphere,ellipsoid radius_polar=radius radius_equatorial=radius |
---|
135 | |
---|
136 | # model timing test requires multiple evals to perform the estimate |
---|
137 | sascomp pringle -engine=single,double -timing=100,100 -noplot |
---|
138 | """ |
---|
139 | |
---|
140 | # Update docs with command line usage string. This is separate from the usual |
---|
141 | # doc string so that we can display it at run time if there is an error. |
---|
142 | # lin |
---|
143 | __doc__ = (__doc__ # pylint: disable=redefined-builtin |
---|
144 | + """ |
---|
145 | Program description |
---|
146 | ------------------- |
---|
147 | |
---|
148 | """ + USAGE) |
---|
149 | |
---|
150 | kerneldll.ALLOW_SINGLE_PRECISION_DLLS = True |
---|
151 | |
---|
152 | # list of math functions for use in evaluating parameters |
---|
153 | MATH = dict((k,getattr(math, k)) for k in dir(math) if not k.startswith('_')) |
---|
154 | |
---|
155 | # CRUFT python 2.6 |
---|
156 | if not hasattr(datetime.timedelta, 'total_seconds'): |
---|
157 | def delay(dt): |
---|
158 | """Return number date-time delta as number seconds""" |
---|
159 | return dt.days * 86400 + dt.seconds + 1e-6 * dt.microseconds |
---|
160 | else: |
---|
161 | def delay(dt): |
---|
162 | """Return number date-time delta as number seconds""" |
---|
163 | return dt.total_seconds() |
---|
164 | |
---|
165 | |
---|
166 | class push_seed(object): |
---|
167 | """ |
---|
168 | Set the seed value for the random number generator. |
---|
169 | |
---|
170 | When used in a with statement, the random number generator state is |
---|
171 | restored after the with statement is complete. |
---|
172 | |
---|
173 | :Parameters: |
---|
174 | |
---|
175 | *seed* : int or array_like, optional |
---|
176 | Seed for RandomState |
---|
177 | |
---|
178 | :Example: |
---|
179 | |
---|
180 | Seed can be used directly to set the seed:: |
---|
181 | |
---|
182 | >>> from numpy.random import randint |
---|
183 | >>> push_seed(24) |
---|
184 | <...push_seed object at...> |
---|
185 | >>> print(randint(0,1000000,3)) |
---|
186 | [242082 899 211136] |
---|
187 | |
---|
188 | Seed can also be used in a with statement, which sets the random |
---|
189 | number generator state for the enclosed computations and restores |
---|
190 | it to the previous state on completion:: |
---|
191 | |
---|
192 | >>> with push_seed(24): |
---|
193 | ... print(randint(0,1000000,3)) |
---|
194 | [242082 899 211136] |
---|
195 | |
---|
196 | Using nested contexts, we can demonstrate that state is indeed |
---|
197 | restored after the block completes:: |
---|
198 | |
---|
199 | >>> with push_seed(24): |
---|
200 | ... print(randint(0,1000000)) |
---|
201 | ... with push_seed(24): |
---|
202 | ... print(randint(0,1000000,3)) |
---|
203 | ... print(randint(0,1000000)) |
---|
204 | 242082 |
---|
205 | [242082 899 211136] |
---|
206 | 899 |
---|
207 | |
---|
208 | The restore step is protected against exceptions in the block:: |
---|
209 | |
---|
210 | >>> with push_seed(24): |
---|
211 | ... print(randint(0,1000000)) |
---|
212 | ... try: |
---|
213 | ... with push_seed(24): |
---|
214 | ... print(randint(0,1000000,3)) |
---|
215 | ... raise Exception() |
---|
216 | ... except Exception: |
---|
217 | ... print("Exception raised") |
---|
218 | ... print(randint(0,1000000)) |
---|
219 | 242082 |
---|
220 | [242082 899 211136] |
---|
221 | Exception raised |
---|
222 | 899 |
---|
223 | """ |
---|
224 | def __init__(self, seed=None): |
---|
225 | # type: (Optional[int]) -> None |
---|
226 | self._state = np.random.get_state() |
---|
227 | np.random.seed(seed) |
---|
228 | |
---|
229 | def __enter__(self): |
---|
230 | # type: () -> None |
---|
231 | pass |
---|
232 | |
---|
233 | def __exit__(self, exc_type, exc_value, traceback): |
---|
234 | # type: (Any, BaseException, Any) -> None |
---|
235 | # TODO: better typing for __exit__ method |
---|
236 | np.random.set_state(self._state) |
---|
237 | |
---|
238 | def tic(): |
---|
239 | # type: () -> Callable[[], float] |
---|
240 | """ |
---|
241 | Timer function. |
---|
242 | |
---|
243 | Use "toc=tic()" to start the clock and "toc()" to measure |
---|
244 | a time interval. |
---|
245 | """ |
---|
246 | then = datetime.datetime.now() |
---|
247 | return lambda: delay(datetime.datetime.now() - then) |
---|
248 | |
---|
249 | |
---|
250 | def set_beam_stop(data, radius, outer=None): |
---|
251 | # type: (Data, float, float) -> None |
---|
252 | """ |
---|
253 | Add a beam stop of the given *radius*. If *outer*, make an annulus. |
---|
254 | |
---|
255 | Note: this function does not require sasview |
---|
256 | """ |
---|
257 | if hasattr(data, 'qx_data'): |
---|
258 | q = np.sqrt(data.qx_data**2 + data.qy_data**2) |
---|
259 | data.mask = (q < radius) |
---|
260 | if outer is not None: |
---|
261 | data.mask |= (q >= outer) |
---|
262 | else: |
---|
263 | data.mask = (data.x < radius) |
---|
264 | if outer is not None: |
---|
265 | data.mask |= (data.x >= outer) |
---|
266 | |
---|
267 | |
---|
268 | def parameter_range(p, v): |
---|
269 | # type: (str, float) -> Tuple[float, float] |
---|
270 | """ |
---|
271 | Choose a parameter range based on parameter name and initial value. |
---|
272 | """ |
---|
273 | # process the polydispersity options |
---|
274 | if p.endswith('_pd_n'): |
---|
275 | return 0., 100. |
---|
276 | elif p.endswith('_pd_nsigma'): |
---|
277 | return 0., 5. |
---|
278 | elif p.endswith('_pd_type'): |
---|
279 | raise ValueError("Cannot return a range for a string value") |
---|
280 | elif any(s in p for s in ('theta', 'phi', 'psi')): |
---|
281 | # orientation in [-180,180], orientation pd in [0,45] |
---|
282 | if p.endswith('_pd'): |
---|
283 | return 0., 180. |
---|
284 | else: |
---|
285 | return -180., 180. |
---|
286 | elif p.endswith('_pd'): |
---|
287 | return 0., 1. |
---|
288 | elif 'sld' in p: |
---|
289 | return -0.5, 10. |
---|
290 | elif p == 'background': |
---|
291 | return 0., 10. |
---|
292 | elif p == 'scale': |
---|
293 | return 0., 1.e3 |
---|
294 | elif v < 0.: |
---|
295 | return 2.*v, -2.*v |
---|
296 | else: |
---|
297 | return 0., (2.*v if v > 0. else 1.) |
---|
298 | |
---|
299 | |
---|
300 | def _randomize_one(model_info, name, value): |
---|
301 | # type: (ModelInfo, str, float) -> float |
---|
302 | # type: (ModelInfo, str, str) -> str |
---|
303 | """ |
---|
304 | Randomize a single parameter. |
---|
305 | """ |
---|
306 | # Set the amount of polydispersity/angular dispersion, but by default pd_n |
---|
307 | # is zero so there is no polydispersity. This allows us to turn on/off |
---|
308 | # pd by setting pd_n, and still have randomly generated values |
---|
309 | if name.endswith('_pd'): |
---|
310 | par = model_info.parameters[name[:-3]] |
---|
311 | if par.type == 'orientation': |
---|
312 | # Let oriention variation peak around 13 degrees; 95% < 42 degrees |
---|
313 | return 180*np.random.beta(2.5, 20) |
---|
314 | else: |
---|
315 | # Let polydispersity peak around 15%; 95% < 0.4; max=100% |
---|
316 | return np.random.beta(1.5, 7) |
---|
317 | |
---|
318 | # pd is selected globally rather than per parameter, so set to 0 for no pd |
---|
319 | # In particular, when multiple pd dimensions, want to decrease the number |
---|
320 | # of points per dimension for faster computation |
---|
321 | if name.endswith('_pd_n'): |
---|
322 | return 0 |
---|
323 | |
---|
324 | # Don't mess with distribution type for now |
---|
325 | if name.endswith('_pd_type'): |
---|
326 | return 'gaussian' |
---|
327 | |
---|
328 | # type-dependent value of number of sigmas; for gaussian use 3. |
---|
329 | if name.endswith('_pd_nsigma'): |
---|
330 | return 3. |
---|
331 | |
---|
332 | # background in the range [0.01, 1] |
---|
333 | if name == 'background': |
---|
334 | return 10**np.random.uniform(-2, 0) |
---|
335 | |
---|
336 | # scale defaults to 0.1% to 30% volume fraction |
---|
337 | if name == 'scale': |
---|
338 | return 10**np.random.uniform(-3, -0.5) |
---|
339 | |
---|
340 | # If it is a list of choices, pick one at random with equal probability |
---|
341 | # In practice, the model specific random generator will override. |
---|
342 | par = model_info.parameters[name] |
---|
343 | if len(par.limits) > 2: # choice list |
---|
344 | return np.random.randint(len(par.limits)) |
---|
345 | |
---|
346 | # If it is a fixed range, pick from it with equal probability. |
---|
347 | # For logarithmic ranges, the model will have to override. |
---|
348 | if np.isfinite(par.limits).all(): |
---|
349 | return np.random.uniform(*par.limits) |
---|
350 | |
---|
351 | # If the paramter is marked as an sld use the range of neutron slds |
---|
352 | # TODO: ought to randomly contrast match a pair of SLDs |
---|
353 | if par.type == 'sld': |
---|
354 | return np.random.uniform(-0.5, 12) |
---|
355 | |
---|
356 | # Limit magnetic SLDs to a smaller range, from zero to iron=5/A^2 |
---|
357 | if par.name.startswith('M0:'): |
---|
358 | return np.random.uniform(0, 5) |
---|
359 | |
---|
360 | # Guess at the random length/radius/thickness. In practice, all models |
---|
361 | # are going to set their own reasonable ranges. |
---|
362 | if par.type == 'volume': |
---|
363 | if ('length' in par.name or |
---|
364 | 'radius' in par.name or |
---|
365 | 'thick' in par.name): |
---|
366 | return 10**np.random.uniform(2, 4) |
---|
367 | |
---|
368 | # In the absence of any other info, select a value in [0, 2v], or |
---|
369 | # [-2|v|, 2|v|] if v is negative, or [0, 1] if v is zero. Mostly the |
---|
370 | # model random parameter generators will override this default. |
---|
371 | low, high = parameter_range(par.name, value) |
---|
372 | limits = (max(par.limits[0], low), min(par.limits[1], high)) |
---|
373 | return np.random.uniform(*limits) |
---|
374 | |
---|
375 | def _random_pd(model_info, pars): |
---|
376 | pd = [p for p in model_info.parameters.kernel_parameters if p.polydisperse] |
---|
377 | pd_volume = [] |
---|
378 | pd_oriented = [] |
---|
379 | for p in pd: |
---|
380 | if p.type == 'orientation': |
---|
381 | pd_oriented.append(p.name) |
---|
382 | elif p.length_control is not None: |
---|
383 | n = int(pars.get(p.length_control, 1) + 0.5) |
---|
384 | pd_volume.extend(p.name+str(k+1) for k in range(n)) |
---|
385 | elif p.length > 1: |
---|
386 | pd_volume.extend(p.name+str(k+1) for k in range(p.length)) |
---|
387 | else: |
---|
388 | pd_volume.append(p.name) |
---|
389 | u = np.random.rand() |
---|
390 | n = len(pd_volume) |
---|
391 | if u < 0.01 or n < 1: |
---|
392 | pass # 1% chance of no polydispersity |
---|
393 | elif u < 0.86 or n < 2: |
---|
394 | pars[np.random.choice(pd_volume)+"_pd_n"] = 35 |
---|
395 | elif u < 0.99 or n < 3: |
---|
396 | choices = np.random.choice(len(pd_volume), size=2) |
---|
397 | pars[pd_volume[choices[0]]+"_pd_n"] = 25 |
---|
398 | pars[pd_volume[choices[1]]+"_pd_n"] = 10 |
---|
399 | else: |
---|
400 | choices = np.random.choice(len(pd_volume), size=3) |
---|
401 | pars[pd_volume[choices[0]]+"_pd_n"] = 25 |
---|
402 | pars[pd_volume[choices[1]]+"_pd_n"] = 10 |
---|
403 | pars[pd_volume[choices[2]]+"_pd_n"] = 5 |
---|
404 | if pd_oriented: |
---|
405 | pars['theta_pd_n'] = 20 |
---|
406 | if np.random.rand() < 0.1: |
---|
407 | pars['phi_pd_n'] = 5 |
---|
408 | if np.random.rand() < 0.1: |
---|
409 | if any(p.name == 'psi' for p in model_info.parameters.kernel_parameters): |
---|
410 | #print("generating psi_pd_n") |
---|
411 | pars['psi_pd_n'] = 5 |
---|
412 | |
---|
413 | ## Show selected polydispersity |
---|
414 | #for name, value in pars.items(): |
---|
415 | # if name.endswith('_pd_n') and value > 0: |
---|
416 | # print(name, value, pars.get(name[:-5], 0), pars.get(name[:-2], 0)) |
---|
417 | |
---|
418 | |
---|
419 | def randomize_pars(model_info, pars): |
---|
420 | # type: (ModelInfo, ParameterSet) -> ParameterSet |
---|
421 | """ |
---|
422 | Generate random values for all of the parameters. |
---|
423 | |
---|
424 | Valid ranges for the random number generator are guessed from the name of |
---|
425 | the parameter; this will not account for constraints such as cap radius |
---|
426 | greater than cylinder radius in the capped_cylinder model, so |
---|
427 | :func:`constrain_pars` needs to be called afterward.. |
---|
428 | """ |
---|
429 | # Note: the sort guarantees order of calls to random number generator |
---|
430 | random_pars = dict((p, _randomize_one(model_info, p, v)) |
---|
431 | for p, v in sorted(pars.items())) |
---|
432 | if model_info.random is not None: |
---|
433 | random_pars.update(model_info.random()) |
---|
434 | _random_pd(model_info, random_pars) |
---|
435 | return random_pars |
---|
436 | |
---|
437 | |
---|
438 | def limit_dimensions(model_info, pars, maxdim): |
---|
439 | # type: (ModelInfo, ParameterSet, float) -> None |
---|
440 | """ |
---|
441 | Limit parameters of units of Ang to maxdim. |
---|
442 | """ |
---|
443 | for p in model_info.parameters.call_parameters: |
---|
444 | value = pars[p.name] |
---|
445 | if p.units == 'Ang' and value > maxdim: |
---|
446 | pars[p.name] = maxdim*10**np.random.uniform(-3,0) |
---|
447 | |
---|
448 | def constrain_pars(model_info, pars): |
---|
449 | # type: (ModelInfo, ParameterSet) -> None |
---|
450 | """ |
---|
451 | Restrict parameters to valid values. |
---|
452 | |
---|
453 | This includes model specific code for models such as capped_cylinder |
---|
454 | which need to support within model constraints (cap radius more than |
---|
455 | cylinder radius in this case). |
---|
456 | |
---|
457 | Warning: this updates the *pars* dictionary in place. |
---|
458 | """ |
---|
459 | # TODO: move the model specific code to the individual models |
---|
460 | name = model_info.id |
---|
461 | # if it is a product model, then just look at the form factor since |
---|
462 | # none of the structure factors need any constraints. |
---|
463 | if '*' in name: |
---|
464 | name = name.split('*')[0] |
---|
465 | |
---|
466 | # Suppress magnetism for python models (not yet implemented) |
---|
467 | if callable(model_info.Iq): |
---|
468 | pars.update(suppress_magnetism(pars)) |
---|
469 | |
---|
470 | if name == 'barbell': |
---|
471 | if pars['radius_bell'] < pars['radius']: |
---|
472 | pars['radius'], pars['radius_bell'] = pars['radius_bell'], pars['radius'] |
---|
473 | |
---|
474 | elif name == 'capped_cylinder': |
---|
475 | if pars['radius_cap'] < pars['radius']: |
---|
476 | pars['radius'], pars['radius_cap'] = pars['radius_cap'], pars['radius'] |
---|
477 | |
---|
478 | elif name == 'guinier': |
---|
479 | # Limit guinier to an Rg such that Iq > 1e-30 (single precision cutoff) |
---|
480 | # I(q) = A e^-(Rg^2 q^2/3) > e^-(30 ln 10) |
---|
481 | # => ln A - (Rg^2 q^2/3) > -30 ln 10 |
---|
482 | # => Rg^2 q^2/3 < 30 ln 10 + ln A |
---|
483 | # => Rg < sqrt(90 ln 10 + 3 ln A)/q |
---|
484 | #q_max = 0.2 # mid q maximum |
---|
485 | q_max = 1.0 # high q maximum |
---|
486 | rg_max = np.sqrt(90*np.log(10) + 3*np.log(pars['scale']))/q_max |
---|
487 | pars['rg'] = min(pars['rg'], rg_max) |
---|
488 | |
---|
489 | elif name == 'pearl_necklace': |
---|
490 | if pars['radius'] < pars['thick_string']: |
---|
491 | pars['radius'], pars['thick_string'] = pars['thick_string'], pars['radius'] |
---|
492 | pass |
---|
493 | |
---|
494 | elif name == 'rpa': |
---|
495 | # Make sure phi sums to 1.0 |
---|
496 | if pars['case_num'] < 2: |
---|
497 | pars['Phi1'] = 0. |
---|
498 | pars['Phi2'] = 0. |
---|
499 | elif pars['case_num'] < 5: |
---|
500 | pars['Phi1'] = 0. |
---|
501 | total = sum(pars['Phi'+c] for c in '1234') |
---|
502 | for c in '1234': |
---|
503 | pars['Phi'+c] /= total |
---|
504 | |
---|
505 | def parlist(model_info, pars, is2d): |
---|
506 | # type: (ModelInfo, ParameterSet, bool) -> str |
---|
507 | """ |
---|
508 | Format the parameter list for printing. |
---|
509 | """ |
---|
510 | is2d = True |
---|
511 | lines = [] |
---|
512 | parameters = model_info.parameters |
---|
513 | magnetic = False |
---|
514 | magnetic_pars = [] |
---|
515 | for p in parameters.user_parameters(pars, is2d): |
---|
516 | if any(p.id.startswith(x) for x in ('M0:', 'mtheta:', 'mphi:')): |
---|
517 | continue |
---|
518 | if p.id.startswith('up:'): |
---|
519 | magnetic_pars.append("%s=%s"%(p.id, pars.get(p.id, p.default))) |
---|
520 | continue |
---|
521 | fields = dict( |
---|
522 | value=pars.get(p.id, p.default), |
---|
523 | pd=pars.get(p.id+"_pd", 0.), |
---|
524 | n=int(pars.get(p.id+"_pd_n", 0)), |
---|
525 | nsigma=pars.get(p.id+"_pd_nsgima", 3.), |
---|
526 | pdtype=pars.get(p.id+"_pd_type", 'gaussian'), |
---|
527 | relative_pd=p.relative_pd, |
---|
528 | M0=pars.get('M0:'+p.id, 0.), |
---|
529 | mphi=pars.get('mphi:'+p.id, 0.), |
---|
530 | mtheta=pars.get('mtheta:'+p.id, 0.), |
---|
531 | ) |
---|
532 | lines.append(_format_par(p.name, **fields)) |
---|
533 | magnetic = magnetic or fields['M0'] != 0. |
---|
534 | if magnetic and magnetic_pars: |
---|
535 | lines.append(" ".join(magnetic_pars)) |
---|
536 | return "\n".join(lines) |
---|
537 | |
---|
538 | #return "\n".join("%s: %s"%(p, v) for p, v in sorted(pars.items())) |
---|
539 | |
---|
540 | def _format_par(name, value=0., pd=0., n=0, nsigma=3., pdtype='gaussian', |
---|
541 | relative_pd=False, M0=0., mphi=0., mtheta=0.): |
---|
542 | # type: (str, float, float, int, float, str) -> str |
---|
543 | line = "%s: %g"%(name, value) |
---|
544 | if pd != 0. and n != 0: |
---|
545 | if relative_pd: |
---|
546 | pd *= value |
---|
547 | line += " +/- %g (%d points in [-%g,%g] sigma %s)"\ |
---|
548 | % (pd, n, nsigma, nsigma, pdtype) |
---|
549 | if M0 != 0.: |
---|
550 | line += " M0:%.3f mtheta:%.1f mphi:%.1f" % (M0, mtheta, mphi) |
---|
551 | return line |
---|
552 | |
---|
553 | def suppress_pd(pars, suppress=True): |
---|
554 | # type: (ParameterSet) -> ParameterSet |
---|
555 | """ |
---|
556 | If suppress is True complete eliminate polydispersity of the model to test |
---|
557 | models more quickly. If suppress is False, make sure at least one |
---|
558 | parameter is polydisperse, setting the first polydispersity parameter to |
---|
559 | 15% if no polydispersity is given (with no explicit demo parameters given |
---|
560 | in the model, there will be no default polydispersity). |
---|
561 | """ |
---|
562 | pars = pars.copy() |
---|
563 | #print("pars=", pars) |
---|
564 | if suppress: |
---|
565 | for p in pars: |
---|
566 | if p.endswith("_pd_n"): |
---|
567 | pars[p] = 0 |
---|
568 | else: |
---|
569 | any_pd = False |
---|
570 | first_pd = None |
---|
571 | for p in pars: |
---|
572 | if p.endswith("_pd_n"): |
---|
573 | pd = pars.get(p[:-2], 0.) |
---|
574 | any_pd |= (pars[p] != 0 and pd != 0.) |
---|
575 | if first_pd is None: |
---|
576 | first_pd = p |
---|
577 | if not any_pd and first_pd is not None: |
---|
578 | if pars[first_pd] == 0: |
---|
579 | pars[first_pd] = 35 |
---|
580 | if first_pd[:-2] not in pars or pars[first_pd[:-2]] == 0: |
---|
581 | pars[first_pd[:-2]] = 0.15 |
---|
582 | return pars |
---|
583 | |
---|
584 | def suppress_magnetism(pars, suppress=True): |
---|
585 | # type: (ParameterSet) -> ParameterSet |
---|
586 | """ |
---|
587 | If suppress is True complete eliminate magnetism of the model to test |
---|
588 | models more quickly. If suppress is False, make sure at least one sld |
---|
589 | parameter is magnetic, setting the first parameter to have a strong |
---|
590 | magnetic sld (8/A^2) at 60 degrees (with no explicit demo parameters given |
---|
591 | in the model, there will be no default magnetism). |
---|
592 | """ |
---|
593 | pars = pars.copy() |
---|
594 | if suppress: |
---|
595 | for p in pars: |
---|
596 | if p.startswith("M0:"): |
---|
597 | pars[p] = 0 |
---|
598 | else: |
---|
599 | any_mag = False |
---|
600 | first_mag = None |
---|
601 | for p in pars: |
---|
602 | if p.startswith("M0:"): |
---|
603 | any_mag |= (pars[p] != 0) |
---|
604 | if first_mag is None: |
---|
605 | first_mag = p |
---|
606 | if not any_mag and first_mag is not None: |
---|
607 | pars[first_mag] = 8. |
---|
608 | return pars |
---|
609 | |
---|
610 | def eval_sasview(model_info, data): |
---|
611 | # type: (Modelinfo, Data) -> Calculator |
---|
612 | """ |
---|
613 | Return a model calculator using the pre-4.0 SasView models. |
---|
614 | """ |
---|
615 | # importing sas here so that the error message will be that sas failed to |
---|
616 | # import rather than the more obscure smear_selection not imported error |
---|
617 | import sas |
---|
618 | import sas.models |
---|
619 | from sas.models.qsmearing import smear_selection |
---|
620 | from sas.models.MultiplicationModel import MultiplicationModel |
---|
621 | from sas.models.dispersion_models import models as dispersers |
---|
622 | |
---|
623 | def get_model_class(name): |
---|
624 | # type: (str) -> "sas.models.BaseComponent" |
---|
625 | #print("new",sorted(_pars.items())) |
---|
626 | __import__('sas.models.' + name) |
---|
627 | ModelClass = getattr(getattr(sas.models, name, None), name, None) |
---|
628 | if ModelClass is None: |
---|
629 | raise ValueError("could not find model %r in sas.models"%name) |
---|
630 | return ModelClass |
---|
631 | |
---|
632 | # WARNING: ugly hack when handling model! |
---|
633 | # Sasview models with multiplicity need to be created with the target |
---|
634 | # multiplicity, so we cannot create the target model ahead of time for |
---|
635 | # for multiplicity models. Instead we store the model in a list and |
---|
636 | # update the first element of that list with the new multiplicity model |
---|
637 | # every time we evaluate. |
---|
638 | |
---|
639 | # grab the sasview model, or create it if it is a product model |
---|
640 | if model_info.composition: |
---|
641 | composition_type, parts = model_info.composition |
---|
642 | if composition_type == 'product': |
---|
643 | P, S = [get_model_class(revert_name(p))() for p in parts] |
---|
644 | model = [MultiplicationModel(P, S)] |
---|
645 | else: |
---|
646 | raise ValueError("sasview mixture models not supported by compare") |
---|
647 | else: |
---|
648 | old_name = revert_name(model_info) |
---|
649 | if old_name is None: |
---|
650 | raise ValueError("model %r does not exist in old sasview" |
---|
651 | % model_info.id) |
---|
652 | ModelClass = get_model_class(old_name) |
---|
653 | model = [ModelClass()] |
---|
654 | model[0].disperser_handles = {} |
---|
655 | |
---|
656 | # build a smearer with which to call the model, if necessary |
---|
657 | smearer = smear_selection(data, model=model) |
---|
658 | if hasattr(data, 'qx_data'): |
---|
659 | q = np.sqrt(data.qx_data**2 + data.qy_data**2) |
---|
660 | index = ((~data.mask) & (~np.isnan(data.data)) |
---|
661 | & (q >= data.qmin) & (q <= data.qmax)) |
---|
662 | if smearer is not None: |
---|
663 | smearer.model = model # because smear_selection has a bug |
---|
664 | smearer.accuracy = data.accuracy |
---|
665 | smearer.set_index(index) |
---|
666 | def _call_smearer(): |
---|
667 | smearer.model = model[0] |
---|
668 | return smearer.get_value() |
---|
669 | theory = _call_smearer |
---|
670 | else: |
---|
671 | theory = lambda: model[0].evalDistribution([data.qx_data[index], |
---|
672 | data.qy_data[index]]) |
---|
673 | elif smearer is not None: |
---|
674 | theory = lambda: smearer(model[0].evalDistribution(data.x)) |
---|
675 | else: |
---|
676 | theory = lambda: model[0].evalDistribution(data.x) |
---|
677 | |
---|
678 | def calculator(**pars): |
---|
679 | # type: (float, ...) -> np.ndarray |
---|
680 | """ |
---|
681 | Sasview calculator for model. |
---|
682 | """ |
---|
683 | oldpars = revert_pars(model_info, pars) |
---|
684 | # For multiplicity models, create a model with the correct multiplicity |
---|
685 | control = oldpars.pop("CONTROL", None) |
---|
686 | if control is not None: |
---|
687 | # sphericalSLD has one fewer multiplicity. This update should |
---|
688 | # happen in revert_pars, but it hasn't been called yet. |
---|
689 | model[0] = ModelClass(control) |
---|
690 | # paying for parameter conversion each time to keep life simple, if not fast |
---|
691 | for k, v in oldpars.items(): |
---|
692 | if k.endswith('.type'): |
---|
693 | par = k[:-5] |
---|
694 | if v == 'gaussian': continue |
---|
695 | cls = dispersers[v if v != 'rectangle' else 'rectangula'] |
---|
696 | handle = cls() |
---|
697 | model[0].disperser_handles[par] = handle |
---|
698 | try: |
---|
699 | model[0].set_dispersion(par, handle) |
---|
700 | except Exception: |
---|
701 | exception.annotate_exception("while setting %s to %r" |
---|
702 | %(par, v)) |
---|
703 | raise |
---|
704 | |
---|
705 | |
---|
706 | #print("sasview pars",oldpars) |
---|
707 | for k, v in oldpars.items(): |
---|
708 | name_attr = k.split('.') # polydispersity components |
---|
709 | if len(name_attr) == 2: |
---|
710 | par, disp_par = name_attr |
---|
711 | model[0].dispersion[par][disp_par] = v |
---|
712 | else: |
---|
713 | model[0].setParam(k, v) |
---|
714 | return theory() |
---|
715 | |
---|
716 | calculator.engine = "sasview" |
---|
717 | return calculator |
---|
718 | |
---|
719 | DTYPE_MAP = { |
---|
720 | 'half': '16', |
---|
721 | 'fast': 'fast', |
---|
722 | 'single': '32', |
---|
723 | 'double': '64', |
---|
724 | 'quad': '128', |
---|
725 | 'f16': '16', |
---|
726 | 'f32': '32', |
---|
727 | 'f64': '64', |
---|
728 | 'float16': '16', |
---|
729 | 'float32': '32', |
---|
730 | 'float64': '64', |
---|
731 | 'float128': '128', |
---|
732 | 'longdouble': '128', |
---|
733 | } |
---|
734 | def eval_opencl(model_info, data, dtype='single', cutoff=0.): |
---|
735 | # type: (ModelInfo, Data, str, float) -> Calculator |
---|
736 | """ |
---|
737 | Return a model calculator using the OpenCL calculation engine. |
---|
738 | """ |
---|
739 | if not core.HAVE_OPENCL: |
---|
740 | raise RuntimeError("OpenCL not available") |
---|
741 | model = core.build_model(model_info, dtype=dtype, platform="ocl") |
---|
742 | calculator = DirectModel(data, model, cutoff=cutoff) |
---|
743 | calculator.engine = "OCL%s"%DTYPE_MAP[str(model.dtype)] |
---|
744 | return calculator |
---|
745 | |
---|
746 | def eval_ctypes(model_info, data, dtype='double', cutoff=0.): |
---|
747 | # type: (ModelInfo, Data, str, float) -> Calculator |
---|
748 | """ |
---|
749 | Return a model calculator using the DLL calculation engine. |
---|
750 | """ |
---|
751 | model = core.build_model(model_info, dtype=dtype, platform="dll") |
---|
752 | calculator = DirectModel(data, model, cutoff=cutoff) |
---|
753 | calculator.engine = "OMP%s"%DTYPE_MAP[str(model.dtype)] |
---|
754 | return calculator |
---|
755 | |
---|
756 | def time_calculation(calculator, pars, evals=1): |
---|
757 | # type: (Calculator, ParameterSet, int) -> Tuple[np.ndarray, float] |
---|
758 | """ |
---|
759 | Compute the average calculation time over N evaluations. |
---|
760 | |
---|
761 | An additional call is generated without polydispersity in order to |
---|
762 | initialize the calculation engine, and make the average more stable. |
---|
763 | """ |
---|
764 | # initialize the code so time is more accurate |
---|
765 | if evals > 1: |
---|
766 | calculator(**suppress_pd(pars)) |
---|
767 | toc = tic() |
---|
768 | # make sure there is at least one eval |
---|
769 | value = calculator(**pars) |
---|
770 | for _ in range(evals-1): |
---|
771 | value = calculator(**pars) |
---|
772 | average_time = toc()*1000. / evals |
---|
773 | #print("I(q)",value) |
---|
774 | return value, average_time |
---|
775 | |
---|
776 | def make_data(opts): |
---|
777 | # type: (Dict[str, Any]) -> Tuple[Data, np.ndarray] |
---|
778 | """ |
---|
779 | Generate an empty dataset, used with the model to set Q points |
---|
780 | and resolution. |
---|
781 | |
---|
782 | *opts* contains the options, with 'qmax', 'nq', 'res', |
---|
783 | 'accuracy', 'is2d' and 'view' parsed from the command line. |
---|
784 | """ |
---|
785 | qmin, qmax, nq, res = opts['qmin'], opts['qmax'], opts['nq'], opts['res'] |
---|
786 | if opts['is2d']: |
---|
787 | q = np.linspace(-qmax, qmax, nq) # type: np.ndarray |
---|
788 | data = empty_data2D(q, resolution=res) |
---|
789 | data.accuracy = opts['accuracy'] |
---|
790 | set_beam_stop(data, 0.0004) |
---|
791 | index = ~data.mask |
---|
792 | else: |
---|
793 | if opts['view'] == 'log' and not opts['zero']: |
---|
794 | q = np.logspace(math.log10(qmin), math.log10(qmax), nq) |
---|
795 | else: |
---|
796 | q = np.linspace(qmin, qmax, nq) |
---|
797 | if opts['zero']: |
---|
798 | q = np.hstack((0, q)) |
---|
799 | data = empty_data1D(q, resolution=res) |
---|
800 | index = slice(None, None) |
---|
801 | return data, index |
---|
802 | |
---|
803 | def make_engine(model_info, data, dtype, cutoff): |
---|
804 | # type: (ModelInfo, Data, str, float) -> Calculator |
---|
805 | """ |
---|
806 | Generate the appropriate calculation engine for the given datatype. |
---|
807 | |
---|
808 | Datatypes with '!' appended are evaluated using external C DLLs rather |
---|
809 | than OpenCL. |
---|
810 | """ |
---|
811 | if dtype == 'sasview': |
---|
812 | return eval_sasview(model_info, data) |
---|
813 | elif dtype is None or not dtype.endswith('!'): |
---|
814 | return eval_opencl(model_info, data, dtype=dtype, cutoff=cutoff) |
---|
815 | else: |
---|
816 | return eval_ctypes(model_info, data, dtype=dtype[:-1], cutoff=cutoff) |
---|
817 | |
---|
818 | def _show_invalid(data, theory): |
---|
819 | # type: (Data, np.ma.ndarray) -> None |
---|
820 | """ |
---|
821 | Display a list of the non-finite values in theory. |
---|
822 | """ |
---|
823 | if not theory.mask.any(): |
---|
824 | return |
---|
825 | |
---|
826 | if hasattr(data, 'x'): |
---|
827 | bad = zip(data.x[theory.mask], theory[theory.mask]) |
---|
828 | print(" *** ", ", ".join("I(%g)=%g"%(x, y) for x, y in bad)) |
---|
829 | |
---|
830 | |
---|
831 | def compare(opts, limits=None, maxdim=np.inf): |
---|
832 | # type: (Dict[str, Any], Optional[Tuple[float, float]]) -> Tuple[float, float] |
---|
833 | """ |
---|
834 | Preform a comparison using options from the command line. |
---|
835 | |
---|
836 | *limits* are the limits on the values to use, either to set the y-axis |
---|
837 | for 1D or to set the colormap scale for 2D. If None, then they are |
---|
838 | inferred from the data and returned. When exploring using Bumps, |
---|
839 | the limits are set when the model is initially called, and maintained |
---|
840 | as the values are adjusted, making it easier to see the effects of the |
---|
841 | parameters. |
---|
842 | |
---|
843 | *maxdim* is the maximum value for any parameter with units of Angstrom. |
---|
844 | """ |
---|
845 | for k in range(opts['sets']): |
---|
846 | if k > 1: |
---|
847 | # print a separate seed for each dataset for better reproducibility |
---|
848 | new_seed = np.random.randint(1000000) |
---|
849 | print("Set %d uses -random=%i"%(k+1,new_seed)) |
---|
850 | np.random.seed(new_seed) |
---|
851 | opts['pars'] = parse_pars(opts, maxdim=maxdim) |
---|
852 | if opts['pars'] is None: |
---|
853 | return |
---|
854 | result = run_models(opts, verbose=True) |
---|
855 | if opts['plot']: |
---|
856 | limits = plot_models(opts, result, limits=limits, setnum=k) |
---|
857 | if opts['show_weights']: |
---|
858 | base, _ = opts['engines'] |
---|
859 | base_pars, _ = opts['pars'] |
---|
860 | model_info = base._kernel.info |
---|
861 | dim = base._kernel.dim |
---|
862 | plot_weights(model_info, get_mesh(model_info, base_pars, dim=dim)) |
---|
863 | if opts['plot']: |
---|
864 | import matplotlib.pyplot as plt |
---|
865 | plt.show() |
---|
866 | return limits |
---|
867 | |
---|
868 | def run_models(opts, verbose=False): |
---|
869 | # type: (Dict[str, Any]) -> Dict[str, Any] |
---|
870 | |
---|
871 | base, comp = opts['engines'] |
---|
872 | base_n, comp_n = opts['count'] |
---|
873 | base_pars, comp_pars = opts['pars'] |
---|
874 | data = opts['data'] |
---|
875 | |
---|
876 | comparison = comp is not None |
---|
877 | |
---|
878 | base_time = comp_time = None |
---|
879 | base_value = comp_value = resid = relerr = None |
---|
880 | |
---|
881 | # Base calculation |
---|
882 | try: |
---|
883 | base_raw, base_time = time_calculation(base, base_pars, base_n) |
---|
884 | base_value = np.ma.masked_invalid(base_raw) |
---|
885 | if verbose: |
---|
886 | print("%s t=%.2f ms, intensity=%.0f" |
---|
887 | % (base.engine, base_time, base_value.sum())) |
---|
888 | _show_invalid(data, base_value) |
---|
889 | except ImportError: |
---|
890 | traceback.print_exc() |
---|
891 | |
---|
892 | # Comparison calculation |
---|
893 | if comparison: |
---|
894 | try: |
---|
895 | comp_raw, comp_time = time_calculation(comp, comp_pars, comp_n) |
---|
896 | comp_value = np.ma.masked_invalid(comp_raw) |
---|
897 | if verbose: |
---|
898 | print("%s t=%.2f ms, intensity=%.0f" |
---|
899 | % (comp.engine, comp_time, comp_value.sum())) |
---|
900 | _show_invalid(data, comp_value) |
---|
901 | except ImportError: |
---|
902 | traceback.print_exc() |
---|
903 | |
---|
904 | # Compare, but only if computing both forms |
---|
905 | if comparison: |
---|
906 | resid = (base_value - comp_value) |
---|
907 | relerr = resid/np.where(comp_value != 0., abs(comp_value), 1.0) |
---|
908 | if verbose: |
---|
909 | _print_stats("|%s-%s|" |
---|
910 | % (base.engine, comp.engine) + (" "*(3+len(comp.engine))), |
---|
911 | resid) |
---|
912 | _print_stats("|(%s-%s)/%s|" |
---|
913 | % (base.engine, comp.engine, comp.engine), |
---|
914 | relerr) |
---|
915 | |
---|
916 | return dict(base_value=base_value, comp_value=comp_value, |
---|
917 | base_time=base_time, comp_time=comp_time, |
---|
918 | resid=resid, relerr=relerr) |
---|
919 | |
---|
920 | |
---|
921 | def _print_stats(label, err): |
---|
922 | # type: (str, np.ma.ndarray) -> None |
---|
923 | # work with trimmed data, not the full set |
---|
924 | sorted_err = np.sort(abs(err.compressed())) |
---|
925 | if len(sorted_err) == 0.: |
---|
926 | print(label + " no valid values") |
---|
927 | return |
---|
928 | |
---|
929 | p50 = int((len(sorted_err)-1)*0.50) |
---|
930 | p98 = int((len(sorted_err)-1)*0.98) |
---|
931 | data = [ |
---|
932 | "max:%.3e"%sorted_err[-1], |
---|
933 | "median:%.3e"%sorted_err[p50], |
---|
934 | "98%%:%.3e"%sorted_err[p98], |
---|
935 | "rms:%.3e"%np.sqrt(np.mean(sorted_err**2)), |
---|
936 | "zero-offset:%+.3e"%np.mean(sorted_err), |
---|
937 | ] |
---|
938 | print(label+" "+" ".join(data)) |
---|
939 | |
---|
940 | |
---|
941 | def plot_models(opts, result, limits=None, setnum=0): |
---|
942 | # type: (Dict[str, Any], Dict[str, Any], Optional[Tuple[float, float]]) -> Tuple[float, float] |
---|
943 | import matplotlib.pyplot as plt |
---|
944 | |
---|
945 | base_value, comp_value = result['base_value'], result['comp_value'] |
---|
946 | base_time, comp_time = result['base_time'], result['comp_time'] |
---|
947 | resid, relerr = result['resid'], result['relerr'] |
---|
948 | |
---|
949 | have_base, have_comp = (base_value is not None), (comp_value is not None) |
---|
950 | base, comp = opts['engines'] |
---|
951 | data = opts['data'] |
---|
952 | use_data = (opts['datafile'] is not None) and (have_base ^ have_comp) |
---|
953 | |
---|
954 | # Plot if requested |
---|
955 | view = opts['view'] |
---|
956 | if limits is None: |
---|
957 | vmin, vmax = np.inf, -np.inf |
---|
958 | if have_base: |
---|
959 | vmin = min(vmin, base_value.min()) |
---|
960 | vmax = max(vmax, base_value.max()) |
---|
961 | if have_comp: |
---|
962 | vmin = min(vmin, comp_value.min()) |
---|
963 | vmax = max(vmax, comp_value.max()) |
---|
964 | limits = vmin, vmax |
---|
965 | |
---|
966 | if have_base: |
---|
967 | if have_comp: |
---|
968 | plt.subplot(131) |
---|
969 | plot_theory(data, base_value, view=view, use_data=use_data, limits=limits) |
---|
970 | plt.title("%s t=%.2f ms"%(base.engine, base_time)) |
---|
971 | #cbar_title = "log I" |
---|
972 | if have_comp: |
---|
973 | if have_base: |
---|
974 | plt.subplot(132) |
---|
975 | if not opts['is2d'] and have_base: |
---|
976 | plot_theory(data, base_value, view=view, use_data=use_data, limits=limits) |
---|
977 | plot_theory(data, comp_value, view=view, use_data=use_data, limits=limits) |
---|
978 | plt.title("%s t=%.2f ms"%(comp.engine, comp_time)) |
---|
979 | #cbar_title = "log I" |
---|
980 | if have_base and have_comp: |
---|
981 | plt.subplot(133) |
---|
982 | if not opts['rel_err']: |
---|
983 | err, errstr, errview = resid, "abs err", "linear" |
---|
984 | else: |
---|
985 | err, errstr, errview = abs(relerr), "rel err", "log" |
---|
986 | if (err == 0.).all(): |
---|
987 | errview = 'linear' |
---|
988 | if 0: # 95% cutoff |
---|
989 | sorted = np.sort(err.flatten()) |
---|
990 | cutoff = sorted[int(sorted.size*0.95)] |
---|
991 | err[err > cutoff] = cutoff |
---|
992 | #err,errstr = base/comp,"ratio" |
---|
993 | plot_theory(data, None, resid=err, view=errview, use_data=use_data) |
---|
994 | plt.xscale('log' if view == 'log' and not opts['is2d'] else 'linear') |
---|
995 | plt.legend(['P%d'%(k+1) for k in range(setnum+1)], loc='best') |
---|
996 | plt.title("max %s = %.3g"%(errstr, abs(err).max())) |
---|
997 | #cbar_title = errstr if errview=="linear" else "log "+errstr |
---|
998 | #if is2D: |
---|
999 | # h = plt.colorbar() |
---|
1000 | # h.ax.set_title(cbar_title) |
---|
1001 | fig = plt.gcf() |
---|
1002 | extra_title = ' '+opts['title'] if opts['title'] else '' |
---|
1003 | fig.suptitle(":".join(opts['name']) + extra_title) |
---|
1004 | |
---|
1005 | if have_base and have_comp and opts['show_hist']: |
---|
1006 | plt.figure() |
---|
1007 | v = relerr |
---|
1008 | v[v == 0] = 0.5*np.min(np.abs(v[v != 0])) |
---|
1009 | plt.hist(np.log10(np.abs(v)), normed=1, bins=50) |
---|
1010 | plt.xlabel('log10(err), err = |(%s - %s) / %s|' |
---|
1011 | % (base.engine, comp.engine, comp.engine)) |
---|
1012 | plt.ylabel('P(err)') |
---|
1013 | plt.title('Distribution of relative error between calculation engines') |
---|
1014 | |
---|
1015 | return limits |
---|
1016 | |
---|
1017 | |
---|
1018 | # =========================================================================== |
---|
1019 | # |
---|
1020 | |
---|
1021 | # Set of command line options. |
---|
1022 | # Normal options such as -plot/-noplot are specified as 'name'. |
---|
1023 | # For options such as -nq=500 which require a value use 'name='. |
---|
1024 | # |
---|
1025 | OPTIONS = [ |
---|
1026 | # Plotting |
---|
1027 | 'plot', 'noplot', 'weights', |
---|
1028 | 'linear', 'log', 'q4', |
---|
1029 | 'rel', 'abs', |
---|
1030 | 'hist', 'nohist', |
---|
1031 | 'title=', |
---|
1032 | |
---|
1033 | # Data generation |
---|
1034 | 'data=', 'noise=', 'res=', 'nq=', 'q=', |
---|
1035 | 'lowq', 'midq', 'highq', 'exq', 'zero', |
---|
1036 | '2d', '1d', |
---|
1037 | |
---|
1038 | # Parameter set |
---|
1039 | 'preset', 'random', 'random=', 'sets=', |
---|
1040 | 'demo', 'default', # TODO: remove demo/default |
---|
1041 | 'nopars', 'pars', |
---|
1042 | 'sphere', 'sphere=', # integrate over a sphere in 2d with n points |
---|
1043 | |
---|
1044 | # Calculation options |
---|
1045 | 'poly', 'mono', 'cutoff=', |
---|
1046 | 'magnetic', 'nonmagnetic', |
---|
1047 | 'accuracy=', |
---|
1048 | 'neval=', # for timing... |
---|
1049 | |
---|
1050 | # Precision options |
---|
1051 | 'engine=', |
---|
1052 | 'half', 'fast', 'single', 'double', 'single!', 'double!', 'quad!', |
---|
1053 | 'sasview', # TODO: remove sasview 3.x support |
---|
1054 | |
---|
1055 | # Output options |
---|
1056 | 'help', 'html', 'edit', |
---|
1057 | ] |
---|
1058 | |
---|
1059 | NAME_OPTIONS = set(k for k in OPTIONS if not k.endswith('=')) |
---|
1060 | VALUE_OPTIONS = [k[:-1] for k in OPTIONS if k.endswith('=')] |
---|
1061 | |
---|
1062 | |
---|
1063 | def columnize(items, indent="", width=79): |
---|
1064 | # type: (List[str], str, int) -> str |
---|
1065 | """ |
---|
1066 | Format a list of strings into columns. |
---|
1067 | |
---|
1068 | Returns a string with carriage returns ready for printing. |
---|
1069 | """ |
---|
1070 | column_width = max(len(w) for w in items) + 1 |
---|
1071 | num_columns = (width - len(indent)) // column_width |
---|
1072 | num_rows = len(items) // num_columns |
---|
1073 | items = items + [""] * (num_rows * num_columns - len(items)) |
---|
1074 | columns = [items[k*num_rows:(k+1)*num_rows] for k in range(num_columns)] |
---|
1075 | lines = [" ".join("%-*s"%(column_width, entry) for entry in row) |
---|
1076 | for row in zip(*columns)] |
---|
1077 | output = indent + ("\n"+indent).join(lines) |
---|
1078 | return output |
---|
1079 | |
---|
1080 | |
---|
1081 | def get_pars(model_info, use_demo=False): |
---|
1082 | # type: (ModelInfo, bool) -> ParameterSet |
---|
1083 | """ |
---|
1084 | Extract demo parameters from the model definition. |
---|
1085 | """ |
---|
1086 | # Get the default values for the parameters |
---|
1087 | pars = {} |
---|
1088 | for p in model_info.parameters.call_parameters: |
---|
1089 | parts = [('', p.default)] |
---|
1090 | if p.polydisperse: |
---|
1091 | parts.append(('_pd', 0.0)) |
---|
1092 | parts.append(('_pd_n', 0)) |
---|
1093 | parts.append(('_pd_nsigma', 3.0)) |
---|
1094 | parts.append(('_pd_type', "gaussian")) |
---|
1095 | for ext, val in parts: |
---|
1096 | if p.length > 1: |
---|
1097 | dict(("%s%d%s" % (p.id, k, ext), val) |
---|
1098 | for k in range(1, p.length+1)) |
---|
1099 | else: |
---|
1100 | pars[p.id + ext] = val |
---|
1101 | |
---|
1102 | # Plug in values given in demo |
---|
1103 | if use_demo and model_info.demo: |
---|
1104 | pars.update(model_info.demo) |
---|
1105 | return pars |
---|
1106 | |
---|
1107 | INTEGER_RE = re.compile("^[+-]?[1-9][0-9]*$") |
---|
1108 | def isnumber(str): |
---|
1109 | match = FLOAT_RE.match(str) |
---|
1110 | isfloat = (match and not str[match.end():]) |
---|
1111 | return isfloat or INTEGER_RE.match(str) |
---|
1112 | |
---|
1113 | # For distinguishing pairs of models for comparison |
---|
1114 | # key-value pair separator = |
---|
1115 | # shell characters | & ; <> $ % ' " \ # ` |
---|
1116 | # model and parameter names _ |
---|
1117 | # parameter expressions - + * / . ( ) |
---|
1118 | # path characters including tilde expansion and windows drive ~ / : |
---|
1119 | # not sure about brackets [] {} |
---|
1120 | # maybe one of the following @ ? ^ ! , |
---|
1121 | PAR_SPLIT = ',' |
---|
1122 | def parse_opts(argv): |
---|
1123 | # type: (List[str]) -> Dict[str, Any] |
---|
1124 | """ |
---|
1125 | Parse command line options. |
---|
1126 | """ |
---|
1127 | MODELS = core.list_models() |
---|
1128 | flags = [arg for arg in argv |
---|
1129 | if arg.startswith('-')] |
---|
1130 | values = [arg for arg in argv |
---|
1131 | if not arg.startswith('-') and '=' in arg] |
---|
1132 | positional_args = [arg for arg in argv |
---|
1133 | if not arg.startswith('-') and '=' not in arg] |
---|
1134 | models = "\n ".join("%-15s"%v for v in MODELS) |
---|
1135 | if len(positional_args) == 0: |
---|
1136 | print(USAGE) |
---|
1137 | print("\nAvailable models:") |
---|
1138 | print(columnize(MODELS, indent=" ")) |
---|
1139 | return None |
---|
1140 | |
---|
1141 | invalid = [o[1:] for o in flags |
---|
1142 | if o[1:] not in NAME_OPTIONS |
---|
1143 | and not any(o.startswith('-%s='%t) for t in VALUE_OPTIONS)] |
---|
1144 | if invalid: |
---|
1145 | print("Invalid options: %s"%(", ".join(invalid))) |
---|
1146 | return None |
---|
1147 | |
---|
1148 | name = positional_args[-1] |
---|
1149 | |
---|
1150 | # pylint: disable=bad-whitespace |
---|
1151 | # Interpret the flags |
---|
1152 | opts = { |
---|
1153 | 'plot' : True, |
---|
1154 | 'view' : 'log', |
---|
1155 | 'is2d' : False, |
---|
1156 | 'qmin' : None, |
---|
1157 | 'qmax' : 0.05, |
---|
1158 | 'nq' : 128, |
---|
1159 | 'res' : 0.0, |
---|
1160 | 'noise' : 0.0, |
---|
1161 | 'accuracy' : 'Low', |
---|
1162 | 'cutoff' : '0.0', |
---|
1163 | 'seed' : -1, # default to preset |
---|
1164 | 'mono' : True, |
---|
1165 | # Default to magnetic a magnetic moment is set on the command line |
---|
1166 | 'magnetic' : False, |
---|
1167 | 'show_pars' : False, |
---|
1168 | 'show_hist' : False, |
---|
1169 | 'rel_err' : True, |
---|
1170 | 'explore' : False, |
---|
1171 | 'use_demo' : True, |
---|
1172 | 'zero' : False, |
---|
1173 | 'html' : False, |
---|
1174 | 'title' : None, |
---|
1175 | 'datafile' : None, |
---|
1176 | 'sets' : 0, |
---|
1177 | 'engine' : 'default', |
---|
1178 | 'count' : '1', |
---|
1179 | 'show_weights' : False, |
---|
1180 | 'sphere' : 0, |
---|
1181 | } |
---|
1182 | for arg in flags: |
---|
1183 | if arg == '-noplot': opts['plot'] = False |
---|
1184 | elif arg == '-plot': opts['plot'] = True |
---|
1185 | elif arg == '-linear': opts['view'] = 'linear' |
---|
1186 | elif arg == '-log': opts['view'] = 'log' |
---|
1187 | elif arg == '-q4': opts['view'] = 'q4' |
---|
1188 | elif arg == '-1d': opts['is2d'] = False |
---|
1189 | elif arg == '-2d': opts['is2d'] = True |
---|
1190 | elif arg == '-exq': opts['qmax'] = 10.0 |
---|
1191 | elif arg == '-highq': opts['qmax'] = 1.0 |
---|
1192 | elif arg == '-midq': opts['qmax'] = 0.2 |
---|
1193 | elif arg == '-lowq': opts['qmax'] = 0.05 |
---|
1194 | elif arg == '-zero': opts['zero'] = True |
---|
1195 | elif arg.startswith('-nq='): opts['nq'] = int(arg[4:]) |
---|
1196 | elif arg.startswith('-q='): |
---|
1197 | opts['qmin'], opts['qmax'] = [float(v) for v in arg[3:].split(':')] |
---|
1198 | elif arg.startswith('-res='): opts['res'] = float(arg[5:]) |
---|
1199 | elif arg.startswith('-noise='): opts['noise'] = float(arg[7:]) |
---|
1200 | elif arg.startswith('-sets='): opts['sets'] = int(arg[6:]) |
---|
1201 | elif arg.startswith('-accuracy='): opts['accuracy'] = arg[10:] |
---|
1202 | elif arg.startswith('-cutoff='): opts['cutoff'] = arg[8:] |
---|
1203 | elif arg.startswith('-title='): opts['title'] = arg[7:] |
---|
1204 | elif arg.startswith('-data='): opts['datafile'] = arg[6:] |
---|
1205 | elif arg.startswith('-engine='): opts['engine'] = arg[8:] |
---|
1206 | elif arg.startswith('-neval='): opts['count'] = arg[7:] |
---|
1207 | elif arg.startswith('-random='): |
---|
1208 | opts['seed'] = int(arg[8:]) |
---|
1209 | opts['sets'] = 0 |
---|
1210 | elif arg == '-random': |
---|
1211 | opts['seed'] = np.random.randint(1000000) |
---|
1212 | opts['sets'] = 0 |
---|
1213 | elif arg.startswith('-sphere'): |
---|
1214 | opts['sphere'] = int(arg[8:]) if len(arg) > 7 else 150 |
---|
1215 | opts['is2d'] = True |
---|
1216 | elif arg == '-preset': opts['seed'] = -1 |
---|
1217 | elif arg == '-mono': opts['mono'] = True |
---|
1218 | elif arg == '-poly': opts['mono'] = False |
---|
1219 | elif arg == '-magnetic': opts['magnetic'] = True |
---|
1220 | elif arg == '-nonmagnetic': opts['magnetic'] = False |
---|
1221 | elif arg == '-pars': opts['show_pars'] = True |
---|
1222 | elif arg == '-nopars': opts['show_pars'] = False |
---|
1223 | elif arg == '-hist': opts['show_hist'] = True |
---|
1224 | elif arg == '-nohist': opts['show_hist'] = False |
---|
1225 | elif arg == '-rel': opts['rel_err'] = True |
---|
1226 | elif arg == '-abs': opts['rel_err'] = False |
---|
1227 | elif arg == '-half': opts['engine'] = 'half' |
---|
1228 | elif arg == '-fast': opts['engine'] = 'fast' |
---|
1229 | elif arg == '-single': opts['engine'] = 'single' |
---|
1230 | elif arg == '-double': opts['engine'] = 'double' |
---|
1231 | elif arg == '-single!': opts['engine'] = 'single!' |
---|
1232 | elif arg == '-double!': opts['engine'] = 'double!' |
---|
1233 | elif arg == '-quad!': opts['engine'] = 'quad!' |
---|
1234 | elif arg == '-sasview': opts['engine'] = 'sasview' |
---|
1235 | elif arg == '-edit': opts['explore'] = True |
---|
1236 | elif arg == '-demo': opts['use_demo'] = True |
---|
1237 | elif arg == '-default': opts['use_demo'] = False |
---|
1238 | elif arg == '-weights': opts['show_weights'] = True |
---|
1239 | elif arg == '-html': opts['html'] = True |
---|
1240 | elif arg == '-help': opts['html'] = True |
---|
1241 | # pylint: enable=bad-whitespace |
---|
1242 | |
---|
1243 | # Magnetism forces 2D for now |
---|
1244 | if opts['magnetic']: |
---|
1245 | opts['is2d'] = True |
---|
1246 | |
---|
1247 | # Force random if sets is used |
---|
1248 | if opts['sets'] >= 1 and opts['seed'] < 0: |
---|
1249 | opts['seed'] = np.random.randint(1000000) |
---|
1250 | if opts['sets'] == 0: |
---|
1251 | opts['sets'] = 1 |
---|
1252 | |
---|
1253 | # Create the computational engines |
---|
1254 | if opts['qmin'] is None: |
---|
1255 | opts['qmin'] = 0.001*opts['qmax'] |
---|
1256 | if opts['datafile'] is not None: |
---|
1257 | data = load_data(os.path.expanduser(opts['datafile'])) |
---|
1258 | else: |
---|
1259 | data, _ = make_data(opts) |
---|
1260 | |
---|
1261 | comparison = any(PAR_SPLIT in v for v in values) |
---|
1262 | if PAR_SPLIT in name: |
---|
1263 | names = name.split(PAR_SPLIT, 2) |
---|
1264 | comparison = True |
---|
1265 | else: |
---|
1266 | names = [name]*2 |
---|
1267 | try: |
---|
1268 | model_info = [core.load_model_info(k) for k in names] |
---|
1269 | except ImportError as exc: |
---|
1270 | print(str(exc)) |
---|
1271 | print("Could not find model; use one of:\n " + models) |
---|
1272 | return None |
---|
1273 | |
---|
1274 | if PAR_SPLIT in opts['engine']: |
---|
1275 | opts['engine'] = opts['engine'].split(PAR_SPLIT, 2) |
---|
1276 | comparison = True |
---|
1277 | else: |
---|
1278 | opts['engine'] = [opts['engine']]*2 |
---|
1279 | |
---|
1280 | if PAR_SPLIT in opts['count']: |
---|
1281 | opts['count'] = [int(k) for k in opts['count'].split(PAR_SPLIT, 2)] |
---|
1282 | comparison = True |
---|
1283 | else: |
---|
1284 | opts['count'] = [int(opts['count'])]*2 |
---|
1285 | |
---|
1286 | if PAR_SPLIT in opts['cutoff']: |
---|
1287 | opts['cutoff'] = [float(k) for k in opts['cutoff'].split(PAR_SPLIT, 2)] |
---|
1288 | comparison = True |
---|
1289 | else: |
---|
1290 | opts['cutoff'] = [float(opts['cutoff'])]*2 |
---|
1291 | |
---|
1292 | base = make_engine(model_info[0], data, opts['engine'][0], opts['cutoff'][0]) |
---|
1293 | if comparison: |
---|
1294 | comp = make_engine(model_info[1], data, opts['engine'][1], opts['cutoff'][1]) |
---|
1295 | else: |
---|
1296 | comp = None |
---|
1297 | |
---|
1298 | # pylint: disable=bad-whitespace |
---|
1299 | # Remember it all |
---|
1300 | opts.update({ |
---|
1301 | 'data' : data, |
---|
1302 | 'name' : names, |
---|
1303 | 'info' : model_info, |
---|
1304 | 'engines' : [base, comp], |
---|
1305 | 'values' : values, |
---|
1306 | }) |
---|
1307 | # pylint: enable=bad-whitespace |
---|
1308 | |
---|
1309 | # Set the integration parameters to the half sphere |
---|
1310 | if opts['sphere'] > 0: |
---|
1311 | set_spherical_integration_parameters(opts, opts['sphere']) |
---|
1312 | |
---|
1313 | return opts |
---|
1314 | |
---|
1315 | def set_spherical_integration_parameters(opts, steps): |
---|
1316 | """ |
---|
1317 | Set integration parameters for spherical integration over the entire |
---|
1318 | surface in theta-phi coordinates. |
---|
1319 | """ |
---|
1320 | # Set the integration parameters to the half sphere |
---|
1321 | opts['values'].extend([ |
---|
1322 | #'theta=90', |
---|
1323 | 'theta_pd=%g'%(90/np.sqrt(3)), |
---|
1324 | 'theta_pd_n=%d'%steps, |
---|
1325 | 'theta_pd_type=rectangle', |
---|
1326 | #'phi=0', |
---|
1327 | 'phi_pd=%g'%(180/np.sqrt(3)), |
---|
1328 | 'phi_pd_n=%d'%(2*steps), |
---|
1329 | 'phi_pd_type=rectangle', |
---|
1330 | #'background=0', |
---|
1331 | ]) |
---|
1332 | if 'psi' in opts['info'][0].parameters: |
---|
1333 | opts['values'].extend([ |
---|
1334 | #'psi=0', |
---|
1335 | 'psi_pd=%g'%(180/np.sqrt(3)), |
---|
1336 | 'psi_pd_n=%d'%(2*steps), |
---|
1337 | 'psi_pd_type=rectangle', |
---|
1338 | ]) |
---|
1339 | pass |
---|
1340 | |
---|
1341 | def parse_pars(opts, maxdim=np.inf): |
---|
1342 | model_info, model_info2 = opts['info'] |
---|
1343 | |
---|
1344 | # Get demo parameters from model definition, or use default parameters |
---|
1345 | # if model does not define demo parameters |
---|
1346 | pars = get_pars(model_info, opts['use_demo']) |
---|
1347 | pars2 = get_pars(model_info2, opts['use_demo']) |
---|
1348 | pars2.update((k, v) for k, v in pars.items() if k in pars2) |
---|
1349 | # randomize parameters |
---|
1350 | #pars.update(set_pars) # set value before random to control range |
---|
1351 | if opts['seed'] > -1: |
---|
1352 | pars = randomize_pars(model_info, pars) |
---|
1353 | limit_dimensions(model_info, pars, maxdim) |
---|
1354 | if model_info != model_info2: |
---|
1355 | pars2 = randomize_pars(model_info2, pars2) |
---|
1356 | limit_dimensions(model_info, pars2, maxdim) |
---|
1357 | # Share values for parameters with the same name |
---|
1358 | for k, v in pars.items(): |
---|
1359 | if k in pars2: |
---|
1360 | pars2[k] = v |
---|
1361 | else: |
---|
1362 | pars2 = pars.copy() |
---|
1363 | constrain_pars(model_info, pars) |
---|
1364 | constrain_pars(model_info2, pars2) |
---|
1365 | pars = suppress_pd(pars, opts['mono']) |
---|
1366 | pars2 = suppress_pd(pars2, opts['mono']) |
---|
1367 | pars = suppress_magnetism(pars, not opts['magnetic']) |
---|
1368 | pars2 = suppress_magnetism(pars2, not opts['magnetic']) |
---|
1369 | |
---|
1370 | # Fill in parameters given on the command line |
---|
1371 | presets = {} |
---|
1372 | presets2 = {} |
---|
1373 | for arg in opts['values']: |
---|
1374 | k, v = arg.split('=', 1) |
---|
1375 | if k not in pars and k not in pars2: |
---|
1376 | # extract base name without polydispersity info |
---|
1377 | s = set(p.split('_pd')[0] for p in pars) |
---|
1378 | print("%r invalid; parameters are: %s"%(k, ", ".join(sorted(s)))) |
---|
1379 | return None |
---|
1380 | v1, v2 = v.split(PAR_SPLIT, 2) if PAR_SPLIT in v else (v,v) |
---|
1381 | if v1 and k in pars: |
---|
1382 | presets[k] = float(v1) if isnumber(v1) else v1 |
---|
1383 | if v2 and k in pars2: |
---|
1384 | presets2[k] = float(v2) if isnumber(v2) else v2 |
---|
1385 | |
---|
1386 | # If pd given on the command line, default pd_n to 35 |
---|
1387 | for k, v in list(presets.items()): |
---|
1388 | if k.endswith('_pd'): |
---|
1389 | presets.setdefault(k+'_n', 35.) |
---|
1390 | for k, v in list(presets2.items()): |
---|
1391 | if k.endswith('_pd'): |
---|
1392 | presets2.setdefault(k+'_n', 35.) |
---|
1393 | |
---|
1394 | # Evaluate preset parameter expressions |
---|
1395 | context = MATH.copy() |
---|
1396 | context['np'] = np |
---|
1397 | context.update(pars) |
---|
1398 | context.update((k, v) for k, v in presets.items() if isinstance(v, float)) |
---|
1399 | for k, v in presets.items(): |
---|
1400 | if not isinstance(v, float) and not k.endswith('_type'): |
---|
1401 | presets[k] = eval(v, context) |
---|
1402 | context.update(presets) |
---|
1403 | context.update((k, v) for k, v in presets2.items() if isinstance(v, float)) |
---|
1404 | for k, v in presets2.items(): |
---|
1405 | if not isinstance(v, float) and not k.endswith('_type'): |
---|
1406 | presets2[k] = eval(v, context) |
---|
1407 | |
---|
1408 | # update parameters with presets |
---|
1409 | pars.update(presets) # set value after random to control value |
---|
1410 | pars2.update(presets2) # set value after random to control value |
---|
1411 | #import pprint; pprint.pprint(model_info) |
---|
1412 | |
---|
1413 | if opts['show_pars']: |
---|
1414 | if model_info.name != model_info2.name or pars != pars2: |
---|
1415 | print("==== %s ====="%model_info.name) |
---|
1416 | print(str(parlist(model_info, pars, opts['is2d']))) |
---|
1417 | print("==== %s ====="%model_info2.name) |
---|
1418 | print(str(parlist(model_info2, pars2, opts['is2d']))) |
---|
1419 | else: |
---|
1420 | print(str(parlist(model_info, pars, opts['is2d']))) |
---|
1421 | |
---|
1422 | return pars, pars2 |
---|
1423 | |
---|
1424 | def show_docs(opts): |
---|
1425 | # type: (Dict[str, Any]) -> None |
---|
1426 | """ |
---|
1427 | show html docs for the model |
---|
1428 | """ |
---|
1429 | import os |
---|
1430 | from .generate import make_html |
---|
1431 | from . import rst2html |
---|
1432 | |
---|
1433 | info = opts['info'][0] |
---|
1434 | html = make_html(info) |
---|
1435 | path = os.path.dirname(info.filename) |
---|
1436 | url = "file://"+path.replace("\\","/")[2:]+"/" |
---|
1437 | rst2html.view_html_qtapp(html, url) |
---|
1438 | |
---|
1439 | def explore(opts): |
---|
1440 | # type: (Dict[str, Any]) -> None |
---|
1441 | """ |
---|
1442 | explore the model using the bumps gui. |
---|
1443 | """ |
---|
1444 | import wx # type: ignore |
---|
1445 | from bumps.names import FitProblem # type: ignore |
---|
1446 | from bumps.gui.app_frame import AppFrame # type: ignore |
---|
1447 | from bumps.gui import signal |
---|
1448 | |
---|
1449 | is_mac = "cocoa" in wx.version() |
---|
1450 | # Create an app if not running embedded |
---|
1451 | app = wx.App() if wx.GetApp() is None else None |
---|
1452 | model = Explore(opts) |
---|
1453 | problem = FitProblem(model) |
---|
1454 | frame = AppFrame(parent=None, title="explore", size=(1000, 700)) |
---|
1455 | if not is_mac: |
---|
1456 | frame.Show() |
---|
1457 | frame.panel.set_model(model=problem) |
---|
1458 | frame.panel.Layout() |
---|
1459 | frame.panel.aui.Split(0, wx.TOP) |
---|
1460 | def reset_parameters(event): |
---|
1461 | model.revert_values() |
---|
1462 | signal.update_parameters(problem) |
---|
1463 | frame.Bind(wx.EVT_TOOL, reset_parameters, frame.ToolBar.GetToolByPos(1)) |
---|
1464 | if is_mac: frame.Show() |
---|
1465 | # If running withing an app, start the main loop |
---|
1466 | if app: |
---|
1467 | app.MainLoop() |
---|
1468 | |
---|
1469 | class Explore(object): |
---|
1470 | """ |
---|
1471 | Bumps wrapper for a SAS model comparison. |
---|
1472 | |
---|
1473 | The resulting object can be used as a Bumps fit problem so that |
---|
1474 | parameters can be adjusted in the GUI, with plots updated on the fly. |
---|
1475 | """ |
---|
1476 | def __init__(self, opts): |
---|
1477 | # type: (Dict[str, Any]) -> None |
---|
1478 | from bumps.cli import config_matplotlib # type: ignore |
---|
1479 | from . import bumps_model |
---|
1480 | config_matplotlib() |
---|
1481 | self.opts = opts |
---|
1482 | opts['pars'] = list(opts['pars']) |
---|
1483 | p1, p2 = opts['pars'] |
---|
1484 | m1, m2 = opts['info'] |
---|
1485 | self.fix_p2 = m1 != m2 or p1 != p2 |
---|
1486 | model_info = m1 |
---|
1487 | pars, pd_types = bumps_model.create_parameters(model_info, **p1) |
---|
1488 | # Initialize parameter ranges, fixing the 2D parameters for 1D data. |
---|
1489 | if not opts['is2d']: |
---|
1490 | for p in model_info.parameters.user_parameters({}, is2d=False): |
---|
1491 | for ext in ['', '_pd', '_pd_n', '_pd_nsigma']: |
---|
1492 | k = p.name+ext |
---|
1493 | v = pars.get(k, None) |
---|
1494 | if v is not None: |
---|
1495 | v.range(*parameter_range(k, v.value)) |
---|
1496 | else: |
---|
1497 | for k, v in pars.items(): |
---|
1498 | v.range(*parameter_range(k, v.value)) |
---|
1499 | |
---|
1500 | self.pars = pars |
---|
1501 | self.starting_values = dict((k, v.value) for k, v in pars.items()) |
---|
1502 | self.pd_types = pd_types |
---|
1503 | self.limits = None |
---|
1504 | |
---|
1505 | def revert_values(self): |
---|
1506 | for k, v in self.starting_values.items(): |
---|
1507 | self.pars[k].value = v |
---|
1508 | |
---|
1509 | def model_update(self): |
---|
1510 | pass |
---|
1511 | |
---|
1512 | def numpoints(self): |
---|
1513 | # type: () -> int |
---|
1514 | """ |
---|
1515 | Return the number of points. |
---|
1516 | """ |
---|
1517 | return len(self.pars) + 1 # so dof is 1 |
---|
1518 | |
---|
1519 | def parameters(self): |
---|
1520 | # type: () -> Any # Dict/List hierarchy of parameters |
---|
1521 | """ |
---|
1522 | Return a dictionary of parameters. |
---|
1523 | """ |
---|
1524 | return self.pars |
---|
1525 | |
---|
1526 | def nllf(self): |
---|
1527 | # type: () -> float |
---|
1528 | """ |
---|
1529 | Return cost. |
---|
1530 | """ |
---|
1531 | # pylint: disable=no-self-use |
---|
1532 | return 0. # No nllf |
---|
1533 | |
---|
1534 | def plot(self, view='log'): |
---|
1535 | # type: (str) -> None |
---|
1536 | """ |
---|
1537 | Plot the data and residuals. |
---|
1538 | """ |
---|
1539 | pars = dict((k, v.value) for k, v in self.pars.items()) |
---|
1540 | pars.update(self.pd_types) |
---|
1541 | self.opts['pars'][0] = pars |
---|
1542 | if not self.fix_p2: |
---|
1543 | self.opts['pars'][1] = pars |
---|
1544 | result = run_models(self.opts) |
---|
1545 | limits = plot_models(self.opts, result, limits=self.limits) |
---|
1546 | if self.limits is None: |
---|
1547 | vmin, vmax = limits |
---|
1548 | self.limits = vmax*1e-7, 1.3*vmax |
---|
1549 | import pylab; pylab.clf() |
---|
1550 | plot_models(self.opts, result, limits=self.limits) |
---|
1551 | |
---|
1552 | |
---|
1553 | def main(*argv): |
---|
1554 | # type: (*str) -> None |
---|
1555 | """ |
---|
1556 | Main program. |
---|
1557 | """ |
---|
1558 | opts = parse_opts(argv) |
---|
1559 | if opts is not None: |
---|
1560 | if opts['seed'] > -1: |
---|
1561 | print("Randomize using -random=%i"%opts['seed']) |
---|
1562 | np.random.seed(opts['seed']) |
---|
1563 | if opts['html']: |
---|
1564 | show_docs(opts) |
---|
1565 | elif opts['explore']: |
---|
1566 | opts['pars'] = parse_pars(opts) |
---|
1567 | if opts['pars'] is None: |
---|
1568 | return |
---|
1569 | explore(opts) |
---|
1570 | else: |
---|
1571 | compare(opts) |
---|
1572 | |
---|
1573 | if __name__ == "__main__": |
---|
1574 | main(*sys.argv[1:]) |
---|