repo_name
stringlengths
7
65
path
stringlengths
5
185
copies
stringlengths
1
4
size
stringlengths
4
6
content
stringlengths
977
990k
license
stringclasses
14 values
hash
stringlengths
32
32
line_mean
float64
7.18
99.4
line_max
int64
31
999
alpha_frac
float64
0.25
0.95
ratio
float64
1.5
7.84
autogenerated
bool
1 class
config_or_test
bool
2 classes
has_no_keywords
bool
2 classes
has_few_assignments
bool
1 class
sergiocorreia/panflute
panflute/base.py
1
8673
""" Base classes and methods of all Pandoc elements """ # --------------------------- # Imports # --------------------------- from operator import attrgetter from collections.abc import MutableSequence, MutableMapping from .containers import ListContainer, DictContainer from .utils import check_type, encode_dict # check_group # --------------------------- # Meta Classes # --------------------------- class Element(object): """ Base class of all Pandoc elements """ __slots__ = ['parent', 'location', 'index'] _children = [] def __new__(cls, *args, **kwargs): # This is just to initialize self.parent to None element = object.__new__(cls) element.parent = None element.location = None element.index = None return element @property def tag(self): tag = type(self).__name__ return tag def __eq__(self, other): # Doc has a different method b/c it uses __dict__ instead of slots if type(self) != type(other): return False for key in self.__slots__: if getattr(self, key) != getattr(other, key): return False return True # --------------------------- # Base methods # --------------------------- # Should be overridden except for trivial elements (Space, Null, etc.) def __repr__(self): # This is just a convenience method # Override it for more complex elements extra = [] for key in self.__slots__: if not key.startswith('_') and key != 'text': val = getattr(self, key) if val not in ([], dict(), ''): extra.append([key, val]) if extra: extra = ('{}={}'.format(k, repr(v)) for k, v in extra) extra = '; ' + ', '.join(x for x in extra) else: extra = '' if '_content' in self.__slots__: content = ' '.join(repr(x) for x in self.content) return '{}({}{})'.format(self.tag, content, extra) elif 'text' in self.__slots__: return '{}({}{})'.format(self.tag, self.text, extra) else: return self.tag def to_json(self): return encode_dict(self.tag, self._slots_to_json()) def _slots_to_json(self): # Default when the element contains nothing return [] # --------------------------- # .identifier .classes .attributes # --------------------------- def _set_ica(self, identifier, classes, attributes): self.identifier = check_type(identifier, str) self.classes = [check_type(cl, str) for cl in classes] self.attributes = dict(attributes) def _ica_to_json(self): return [self.identifier, self.classes, list(self.attributes.items())] # --------------------------- # .content (setter and getter) # --------------------------- @property def content(self): """ Sequence of :class:`Element` objects (usually either :class:`Block` or :class:`Inline`) that are "children" of the current element. Only available for elements that accept ``*args``. Note: some elements have children in attributes other than ``content`` (such as :class:`.Table` that has children in the header and caption attributes). """ return self._content @content.setter def content(self, value): oktypes = self._content.oktypes value = value.list if isinstance(value, ListContainer) else list(value) self._content = ListContainer(*value, oktypes=oktypes, parent=self) def _set_content(self, value, oktypes): """ Similar to content.setter but when there are no existing oktypes """ if value is None: value = [] self._content = ListContainer(*value, oktypes=oktypes, parent=self) # --------------------------- # Navigation # --------------------------- @property def container(self): """ Rarely used attribute that returns the ``ListContainer`` or ``DictContainer`` that contains the element (or returns None if no such container exist) :rtype: ``ListContainer`` | ``DictContainer`` | ``None`` """ if self.parent is None: return None elif self.location is None: return self.parent.content else: container = getattr(self.parent, self.location) if isinstance(container, (ListContainer, DictContainer)): return container else: assert self is container # id(self) == id(container) def offset(self, n): """ Return a sibling element offset by n :rtype: :class:`Element` | ``None`` """ idx = self.index if idx is not None: sibling = idx + n container = self.container if 0 <= sibling < len(container): return container[sibling] @property def next(self): """ Return the next sibling. Note that ``elem.offset(1) == elem.next`` :rtype: :class:`Element` | ``None`` """ return self.offset(1) @property def prev(self): """ Return the previous sibling. Note that ``elem.offset(-1) == elem.prev`` :rtype: :class:`Element` | ``None`` """ return self.offset(-1) def ancestor(self, n): """ Return the n-th ancestor. Note that ``elem.ancestor(1) == elem.parent`` :rtype: :class:`Element` | ``None`` """ if not isinstance(n, int) or n < 1: raise TypeError('Ancestor needs to be positive, received', n) if n == 1 or self.parent is None: return self.parent else: return self.parent.ancestor(n-1) # --------------------------- # Walking # --------------------------- @property def doc(self): """ Return the root Doc element (if there is one) """ guess = self while guess is not None and guess.tag != 'Doc': guess = guess.parent # If no parent, this will be None return guess # Returns either Doc or None def walk(self, action, doc=None, stop_if=None): """ Walk through the element and all its children (sub-elements), applying the provided function ``action``. A trivial example would be: .. code-block:: python from panflute import * def no_action(elem, doc): pass doc = Doc(Para(Str('a'))) altered = doc.walk(no_action) :param action: function that takes (element, doc) as arguments. :type action: :class:`function` :param doc: root document; used to access metadata, the output format (in ``.format``, other elements, and other variables). Only use this variable if for some reason you don't want to use the current document of an element. :type doc: :class:`.Doc` :param stop_if: function that takes (element) as argument. :type stop_if: :class:`function`, optional :rtype: :class:`Element` | ``[]`` | ``None`` """ # Infer the document thanks to .parent magic if doc is None: doc = self.doc # First iterate over children; unless the stop condition is met if stop_if is None or not stop_if(self): # self._children has property *names* so we need a bit of getattr/setattr magic to modify the objects themselves children = ((child_name, getattr(self, child_name)) for child_name in self._children) for child_name, child in children: if isinstance(child, (Element, ListContainer, DictContainer)): child = child.walk(action, doc, stop_if) elif child is None: child = None # Empty table headers or captions else: raise TypeError(type(child)) setattr(self, child_name, child) # Then apply the action() to the root element altered = action(self, doc) return self if altered is None else altered class Inline(Element): """ Base class of all inline elements """ __slots__ = [] class Block(Element): """ Base class of all block elements """ __slots__ = [] class MetaValue(Element): """ Base class of all metadata elements """ __slots__ = []
bsd-3-clause
88d233e663f8adbe629cade678cffd3e
28.5
124
0.531189
4.484488
false
false
false
false
morepath/morepath
morepath/tests/test_predicates.py
1
1348
from morepath.app import App from webtest import TestApp as Client import morepath from reg import KeyIndex def test_view_predicates(): class app(App): pass @app.path(path="") class Root: pass @app.view(model=Root, name="foo", request_method="GET") def get(self, request): return "GET" @app.view(model=Root, name="foo", request_method="POST") def post(self, request): return "POST" c = Client(app()) response = c.get("/foo") assert response.body == b"GET" response = c.post("/foo") assert response.body == b"POST" def test_extra_predicates(): class app(App): pass @app.path(path="{id}") class Model: def __init__(self, id): self.id = id @app.view(model=Model, name="foo", id="a") def get_a(self, request): return "a" @app.view(model=Model, name="foo", id="b") def get_b(self, request): return "b" @app.predicate( morepath.App.get_view, name="id", default="", index=KeyIndex, after=morepath.request_method_predicate, ) def id_predicate(self, obj, request): return obj.id c = Client(app()) response = c.get("/a/foo") assert response.body == b"a" response = c.get("/b/foo") assert response.body == b"b"
bsd-3-clause
09d0ef727a506341014df61aaa81d039
20.0625
60
0.568991
3.42132
false
true
false
false
morepath/morepath
morepath/predicate.py
1
4328
""" The :meth:`morepath.App.predicate` directive lets you install predicates for function that use :func:`reg.dispatch_method`. This is used by :mod:`morepath.core` to install the view predicates, and you can also use it for your own functions. This implements the functionality that drives Reg to install these predicates. See also :class:`morepath.directive.PredicateRegistry` """ from reg import Predicate from .toposort import toposorted, Info from collections import defaultdict class PredicateRegistry: """A registry of what predicates are registered for which functions. It also keeps track of how predicates are to be ordered. """ app_class_arg = True def __init__(self, app_class): self.app_class = app_class self._predicate_infos = defaultdict(list) self._predicate_fallbacks = defaultdict(dict) def register_predicate( self, func, dispatch, name, default, index, before, after ): """Register a predicate for installation into the reg registry. See :meth:`morepath.App.predicate` for details. :param func: the function that implements the predicate. :param dispatch: the dispatch function to register the predicate on. :param name: name of the predicate. :param default: default value. :param index: index to use. :param before: predicate function to have priority over. :param after: predicate function that has priority over this one. """ info = PredicateInfo(func, name, default, index, before, after) self._predicate_infos[dispatch].append(info) def register_predicate_fallback(self, dispatch, func, fallback_func): """Register a predicate fallback for installation into reg registry. See :meth:`morepath.App.predicate_fallback` for details. :param dispatch: the dispatch function to register fallback on. :param func: the predicate function to register fallback for. :param fallback_func: the fallback function. """ self._predicate_fallbacks[dispatch][func] = fallback_func def install_predicates(self): """Install the predicates with reg. This should be called during configuration once all predicates and fallbacks are known. Uses :meth:`PredicateRegistry.get_predicates` to get out the predicates in the correct order. """ for dispatch in self._predicate_infos.keys(): getattr(self.app_class, dispatch.__name__).add_predicates( self.get_predicates(dispatch) ) def get_predicates(self, dispatch): """Create Reg predicates. This creates :class:`reg.Predicate` objects for a particular dispatch function. Uses :meth:`PredicateRegistry.sorted_predicate_infos` to sort the predicate infos. :param dispatch: the dispatch function to create the predicates for. :return: a list of :class:`reg.Predicate` instances in the correct order. """ infos = self.sorted_predicate_infos(dispatch) result = [] for info in infos: fallback = self._predicate_fallbacks[dispatch].get(info.func) predicate = Predicate( info.name, info.index, adapt(info.func), fallback=fallback, default=info.default, ) result.append(predicate) return result def sorted_predicate_infos(self, dispatch): """Topologically sort predicate infos for a dispatch function. :param dispatch: the dispatch function to sort for. :return: a list of sorted :class:`PredicateInfo` instances. """ return toposorted(self._predicate_infos[dispatch]) def adapt(func): def wrapper(d): return func(**d) return wrapper class PredicateInfo(Info): """Used by :class:`PredicateRegistry` internally. Is used to store registration information on a predicate before it is registered with Reg. """ def __init__(self, func, name, default, index, before, after): super().__init__(func, before, after) self.func = func self.name = name self.default = default self.index = index
bsd-3-clause
0e9cfceb537cb828fcec49f6f77defd2
32.292308
76
0.65134
4.536688
false
false
false
false
scikit-hep/root_numpy
examples/core/plot_bootstrap.py
3
2604
#!/usr/bin/env python """ ============================ Bootstrap a TTree with NumPy ============================ This example demonstrates how to sample entries in a TTree with replacement with the help of NumPy and root_numpy. This example depends on `rootpy <http://www.rootpy.org/>`_ which can be installed with pip:: pip install --user rootpy """ from rootpy.extern.six.moves import range from rootpy.tree import Tree, TreeModel, FloatCol from rootpy.plotting import Canvas, Hist2D, set_style from rootpy.io import root_open from root_numpy import root2array, array2tree, rec2array, fill_hist import ROOT import numpy as np from random import gauss import random import os ROOT.gROOT.SetBatch() set_style('ATLAS') np.random.seed(0) random.seed(0) # create an example TTree dataset class Sample(TreeModel): x = FloatCol() y = FloatCol() with root_open('sample.root', 'recreate'): # generate toy data in a TTree tree = Tree('sample', model=Sample) for i in range(500): tree.x = gauss(0, 1) tree.y = gauss(0, 1) tree.Fill() tree.write() # read in the TTree as a NumPy array array = root2array('sample.root', 'sample') if os.path.exists('bootstrap.gif'): os.remove('bootstrap.gif') # Canvas name is set here to aid the automatic documentation generation # It needs to take the GIF already saved instead of saving a png of the last # frame. canvas = Canvas(width=500, height=400, name='bootstrap.gif') hist = Hist2D(10, -3, 3, 10, -3, 3, drawstyle='LEGO2') output = root_open('bootstrap.root', 'recreate') # bootstrap 10 times for bootstrap_idx in range(10): # sample with replacement # http://docs.scipy.org/doc/numpy-dev/reference/generated/numpy.random.choice.html sample_idx = np.random.choice(len(array), size=len(array), replace=True) array_bootstrapped = array[sample_idx] # convert back to a TTree and write it out tree_bootstrapped = array2tree( array_bootstrapped, name='bootstrap_{0}'.format(bootstrap_idx)) tree_bootstrapped.Write() tree_bootstrapped.Delete() # fill the ROOT histogram with the numpy array hist.Reset() fill_hist(hist, rec2array(array_bootstrapped)) hist.Draw() hist.xaxis.title = 'x' hist.yaxis.title = 'y' hist.zaxis.title = 'Events' hist.xaxis.limits = (-2.5, 2.5) hist.yaxis.limits = (-2.5, 2.5) hist.zaxis.range_user = (0, 30) hist.xaxis.divisions = 5 hist.yaxis.divisions = 5 hist.zaxis.divisions = 5 canvas.Print('bootstrap.gif+50') # loop the gif canvas.Print('bootstrap.gif++') output.Close()
bsd-3-clause
2826b7b72f7be9df63d2e2c4e0d7e51c
28.258427
86
0.679339
3.304569
false
false
false
false
flutter/buildroot
build/linux/unbundle/replace_gyp_files.py
1
2938
#!/usr/bin/env python3 # # Copyright 2013 The Chromium Authors. All rights reserved. # Use of this source code is governed by a BSD-style license that can be # found in the LICENSE file. """ Replaces gyp files in tree with files from here that make the build use system libraries. """ import optparse import os.path import shutil import sys REPLACEMENTS = { 'use_system_expat': 'third_party/expat/expat.gyp', 'use_system_ffmpeg': 'third_party/ffmpeg/ffmpeg.gyp', 'use_system_flac': 'third_party/flac/flac.gyp', 'use_system_harfbuzz': 'third_party/harfbuzz-ng/harfbuzz.gyp', 'use_system_icu': 'third_party/icu/icu.gyp', 'use_system_jsoncpp': 'third_party/jsoncpp/jsoncpp.gyp', 'use_system_libevent': 'third_party/libevent/libevent.gyp', 'use_system_libjpeg': 'third_party/libjpeg/libjpeg.gyp', 'use_system_libpng': 'third_party/libpng/libpng.gyp', 'use_system_libusb': 'third_party/libusb/libusb.gyp', 'use_system_libvpx': 'third_party/libvpx/libvpx.gyp', 'use_system_libwebp': 'third_party/libwebp/libwebp.gyp', 'use_system_libxml': 'third_party/libxml/libxml.gyp', 'use_system_libxnvctrl' : 'third_party/libXNVCtrl/libXNVCtrl.gyp', 'use_system_libxslt': 'third_party/libxslt/libxslt.gyp', 'use_system_opus': 'third_party/opus/opus.gyp', 'use_system_protobuf': 'third_party/protobuf/protobuf.gyp', 'use_system_re2': 'third_party/re2/re2.gyp', 'use_system_snappy': 'third_party/snappy/snappy.gyp', 'use_system_speex': 'third_party/speex/speex.gyp', 'use_system_sqlite': 'third_party/sqlite/sqlite.gyp', 'use_system_v8': 'v8/tools/gyp/v8.gyp', 'use_system_zlib': 'third_party/zlib/zlib.gyp', } def DoMain(argv): my_dirname = os.path.dirname(__file__) source_tree_root = os.path.abspath( os.path.join(my_dirname, '..', '..', '..')) parser = optparse.OptionParser() # Accept arguments in gyp command-line syntax, so that the caller can re-use # command-line for this script and gyp. parser.add_option('-D', dest='defines', action='append') parser.add_option('--undo', action='store_true') options, args = parser.parse_args(argv) for flag, path in list(REPLACEMENTS.items()): if '%s=1' % flag not in options.defines: continue if options.undo: # Restore original file, and also remove the backup. # This is meant to restore the source tree to its original state. os.rename(os.path.join(source_tree_root, path + '.orig'), os.path.join(source_tree_root, path)) else: # Create a backup copy for --undo. shutil.copyfile(os.path.join(source_tree_root, path), os.path.join(source_tree_root, path + '.orig')) # Copy the gyp file from directory of this script to target path. shutil.copyfile(os.path.join(my_dirname, os.path.basename(path)), os.path.join(source_tree_root, path)) return 0 if __name__ == '__main__': sys.exit(DoMain(sys.argv))
bsd-3-clause
bf8368a7feb743986b00e72b3402d021
34.39759
78
0.679374
3.007165
false
false
false
false
flutter/buildroot
build/fuchsia/fidl_gen_cpp.py
1
3070
#!/usr/bin/env python3 # # Copyright 2013 The Flutter Authors. All rights reserved. # Use of this source code is governed by a BSD-style license that can be # found in the LICENSE file. """ Generate C/C++ headers and source files from the set of FIDL files specified in the meta.json manifest. """ import argparse import collections import json import os import subprocess import sys def GetFIDLFilesRecursive(libraries, sdk_base, path): with open(path) as json_file: parsed = json.load(json_file) result = [] deps = parsed['deps'] for dep in deps: dep_meta_json = os.path.abspath('%s/fidl/%s/meta.json' % (sdk_base, dep)) GetFIDLFilesRecursive(libraries, sdk_base, dep_meta_json) libraries[parsed['name']] = result + parsed['sources'] def GetFIDLFilesByLibraryName(sdk_base, root): libraries = collections.OrderedDict() GetFIDLFilesRecursive(libraries, sdk_base, root) return libraries def main(): parser = argparse.ArgumentParser(); parser.add_argument('--fidlc-bin', dest='fidlc_bin', action='store', required=True) parser.add_argument('--fidlgen-bin', dest='fidlgen_bin', action='store', required=False) parser.add_argument('--sdk-base', dest='sdk_base', action='store', required=True) parser.add_argument('--root', dest='root', action='store', required=True) parser.add_argument('--json', dest='json', action='store', required=True) parser.add_argument('--fidlgen-output-root', dest='fidlgen_output_root', action='store', required=False) parser.add_argument('--output-c-tables', dest='output_c_tables', action='store', required=True) parser.add_argument('--target-api-level', dest='target_api_level', action='store', required=False) args = parser.parse_args() assert os.path.exists(args.fidlc_bin) # --fidlgen-bin and --fidlgen-output-root should be passed in together. assert os.path.exists(args.fidlgen_bin or '') == bool(args.fidlgen_output_root) fidl_files_by_name = GetFIDLFilesByLibraryName(args.sdk_base, args.root) fidlc_command = [ args.fidlc_bin, '--tables', args.output_c_tables, '--json', args.json ] if args.target_api_level: fidlc_command += [ '--available', 'fuchsia:{api_level}'.format(api_level=args.target_api_level), ] # Create an iterator that works on both python3 and python2 try: fidl_files_by_name_iter = list(fidl_files_by_name.items()) except AttributeError: fidl_files_by_name_iter = iter(fidl_files_by_name.items()) for _, fidl_files in fidl_files_by_name_iter: fidlc_command.append('--files') for fidl_file in fidl_files: fidl_abspath = os.path.abspath('%s/%s' % (args.sdk_base, fidl_file)) fidlc_command.append(fidl_abspath) subprocess.check_call(fidlc_command) if args.fidlgen_output_root: assert os.path.exists(args.json) fidlgen_command = [ args.fidlgen_bin, '-json', args.json, '-root', args.fidlgen_output_root ] subprocess.check_call(fidlgen_command) return 0 if __name__ == '__main__': sys.exit(main())
bsd-3-clause
fcdbdce2c158a6d4681ff25757012829
30.326531
106
0.685016
3.234984
false
false
false
false
flutter/buildroot
build/symlink.py
1
1176
#!/usr/bin/env python3 # Copyright (c) 2013 The Chromium Authors. All rights reserved. # Use of this source code is governed by a BSD-style license that can be # found in the LICENSE file. """Make a symlink and optionally touch a file (to handle dependencies).""" import errno import optparse import os.path import shutil import sys def Main(argv): parser = optparse.OptionParser() parser.add_option('-f', '--force', action='store_true') parser.add_option('--touch') options, args = parser.parse_args(argv[1:]) if len(args) < 2: parser.error('at least two arguments required.') target = args[-1] sources = args[:-1] for s in sources: t = os.path.join(target, os.path.basename(s)) if len(sources) == 1 and not os.path.isdir(target): t = target try: os.symlink(s, t) except OSError as e: if e.errno == errno.EEXIST and options.force: if os.path.isdir(t): shutil.rmtree(t, ignore_errors=True) else: os.remove(t) os.symlink(s, t) else: raise if options.touch: with open(options.touch, 'w') as f: pass if __name__ == '__main__': sys.exit(Main(sys.argv))
bsd-3-clause
247e7a324be9553912698a19e923e242
28.4
74
0.636054
3.458824
false
false
false
false
flutter/buildroot
build/get_syzygy_binaries.py
1
18089
#!/usr/bin/env python3 # # Copyright 2014 The Chromium Authors. All rights reserved. # Use of this source code is governed by a BSD-style license that can be # found in the LICENSE file. """A utility script for downloading versioned Syzygy binaries.""" import hashlib import errno import json import logging import optparse import os import re import shutil import stat import sys import subprocess import tempfile import time import zipfile _LOGGER = logging.getLogger(os.path.basename(__file__)) # The relative path where official builds are archived in their GS bucket. _SYZYGY_ARCHIVE_PATH = ('/builds/official/%(revision)s') # A JSON file containing the state of the download directory. If this file and # directory state do not agree, then the binaries will be downloaded and # installed again. _STATE = '.state' # This matches an integer (an SVN revision number) or a SHA1 value (a GIT hash). # The archive exclusively uses lowercase GIT hashes. _REVISION_RE = re.compile('^(?:\d+|[a-f0-9]{40})$') # This matches an MD5 hash. _MD5_RE = re.compile('^[a-f0-9]{32}$') # List of reources to be downloaded and installed. These are tuples with the # following format: # (basename, logging name, relative installation path, extraction filter) _RESOURCES = [ ('benchmark.zip', 'benchmark', '', None), ('binaries.zip', 'binaries', 'exe', None), ('symbols.zip', 'symbols', 'exe', lambda x: x.filename.endswith('.dll.pdb'))] def _LoadState(output_dir): """Loads the contents of the state file for a given |output_dir|, returning None if it doesn't exist. """ path = os.path.join(output_dir, _STATE) if not os.path.exists(path): _LOGGER.debug('No state file found.') return None with open(path, 'rb') as f: _LOGGER.debug('Reading state file: %s', path) try: return json.load(f) except ValueError: _LOGGER.debug('Invalid state file.') return None def _SaveState(output_dir, state, dry_run=False): """Saves the |state| dictionary to the given |output_dir| as a JSON file.""" path = os.path.join(output_dir, _STATE) _LOGGER.debug('Writing state file: %s', path) if dry_run: return with open(path, 'wb') as f: f.write(json.dumps(state, sort_keys=True, indent=2)) def _Md5(path): """Returns the MD5 hash of the file at |path|, which must exist.""" return hashlib.md5(open(path, 'rb').read()).hexdigest() def _StateIsValid(state): """Returns true if the given state structure is valid.""" if not isinstance(state, dict): _LOGGER.debug('State must be a dict.') return False r = state.get('revision', None) if not isinstance(r, str) or not _REVISION_RE.match(r): _LOGGER.debug('State contains an invalid revision.') return False c = state.get('contents', None) if not isinstance(c, dict): _LOGGER.debug('State must contain a contents dict.') return False for (relpath, md5) in c.items(): if not isinstance(relpath, str) or len(relpath) == 0: _LOGGER.debug('State contents dict contains an invalid path.') return False if not isinstance(md5, str) or not _MD5_RE.match(md5): _LOGGER.debug('State contents dict contains an invalid MD5 digest.') return False return True def _BuildActualState(stored, revision, output_dir): """Builds the actual state using the provided |stored| state as a template. Only examines files listed in the stored state, causing the script to ignore files that have been added to the directories locally. |stored| must be a valid state dictionary. """ contents = {} state = { 'revision': revision, 'contents': contents } for relpath, md5 in stored['contents'].items(): abspath = os.path.abspath(os.path.join(output_dir, relpath)) if os.path.isfile(abspath): m = _Md5(abspath) contents[relpath] = m return state def _StatesAreConsistent(stored, actual): """Validates whether two state dictionaries are consistent. Both must be valid state dictionaries. Additional entries in |actual| are ignored. """ if stored['revision'] != actual['revision']: _LOGGER.debug('Mismatched revision number.') return False cont_stored = stored['contents'] cont_actual = actual['contents'] for relpath, md5 in cont_stored.items(): if relpath not in cont_actual: _LOGGER.debug('Missing content: %s', relpath) return False if md5 != cont_actual[relpath]: _LOGGER.debug('Modified content: %s', relpath) return False return True def _GetCurrentState(revision, output_dir): """Loads the current state and checks to see if it is consistent. Returns a tuple (state, bool). The returned state will always be valid, even if an invalid state is present on disk. """ stored = _LoadState(output_dir) if not _StateIsValid(stored): _LOGGER.debug('State is invalid.') # Return a valid but empty state. return ({'revision': '0', 'contents': {}}, False) actual = _BuildActualState(stored, revision, output_dir) # If the script has been modified consider the state invalid. path = os.path.join(output_dir, _STATE) if os.path.getmtime(__file__) > os.path.getmtime(path): return (stored, False) # Otherwise, explicitly validate the state. if not _StatesAreConsistent(stored, actual): return (stored, False) return (stored, True) def _DirIsEmpty(path): """Returns true if the given directory is empty, false otherwise.""" for root, dirs, files in os.walk(path): return not dirs and not files def _RmTreeHandleReadOnly(func, path, exc): """An error handling function for use with shutil.rmtree. This will detect failures to remove read-only files, and will change their properties prior to removing them. This is necessary on Windows as os.remove will return an access error for read-only files, and git repos contain read-only pack/index files. """ excvalue = exc[1] if func in (os.rmdir, os.remove) and excvalue.errno == errno.EACCES: _LOGGER.debug('Removing read-only path: %s', path) os.chmod(path, stat.S_IRWXU | stat.S_IRWXG | stat.S_IRWXO) func(path) else: raise def _RmTree(path): """A wrapper of shutil.rmtree that handles read-only files.""" shutil.rmtree(path, ignore_errors=False, onerror=_RmTreeHandleReadOnly) def _CleanState(output_dir, state, dry_run=False): """Cleans up files/directories in |output_dir| that are referenced by the given |state|. Raises an error if there are local changes. Returns a dictionary of files that were deleted. """ _LOGGER.debug('Deleting files from previous installation.') deleted = {} # Generate a list of files to delete, relative to |output_dir|. contents = state['contents'] files = sorted(contents.keys()) # Try to delete the files. Keep track of directories to delete as well. dirs = {} for relpath in files: fullpath = os.path.join(output_dir, relpath) fulldir = os.path.dirname(fullpath) dirs[fulldir] = True if os.path.exists(fullpath): # If somehow the file has become a directory complain about it. if os.path.isdir(fullpath): raise Exception('Directory exists where file expected: %s' % fullpath) # Double check that the file doesn't have local changes. If it does # then refuse to delete it. if relpath in contents: stored_md5 = contents[relpath] actual_md5 = _Md5(fullpath) if actual_md5 != stored_md5: raise Exception('File has local changes: %s' % fullpath) # The file is unchanged so it can safely be deleted. _LOGGER.debug('Deleting file "%s".', fullpath) deleted[relpath] = True if not dry_run: os.unlink(fullpath) # Sort directories from longest name to shortest. This lets us remove empty # directories from the most nested paths first. dirs = sorted(list(dirs.keys()), key=lambda x: len(x), reverse=True) for p in dirs: if os.path.exists(p) and _DirIsEmpty(p): _LOGGER.debug('Deleting empty directory "%s".', p) if not dry_run: _RmTree(p) return deleted def _FindGsUtil(): """Looks for depot_tools and returns the absolute path to gsutil.py.""" for path in os.environ['PATH'].split(os.pathsep): path = os.path.abspath(path) git_cl = os.path.join(path, 'git_cl.py') gs_util = os.path.join(path, 'gsutil.py') if os.path.exists(git_cl) and os.path.exists(gs_util): return gs_util return None def _GsUtil(*cmd): """Runs the given command in gsutil with exponential backoff and retries.""" gs_util = _FindGsUtil() cmd = [sys.executable, gs_util] + list(cmd) retries = 3 timeout = 4 # Seconds. while True: _LOGGER.debug('Running %s', cmd) prog = subprocess.Popen(cmd, shell=False) prog.communicate() # Stop retrying on success. if prog.returncode == 0: return # Raise a permanent failure if retries have been exhausted. if retries == 0: raise RuntimeError('Command "%s" returned %d.' % (cmd, prog.returncode)) _LOGGER.debug('Sleeping %d seconds and trying again.', timeout) time.sleep(timeout) retries -= 1 timeout *= 2 def _Download(resource): """Downloads the given GS resource to a temporary file, returning its path.""" tmp = tempfile.mkstemp(suffix='syzygy_archive') os.close(tmp[0]) url = 'gs://syzygy-archive' + resource _GsUtil('cp', url, tmp[1]) return tmp[1] def _InstallBinaries(options, deleted={}): """Installs Syzygy binaries. This assumes that the output directory has already been cleaned, as it will refuse to overwrite existing files.""" contents = {} state = { 'revision': options.revision, 'contents': contents } archive_path = _SYZYGY_ARCHIVE_PATH % { 'revision': options.revision } if options.resources: resources = [(resource, resource, '', None) for resource in options.resources] else: resources = _RESOURCES for (base, name, subdir, filt) in resources: # Create the output directory if it doesn't exist. fulldir = os.path.join(options.output_dir, subdir) if os.path.isfile(fulldir): raise Exception('File exists where a directory needs to be created: %s' % fulldir) if not os.path.exists(fulldir): _LOGGER.debug('Creating directory: %s', fulldir) if not options.dry_run: os.makedirs(fulldir) # Download and read the archive. resource = archive_path + '/' + base _LOGGER.debug('Retrieving %s archive at "%s".', name, resource) path = _Download(resource) _LOGGER.debug('Unzipping %s archive.', name) with open(path, 'rb') as data: archive = zipfile.ZipFile(data) for entry in archive.infolist(): if not filt or filt(entry): fullpath = os.path.normpath(os.path.join(fulldir, entry.filename)) relpath = os.path.relpath(fullpath, options.output_dir) if os.path.exists(fullpath): # If in a dry-run take into account the fact that the file *would* # have been deleted. if options.dry_run and relpath in deleted: pass else: raise Exception('Path already exists: %s' % fullpath) # Extract the file and update the state dictionary. _LOGGER.debug('Extracting "%s".', fullpath) if not options.dry_run: archive.extract(entry.filename, fulldir) md5 = _Md5(fullpath) contents[relpath] = md5 if sys.platform == 'cygwin': os.chmod(fullpath, os.stat(fullpath).st_mode | stat.S_IXUSR) _LOGGER.debug('Removing temporary file "%s".', path) os.remove(path) return state def _ParseCommandLine(): """Parses the command-line and returns an options structure.""" option_parser = optparse.OptionParser() option_parser.add_option('--dry-run', action='store_true', default=False, help='If true then will simply list actions that would be performed.') option_parser.add_option('--force', action='store_true', default=False, help='Force an installation even if the binaries are up to date.') option_parser.add_option('--no-cleanup', action='store_true', default=False, help='Allow installation on non-Windows platforms, and skip the forced ' 'cleanup step.') option_parser.add_option('--output-dir', type='string', help='The path where the binaries will be replaced. Existing binaries ' 'will only be overwritten if not up to date.') option_parser.add_option('--overwrite', action='store_true', default=False, help='If specified then the installation will happily delete and rewrite ' 'the entire output directory, blasting any local changes.') option_parser.add_option('--revision', type='string', help='The SVN revision or GIT hash associated with the required version.') option_parser.add_option('--revision-file', type='string', help='A text file containing an SVN revision or GIT hash.') option_parser.add_option('--resource', type='string', action='append', dest='resources', help='A resource to be downloaded.') option_parser.add_option('--verbose', dest='log_level', action='store_const', default=logging.INFO, const=logging.DEBUG, help='Enables verbose logging.') option_parser.add_option('--quiet', dest='log_level', action='store_const', default=logging.INFO, const=logging.ERROR, help='Disables all output except for errors.') options, args = option_parser.parse_args() if args: option_parser.error('Unexpected arguments: %s' % args) if not options.output_dir: option_parser.error('Must specify --output-dir.') if not options.revision and not options.revision_file: option_parser.error('Must specify one of --revision or --revision-file.') if options.revision and options.revision_file: option_parser.error('Must not specify both --revision and --revision-file.') # Configure logging. logging.basicConfig(level=options.log_level) # If a revision file has been specified then read it. if options.revision_file: options.revision = open(options.revision_file, 'rb').read().strip() _LOGGER.debug('Parsed revision "%s" from file "%s".', options.revision, options.revision_file) # Ensure that the specified SVN revision or GIT hash is valid. if not _REVISION_RE.match(options.revision): option_parser.error('Must specify a valid SVN or GIT revision.') # This just makes output prettier to read. options.output_dir = os.path.normpath(options.output_dir) return options def _RemoveOrphanedFiles(options): """This is run on non-Windows systems to remove orphaned files that may have been downloaded by a previous version of this script. """ # Reconfigure logging to output info messages. This will allow inspection of # cleanup status on non-Windows buildbots. _LOGGER.setLevel(logging.INFO) output_dir = os.path.abspath(options.output_dir) # We only want to clean up the folder in 'src/third_party/syzygy', and we # expect to be called with that as an output directory. This is an attempt to # not start deleting random things if the script is run from an alternate # location, or not called from the gclient hooks. expected_syzygy_dir = os.path.abspath(os.path.join( os.path.dirname(__file__), '..', 'third_party', 'syzygy')) expected_output_dir = os.path.join(expected_syzygy_dir, 'binaries') if expected_output_dir != output_dir: _LOGGER.info('Unexpected output directory, skipping cleanup.') return if not os.path.isdir(expected_syzygy_dir): _LOGGER.info('Output directory does not exist, skipping cleanup.') return def OnError(function, path, excinfo): """Logs error encountered by shutil.rmtree.""" _LOGGER.error('Error when running %s(%s)', function, path, exc_info=excinfo) _LOGGER.info('Removing orphaned files from %s', expected_syzygy_dir) if not options.dry_run: shutil.rmtree(expected_syzygy_dir, True, OnError) def main(): options = _ParseCommandLine() if options.dry_run: _LOGGER.debug('Performing a dry-run.') # We only care about Windows platforms, as the Syzygy binaries aren't used # elsewhere. However, there was a short period of time where this script # wasn't gated on OS types, and those OSes downloaded and installed binaries. # This will cleanup orphaned files on those operating systems. if sys.platform not in ('win32', 'cygwin'): if options.no_cleanup: _LOGGER.debug('Skipping usual cleanup for non-Windows platforms.') else: return _RemoveOrphanedFiles(options) # Load the current installation state, and validate it against the # requested installation. state, is_consistent = _GetCurrentState(options.revision, options.output_dir) # Decide whether or not an install is necessary. if options.force: _LOGGER.debug('Forcing reinstall of binaries.') elif is_consistent: # Avoid doing any work if the contents of the directory are consistent. _LOGGER.debug('State unchanged, no reinstall necessary.') return # Under normal logging this is the only only message that will be reported. _LOGGER.info('Installing revision %s Syzygy binaries.', options.revision[0:12]) # Clean up the old state to begin with. deleted = [] if options.overwrite: if os.path.exists(options.output_dir): # If overwrite was specified then take a heavy-handed approach. _LOGGER.debug('Deleting entire installation directory.') if not options.dry_run: _RmTree(options.output_dir) else: # Otherwise only delete things that the previous installation put in place, # and take care to preserve any local changes. deleted = _CleanState(options.output_dir, state, options.dry_run) # Install the new binaries. In a dry-run this will actually download the # archives, but it won't write anything to disk. state = _InstallBinaries(options, deleted) # Build and save the state for the directory. _SaveState(options.output_dir, state, options.dry_run) if __name__ == '__main__': main()
bsd-3-clause
c2f4a9787dad59c23a43b5eecc5d995b
36.067623
80
0.686273
3.807409
false
false
false
false
flutter/buildroot
build/landmine_utils.py
1
3015
# Copyright 2013 The Chromium Authors. All rights reserved. # Use of this source code is governed by a BSD-style license that can be # found in the LICENSE file. import functools import logging import os import shlex import sys def memoize(default=None): """This decorator caches the return value of a parameterless pure function""" def memoizer(func): val = [] @functools.wraps(func) def inner(): if not val: ret = func() val.append(ret if ret is not None else default) if logging.getLogger().isEnabledFor(logging.INFO): print('%s -> %r' % (func.__name__, val[0])) return val[0] return inner return memoizer @memoize() def IsWindows(): return sys.platform in ['win32', 'cygwin'] @memoize() def IsLinux(): return sys.platform.startswith(('linux', 'freebsd', 'openbsd')) @memoize() def IsMac(): return sys.platform == 'darwin' @memoize() def gyp_defines(): """Parses and returns GYP_DEFINES env var as a dictionary.""" return dict(arg.split('=', 1) for arg in shlex.split(os.environ.get('GYP_DEFINES', ''))) @memoize() def gyp_generator_flags(): """Parses and returns GYP_GENERATOR_FLAGS env var as a dictionary.""" return dict(arg.split('=', 1) for arg in shlex.split(os.environ.get('GYP_GENERATOR_FLAGS', ''))) @memoize() def gyp_msvs_version(): return os.environ.get('GYP_MSVS_VERSION', '') @memoize() def distributor(): """ Returns a string which is the distributed build engine in use (if any). Possible values: 'goma', 'ib', '' """ if 'goma' in gyp_defines(): return 'goma' elif IsWindows(): if 'CHROME_HEADLESS' in os.environ: return 'ib' # use (win and !goma and headless) as approximation of ib @memoize() def platform(): """ Returns a string representing the platform this build is targetted for. Possible values: 'win', 'mac', 'linux', 'ios', 'android' """ if 'OS' in gyp_defines(): if 'android' in gyp_defines()['OS']: return 'android' else: return gyp_defines()['OS'] elif IsWindows(): return 'win' elif IsLinux(): return 'linux' else: return 'mac' @memoize() def builder(): """ Returns a string representing the build engine (not compiler) to use. Possible values: 'make', 'ninja', 'xcode', 'msvs', 'scons' """ if 'GYP_GENERATORS' in os.environ: # for simplicity, only support the first explicit generator generator = os.environ['GYP_GENERATORS'].split(',')[0] if generator.endswith('-android'): return generator.split('-')[0] elif generator.endswith('-ninja'): return 'ninja' else: return generator else: if platform() == 'android': # Good enough for now? Do any android bots use make? return 'ninja' elif platform() == 'ios': return 'xcode' elif IsWindows(): return 'ninja' elif IsLinux(): return 'ninja' elif IsMac(): return 'ninja' else: assert False, 'Don\'t know what builder we\'re using!'
bsd-3-clause
a20e3ec2207e7828493ea344c2effe91
24.125
79
0.636816
3.663426
false
false
false
false
flutter/buildroot
tools/dart/dart_roll_helper.py
1
13149
#!/usr/bin/env python3 -u # # Copyright 2018 The Dart project authors. All rights reserved. # Use of this source code is governed by a BSD-style license that can be # found in the LICENSE file. # This script automates the Dart SDK roll steps, including: # - Updating the Dart revision in DEPS # - Updating the Dart dependencies in DEPS # - Syncing dependencies with 'gclient sync' # - Generating GN files for relevant engine configurations # - Building relevant engine configurations # - Running tests in 'example/flutter_gallery' and 'packages/flutter' # - Launching flutter_gallery in release and debug mode # - Running license.sh and updating license files in # 'flutter/ci/licenses_golden' # - Generating a commit with relevant Dart SDK commit logs (optional) # # The following environment variables can be set instead of being passed as # arguments: # - FLUTTER_HOME: the absolute path to the 'flutter' directory # - ENGINE_HOME: the absolute path to the 'engine/src' directory # - DART_SDK_HOME: the absolute path to the root of a Dart SDK project from dart_roll_utils import * import argparse import atexit import datetime import fileinput import os import platform import signal import shutil import subprocess import sys DART_REVISION_ENTRY = 'dart_revision' FLUTTER = '{}/bin/flutter'.format(flutter_home()) FLUTTER_DOCTOR = [FLUTTER, 'doctor'] FLUTTER_RUN = [FLUTTER, 'run'] FLUTTER_TEST = [FLUTTER, 'test'] MAX_GCLIENT_RETRIES = 3 # Returned when licenses do not require updating. LICENSE_SCRIPT_OKAY = 0 # Returned when licenses require updating. LICENSE_SCRIPT_UPDATES = 1 # Returned when either 'pub' or 'dart' isn't in the path. LICENSE_SCRIPT_EXIT_ERROR = 127 CURRENT_SUBPROCESS = None def update_dart_revision(dart_revision): original_revision = '' print_status('Updating Dart revision to {}'.format(dart_revision)) content = get_deps() for idx, line in enumerate(content): if DART_REVISION_ENTRY in line: original_revision = line.strip().split(' ')[1][1:-2] if not is_ancestor_commit(original_revision, dart_revision, dart_sdk_home()): print_error('Dart revision {} is older than existing revision, {}.' + ' Aborting roll.'.format(dart_revision, original_revision)) sys.exit(ERROR_OLD_COMMIT_PROVIDED) content[idx] = " 'dart_revision': '" + dart_revision + "',\n" break write_deps(content) return original_revision def run_process(args, cwd=None, stdout=None): global CURRENT_SUBPROCESS CURRENT_SUBPROCESS = subprocess.Popen(args, cwd=cwd, stdout=stdout) exit_code = CURRENT_SUBPROCESS.wait() CURRENT_SUBPROCESS = None return exit_code def gclient_sync(): global CURRENT_SUBPROCESS exit_code = None num_retries = 0 while ((exit_code != 0) and not (num_retries >= MAX_GCLIENT_RETRIES)): print_status('Running gclient sync (Attempt {}/{})' .format(num_retries + 1, MAX_GCLIENT_RETRIES)) exit_code = run_process(['gclient', 'sync', '--delete_unversioned_trees'], cwd=engine_home()) if exit_code != 0: num_retries += 1 if num_retries == MAX_GCLIENT_RETRIES: print_error('Max number of gclient sync retries attempted. Aborting roll.') sys.exit(ERROR_GCLIENT_SYNC_FAILED) def get_deps(): with open(flutter_deps_path(), 'r') as f: content = f.readlines() return content def update_deps(): print_status('Updating Dart dependencies') run_process([update_dart_deps_path()], cwd=engine_home()) def write_deps(newdeps): with open(flutter_deps_path(), 'w') as f: f.write(''.join(newdeps)) def run_gn(): print_status('Generating build files') common = [os.path.join('flutter', 'tools', 'gn'), '--goma', '--full-dart-sdk'] debug = ['--runtime-mode=debug'] profile = ['--runtime-mode=profile'] release = ['--runtime-mode=release'] runtime_modes = [debug, profile, release] unopt = ['--unoptimized'] android = ['--android'] fuchsia = ['--fuchsia'] for mode in runtime_modes: if set(mode) != set(release): run_process(common + android + unopt + mode, cwd=engine_home()) run_process(common + android + mode, cwd=engine_home()) host = common[:] if set(mode) == set(debug): host += unopt run_process(host + mode, cwd=engine_home()) run_process(common + fuchsia + debug + unopt, cwd=engine_home()) def build(): print_status('Building Flutter engine') command = ['ninja', '-j1000'] configs = [ 'host_debug_unopt', 'host_release', 'host_profile', 'android_debug_unopt', 'android_profile', 'android_release', 'fuchsia_debug_unopt_x64', ] build_dir = 'out' if platform.system() == 'Darwin': build_dir = 'xcodebuild' for config in configs: error_code = run_process(command + ['-C', os.path.join(build_dir, config)], cwd=engine_home()) if error_code != 0: print_error('Build failure for configuration "' + config + '". Aborting roll.') sys.exit(ERROR_BUILD_FAILED) def run_flutter_doctor(): print_status('Running flutter doctor') engine_src_path = '--local-engine-src-path={}'.format(engine_home()) result = run_process(FLUTTER_DOCTOR + ['--local-engine=host_debug_unopt', engine_src_path], cwd=package_flutter_path()) if result != 0: print_error('flutter doctor failed. Aborting roll.') sys.exit(ERROR_FLUTTER_DOCTOR_FAILED) def run_tests(): print_status('Running tests in packages/flutter') engine_src_path = '--local-engine-src-path={}'.format(engine_home()) result = run_process(FLUTTER_TEST + ['--local-engine=host_debug_unopt', engine_src_path], cwd=package_flutter_path()) if result != 0: print_error('package/flutter tests failed. Aborting roll.') sys.exit(ERROR_PKG_FLUTTER_FAILED) print_status('Running tests in examples/flutter_gallery') result = run_process(FLUTTER_TEST + ['--local-engine=host_debug_unopt', engine_src_path, '--disable-service-auth-codes'], cwd=flutter_gallery_path()); if result != 0: print_error('flutter_gallery tests failed. Aborting roll.') sys.exit(ERROR_FLUTTER_GALLERY_FAILED) def run_hot_reload_configurations(): print_status('Running flutter gallery release') engine_src_path = '--local-engine-src-path={}'.format(engine_home()) run_process(FLUTTER_RUN + ['--release', '--local-engine=android_release', engine_src_path], cwd=flutter_gallery_path()) print_status('Running flutter gallery debug') run_process(FLUTTER_RUN + ['--local-engine=android_debug_unopt', engine_src_path], cwd=flutter_gallery_path()) def update_licenses(): print_status('Updating Flutter licenses') result = run_process([engine_license_script_path()], cwd=engine_home()) if result == LICENSE_SCRIPT_EXIT_ERROR: print_error('License script failed to run. Is the Dart SDK (specifically' + ' dart and pub) in your path? Aborting roll.') sys.exit(ERROR_LICENSE_SCRIPT_FAILED) elif (result != LICENSE_SCRIPT_OKAY) and (result != LICENSE_SCRIPT_UPDATES): print_error('Unknown license script error: {}. Aborting roll.' .format(result)) sys.exit(ERROR_LICENSE_SCRIPT_FAILED) # Ignore 'licenses_skia' as they shouldn't change during a Dart SDK roll. src_files = ['licenses_flutter', 'licenses_third_party', 'tool_signature'] for f in src_files: path = os.path.join(license_script_output_path(), f) if os.path.isfile(path): shutil.copy(path, engine_golden_licenses_path()) run_process(['pub', 'get'], cwd=engine_license_script_package_path()) gclient_sync() # Update the LICENSE file. with open(sky_license_file_path(), 'w') as sky_license: run_process(['dart', os.path.join('lib', 'main.dart'), '--release', '--src', engine_home(), '--out', engine_license_script_output_path()], cwd=engine_license_script_package_path(), stdout=sky_license) def get_commit_range(start, finish): range_str = '{}..{}'.format(start, finish) command = ['git', 'log', '--oneline', range_str] orig_dir = os.getcwd() os.chdir(dart_sdk_home()) result = subprocess.check_output(command).decode('utf-8') result = '\n'.join(['dart-lang/sdk@' + l for l in result.splitlines()]) os.chdir(orig_dir) return result def get_short_rev(rev): command = ['git', 'rev-parse', '--short', rev] orig_dir = os.getcwd() os.chdir(dart_sdk_home()) result = subprocess.check_output(command).decode('utf-8').rstrip() os.chdir(orig_dir) return result def git_commit(original_revision, updated_revision): print_status('Committing Dart SDK roll') current_date = datetime.date.today() sdk_log = get_commit_range(original_revision, updated_revision) num_commits = len(sdk_log.splitlines()) commit_msg = ('Roll src/third_party/dart {}..{} ({} commits)' .format(get_short_rev(original_revision), get_short_rev(updated_revision), num_commits)) commit_msg += '\n\n' + sdk_log commit_cmd = ['git', 'commit', '-a', '-m', commit_msg] run_process(commit_cmd, cwd=engine_flutter_path()) def update_roots(args): if args.flutter_home: set_flutter_home(args.flutter_home) if args.engine_home: set_engine_home(args.engine_home) if args.dart_sdk_home: set_dart_sdk_home(args.dart_sdk_home) if flutter_home() == '': print_error('Either "--flutter-home" must be provided or FLUTTER_HOME must' + ' be set. Aborting roll.') sys.exit(ERROR_MISSING_ROOTS) if engine_home() == '': print_error('Either "--engine-home" must be provided or ENGINE_HOME must' + ' be set. Aborting roll.') sys.exit(ERROR_MISSING_ROOTS) if dart_sdk_home() == '': print_error('Either "--dart-sdk-home" must be provided or DART_SDK_HOME ' + 'must be set. Aborting roll.') sys.exit(ERROR_MISSING_ROOTS) def sys_exit(signal, frame): sys.exit() def cleanup_children(): if CURRENT_SUBPROCESS != None: CURRENT_SUBPROCESS.terminate() def main(): parser = argparse.ArgumentParser(description='Automate most Dart SDK roll tasks.') parser.add_argument('--dart-sdk-home', help='Path to the Dart SDK ' + 'repository. Overrides DART_SDK_HOME environment variable') parser.add_argument('dart_sdk_revision', help='Target Dart SDK revision') parser.add_argument('--create-commit', action='store_true', help='Create the engine commit with Dart SDK commit log') parser.add_argument('--engine-home', help='Path to the Flutter engine ' + 'repository. Overrides ENGINE_HOME environment variable') parser.add_argument('--flutter-home', help='Path to the Flutter framework ' + 'repository. Overrides FLUTTER_HOME environment variable') parser.add_argument('--no-build', action='store_true', help='Skip rebuilding the Flutter engine') parser.add_argument('--no-hot-reload', action='store_true', help="Skip hot reload testing") parser.add_argument('--no-test', action='store_true', help='Skip running host tests for package/flutter and ' + 'flutter_gallery') parser.add_argument('--no-update-deps', action='store_true', help='Skip updating DEPS file') parser.add_argument('--no-update-licenses', action='store_true', help='Skip updating licenses') args = parser.parse_args() atexit.register(cleanup_children) signal.signal(signal.SIGTERM, sys_exit) original_revision = None updated_revision = args.dart_sdk_revision # Disable buffering of log output os.environ["PYTHONUNBUFFERED"] = "1" update_roots(args) print_status('Starting Dart SDK roll') if not args.no_update_deps: original_revision = update_dart_revision(updated_revision) gclient_sync() update_deps() gclient_sync() if not args.no_build: run_gn() build() if ((not args.no_test) or (not args.no_hot_reload)): run_flutter_doctor() if not args.no_test: run_tests() if not args.no_hot_reload: run_hot_reload_configurations() if not args.no_update_licenses: update_licenses() if args.create_commit: if original_revision == None: print_warning('"original_revision" not specified. Skipping commit.') print_warning('This happens when the "--no_update_deps" argument is ' + 'provided') else: git_commit(original_revision, updated_revision) print_status('Dart SDK roll complete!') if __name__ == '__main__': main()
bsd-3-clause
1bd0a4afdfba0605df11f46601cb0baf
34.92623
84
0.63617
3.594587
false
false
false
false
niwinz/django-jinja
django_jinja/views/__init__.py
2
1916
from django import http from django.template import loader from django.views import View from ..base import get_match_extension class GenericView(View): response_cls = http.HttpResponse content_type = "text/html" tmpl_name = None def get_context_data(self): return {"view": self} def get(self, request, *args, **kwargs): context = self.get_context_data() template_name = callable(self.tmpl_name) and self.tmpl_name() or self.tmpl_name output = loader.render_to_string(template_name, context, request=request) return self.response_cls(output, content_type=self.content_type) class ErrorView(GenericView): def head(self, request, *args, **kwargs): return self.get(request, *args, **kwargs) def post(self, request, *args, **kwargs): return self.get(request, *args, **kwargs) def options(self, request, *args, **kwargs): return self.get(request, *args, **kwargs) def delete(self, request, *args, **kwargs): return self.get(request, *args, **kwargs) def put(self, request, *args, **kwargs): return self.get(request, *args, **kwargs) def patch(self, request, *args, **kwargs): return self.get(request, *args, **kwargs) class PageNotFound(ErrorView): response_cls = http.HttpResponseNotFound def tmpl_name(self): return "404" + (get_match_extension() or ".jinja") class PermissionDenied(ErrorView): response_cls = http.HttpResponseForbidden def tmpl_name(self): return "403" + (get_match_extension() or ".jinja") class BadRequest(ErrorView): response_cls = http.HttpResponseBadRequest def tmpl_name(self): return "400" + (get_match_extension() or ".jinja") class ServerError(ErrorView): response_cls = http.HttpResponseServerError def tmpl_name(self): return "500" + (get_match_extension() or ".jinja")
bsd-3-clause
7be6ed9492694f9f77825a707104383c
27.176471
87
0.66023
3.691715
false
false
false
false
dask/dask-ml
tests/test_spectral_clustering.py
1
3325
from functools import partial import numpy as np import pytest import sklearn.cluster from dask_ml import metrics from dask_ml.cluster import SpectralClustering from dask_ml.datasets import make_blobs X, y = make_blobs(n_samples=200, chunks=100, random_state=0) @pytest.mark.parametrize("as_ndarray", [False, True]) @pytest.mark.parametrize("persist_embedding", [True, False]) def test_basic(as_ndarray, persist_embedding): sc = SpectralClustering( n_components=25, random_state=0, persist_embedding=persist_embedding ) if as_ndarray: X_ = X.compute() else: X_ = X sc.fit(X_) assert len(sc.labels_) == len(X_) @pytest.mark.parametrize( "assign_labels", [sklearn.cluster.KMeans(n_init=2), "sklearn-kmeans"] ) def test_sklearn_kmeans(assign_labels): sc = SpectralClustering( n_components=25, random_state=0, assign_labels=assign_labels, kmeans_params={"n_clusters": 8}, ) sc.fit(X) assert isinstance(sc.assign_labels_, sklearn.cluster.KMeans) @pytest.mark.skip(reason="Can't reproduce CI failure.") def test_callable_affinity(): affinity = partial( metrics.pairwise.pairwise_kernels, metric="rbf", filter_params=True, gamma=1.0 / len(X), ) sc = SpectralClustering(affinity=affinity, gamma=None) sc.fit(X) def test_n_components_raises(): sc = SpectralClustering(n_components=len(X)) with pytest.raises(ValueError) as m: sc.fit(X) assert m.match("n_components") def test_assign_labels_raises(): sc = SpectralClustering(assign_labels="foo") with pytest.raises(ValueError) as m: sc.fit(X) assert m.match("Unknown 'assign_labels' 'foo'") sc = SpectralClustering(assign_labels=dict()) with pytest.raises(TypeError) as m: sc.fit(X) assert m.match("Invalid type ") def test_affinity_raises(): sc = SpectralClustering(affinity="foo") with pytest.raises(ValueError) as m: sc.fit(X) assert m.match("Unknown affinity metric name 'foo'") sc = SpectralClustering(affinity=np.array([])) with pytest.raises(TypeError) as m: sc.fit(X) assert m.match("Unexpected type for affinity 'ndarray'") def test_spectral_clustering(Xl_blobs_easy): X, y = Xl_blobs_easy X = (X - X.mean(0)) / X.std(0) model = SpectralClustering( random_state=0, n_clusters=3, n_components=5, gamma=None ).fit(X) labels = model.labels_.compute() y = y.compute() idx = [(y == i).argmax() for i in range(3)] grouped_idx = [np.where(y == y[idx[i]])[0] for i in range(3)] for indices in grouped_idx: assert len(set(labels[indices])) == 1 @pytest.mark.parametrize("keep", [[4, 7], [4, 5], [0, 3], [1, 9], [0, 1, 5, 8, 9]]) def test_slice_mostly_sorted(keep): import dask.array as da import numpy as np from dask.array.utils import assert_eq from dask_ml.cluster.spectral import _slice_mostly_sorted X = np.arange(10).reshape(-1, 1) dX = da.from_array(X, chunks=5) keep = np.array(keep).ravel() rest = ~np.isin(X, keep).ravel() array = dX[np.concatenate([X[keep], X[rest]]).ravel()] result = _slice_mostly_sorted(array, keep, rest) assert_eq(result, X) assert all(x > 0 for x in result.chunks[0])
bsd-3-clause
8d28f150874ae3f8e3a10d4d23eb7c40
26.254098
83
0.64391
3.234436
false
true
false
false
dials/dials
src/dials/algorithms/merging/merge.py
1
16432
"""Merging functions for experiment lists and reflection tables.""" from __future__ import annotations import logging from contextlib import contextmanager from io import StringIO from typing import Optional, Tuple import numpy as np from dxtbx.model import ExperimentList from iotbx import mtz, phil from mmtbx.scaling import data_statistics from dials.algorithms.merging.reporting import ( MergeJSONCollector, MergingStatisticsData, make_dano_table, ) from dials.algorithms.scaling.Ih_table import ( _reflection_table_to_iobs, map_indices_to_asu, ) from dials.algorithms.scaling.scaling_library import ( merging_stats_from_scaled_array, scaled_data_as_miller_array, ) from dials.algorithms.scaling.scaling_utilities import DialsMergingStatisticsError from dials.algorithms.symmetry.absences.run_absences_checks import ( run_systematic_absences_checks, ) from dials.array_family import flex from dials.util.export_mtz import MADMergedMTZWriter, MergedMTZWriter from dials.util.filter_reflections import filter_reflection_table from .french_wilson import french_wilson logger = logging.getLogger("dials") @contextmanager def collect_html_data_from_merge(): try: html_collector = MergeJSONCollector() yield html_collector finally: html_collector.reset() def prepare_merged_reflection_table( experiments, reflection_table, d_min=None, d_max=None, partiality_threshold=0.99, ): """Filter the data and prepare a reflection table with merged data.""" if ( "inverse_scale_factor" in reflection_table and "intensity.scale.value" in reflection_table ): logger.info("Performing systematic absence checks on scaled data") reflections = filter_reflection_table( reflection_table, intensity_choice=["scale"], d_min=d_min, d_max=d_max, partiality_threshold=partiality_threshold, ) reflections["intensity"] = reflections["intensity.scale.value"] reflections["variance"] = reflections["intensity.scale.variance"] elif "intensity.prf.value" in reflection_table: logger.info( "Performing systematic absence checks on unscaled profile-integrated data" ) reflections = filter_reflection_table( reflection_table, intensity_choice=["profile"], d_min=d_min, d_max=d_max, partiality_threshold=partiality_threshold, ) reflections["intensity"] = reflections["intensity.prf.value"] reflections["variance"] = reflections["intensity.prf.variance"] else: logger.info( "Performing systematic absence checks on unscaled summation-integrated data" ) reflections = filter_reflection_table( reflection_table, intensity_choice=["sum"], d_min=d_min, d_max=d_max, partiality_threshold=partiality_threshold, ) reflections["intensity"] = reflections["intensity.sum.value"] reflections["variance"] = reflections["intensity.sum.variance"] # now merge space_group = experiments[0].crystal.get_space_group() reflections["asu_miller_index"] = map_indices_to_asu( reflections["miller_index"], space_group ) reflections["inverse_scale_factor"] = flex.double(reflections.size(), 1.0) merged = ( _reflection_table_to_iobs( reflections, experiments[0].crystal.get_unit_cell(), space_group ) .merge_equivalents(use_internal_variance=False) .array() ) merged_reflections = flex.reflection_table() merged_reflections["intensity"] = merged.data() merged_reflections["variance"] = flex.pow2(merged.sigmas()) merged_reflections["miller_index"] = merged.indices() return merged_reflections class MTZDataClass: """Container class (i.e. Python 3.7 dataclass) for per-wavelength mtz dataset.""" def __init__( self, wavelength=0.0, project_name="AUTOMATIC", dataset_name="NATIVE", crystal_name="XTAL", merged_array=None, merged_anomalous_array=None, amplitudes=None, anomalous_amplitudes=None, dano=None, multiplicities=None, ): self.wavelength = wavelength self.project_name = project_name self.dataset_name = dataset_name self.crystal_name = crystal_name self.merged_array = merged_array self.merged_anomalous_array = merged_anomalous_array self.amplitudes = amplitudes self.anomalous_amplitudes = anomalous_amplitudes self.dano = dano self.multiplicities = multiplicities def make_merged_mtz_file(mtz_datasets): """ Make an mtz file object for the data, adding the date, time and program. For multi-wavelength data, each wavelength is added as a new crystal. Args: mtz_datasets: A list of MTZDataClass objects, one per wavelength of the experiment. Returns: An iotbx mtz file object. """ if len(mtz_datasets) > 1: writer = MADMergedMTZWriter else: writer = MergedMTZWriter mtz_writer = writer( mtz_datasets[0].merged_array.space_group(), mtz_datasets[0].merged_array.unit_cell(), ) #### Add each wavelength as a new crystal. for dataset in mtz_datasets: mtz_writer.add_crystal( crystal_name=dataset.crystal_name, project_name=dataset.project_name ) mtz_writer.add_empty_dataset(dataset.wavelength, name=dataset.dataset_name) mtz_writer.add_dataset( dataset.merged_array, dataset.merged_anomalous_array, dataset.amplitudes, dataset.anomalous_amplitudes, dataset.dano, dataset.multiplicities, ) return mtz_writer.mtz_file def merge_scaled_array( experiments, scaled_array, anomalous=True, use_internal_variance=False, assess_space_group=False, n_bins=20, ): # assumes filtering already done and converted to combined scaled array # Note, merge_equivalents does not raise an error if data is unique. merged = scaled_array.merge_equivalents(use_internal_variance=use_internal_variance) merged_anom = None if anomalous: anomalous_scaled = scaled_array.as_anomalous_array() merged_anom = anomalous_scaled.merge_equivalents( use_internal_variance=use_internal_variance ) # Before merge, do assessment of the space_group if assess_space_group: merged_reflections = flex.reflection_table() merged_reflections["intensity"] = merged.array().data() merged_reflections["variance"] = flex.pow2(merged.array().sigmas()) merged_reflections["miller_index"] = merged.array().indices() logger.info("Running systematic absences check") run_systematic_absences_checks(experiments, merged_reflections) stats_data = MergingStatisticsData(experiments, scaled_array) try: stats, anom_stats = merging_stats_from_scaled_array( scaled_array, n_bins, use_internal_variance, ) except DialsMergingStatisticsError as e: logger.error(e, exc_info=True) else: stats_data.merging_statistics_result = stats stats_data.anom_merging_statistics_result = anom_stats return merged, merged_anom, stats_data def merge( experiments, reflections, d_min=None, d_max=None, combine_partials=True, partiality_threshold=0.4, best_unit_cell=None, anomalous=True, use_internal_variance=False, assess_space_group=False, n_bins=20, ): """ Merge reflection table data and generate a summary of the merging statistics. This procedure filters the input data, merges the data (normal and optionally anomalous), assesses the space group symmetry and generates a summary of the merging statistics. Returns two merge_equivalents objects and a statistics summary. """ logger.info("\nMerging scaled reflection data\n") # first filter bad reflections using dials.util.filter methods reflections = filter_reflection_table( reflections, intensity_choice=["scale"], d_min=d_min, d_max=d_max, combine_partials=combine_partials, partiality_threshold=partiality_threshold, ) # ^ scale factor has been applied, so now set to 1.0 - okay as not # going to output scale factor in merged mtz. reflections["inverse_scale_factor"] = flex.double(reflections.size(), 1.0) scaled_array = scaled_data_as_miller_array( [reflections], experiments, best_unit_cell ) return merge_scaled_array( experiments, scaled_array, anomalous, use_internal_variance, assess_space_group, n_bins, ) def show_wilson_scaling_analysis(merged_intensities, n_residues=200): """ Report the wilson statistics for a merged intensity array Args: merged_intensities: A merged miller intensity array. n_residues: The number of residues to use for the wilson analysis. """ if not merged_intensities.space_group().is_centric(): try: wilson_scaling = data_statistics.wilson_scaling( miller_array=merged_intensities, n_residues=n_residues ) except (IndexError, RuntimeError) as e: logger.error( "\n" "Error encountered during Wilson statistics calculation:\n" "Perhaps there are too few unique reflections.\n" "%s", e, exc_info=True, ) else: # Divert output through logger - do with StringIO rather than # info_handle else get way too much whitespace in output. out = StringIO() wilson_scaling.show(out=out) logger.info(out.getvalue()) return wilson_scaling.iso_b_wilson return None def truncate( merged_intensities, implementation: str = "dials", min_reflections: int = 200, fallback_to_flat_prior: bool = True, ): """ Perform French-Wilson truncation procedure on merged intensities. Args: merged_intensities (miller.array): A merged miller intensity array (normal or anomalous) implementation (str): Choice of implementation of French & Wilson algorithm, either "dials" or "cctbx" min_reflections (int): Minimum number of reflections to perform the French & Wilson procedure fallback_to_flat_prior (bool): Fallback to assumption of a flat, positive prior, if the number of reflections are fewer than min_reflections, i.e. |F| = sqrt((Io+sqrt(Io**2 +2sigma**2))/2.0) Returns: (tuple): tuple containing: amplitudes: A normal all-positive miller amplitude array anom_amplitudes: An anomalous all-positive amplitude array, if the input array has the anomalous_flag set, else None. dano: The array of anomalous differences, if the input array has the anomalous_flag set, else None. """ logger.info("\nPerforming French-Wilson treatment of scaled intensities") out = StringIO() n_refl = merged_intensities.size() if n_refl < min_reflections and fallback_to_flat_prior: logger.info( "Insufficient reflections for French & Wilson procedure, " "falling back to assumption of a flat, positive prior, i.e.: " " |F| = sqrt((Io+sqrt(Io**2 +2sigma**2))/2.0)" ) do_french_wilson = lambda ma: ma.enforce_positive_amplitudes() elif n_refl < min_reflections: raise ValueError( "Insufficient reflections for French & Wilson procedure. " "Either set fallback_to_flat_prior=True or truncate=False." ) elif implementation == "cctbx": do_french_wilson = lambda ma: ma.french_wilson(log=out) else: do_french_wilson = french_wilson if merged_intensities.anomalous_flag(): anom_amplitudes = do_french_wilson(merged_intensities) n_removed = merged_intensities.size() - anom_amplitudes.size() assert anom_amplitudes.is_xray_amplitude_array() amplitudes = anom_amplitudes.as_non_anomalous_array() amplitudes = amplitudes.merge_equivalents(use_internal_variance=False).array() dano = anom_amplitudes.anomalous_differences() else: anom_amplitudes = None dano = None amplitudes = do_french_wilson(merged_intensities) n_removed = merged_intensities.size() - amplitudes.size() logger.info("Total number of rejected intensities %s", n_removed) logger.debug(out.getvalue()) return amplitudes, anom_amplitudes, dano def merge_scaled_array_to_mtz_with_report_collection( params: phil.scope_extract, experiments: ExperimentList, scaled_array, wavelength: Optional[float] = None, ) -> Tuple[mtz.object, dict]: if wavelength is None: wavelength = np.mean( np.array( [expt.beam.get_wavelength() for expt in experiments], dtype=np.float ) ) with collect_html_data_from_merge() as collector: mtz_dataset = MTZDataClass( wavelength=wavelength, project_name=params.output.project_name, dataset_name=params.output.dataset_names[0], crystal_name=params.output.crystal_names[0], ) merged, merged_anomalous, stats_summary = merge_scaled_array( experiments, scaled_array, anomalous=params.anomalous, assess_space_group=params.assess_space_group, n_bins=params.merging.n_bins, use_internal_variance=params.merging.use_internal_variance, ) process_merged_data( params, mtz_dataset, merged, merged_anomalous, stats_summary ) mtz = make_merged_mtz_file([mtz_dataset]) json_data = collector.create_json() return mtz, json_data def process_merged_data(params, mtz_dataset, merged, merged_anomalous, stats_summary): merged_array = merged.array() # Save the relevant data in the mtz_dataset dataclass # This will add the data for IMEAN/SIGIMEAN mtz_dataset.merged_array = merged_array if merged_anomalous: merged_anomalous_array = merged_anomalous.array() # This will add the data for I(+), I(-), SIGI(+), SIGI(-), N(+), N(-) mtz_dataset.merged_anomalous_array = merged_anomalous_array mtz_dataset.multiplicities = merged_anomalous.redundancies() else: merged_anomalous_array = None # This will add the data for N mtz_dataset.multiplicities = merged.redundancies() if params.anomalous: merged_intensities = merged_anomalous_array else: merged_intensities = merged_array anom_amplitudes = None if params.truncate: amplitudes, anom_amplitudes, dano = truncate( merged_intensities, implementation=params.french_wilson.implementation, min_reflections=params.french_wilson.min_reflections, fallback_to_flat_prior=params.french_wilson.fallback_to_flat_prior, ) # This will add the data for F, SIGF mtz_dataset.amplitudes = amplitudes # This will add the data for F(+), F(-), SIGF(+), SIGF(-) mtz_dataset.anomalous_amplitudes = anom_amplitudes # This will add the data for DANO, SIGDANO mtz_dataset.dano = dano # print out analysis statistics try: B_iso = show_wilson_scaling_analysis(merged_intensities) except Exception as e: logger.info(e) else: stats_summary.Wilson_B_iso = B_iso if anom_amplitudes: logger.info(make_dano_table(anom_amplitudes)) if stats_summary.merging_statistics_result: logger.info(stats_summary) if MergeJSONCollector.initiated: stats_summary.anomalous_amplitudes = anom_amplitudes MergeJSONCollector.data[mtz_dataset.wavelength] = stats_summary
bsd-3-clause
e18aa00b4f16c3a00a6a5d5d5a3e3148
34.111111
99
0.645691
3.679355
false
false
false
false
dials/dials
tests/algorithms/integration/test_interface.py
1
18674
from __future__ import annotations import random import pytest def test_split_blocks_1_frame(): from dials.algorithms.integration.integrator import JobList from dials.array_family import flex r = flex.reflection_table() r["value1"] = flex.double() r["value2"] = flex.int() r["value3"] = flex.double() r["bbox"] = flex.int6() r["id"] = flex.int() expected = [] for i in range(100): x0 = random.randint(0, 100) x1 = x0 + random.randint(1, 10) y0 = random.randint(0, 100) y1 = y0 + random.randint(1, 10) z0 = random.randint(0, 100) z1 = z0 + random.randint(1, 10) v1 = random.uniform(0, 100) v2 = random.randint(0, 100) v3 = random.uniform(0, 100) r.append( { "id": 0, "value1": v1, "value2": v2, "value3": v3, "bbox": (x0, x1, y0, y1, z0, z1), } ) for z in range(z0, z1): expected.append( { "id": 0, "value1": v1, "value2": v2, "value3": v3, "bbox": (x0, x1, y0, y1, z, z + 1), "partial_id": i, } ) jobs = JobList() jobs.add((0, 1), (0, 111), 1, 0) jobs.split(r) assert len(r) == len(expected) EPS = 1e-7 for r1, r2 in zip(r.rows(), expected): assert r1["bbox"] == r2["bbox"] assert r1["partial_id"] == r2["partial_id"] assert abs(r1["value1"] - r2["value1"]) < EPS assert r1["value2"] == r2["value2"] assert abs(r1["value3"] - r2["value3"]) < EPS def test_split_blocks_non_overlapping(): from scitbx.array_family import shared from dials.algorithms.integration.integrator import JobList from dials.array_family import flex blocks = shared.tiny_int_2( [ (0, 10), (10, 20), (20, 30), (30, 35), (35, 40), (40, 50), (50, 60), (60, 70), (70, 80), (80, 90), (90, 100), (100, 110), ] ) jobs = JobList((0, 1), blocks) r = flex.reflection_table() r["value1"] = flex.double() r["value2"] = flex.int() r["value3"] = flex.double() r["bbox"] = flex.int6() r["id"] = flex.int() expected = [] for i in range(100): x0 = random.randint(0, 100) x1 = x0 + random.randint(1, 10) y0 = random.randint(0, 100) y1 = y0 + random.randint(1, 10) z0 = random.randint(0, 100) z1 = z0 + random.randint(1, 10) v1 = random.uniform(0, 100) v2 = random.randint(0, 100) v3 = random.uniform(0, 100) r.append( { "id": 0, "value1": v1, "value2": v2, "value3": v3, "bbox": (x0, x1, y0, y1, z0, z1), } ) for j in range(len(blocks)): b0 = blocks[j][0] b1 = blocks[j][1] if ( (z0 >= b0 and z1 <= b1) or (z0 < b1 and z1 >= b1) or (z0 < b0 and z1 > b0) ): z00 = max(b0, z0) z11 = min(b1, z1) expected.append( { "id": 0, "value1": v1, "value2": v2, "value3": v3, "bbox": (x0, x1, y0, y1, z00, z11), "partial_id": i, } ) jobs.split(r) assert len(r) == len(expected) EPS = 1e-7 for r1, r2 in zip(r.rows(), expected): assert r1["bbox"] == r2["bbox"] assert r1["partial_id"] == r2["partial_id"] assert abs(r1["value1"] - r2["value1"]) < EPS assert r1["value2"] == r2["value2"] assert abs(r1["value3"] - r2["value3"]) < EPS def test_split_blocks_overlapping(): from scitbx.array_family import shared from dials.algorithms.integration.integrator import JobList from dials.array_family import flex blocks = shared.tiny_int_2( [ (0, 10), (5, 15), (10, 20), (15, 25), (20, 30), (25, 35), (30, 40), (35, 45), (40, 50), (45, 55), (50, 60), (55, 65), (60, 70), (65, 75), (70, 80), (75, 85), (80, 90), (85, 95), (90, 100), (95, 105), (100, 110), ] ) jobs = JobList((0, 1), blocks) r = flex.reflection_table() r["value1"] = flex.double() r["value2"] = flex.int() r["value3"] = flex.double() r["bbox"] = flex.int6() r["id"] = flex.int() expected = [] for i in range(100): x0 = random.randint(0, 100) x1 = x0 + random.randint(1, 10) y0 = random.randint(0, 100) y1 = y0 + random.randint(1, 10) z0 = random.randint(0, 90) z1 = z0 + random.randint(1, 20) v1 = random.uniform(0, 100) v2 = random.randint(0, 100) v3 = random.uniform(0, 100) r.append( { "id": 0, "value1": v1, "value2": v2, "value3": v3, "bbox": (x0, x1, y0, y1, z0, z1), } ) expected.append( { "id": 0, "value1": v1, "value2": v2, "value3": v3, "bbox": (x0, x1, y0, y1, z0, z1), } ) jobs.split(r) assert len(r) > 100 for r1 in r.rows(): v1 = r1["value1"] v2 = r1["value2"] v3 = r1["value3"] bbox = r1["bbox"] pid = r1["partial_id"] z0 = bbox[4] z1 = bbox[5] success = False for i in range(len(blocks)): b0 = blocks[i][0] b1 = blocks[i][1] if z0 >= b0 and z1 <= b1: success = True break assert success v11 = expected[pid]["value1"] v22 = expected[pid]["value2"] v33 = expected[pid]["value3"] bb = expected[pid]["bbox"] assert v11 == v1 assert v22 == v2 assert v33 == v3 assert bb[0] == bbox[0] assert bb[1] == bbox[1] assert bb[2] == bbox[2] assert bb[3] == bbox[3] def test_reflection_manager(): from dials.array_family import flex reflections = flex.reflection_table() reflections["panel"] = flex.size_t() reflections["bbox"] = flex.int6() reflections["miller_index"] = flex.miller_index() reflections["s1"] = flex.vec3_double() reflections["xyzcal.px"] = flex.vec3_double() reflections["xyzcal.mm"] = flex.vec3_double() reflections["entering"] = flex.bool() reflections["id"] = flex.int() reflections["flags"] = flex.size_t() width = 1000 height = 1000 nrefl = 10000 array_range = (0, 130) block_size = 20 block_overlap = 10 random.seed(0) processed = [[] for i in range(12)] for i in range(nrefl): x0 = random.randint(0, width - 10) y0 = random.randint(0, height - 10) zs = random.randint(2, 9) x1 = x0 + random.randint(2, 10) y1 = y0 + random.randint(2, 10) for k, j in enumerate([10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 110, 120]): m = k + i * 12 pos = random.choice(["left", "right", "centre"]) if pos == "left": z0 = j - zs z1 = j elif pos == "right": z0 = j z1 = j + zs else: z0 = j - zs // 2 z1 = j + zs // 2 bbox = (x0, x1, y0, y1, z0, z1) reflections.append( { "panel": random.randint(0, 1), "bbox": bbox, "flags": flex.reflection_table.flags.reference_spot, } ) processed[k].append(m) # Add reflection to ignore # zc = random.choice([10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 110, 120]) # z0 = zc - 11 # z1 = zc + 11 # bbox = (x0, x1, y0, y1, z0, z1) # reflections.append({ # "panel" : randint(0,1), # "bbox" : bbox, # "flags" : flex.reflection_table.flags.reference_spot # }) from dials.algorithms.integration.integrator import JobList, ReflectionManager jobs = JobList() jobs.add((0, 1), array_range, block_size, block_overlap) # Create the executor executor = ReflectionManager(jobs, reflections) # Ensure the tasks make sense jobs = [executor.job(i) for i in range(len(executor))] assert len(executor) == 12 assert not executor.finished() assert len(jobs) == 12 assert jobs[0].frames() == (0, 20) assert jobs[1].frames() == (10, 30) assert jobs[2].frames() == (20, 40) assert jobs[3].frames() == (30, 50) assert jobs[4].frames() == (40, 60) assert jobs[5].frames() == (50, 70) assert jobs[6].frames() == (60, 80) assert jobs[7].frames() == (70, 90) assert jobs[8].frames() == (80, 100) assert jobs[9].frames() == (90, 110) assert jobs[10].frames() == (100, 120) assert jobs[11].frames() == (110, 130) # Get the task specs data0 = executor.split(0) data1 = executor.split(1) data2 = executor.split(2) data3 = executor.split(3) data4 = executor.split(4) data5 = executor.split(5) data6 = executor.split(6) data7 = executor.split(7) data8 = executor.split(8) data9 = executor.split(9) data10 = executor.split(10) data11 = executor.split(11) assert len(data0) == len(processed[0]) assert len(data1) == len(processed[1]) assert len(data2) == len(processed[2]) assert len(data3) == len(processed[3]) assert len(data4) == len(processed[4]) assert len(data5) == len(processed[5]) assert len(data6) == len(processed[6]) assert len(data7) == len(processed[7]) assert len(data8) == len(processed[8]) assert len(data9) == len(processed[9]) assert len(data10) == len(processed[10]) assert len(data11) == len(processed[11]) # Add some results data0["data"] = flex.double(len(data0), 1) data1["data"] = flex.double(len(data1), 2) data2["data"] = flex.double(len(data2), 3) data3["data"] = flex.double(len(data3), 4) data4["data"] = flex.double(len(data4), 5) data5["data"] = flex.double(len(data5), 6) data6["data"] = flex.double(len(data6), 7) data7["data"] = flex.double(len(data7), 8) data8["data"] = flex.double(len(data8), 9) data9["data"] = flex.double(len(data9), 10) data10["data"] = flex.double(len(data10), 11) data11["data"] = flex.double(len(data11), 12) # Accumulate the data again assert not executor.finished() executor.accumulate(0, data0) executor.accumulate(1, data1) executor.accumulate(2, data2) executor.accumulate(3, data3) executor.accumulate(4, data4) executor.accumulate(5, data5) executor.accumulate(6, data6) executor.accumulate(7, data7) executor.accumulate(8, data8) executor.accumulate(9, data9) executor.accumulate(10, data10) executor.accumulate(11, data11) assert executor.finished() # Get results and check they're as expected data = executor.data() result = data["data"] for i in range(len(processed)): for j in range(len(processed[i])): assert result[processed[i][j]] == i + 1 # Test passed @pytest.mark.parametrize("nproc", [1, 2]) def test_integrator_3d(dials_data, nproc): from math import pi from dxtbx.model.experiment_list import ExperimentListFactory from dials.algorithms.profile_model.gaussian_rs import Model from dials.array_family import flex path = dials_data("centroid_test_data", pathlib=True) / "experiments.json" exlist = ExperimentListFactory.from_json_file(path) exlist[0].profile = Model( None, n_sigma=3, sigma_b=0.024 * pi / 180.0, sigma_m=0.044 * pi / 180.0 ) rlist = flex.reflection_table.from_predictions(exlist[0]) rlist["id"] = flex.int(len(rlist), 0) rlist.compute_bbox(exlist) rlist.compute_zeta_multi(exlist) rlist.compute_d(exlist) from libtbx.phil import parse from dials.algorithms.integration.integrator import Integrator3D, phil_scope params = phil_scope.fetch( parse( """ integration.block.size=%d integration.mp.nproc=%d integration.profile_fitting=False """ % (5, nproc) ) ).extract() integrator = Integrator3D(exlist, rlist, params) integrator.integrate() def test_summation(dials_data): from math import pi from dxtbx.model.experiment_list import ExperimentListFactory from dials.algorithms.profile_model.gaussian_rs import Model from dials.array_family import flex path = dials_data("centroid_test_data", pathlib=True) / "experiments.json" exlist = ExperimentListFactory.from_json_file(path) exlist[0].profile = Model( None, n_sigma=3, sigma_b=0.024 * pi / 180.0, sigma_m=0.044 * pi / 180.0 ) rlist = flex.reflection_table.from_predictions(exlist[0]) rlist["id"] = flex.int(len(rlist), 0) def integrate(integrator_type, rlist): from libtbx.phil import parse from dials.algorithms.integration.integrator import create_integrator from dials.algorithms.integration.integrator import ( phil_scope as master_phil_scope, ) rlist = rlist.copy() phil_scope = parse( f""" integration.background.algorithm=null integration.intensity.algorithm=sum integration.intensity.sum.integrator={integrator_type} integration.block.size=0.5 integration.profile_fitting=False """ ) params = master_phil_scope.fetch(source=phil_scope).extract() integrator = create_integrator(params, exlist, rlist) result = integrator.integrate() return result from libtbx.test_utils import approx_equal def approx_equal_dict(a, b, k): return approx_equal(a[k], b[k]) # Do summation by all different methods result1 = integrate("3d", rlist) result2 = integrate("flat3d", rlist) result3 = integrate("2d", rlist) result4 = integrate("single2d", rlist) assert len(result1) >= len(rlist) assert len(result2) >= len(rlist) assert len(result3) >= len(rlist) assert len(result4) >= len(rlist) # result1 and result2 should be the same assert len(result1) == len(result2) for r1, r2 in zip(result1.rows(), result2.rows()): assert r1["partial_id"] == r2["partial_id"] assert r1["bbox"] == r2["bbox"] assert r1["entering"] == r2["entering"] assert r1["flags"] == r2["flags"] assert r1["id"] == r2["id"] assert r1["miller_index"] == r2["miller_index"] assert r1["panel"] == r2["panel"] assert approx_equal_dict(r1, r2, "d") assert approx_equal_dict(r1, r2, "intensity.sum.value") assert approx_equal_dict(r1, r2, "intensity.sum.variance") assert approx_equal_dict(r1, r2, "lp") assert approx_equal_dict(r1, r2, "partiality") assert approx_equal_dict(r1, r2, "s1") assert approx_equal_dict(r1, r2, "xyzcal.mm") assert approx_equal_dict(r1, r2, "xyzcal.px") assert approx_equal_dict(r1, r2, "zeta") # result3 and result4 should be the same assert len(result3) == len(result4) for r3, r4 in zip(result3.rows(), result4.rows()): assert r3["partial_id"] == r4["partial_id"] assert r3["bbox"] == r4["bbox"] assert r3["entering"] == r4["entering"] assert r3["flags"] == r4["flags"] assert r3["id"] == r4["id"] assert r3["miller_index"] == r4["miller_index"] assert r3["panel"] == r4["panel"] assert approx_equal_dict(r3, r4, "d") assert approx_equal_dict(r3, r4, "intensity.sum.value") assert approx_equal_dict(r3, r4, "intensity.sum.variance") assert approx_equal_dict(r3, r4, "lp") assert approx_equal_dict(r3, r4, "partiality") assert approx_equal_dict(r3, r4, "s1") assert approx_equal_dict(r3, r4, "xyzcal.mm") assert approx_equal_dict(r3, r4, "xyzcal.px") assert approx_equal_dict(r3, r4, "xyzobs.px.value") assert approx_equal_dict(r3, r4, "xyzobs.px.variance") assert approx_equal_dict(r3, r4, "zeta") # result3 should add up to result1 assert len(result3) >= len(result1) expected1 = rlist.copy() expected1["intensity.sum.value"] = flex.double(len(rlist), 0) expected1["intensity.sum.variance"] = flex.double(len(rlist), 0) for r1 in result1.rows(): pid = r1["partial_id"] r2 = expected1[pid] assert r1["entering"] == r2["entering"] assert r1["id"] == r2["id"] assert r1["miller_index"] == r2["miller_index"] assert r1["panel"] == r2["panel"] assert approx_equal_dict(r1, r2, "s1") assert approx_equal_dict(r1, r2, "xyzcal.mm") assert approx_equal_dict(r1, r2, "xyzcal.px") expected1["intensity.sum.value"][pid] += r1["intensity.sum.value"] expected1["intensity.sum.variance"][pid] += r1["intensity.sum.variance"] expected3 = rlist.copy() expected3["intensity.sum.value"] = flex.double(len(rlist), 0) expected3["intensity.sum.variance"] = flex.double(len(rlist), 0) for r1 in result3.rows(): pid = r1["partial_id"] r2 = expected3[pid] assert r1["entering"] == r2["entering"] assert r1["id"] == r2["id"] assert r1["miller_index"] == r2["miller_index"] assert r1["panel"] == r2["panel"] assert approx_equal_dict(r1, r2, "s1") assert approx_equal_dict(r1, r2, "xyzcal.mm") assert approx_equal_dict(r1, r2, "xyzcal.px") expected3["intensity.sum.value"][pid] += r1["intensity.sum.value"] expected3["intensity.sum.variance"][pid] += r1["intensity.sum.variance"] for r1, r3 in zip(expected1.rows(), expected3.rows()): assert approx_equal_dict(r1, r3, "intensity.sum.value") assert approx_equal_dict(r1, r3, "intensity.sum.variance")
bsd-3-clause
7906468b4f2b313c98ad1ff8d8097dd3
31.085911
83
0.532826
3.229678
false
false
false
false
dials/dials
tests/algorithms/scaling/test_scaler_factory.py
1
14719
""" Tests for the scaler factory classes and helper functions. """ from __future__ import annotations from unittest.mock import MagicMock, Mock import pytest from dxtbx.model import Crystal from libtbx import phil from dials.algorithms.scaling.error_model.error_model import BasicErrorModel from dials.algorithms.scaling.scaler import ( MultiScaler, NullScaler, SingleScaler, TargetScaler, ) from dials.algorithms.scaling.scaler_factory import ( MultiScalerFactory, SingleScalerFactory, TargetScalerFactory, create_scaler, ) from dials.array_family import flex from dials.util.options import ArgumentParser def generated_refl(not_integrated=False, idval=0): """Generate a test reflection table.""" reflections = flex.reflection_table() reflections["intensity.prf.value"] = flex.double([1.0, 10.0, 100.0, 1.0]) reflections["intensity.prf.variance"] = flex.double([1.0, 10.0, 100.0, 1.0]) reflections["intensity.sum.value"] = flex.double([12.0, 120.0, 1200.0, 21.0]) reflections["intensity.sum.variance"] = flex.double([12.0, 120.0, 2100.0, 1.0]) # reflections['inverse_scale_factor'] = flex.double(4, 1.0) reflections["miller_index"] = flex.miller_index( [(1, 0, 0), (0, 0, 1), (2, 0, 0), (2, 2, 2)] ) # don't change reflections["d"] = flex.double([0.8, 2.0, 2.0, 0.0]) # don't change reflections["partiality"] = flex.double([1.0, 1.0, 1.0, 1.0]) reflections["xyzobs.px.value"] = flex.vec3_double( [(0.0, 0.0, 0.0), (0.0, 0.0, 5.0), (0.0, 0.0, 10.0), (0.0, 0.0, 10.0)] ) reflections["s1"] = flex.vec3_double( [(0.0, 0.1, 1.0), (0.0, 0.1, 1.0), (0.0, 0.1, 1.0), (0.0, 0.1, 1.0)] ) if not_integrated: reflections.set_flags( flex.bool([False, False, False, False]), reflections.flags.integrated ) else: reflections.set_flags( flex.bool([True, True, False, False]), reflections.flags.integrated ) reflections.set_flags( flex.bool([False, False, True, True]), reflections.flags.bad_for_scaling ) reflections["id"] = flex.int(4, idval) reflections.experiment_identifiers()[idval] = str(idval) return reflections @pytest.fixture def refl_to_filter(): """Generate a separate reflection table for filtering""" reflections = flex.reflection_table() reflections["partiality"] = flex.double([0.1, 1.0, 1.0, 1.0, 1.0, 1.0]) reflections["id"] = flex.int(reflections.size(), 0) reflections["intensity.sum.value"] = flex.double([1.0, 1.0, 1.0, 1.0, 1.0, 1.0]) reflections["intensity.sum.variance"] = flex.double([1.0, 1.0, 1.0, 1.0, 1.0, 1.0]) reflections.set_flags( flex.bool([True, False, True, True, True, True]), reflections.flags.integrated_sum, ) return reflections @pytest.fixture def prf_sum_refl_to_filter(): """Generate a separate reflection table for filtering""" reflections = flex.reflection_table() reflections["partiality"] = flex.double(5, 1.0) reflections["id"] = flex.int(reflections.size(), 0) reflections.experiment_identifiers()[0] = "0" reflections["intensity.sum.value"] = flex.double([1.0, 2.0, 3.0, 4.0, 5.0]) reflections["intensity.sum.variance"] = flex.double(5, 1.0) reflections["intensity.prf.value"] = flex.double([11.0, 12.0, 13.0, 14.0, 15.0]) reflections["intensity.prf.variance"] = flex.double(5, 1.0) reflections["miller_index"] = flex.miller_index([(0, 0, 1)] * 5) reflections.set_flags( flex.bool([False, False, True, True, True]), reflections.flags.integrated_sum, ) reflections.set_flags( flex.bool([True, False, False, True, True]), reflections.flags.integrated_prf, ) return reflections def test_refl_and_exp(mock_scaling_component, idval=0): r = test_refl(idval=idval) exp = mock_exp(mock_scaling_component, idval=idval) return r, exp def test_refl_and_exp_list(mock_scaling_component, n=1): rlist = [] explist = [] for i in range(n): rlist.append(test_refl(idval=i)) explist.append(mock_exp(mock_scaling_component, idval=i)) return rlist, explist def test_refl(idval=0): """Generate a test reflection table.""" return generated_refl(idval=idval) @pytest.fixture def refl_list(): """Make a list of three reflection tables.""" refl_list = [generated_refl()] refl_list.append(generated_refl()) refl_list.append(generated_refl()) return refl_list @pytest.fixture def generated_param(): """Generate a param phil scope.""" phil_scope = phil.parse( """ include scope dials.algorithms.scaling.scaling_options.phil_scope """, process_includes=True, ) parser = ArgumentParser(phil=phil_scope, check_format=False) parameters, _ = parser.parse_args(args=[], quick_parse=True, show_diff_phil=False) parameters.__inject__("model", "KB") parameters.scaling_options.free_set_percentage = 50.0 parameters.scaling_options.emax = 0 parameters.reflection_selection.method = "use_all" return parameters @pytest.fixture def mock_scaling_component(): """Mock scaling component to allow creation of a scaling model.""" component = MagicMock() component.n_params = 2 component.inverse_scales = flex.double([0.9, 1.1]) return component def mock_exp(mock_scaling_component, idval=0): """Mock experiments object for initialising a scaler.""" def side_effect_config_table(*args): """Side effect to mock configure reflection table call during initialisation.""" return args[0] exp = MagicMock() exp.identifier = str(idval) exp.scaling_model.components = {"scale": mock_scaling_component} exp.scaling_model.consecutive_refinement_order = ["scale"] exp.scaling_model.is_scaled = False exp.scaling_model.error_model = BasicErrorModel() exp.scaling_model.configure_reflection_table.side_effect = side_effect_config_table exp_dict = { "__id__": "crystal", "real_space_a": [1.0, 0.0, 0.0], "real_space_b": [0.0, 1.0, 0.0], "real_space_c": [0.0, 0.0, 2.0], "space_group_hall_symbol": " C 2y", } exp.crystal = Crystal.from_dict(exp_dict) exp.scan.get_oscillation.return_value = (0, 1.0) exp.beam.get_sample_to_source_direction.return_value = (0.0, 0.0, -1.0) exp.goniometer.get_rotation_axis.return_value = (0.0, 0.0, 1.0) return exp def mock_explist_3exp(mock_scaling_component): """A mock experimentlist, containing one mock exp instance three times.""" exp = [mock_exp(mock_scaling_component)] exp.append(mock_exp(mock_scaling_component)) exp.append(mock_exp(mock_scaling_component)) return exp @pytest.fixture def mock_scaled_exp(): """A mock experiments object with scaling_model.is_scaled = True""" exp = Mock() exp.scaling_model.is_scaled = True return exp @pytest.fixture def mock_unscaled_exp(): """A mock experiments object with scaling_model.is_scaled = False""" exp = Mock() exp.scaling_model.is_scaled = False return exp @pytest.fixture def mock_experimentlist(mock_scaled_exp, mock_unscaled_exp): """A mock experimentlist of mock scaled/unscaled mock exp.""" explist = [ mock_scaled_exp, mock_scaled_exp, mock_unscaled_exp, mock_scaled_exp, mock_unscaled_exp, ] return explist def test_SingleScalerFactory(generated_param, refl_to_filter, mock_scaling_component): """Test the single scaler factory.""" test_refl, exp = test_refl_and_exp(mock_scaling_component) # Test that all required attributes get added with standard params. assert all( (i not in test_refl) for i in ["inverse_scale_factor", "intensity", "variance"] ) # Test default, (no split into free set) ss = SingleScalerFactory.create(generated_param, exp, test_refl) assert isinstance(ss, SingleScaler) assert all( i in ss.reflection_table for i in ["inverse_scale_factor", "intensity", "variance"] ) # Test reflection filtering rt = SingleScalerFactory.filter_bad_reflections(refl_to_filter) assert list(rt.get_flags(rt.flags.excluded_for_scaling)) == [ True, True, False, False, False, False, ] def test_selection_of_profile_or_summation_intensities( generated_param, prf_sum_refl_to_filter, mock_scaling_component ): _, exp = test_refl_and_exp(mock_scaling_component) # Test that all required attributes get added with standard params. assert all( (i not in prf_sum_refl_to_filter) for i in ["inverse_scale_factor", "intensity", "variance"] ) # Test default, (no split into free set) ss = SingleScalerFactory.create(generated_param, exp, prf_sum_refl_to_filter) assert isinstance(ss, SingleScaler) rt = ss.reflection_table assert all(i in rt for i in ["inverse_scale_factor", "intensity", "variance"]) assert list(rt.get_flags(rt.flags.excluded_for_scaling)) == [ False, True, False, False, False, ] # test correct initial intensities have been chosen - should be prf then sum assert list(rt["intensity"]) == [11.0, 2.0, 3.0, 14.0, 15.0] def test_TargetScalerFactory(generated_param, mock_scaling_component): """Test the target scaler factory.""" refl_list, explist = test_refl_and_exp_list(mock_scaling_component, 3) # Test standard initialisation. assert generated_param.scaling_options.use_free_set is False # just to check explist[0].scaling_model.is_scaled = True explist[1].scaling_model.is_scaled = True target = TargetScalerFactory.create(generated_param, explist, refl_list) assert isinstance(target, TargetScaler) assert len(target.single_scalers) == 2 assert len(target.unscaled_scalers) == 1 assert set(target.single_scalers[0].reflection_table["id"]) == {0} assert set(target.single_scalers[1].reflection_table["id"]) == {1} assert set(target.unscaled_scalers[0].reflection_table["id"]) == {2} # Now test converting targetscaler to multiscaler multiscaler = MultiScalerFactory.create_from_targetscaler(target) assert isinstance(multiscaler, MultiScaler) assert len(multiscaler.single_scalers) == 3 # Test for correct initialisation when scaling against a target model. generated_param.scaling_options.reference = True target = TargetScalerFactory.create_for_target_against_reference( generated_param, explist, refl_list ) assert isinstance(target.single_scalers[0], NullScaler) # This time make one dataset bad, and check it gets removed refl_list, explist = test_refl_and_exp_list(mock_scaling_component, 3) generated_param.scaling_options.reference = False refl_list[1].unset_flags(flex.bool(4, True), refl_list[1].flags.integrated_prf) refl_list[1].unset_flags(flex.bool(4, True), refl_list[1].flags.integrated_sum) explist[0].scaling_model.is_scaled = True target = TargetScalerFactory.create(generated_param, explist, refl_list) assert isinstance(target, TargetScaler) assert len(target.single_scalers) == 1 assert len(target.unscaled_scalers) == 1 assert set(target.single_scalers[0].reflection_table["id"]) == {0} assert set(target.unscaled_scalers[0].reflection_table["id"]) == {2} refl_list, explist = test_refl_and_exp_list(mock_scaling_component, 3) refl_list[0].unset_flags(flex.bool(4, True), refl_list[0].flags.integrated_prf) refl_list[0].unset_flags(flex.bool(4, True), refl_list[0].flags.integrated_sum) explist[0].scaling_model.is_scaled = True explist[1].scaling_model.is_scaled = True target = TargetScalerFactory.create(generated_param, explist, refl_list) assert isinstance(target, TargetScaler) assert len(target.single_scalers) == 1 assert len(target.unscaled_scalers) == 1 assert set(target.single_scalers[0].reflection_table["id"]) == {1} assert set(target.unscaled_scalers[0].reflection_table["id"]) == {2} def test_MultiScalerFactory(generated_param, mock_scaling_component, refl_list): """Test the MultiScalerFactory.""" refl_list, explist = test_refl_and_exp_list(mock_scaling_component, 3) multiscaler = MultiScalerFactory.create(generated_param, explist, refl_list) assert isinstance(multiscaler, MultiScaler) assert len(multiscaler.single_scalers) == 3 for i in range(3): assert set(multiscaler.single_scalers[i].reflection_table["id"]) == {i} # This time make one dataset bad, and check it gets removed r1 = generated_refl(not_integrated=True) r2 = generated_refl() r3 = generated_refl() new_list = [r1, r2, r3] multiscaler = MultiScalerFactory.create( generated_param, mock_explist_3exp(mock_scaling_component), new_list ) assert isinstance(multiscaler, MultiScaler) assert len(multiscaler.single_scalers) == 2 r1 = multiscaler.single_scalers[0].reflection_table assert list(r1.get_flags(r1.flags.integrated)) == [True, True, False, False] r2 = multiscaler.single_scalers[1].reflection_table assert list(r2.get_flags(r2.flags.integrated)) == [True, True, False, False] def test_scaler_factory_helper_functions( mock_experimentlist, generated_param, refl_list, mock_scaling_component ): """Test the helper functions.""" test_refl, exp = test_refl_and_exp(mock_scaling_component) # Test create_scaler # Test case for single refl and exp scaler = create_scaler(generated_param, [exp], [test_refl]) assert isinstance(scaler, SingleScaler) # If none or allscaled explist = mock_explist_3exp(mock_scaling_component) scaler = create_scaler(generated_param, explist, refl_list) assert isinstance(scaler, MultiScaler) explist[0].scaling_model.is_scaled = False # ^ changes all in list as same instance of exp. scaler = create_scaler( generated_param, mock_explist_3exp(mock_scaling_component), refl_list ) assert isinstance(scaler, MultiScaler) # If only some scaled explist = [] explist.append(mock_exp(mock_scaling_component)) explist.append(mock_exp(mock_scaling_component)) explist[1].scaling_model.is_scaled = True r1 = generated_refl() r2 = generated_refl() refl_list = [r1, r2] scaler = create_scaler(generated_param, explist, refl_list) assert isinstance(scaler, TargetScaler) # If no reflections passed in. with pytest.raises(ValueError): scaler = create_scaler( generated_param, mock_explist_3exp(mock_scaling_component), [] )
bsd-3-clause
22cd3eced5a8cb9f11878688ef376287
35.889724
87
0.671377
3.279635
false
true
false
false
dials/dials
tests/command_line/test_indexed_as_integrated.py
1
1697
from __future__ import annotations import procrunner import iotbx.merging_statistics def test_indexed_as_integrated(dials_data, tmp_path): data_dir = dials_data("insulin_processed", pathlib=True) refl = data_dir / "indexed.refl" expt = data_dir / "indexed.expt" command = [ "dials.indexed_as_integrated", refl, expt, "output.reflections=output.refl", ] result = procrunner.run(command, working_directory=tmp_path) assert not result.returncode and not result.stderr assert (tmp_path / "output.refl").is_file() # now make sure we can run dials.symmetry and dials.scale successfully sym_command = ["dials.symmetry", tmp_path / "output.refl", expt] result = procrunner.run(sym_command, working_directory=tmp_path) assert not result.returncode and not result.stderr assert (tmp_path / "symmetrized.refl").is_file() assert (tmp_path / "symmetrized.expt").is_file() scale_command = [ "dials.scale", tmp_path / "symmetrized.refl", tmp_path / "symmetrized.expt", "unmerged_mtz=scaled.mtz", "d_min=2.0", ] result = procrunner.run(scale_command, working_directory=tmp_path) assert not result.returncode and not result.stderr assert (tmp_path / "scaled.refl").is_file() assert (tmp_path / "scaled.expt").is_file() assert (tmp_path / "scaled.mtz").is_file() i_obs = iotbx.merging_statistics.select_data(str(tmp_path / "scaled.mtz"), None) result = iotbx.merging_statistics.dataset_statistics( i_obs=i_obs, use_internal_variance=False, eliminate_sys_absent=False, ) assert result.overall.cc_one_half > 0.8
bsd-3-clause
34affdffd1066f520793174a88933d6f
32.94
84
0.657042
3.20794
false
false
false
false
dials/dials
src/dials/algorithms/refinement/parameterisation/scan_varying_crystal_parameters.py
1
7512
from __future__ import annotations from scitbx import matrix from dials.algorithms.refinement.parameterisation.crystal_parameters import ( CrystalOrientationMixin, CrystalUnitCellMixin, ) from dials.algorithms.refinement.parameterisation.scan_varying_model_parameters import ( GaussianSmoother, ScanVaryingModelParameterisation, ScanVaryingParameterSet, ) from dials.algorithms.refinement.refinement_helpers import CrystalOrientationCompose class ScanVaryingCrystalOrientationParameterisation( ScanVaryingModelParameterisation, CrystalOrientationMixin ): """Scan-varying parameterisation for crystal orientation, with angles expressed in mrad""" def __init__(self, crystal, t_range, num_intervals, experiment_ids=None): if experiment_ids is None: experiment_ids = [0] # The state of a scan varying crystal orientation parameterisation # is an orientation # matrix '[U](t)', expressed as a function of image number 't' # in a sequential scan. # # The initial state is a snapshot of the crystal orientation # at the point of initialisation '[U0]', which is independent of # image number. # # Future states are composed by # rotations around axes of the phi-axis frame by Tait-Bryan angles. # # [U](t) = [Phi3](t)[Phi2](t)[Phi1](t)[U0] # Set up the smoother smoother = GaussianSmoother(t_range, num_intervals) nv = smoother.num_values() # Set up the initial state istate = matrix.sqr(crystal.get_U()) self._U_at_t = istate # Factory function to provide to _build_p_list def parameter_type(value, axis, ptype, name): return ScanVaryingParameterSet(value, nv, axis, ptype, name) # Build the parameter list p_list = self._build_p_list(parameter_type) # Set up the base class ScanVaryingModelParameterisation.__init__( self, crystal, istate, p_list, smoother, experiment_ids=experiment_ids ) return def compose(self, t): """calculate state and derivatives for model at image number t""" # Extract orientation from the initial state U0 = self._initial_state # extract parameter sets from the internal list phi1_set, phi2_set, phi3_set = self._param # extract angles and other data at time t using the smoother phi1, phi1_weights, phi1_sumweights = self._smoother.value_weight(t, phi1_set) phi2, phi2_weights, phi2_sumweights = self._smoother.value_weight(t, phi2_set) phi3, phi3_weights, phi3_sumweights = self._smoother.value_weight(t, phi3_set) # calculate derivatives of angles wrt underlying parameters. dphi1_dp = phi1_weights * (1.0 / phi1_sumweights) dphi2_dp = phi2_weights * (1.0 / phi2_sumweights) dphi3_dp = phi3_weights * (1.0 / phi3_sumweights) # calculate state and derivatives using the helper class coc = CrystalOrientationCompose( U0, phi1, phi1_set.axis, phi2, phi2_set.axis, phi3, phi3_set.axis ) self._U_at_t = coc.U() dU_dphi1 = coc.dU_dphi1() dU_dphi2 = coc.dU_dphi2() dU_dphi3 = coc.dU_dphi3() # calculate derivatives of state wrt underlying parameters dU_dp1 = [None] * dphi1_dp.size for (i, v) in dphi1_dp: dU_dp1[i] = dU_dphi1 * v dU_dp2 = [None] * dphi2_dp.size for (i, v) in dphi2_dp: dU_dp2[i] = dU_dphi2 * v dU_dp3 = [None] * dphi3_dp.size for (i, v) in dphi3_dp: dU_dp3[i] = dU_dphi3 * v # store derivatives as list-of-lists self._dstate_dp = [dU_dp1, dU_dp2, dU_dp3] return def get_state(self): """Return crystal orientation matrix [U] at image number t""" # only a single crystal is parameterised here, so no multi_state_elt # argument is allowed return self._U_at_t class ScanVaryingCrystalUnitCellParameterisation( ScanVaryingModelParameterisation, CrystalUnitCellMixin ): """Scan-varying parameterisation for the crystal unit cell""" def __init__( self, crystal, t_range, num_intervals, experiment_ids=None, set_state_uncertainties=False, ): self._set_state_uncertainties = set_state_uncertainties from scitbx import matrix if experiment_ids is None: experiment_ids = [0] # The state of a scan-varying unit cell parameterisation is the # reciprocal space orthogonalisation matrix '[B](t)', expressed as a # function of image number 't' in a sequential scan. # Other comments from CrystalUnitCellParameterisation are relevant here # Set up the smoother smoother = GaussianSmoother(t_range, num_intervals) nv = smoother.num_values() # Set up the initial state istate = None self._B_at_t = matrix.sqr(crystal.get_B()) # Factory function to provide to _build_p_list def parameter_type(value, name): return ScanVaryingParameterSet(value, nv, name=name) # Build the parameter list p_list = self._build_p_list(crystal, parameter_type) # Set up the base class ScanVaryingModelParameterisation.__init__( self, crystal, istate, p_list, smoother, experiment_ids=experiment_ids ) return def compose(self, t): """calculate state and derivatives for model at image number t""" # extract values and weights at time t using the smoother vals, weights, sumweights = zip( *(self._smoother.value_weight(t, pset) for pset in self._param) ) # calculate derivatives of metrical matrix parameters wrt underlying # scan-varying parameters inv_sumw = [1.0 / sw for sw in sumweights] dvals_dp = [e * isw for e, isw in zip(weights, inv_sumw)] # calculate new B and derivatives self._B_at_t, dB_dval = self._compose_core(vals) # calculate derivatives of state wrt underlying parameters self._dstate_dp = [ [b * e for e in a.as_dense_vector()] for a, b in zip(dvals_dp, dB_dval) ] self._dstate_dp = [[None] * e.size for e in dvals_dp] for i, (dv, dB) in enumerate(zip(dvals_dp, dB_dval)): for j, e in dv: self._dstate_dp[i][j] = e * dB return def get_state(self): """Return crystal orthogonalisation matrix [B] at image number t""" # only a single crystal is parameterised here, so no multi_state_elt # argument is allowed return self._B_at_t def set_state_uncertainties(self, var_cov_list): """Send the calculated variance-covariance of the elements of the B matrix for all scan points back to the crystal model, if required """ if not self._set_state_uncertainties: return # Convert list of 9*9 matrices to a 3d array from scitbx.array_family import flex B_cov = flex.double(flex.grid(len(var_cov_list), 9, 9)) for i, v in enumerate(var_cov_list): v = v.as_flex_double_matrix() v.reshape(flex.grid(1, 9, 9)) B_cov[i : (i + 1), :, :] = v # Pass it back to the model self._model.set_B_covariance_at_scan_points(B_cov)
bsd-3-clause
fbf7cc4df184544d16825101321f6b7c
33.617512
88
0.622204
3.500466
false
false
false
false
dials/dials
src/dials/command_line/show.py
1
22945
from __future__ import annotations import os import sys import numpy as np import iotbx.phil from cctbx import uctbx from dxtbx.model.experiment_list import ExperimentListFactory from scitbx.math import five_number_summary import dials.util from dials.array_family import flex from dials.util import Sorry, tabulate help_message = """ Examples:: dials.show models.expt dials.show image_*.cbf dials.show observations.refl """ phil_scope = iotbx.phil.parse( """\ show_scan_varying = False .type = bool .help = "Whether or not to show the crystal at each scan point." show_shared_models = False .type = bool .help = "Show which models are linked to which experiments" show_all_reflection_data = False .type = bool .help = "Whether or not to print individual reflections" show_intensities = False .type = bool show_centroids = False .type = bool show_profile_fit = False .type = bool show_flags = False .type = bool .help = "Show a summary table of reflection flags" show_identifiers = False .type = bool .help = "Show experiment identifiers map if set" image_statistics{ show_corrected = False .type = bool .help = "Show statistics on the distribution of values in each corrected image" show_raw = False .type = bool .help = "Show statistics on the distribution of values in each raw image" } max_reflections = None .type = int .help = "Limit the number of reflections in the output." """, process_includes=True, ) def beam_centre_mm(detector, s0): x, y = (None, None) for panel_id, panel in enumerate(detector): try: x, y = panel.get_ray_intersection(s0) except RuntimeError: continue else: if panel.is_coord_valid_mm((x, y)): break else: x, y = (None, None) return panel_id, (x, y) def beam_centre_raw_image_px(detector, s0): panel_id, (x, y) = beam_centre_mm(detector, s0) panel = detector[panel_id] x_px, y_px = panel.millimeter_to_pixel((x, y)) offset = panel.get_raw_image_offset() return x_px + offset[0], y_px + offset[1] def show_beam(detector, beam): # standard static beam model string s = str(beam) # report whether the beam is scan-varying if beam.num_scan_points > 0: s += " s0 sampled at " + str(beam.num_scan_points) + " scan points\n" # add static model beam centres panel_id, (x, y) = beam_centre_mm(detector, beam.get_s0()) if panel_id >= 0 and x is not None and y is not None: x_px, y_px = detector[panel_id].millimeter_to_pixel((x, y)) if len(detector) > 1: beam_centre_mm_str = " mm: panel %i, (%.2f,%.2f)" % (panel_id, x, y) beam_centre_px_str = " px: panel %i, (%.2f,%.2f)" % ( panel_id, x_px, y_px, ) x_raw_px, y_raw_px = beam_centre_raw_image_px(detector, beam.get_s0()) beam_centre_raw_px_str = " px, raw image: ({:.2f},{:.2f})".format( x_raw_px, y_raw_px, ) x_raw_mm, y_raw_mm = detector[panel_id].pixel_to_millimeter( (x_raw_px, y_raw_px) ) beam_centre_raw_mm_str = " mm, raw image: ({:.2f},{:.2f})".format( x_raw_mm, y_raw_mm, ) else: beam_centre_mm_str = f" mm: ({x:.2f},{y:.2f})" beam_centre_px_str = f" px: ({x_px:.2f},{y_px:.2f})" beam_centre_raw_px_str = "" beam_centre_raw_mm_str = "" s += "\nBeam centre: \n" s += beam_centre_mm_str + "\n" + beam_centre_px_str + "\n" if beam_centre_raw_mm_str: s += beam_centre_raw_mm_str + "\n" if beam_centre_raw_px_str: s += beam_centre_raw_px_str + "\n" # report range of scan-varying model beam centres if beam.num_scan_points > 0: # get scan-varying beam centres, ensuring all on same panel sv_s0 = beam.get_s0_at_scan_points() impacts = [beam_centre_mm(detector, s0) for s0 in sv_s0] pnl, xy = zip(*impacts) uniq_pnls = set(pnl) if len(uniq_pnls) > 1 or min(uniq_pnls) < 0: return s if any(e == (None, None) for e in xy): return s pnl = list(uniq_pnls)[0] x_mm, y_mm = zip(*xy) # convert to pixels xy = [detector[pnl].millimeter_to_pixel(e) for e in xy] x_px, y_px = zip(*xy) s += "Beam centre range (mm): ([{:.2f},{:.2f}],[{:.2f},{:.2f}])\n".format( min(x_mm), max(x_mm), min(y_mm), max(y_mm), ) s += "Beam centre range (px): ([{:.2f},{:.2f}],[{:.2f},{:.2f}])\n".format( min(x_px), max(x_px), min(y_px), max(y_px), ) return s def show_goniometer(goniometer): # standard static goniometer model string s = str(goniometer) # report whether the goniometer is scan-varying if goniometer.num_scan_points > 0: s += ( " Setting rotation sampled at " + str(goniometer.num_scan_points) + " scan points\n" ) return s @dials.util.show_mail_handle_errors() def run(args=None): import dials.util.log dials.util.log.print_banner() from dials.util.options import ( ArgumentParser, reflections_and_experiments_from_files, ) usage = "dials.show [options] models.expt | image_*.cbf" parser = ArgumentParser( usage=usage, phil=phil_scope, read_experiments=True, read_experiments_from_images=True, read_reflections=True, check_format=False, epilog=help_message, ) params, options = parser.parse_args(args=args, show_diff_phil=True) reflections, experiments = reflections_and_experiments_from_files( params.input.reflections, params.input.experiments ) if len(experiments) == 0 and len(reflections) == 0: parser.print_help() exit() if len(experiments): if not all(e.detector for e in experiments): sys.exit("Error: experiment has no detector") if not all(e.beam for e in experiments): sys.exit("Error: experiment has no beam") print(show_experiments(experiments, show_scan_varying=params.show_scan_varying)) if params.image_statistics.show_raw: show_image_statistics(experiments, "raw") if params.image_statistics.show_corrected: show_image_statistics(experiments, "corrected") if params.show_shared_models: print() print(model_connectivity(experiments)) if len(reflections): print( show_reflections( reflections, show_intensities=params.show_intensities, show_profile_fit=params.show_profile_fit, show_centroids=params.show_centroids, show_all_reflection_data=params.show_all_reflection_data, show_flags=params.show_flags, max_reflections=params.max_reflections, show_identifiers=params.show_identifiers, ) ) def show_experiments(experiments, show_scan_varying=False): text = [] for i_expt, expt in enumerate(experiments): text.append("Experiment %i:" % i_expt) if expt.imageset: format_class = expt.imageset.get_format_class() if not format_class.is_abstract(): text.append(f"Format class: {format_class.__name__}") if expt.identifier != "": text.append(f"Experiment identifier: {expt.identifier}") # get_template() only exists for ImageSequence, not ImageSet try: text.append(f"Image template: {expt.imageset.get_template()}") except AttributeError: pass text.append(str(expt.detector)) text.append( "Max resolution (at corners): %f" % (expt.detector.get_max_resolution(expt.beam.get_s0())) ) text.append( "Max resolution (inscribed): %f" % (expt.detector.get_max_inscribed_resolution(expt.beam.get_s0())) ) text.append("") text.append(show_beam(expt.detector, expt.beam)) if expt.scan is not None: text.append(str(expt.scan)) if expt.goniometer is not None: text.append(show_goniometer(expt.goniometer)) if expt.crystal is not None: text.append(expt.crystal.as_str(show_scan_varying=show_scan_varying)) if expt.crystal.num_scan_points: abc = flex.vec3_double() angles = flex.vec3_double() for n in range(expt.crystal.num_scan_points): ( a, b, c, alpha, beta, gamma, ) = expt.crystal.get_unit_cell_at_scan_point(n).parameters() abc.append((a, b, c)) angles.append((alpha, beta, gamma)) a, b, c = abc.mean() alpha, beta, gamma = angles.mean() mean_unit_cell = uctbx.unit_cell((a, b, c, alpha, beta, gamma)) text.append(f" Average unit cell: {mean_unit_cell}") if expt.profile is not None: text.append(str(expt.profile)) if expt.scaling_model is not None: text.append(str(expt.scaling_model)) return "\n".join(text) def show_image_statistics(experiments, im_type): if im_type == "raw": raw = True elif im_type == "corrected": raw = False else: raise ValueError(f"Unknown im_type: {im_type}") # To show image statistics, check_format has to be true. So we have to reinstatiate # the experiment list here try: experiments = ExperimentListFactory.from_json( experiments.as_json(), check_format=True ) except OSError as e: raise Sorry( f"Unable to read image data. Please check {e.filename} is accessible" ) print(f"Five number summary of the {im_type} images") for i_expt, expt in enumerate(experiments): for i in range(len(expt.imageset)): identifier = os.path.basename(expt.imageset.get_image_identifier(i)) if raw: pnl_data = expt.imageset.get_raw_data(i) else: pnl_data = expt.imageset.get_corrected_data(i) if not isinstance(pnl_data, tuple): pnl_data = (pnl_data,) flat_data = pnl_data[0].as_1d() for p in pnl_data[1:]: flat_data.extend(p.as_1d()) fns = five_number_summary(flat_data) print( "{}: Min: {:.1f} Q1: {:.1f} Med: {:.1f} Q3: {:.1f} Max: {:.1f}".format( identifier, *fns ) ) def model_connectivity(experiments): def model_connectivity_impl(experiments, model): text = [""] text.append(f"{model.capitalize()}:") models = getattr(experiments, f"{model}s")() rows = [[""] + [str(j) for j in range(len(models))]] for j, e in enumerate(experiments): row = ["Experiment %d" % j] for m in models: if getattr(e, model) is m: row.append("x") else: row.append(".") rows.append(row) text.append(tabulate(rows, tablefmt="plain")) return text if len(experiments) == 1: return "" text = [] text.append("Experiment / Models") text.extend(model_connectivity_impl(experiments, "detector")) text.extend(model_connectivity_impl(experiments, "crystal")) text.extend(model_connectivity_impl(experiments, "beam")) return "\n".join(text) def _create_flag_count_table(table): """Generate a summary table of flag values in a reflection table. :param table: A reflection table :returns: A string of the formatted flags table """ # Calculate the counts of entries that match each flag numpy_flags = table["flags"].as_numpy_array() flag_count = { flag: np.sum(numpy_flags & value != 0) for value, flag in table.flags.values.items() } # Work out the numeric-value order of the flags flag_order = sorted(table.flags.values.values(), key=lambda x: x.real) # Build the actual table flag_rows = [["Flag", "Count", "%"]] max_count_len = max(5, len(str(max(flag_count.values())))) last_flag = None for flag in flag_order: indent = "" # As a hint for reading, indent any 'summary' flags. # A summary flag is any flag which overlaps with the previous one. if last_flag and (last_flag.real & flag.real): indent = " " last_flag = flag # Add the row to the table we're building flag_rows.append( [ indent + flag.name, "{:{:d}d}".format(flag_count[flag], max_count_len), f"{100 * flag_count[flag] / len(table):5.01f}", ] ) # Build the array of output strings text = [] text.append("Reflection flags:") text.append(tabulate(flag_rows, headers="firstrow")) return "\n".join(text) def show_reflections( reflections, show_intensities=False, show_profile_fit=False, show_centroids=False, show_all_reflection_data=False, show_flags=False, max_reflections=None, show_identifiers=False, ): text = [] from orderedset import OrderedSet formats = { "miller_index": "%i, %i, %i", "d": "%.2f", "qe": "%.3f", "dqe": "%.3f", "id": "%i", "imageset_id": "%i", "panel": "%i", "flags": "%i", "background.mean": "%.1f", "background.dispersion": "%.1f", "background.mse": "%.1f", "background.sum.value": "%.1f", "background.sum.variance": "%.1f", "intensity.prf.value": "%.1f", "intensity.prf.variance": "%.1f", "intensity.sum.value": "%.1f", "intensity.sum.variance": "%.1f", "intensity.cor.value": "%.1f", "intensity.cor.variance": "%.1f", "intensity.scale.value": "%.1f", "intensity.scale.variance": "%.1f", "Ih_values": "%.1f", "lp": "%.3f", "num_pixels.background": "%i", "num_pixels.background_used": "%i", "num_pixels.foreground": "%i", "num_pixels.valid": "%i", "partial_id": "%i", "partiality": "%.4f", "profile.correlation": "%.3f", "profile.rmsd": "%.3f", "xyzcal.mm": "%.2f, %.2f, %.2f", "xyzcal.px": "%.2f, %.2f, %.2f", "delpsical.rad": "%.3f", "delpsical2": "%.3f", "delpsical.weights": "%.3f", "xyzobs.mm.value": "%.2f, %.2f, %.2f", "xyzobs.mm.variance": "%.4e, %.4e, %.4e", "xyzobs.px.value": "%.2f, %.2f, %.2f", "xyzobs.px.variance": "%.4f, %.4f, %.4f", "s1": "%.4f, %.4f, %.4f", "s2": "%.4f, %.4f, %.4f", "shoebox": "%.1f", "rlp": "%.4f, %.4f, %.4f", "zeta": "%.3f", "x_resid": "%.3f", "x_resid2": "%.3f", "y_resid": "%.3f", "y_resid2": "%.3f", "kapton_absorption_correction": "%.3f", "kapton_absorption_correction_sigmas": "%.3f", "inverse_scale_factor": "%.3f", "inverse_scale_factor_variance": "%.3f", } for rlist in reflections: from dials.algorithms.shoebox import MaskCode foreground_valid = MaskCode.Valid | MaskCode.Foreground text.append("") text.append(f"Reflection list contains {len(rlist)} reflections") if len(rlist) == 0: continue rows = [["Column", "min", "max", "mean"]] for k, col in rlist.cols(): if k in formats and "%" not in formats.get(k, "%s"): # Allow blanking out of entries that wouldn't make sense rows.append( [ k, formats.get(k, "%s"), formats.get(k, "%s"), formats.get(k, "%s"), ] ) elif type(col) in (flex.double, flex.int, flex.size_t): if type(col) in (flex.int, flex.size_t): col = col.as_double() rows.append( [ k, formats.get(k, "%s") % flex.min(col), formats.get(k, "%s") % flex.max(col), formats.get(k, "%s") % flex.mean(col), ] ) elif type(col) in (flex.vec3_double, flex.miller_index): if isinstance(col, flex.miller_index): col = col.as_vec3_double() rows.append( [ k, formats.get(k, "%.2f, %.2f, %.2f") % col.min(), formats.get(k, "%.2f, %.2f, %.2f") % col.max(), formats.get(k, "%.2f, %.2f, %.2f") % col.mean(), ] ) elif isinstance(col, flex.shoebox): rows.append([k, "", "", ""]) si = col.summed_intensity().observed_value() rows.append( [ " summed I", formats.get(k, "%s") % flex.min(si), formats.get(k, "%s") % flex.max(si), formats.get(k, "%s") % flex.mean(si), ] ) x1, x2, y1, y2, z1, z2 = col.bounding_boxes().parts() bbox_sizes = ((z2 - z1) * (y2 - y1) * (x2 - x1)).as_double() rows.append( [ " N pix", formats.get(k, "%s") % flex.min(bbox_sizes), formats.get(k, "%s") % flex.max(bbox_sizes), formats.get(k, "%s") % flex.mean(bbox_sizes), ] ) fore_valid = col.count_mask_values(foreground_valid).as_double() rows.append( [ " N valid foreground pix", formats.get(k, "%s") % flex.min(fore_valid), formats.get(k, "%s") % flex.max(fore_valid), formats.get(k, "%s") % flex.mean(fore_valid), ] ) text.append(tabulate(rows, headers="firstrow")) if show_flags: text.append(_create_flag_count_table(rlist)) if show_identifiers: if rlist.experiment_identifiers(): text.append( """Experiment identifiers id-map values:\n%s""" % ( "\n".join( "id:" + str(k) + " -> experiment identifier:" + str(rlist.experiment_identifiers()[k]) for k in rlist.experiment_identifiers().keys() ) ) ) intensity_keys = ( "miller_index", "d", "intensity.prf.value", "intensity.prf.variance", "intensity.sum.value", "intensity.sum.variance", "background.mean", "profile.correlation", "profile.rmsd", ) profile_fit_keys = ("miller_index", "d") centroid_keys = ( "miller_index", "d", "xyzcal.mm", "xyzcal.px", "xyzobs.mm.value", "xyzobs.mm.variance", "xyzobs.px.value", "xyzobs.px.variance", ) keys_to_print = OrderedSet() if show_intensities: for k in intensity_keys: keys_to_print.add(k) if show_profile_fit: for k in profile_fit_keys: keys_to_print.add(k) if show_centroids: for k in centroid_keys: keys_to_print.add(k) if show_all_reflection_data: for k in formats: keys_to_print.add(k) def format_column(key, data, format_strings=None): if isinstance(data, flex.vec3_double): c_strings = [ c.as_string(format_strings[i].strip()) for i, c in enumerate(data.parts()) ] elif isinstance(data, flex.miller_index): c_strings = [ c.as_string(format_strings[i].strip()) for i, c in enumerate(data.as_vec3_double().parts()) ] elif isinstance(data, flex.size_t): c_strings = [data.as_int().as_string(format_strings[0].strip())] elif isinstance(data, flex.shoebox): x1, x2, y1, y2, z1, z2 = data.bounding_boxes().parts() bbox_sizes = ((z2 - z1) * (y2 - y1) * (x2 - x1)).as_double() c_strings = [bbox_sizes.as_string(format_strings[0].strip())] key += " (N pix)" else: c_strings = [data.as_string(format_strings[0].strip())] column = flex.std_string() max_element_lengths = [c.max_element_length() for c in c_strings] for i in range(len(c_strings[0])): column.append( f"%{len(key)}s" % ", ".join( ("%%%is" % max_element_lengths[j]) % c_strings[j][i] for j in range(len(c_strings)) ) ) return column if keys_to_print: keys = [k for k in keys_to_print if k in rlist] if max_reflections is not None: max_reflections = min(len(rlist), max_reflections) else: max_reflections = len(rlist) columns = [] for k in keys: columns.append( format_column(k, rlist[k], format_strings=formats[k].split(",")) ) text.append("") text.append("Printing %i of %i reflections:" % (max_reflections, len(rlist))) line = [] for j in range(len(columns)): key = keys[j] if key == "shoebox": key += " (N pix)" width = max(len(key), columns[j].max_element_length()) line.append("%%%is" % width % key) text.append(" ".join(line)) for i in range(max_reflections): line = (c[i] for c in columns) text.append(" ".join(line)) return "\n".join(text) if __name__ == "__main__": run()
bsd-3-clause
ccb5b7cf5216cdaca4c4f2d9b2cbb8bf
32.014388
88
0.507823
3.544724
false
false
false
false
douban/dpark
dpark/utils/tdigest.py
1
7283
from math import isnan, ceil, pi from six.moves import range class Centroid(object): def __init__(self, x, w=1): self.__mean = float(x) self.__count = float(w) @property def mean(self): return self.__mean @property def count(self): return self.__count def __repr__(self): return """<Centroid: mean==%.8f, count=%d>""" % (self.mean, self.count) def __eq__(self, other): if isinstance(other, Centroid): return self.mean == other.mean and self.count == other.count return False class TDigest(object): def __init__(self, compression=100, size=None): self._min = None self._max = None self.compression = compression self._total_weight = 0 self._weight = [] self._mean = [] self._unmerge_weight = 0 self._tmp_weight = [] self._tmp_mean = [] if size is None: size = int(2 * ceil(compression)) + 10 self._size = size @staticmethod def _weighted_average(x1, w1, x2, w2): a, b = min(x1, x2), max(x1, x2) x = float(x1 * w1 + x2 * w2) / (w1 + w2) return max(a, min(b, x)) def __len__(self): return int(self._total_weight + self._unmerge_weight) def __add__(self, other): if not isinstance(other, TDigest): raise TypeError('Can not add {} with {}'.format( self.__class__.__name__, other.__class__.__name__, )) if len(other) == 0: return self other.compress() self._tmp_mean.extend(other._mean) self._tmp_weight.extend(other._weight) total = sum(other._weight) self._unmerge_weight = total self.compress() return self def add(self, x, w=1): x = float(x) w = float(w) if isnan(x): raise ValueError('Cannot add NaN') if len(self._tmp_weight) + len(self._weight) >= self._size - 1: self.compress() self._tmp_weight.append(w) self._tmp_mean.append(x) self._unmerge_weight += w def compress(self): if self._unmerge_weight > 0: self._merge(self._tmp_weight, self._tmp_mean) self._tmp_weight = [] self._tmp_mean = [] self._unmerge_weight = 0 def _merge(self, incoming_weight, incoming_mean): def _argsort(seq): return sorted(range(len(seq)), key=seq.__getitem__) incoming_weight = incoming_weight + self._weight incoming_mean = incoming_mean + self._mean assert incoming_weight incoming_order = _argsort(incoming_mean) self._total_weight += self._unmerge_weight normalizer = self.compression / (pi * self._total_weight) mean = [] weight = [] mean.append(incoming_mean[incoming_order[0]]) weight.append(incoming_weight[incoming_order[0]]) w_so_far = 0. for ix in incoming_order[1:]: proposed_weight = weight[-1] + incoming_weight[ix] z = proposed_weight * normalizer q0 = w_so_far / self._total_weight q2 = (w_so_far + proposed_weight) / self._total_weight if z * z <= q0 * (1 - q0) and z * z <= q2 * (1 - q2): weight[-1] += incoming_weight[ix] mean[-1] = mean[-1] + (incoming_mean[ix] - mean[-1]) * incoming_weight[ix] / weight[-1] else: w_so_far += weight[-1] mean.append(incoming_mean[ix]) weight.append(incoming_weight[ix]) self._mean = mean self._weight = weight # assert sum(weight) == self._total_weight if self._total_weight > 0: self._min = mean[0] if self._min is None else min(self._min, mean[0]) self._max = mean[-1] if self._max is None else max(self._max, mean[-1]) def quantile(self, q): q = float(q) if not 0 <= q <= 1: raise ValueError('q should be in [0, 1], got {}'.format(q)) self.compress() weight = self._weight mean = self._mean if not weight: return float('nan') elif len(weight) == 1: return mean[0] index = q * self._total_weight if index < weight[0] / 2: return self._min + 2. * index / weight[0] * (mean[0] - self._min) weight_so_far = weight[0] / 2. for i in range(len(weight) - 1): dw = (weight[i] + weight[i + 1]) / 2. if weight_so_far + dw > index: z1 = index - weight_so_far z2 = weight_so_far + dw - index return self._weighted_average(mean[i], z2, mean[i + 1], z1) weight_so_far += dw assert index <= self._total_weight assert index >= self._total_weight - weight[-1] / 2. z1 = index - self._total_weight - weight[-1] / 2. z2 = weight[-1] / 2. - z1 return self._weighted_average(mean[-1], z1, self._max, z2) def cdf(self, x): x = float(x) self.compress() weight = self._weight mean = self._mean if not weight: return float('nan') elif len(weight) == 1: width = self._max - self._min if x < self._min: return 0. elif x > self._max: return 1. elif x - self._min <= width: return 0.5 else: return (x - self._min) / (self._max - self._min) if x < self._min: return 0. if x > self._max: return 1. if x <= mean[0]: if mean[0] - self._min > 0: return (x - self._min) / (mean[0] - self._min) * weight[0] / self._total_weight / 2. else: return 0. if x >= mean[-1]: if self._max - mean[-1] > 0: return 1. - (self._max - x) / (self._max - mean[-1]) * weight[-1] / self._total_weight / 2. else: return 1. weight_so_far = weight[0] / 2. for it in range(len(weight) - 1): if mean[it] == x: w0 = weight_so_far weight_so_far += sum( weight[i] + weight[i + 1] for i in range(it, len(weight) - 1) if mean[i + 1] == x ) return (w0 + weight_so_far) / 2. / self._total_weight if mean[it] <= x < mean[it + 1]: if mean[it + 1] - mean[it] > 0: dw = (weight[it] + weight[it + 1]) / 2. return (weight_so_far + dw * (x - mean[it]) / (mean[it + 1] - mean[it])) / self._total_weight else: dw = (weight[it] + weight[it + 1]) / 2. return weight_so_far + dw / self._total_weight weight_so_far += (weight[it] + weight[it + 1]) / 2. assert False @property def centroids(self): self.compress() weight = self._weight mean = self._mean return [Centroid(mean[i], weight[i]) for i in range(len(self._weight))]
bsd-3-clause
acaa15c58ccbd1eeb1c42a46c511c01c
29.219917
107
0.481395
3.670867
false
false
false
false
dials/dials
src/dials/command_line/cluster_exec.py
1
1592
# LIBTBX_SET_DISPATCHER_NAME cluster.dials.exec from __future__ import annotations import pickle import dials.util def get_cwd(): """ Get the current working directory """ import sys return sys.argv[1] def get_tid(): """ Get the task id """ import os # FIXME At the moment, there is no portable way to know the task id through # drmaa. This is really annoying. So we have to use the SGE_TASK_ID. # Therefore, this will only work for SGE. We can probably add support for # other systems as and when needed. if "SGE_TASK_ID" in os.environ: return os.environ["SGE_TASK_ID"] else: raise KeyError("Could not find task id") @dials.util.show_mail_handle_errors() def run(_=None): import traceback from os.path import exists, join from time import sleep # Get the task id and the current working directory tid = get_tid() cwd = get_cwd() # Set the paths input_fn = join(cwd, f"{tid}.input") output_fn = join(cwd, f"{tid}.output") # Wait until it exists while not exists(input_fn): sleep(1) # Try to run the function, otherwise return an exception try: with open(input_fn, "rb") as infile: function, element = pickle.load(infile) result = function(element) except Exception as e: e.args = [traceback.format_exc()] result = e # Dump the result with open(output_fn, "wb") as outfile: pickle.dump(result, outfile, protocol=pickle.HIGHEST_PROTOCOL) if __name__ == "__main__": run()
bsd-3-clause
1458942fd2c7c727aa66112d75ab01c6
22.072464
79
0.625
3.668203
false
false
false
false
dials/dials
src/dials/util/image_viewer/slip_viewer/calibration_frame.py
1
7837
from __future__ import annotations import os import wx from scitbx.matrix import col class SBSettingsFrame(wx.MiniFrame): def __init__(self, *args, **kwds): super().__init__(*args, **kwds) szr = wx.BoxSizer(wx.VERTICAL) panel = SBSettingsPanel(self) self.SetSizer(szr) szr.Add(panel, 1, wx.EXPAND) szr.Fit(panel) self.panel = panel self.sizer = szr self.Fit() self.Bind(wx.EVT_CLOSE, lambda evt: self.Destroy(), self) # XXX Could have a set_image() function instead of referring back to # the frame all the time? class SBSettingsPanel(wx.Panel): # XXX Names: they're not really settings. XXX Allow for setting # rotation, and provide a hierarchical drop-down menu to play with # detector, panel, sensor and ASIC. def __init__(self, *args, **kwds): super().__init__(*args, **kwds) sizer = wx.BoxSizer(wx.VERTICAL) self.SetSizer(sizer) # Number of decimal digits for distances. self.digits = 2 # Quad translation controls from wx.lib.agw.floatspin import EVT_FLOATSPIN, FloatSpin img = self.GetParent().GetParent().pyslip.tiles.raw_image d = img.get_detector() self._quad_spinners = [] for serial in range(4): fast, slow = d.hierarchy()[serial].get_origin()[0:2] name_quadrant = ["Q0", "Q1", "Q2", "Q3"][serial] box = wx.BoxSizer(wx.HORIZONTAL) for (name_direction, value) in [("fast", fast), ("slow", slow)]: name_ctrl = name_quadrant + "_" + name_direction + "_ctrl" spinner = FloatSpin( self, digits=self.digits, name=name_ctrl, value=value ) self.Bind(EVT_FLOATSPIN, self.OnUpdateQuad, spinner) box.Add( spinner, 0, wx.RIGHT | wx.TOP | wx.BOTTOM | wx.ALIGN_CENTER_VERTICAL, 5, ) box.Add( wx.StaticText(self, label=name_quadrant + " " + name_direction), 0, wx.ALL | wx.ALIGN_CENTER_VERTICAL, 5, ) setattr(self, "_" + name_ctrl, spinner) self._quad_spinners.append(spinner) sizer.Add(box) # Spinner amount control box = wx.BoxSizer(wx.HORIZONTAL) self._spinner_amt_control = FloatSpin( self, digits=self.digits, name="spin_amount", value=1, min_val=0.1, increment=0.1, ) self.Bind(EVT_FLOATSPIN, self.OnSpinAmount, self._spinner_amt_control) box.Add( self._spinner_amt_control, 0, wx.RIGHT | wx.TOP | wx.BOTTOM | wx.ALIGN_CENTER_VERTICAL, 5, ) box.Add( wx.StaticText(self, label="Spinner increment (mm)"), 0, wx.ALL | wx.ALIGN_CENTER_VERTICAL, 5, ) sizer.Add(box) box = wx.BoxSizer(wx.HORIZONTAL) btn = wx.Button(self, label="Restore metrology") box.Add(btn, flag=wx.ALL, border=5) self.Bind(wx.EVT_BUTTON, self.OnRestoreMetrology, btn) btn = wx.Button(self, label="Save current metrology") box.Add(btn, flag=wx.ALL, border=5) self.Bind(wx.EVT_BUTTON, self.OnSaveMetrology, btn) sizer.Add(box, flag=wx.ALIGN_CENTER) # XXX Rename to metrology tool? def OnRestoreMetrology(self, event): print("Not implemented") def OnSaveMetrology(self, event): import pycbf dialog = wx.FileDialog( self, defaultDir=os.curdir, defaultFile="quadrants.def", message="Save metrology file", style=wx.FD_SAVE | wx.FD_OVERWRITE_PROMPT, wildcard="Phil files (*.def)|*.def", ) if dialog.ShowModal() == wx.ID_OK: path = str(dialog.GetPath()) if path != "": # The detector object of the format instance is adjusted when the quadrant calibration # arrows are clicked. Sync those adjustments to a new cbf handle, drop unneeded categories # (categories frame specific but not metrology specific) and write the file. frame = self.GetParent().GetParent() img = frame.pyslip.tiles.raw_image header = img.image_set.get_format_class()(img.full_path) header.sync_detector_to_cbf(img.get_detector()) cbf = header._cbf_handle cbf.find_category("array_data") cbf.remove_category() cbf.find_category("array_structure") cbf.remove_category() cbf.find_category("array_intensities") cbf.remove_category() cbf.find_category("diffrn_radiation") cbf.remove_category() cbf.find_category("diffrn_radiation_wavelength") cbf.remove_category() cbf.find_category("diffrn_measurement") cbf.remove_category() cbf.find_category("diffrn_scan") cbf.remove_category() cbf.find_category("diffrn_scan_frame") cbf.remove_category() cbf.write_widefile( path, pycbf.CBF, pycbf.MIME_HEADERS | pycbf.MSG_DIGEST | pycbf.PAD_4K, 0, ) print("Saved cbf header to", path) def OnUpdateQuad(self, event): # Get the name of the spinner and its delta, the deviation from # the default value. Update the default for the next event. obj = event.EventObject name = obj.GetName() value = obj.GetValue() delta = float(value - obj.GetDefaultValue()) obj.SetDefaultValue(value) # Update the frame's effective metrology parameters. frame = self.GetParent().GetParent() img = frame.pyslip.tiles.raw_image quads = img.get_detector().hierarchy() if name == "Q0_fast_ctrl": quad, delta = (quads[0], col((delta, 0, 0))) elif name == "Q0_slow_ctrl": quad, delta = (quads[0], col((0, delta, 0))) elif name == "Q1_fast_ctrl": quad, delta = (quads[1], col((delta, 0, 0))) elif name == "Q1_slow_ctrl": quad, delta = (quads[1], col((0, delta, 0))) elif name == "Q2_fast_ctrl": quad, delta = (quads[2], col((delta, 0, 0))) elif name == "Q2_slow_ctrl": quad, delta = (quads[2], col((0, delta, 0))) elif name == "Q3_fast_ctrl": quad, delta = (quads[3], col((delta, 0, 0))) elif name == "Q3_slow_ctrl": quad, delta = (quads[3], col((0, delta, 0))) else: raise RuntimeError("Unknown control name " + name) ldm = quad.get_local_d_matrix() fast = (ldm[0], ldm[3], ldm[6]) slow = (ldm[1], ldm[4], ldm[7]) orig = col((ldm[2], ldm[5], ldm[8])) + delta quad.set_local_frame(fast, slow, orig) # Update the view, trigger redraw. tiles = frame.pyslip.tiles tiles.set_image(tiles.raw_image) tiles.flex_image.adjust(color_scheme=tiles.current_color_scheme) tiles.reset_the_cache() tiles.tile_cache = tiles.cache[tiles.zoom_level] tiles.tile_list = tiles.lru[tiles.zoom_level] frame.pyslip.Update() def OnSpinAmount(self, event): obj = event.EventObject for spinner in self._quad_spinners: spinner.SetIncrement(obj.GetValue())
bsd-3-clause
c86963b03f7b8e30a80d62aff02f9b7b
34.301802
107
0.535281
3.769601
false
false
false
false
dials/dials
src/dials/util/resolution_analysis.py
1
25324
""" Algorithms for analysis of resolution limits. """ from __future__ import annotations import enum import logging import math import typing import iotbx.merging_statistics import iotbx.mtz import iotbx.phil from cctbx import miller, uctbx from cctbx.array_family import flex from iotbx.reflection_file_utils import label_table from scitbx.math import curve_fitting, five_number_summary from dials.algorithms.scaling.scaling_library import determine_best_unit_cell from dials.report import plots from dials.util import Sorry, tabulate from dials.util.batch_handling import ( assign_batches_to_reflections, calculate_batch_offsets, ) from dials.util.filter_reflections import filter_reflection_table from dials.util.normalisation import quasi_normalisation logger = logging.getLogger(__name__) class metrics(enum.Enum): """Supported metrics for estimation of resolution limits.""" CC_HALF = "cc_half" CC_REF = "cc_ref" ISIGMA = "unmerged_i_over_sigma_mean" MISIGMA = "i_over_sigma_mean" I_MEAN_OVER_SIGMA_MEAN = "i_mean_over_sigi_mean" RMERGE = "r_merge" COMPLETENESS = "completeness" def polynomial_fit(x, y, degree=5): """ Fit a polynomial to the values y(x) and return this fit x, y should be iterables containing floats of the same size. The order is the order of polynomial to use for this fit. This will be useful for e.g. I/sigma. """ fit = curve_fitting.univariate_polynomial_fit( x, y, degree=degree, max_iterations=100 ) f = curve_fitting.univariate_polynomial(*fit.params) return f(x) def tanh_fit(x, y, iqr_multiplier=None): """ Fit a tanh function to the values y(x) and return this fit x, y should be iterables containing floats of the same size. This is used for fitting a curve to CC½. """ tf = curve_fitting.tanh_fit(x, y) f = curve_fitting.tanh(*tf.params) if iqr_multiplier: assert iqr_multiplier > 0 yc = f(x) dy = y - yc min_x, q1_x, med_x, q3_x, max_x = five_number_summary(dy) iqr_x = q3_x - q1_x cut_x = iqr_multiplier * iqr_x outliers = (dy > q3_x + cut_x) | (dy < q1_x - cut_x) if outliers.count(True) > 0: xo = x.select(~outliers) yo = y.select(~outliers) tf = curve_fitting.tanh_fit(xo, yo) f = curve_fitting.tanh(*tf.params) return f(x) def log_fit(x, y, degree=5): """Fit the values log(y(x)) then return exp() to this fit. x, y should be iterables containing floats of the same size. The order is the order of polynomial to use for this fit. This will be useful for e.g. I/sigma.""" fit = curve_fitting.univariate_polynomial_fit( x, flex.log(y), degree=degree, max_iterations=100 ) f = curve_fitting.univariate_polynomial(*fit.params) return flex.exp(f(x)) def log_inv_fit(x, y, degree=5): """Fit the values log(1 / y(x)) then return the inverse of this fit. x, y should be iterables, the order of the polynomial for the transformed fit needs to be specified. This will be useful for e.g. Rmerge.""" fit = curve_fitting.univariate_polynomial_fit( x, flex.log(1 / y), degree=degree, max_iterations=100 ) f = curve_fitting.univariate_polynomial(*fit.params) return 1 / flex.exp(f(x)) def resolution_fit_from_merging_stats(merging_stats, metric, model, limit, sel=None): """Estimate a resolution limit based on the input `metric` The function defined by `model` will be fit to the selected `metric` which has been pre-calculated by the `merging_stats` object. The estimated resolution limit is chosen as the `d_star_sq` value at which the fitted function equals `limit`. Args: merging_stats (iotbx.merging_statistics.dataset_statistics): Pre-calculated merging statistics object metric (str): The metric to use for estimating a resolution limit. Must be a metric calculated by `iotbx.merging_statistics.merging_stats` and available as an attribute on the `bins` attribute of the input `merging_stats` object. model: The function to fit to the selected metric. Must be callable, taking as input x (d_star_sq) and y (the metric to be fitted) values, returning the fitted y(x) values. limit (float): The resolution limit criterion. sel (scitbx.array_family.flex.bool): An optional selection to apply to the `merging_stats` bins. Returns: The estimated resolution limit in units of Å^-1 """ y_obs = flex.double(getattr(b, metric) for b in merging_stats.bins).reversed() d_star_sq = flex.double( uctbx.d_as_d_star_sq(b.d_min) for b in merging_stats.bins ).reversed() return resolution_fit(d_star_sq, y_obs, model, limit, sel=sel) def resolution_fit(d_star_sq, y_obs, model, limit, sel=None): """Estimate a resolution limit based on the input merging statistics The function defined by `model` will be fit to the input `d_star_sq` and `y_obs`. The estimated resolution limit is chosen as the `d_star_sq` value at which the fitted function equals `limit`. Args: d_star_sq (scitbx.array_family.flex.double): The high resolution limits of the resolution bins in units 1/d*2 y_obs (scitbx.array_family.flex.double): The statistic against which to fit the function `model` model: The function to fit against `y_obs`. Must be callable, taking as input x (d_star_sq) and y (the metric to be fitted) values, returning the fitted y(x) values. limit (float): The resolution limit criterion. sel (scitbx.array_family.flex.bool): An optional selection to apply to the `d_star_sq` and `y_obs` values. Returns: The estimated resolution limit in units of Å^-1 Raises: RuntimeError: Raised if no `y_obs` values remain after application of the selection `sel` """ if not sel: sel = flex.bool(len(d_star_sq), True) sel &= y_obs > 0 y_obs = y_obs.select(sel) d_star_sq = d_star_sq.select(sel) if not len(y_obs): raise RuntimeError("No reflections left for fitting") y_fit = model(d_star_sq, y_obs, 6) logger.debug( tabulate( [("d*2", "d", "obs", "fit")] + [ (ds2, uctbx.d_star_sq_as_d(ds2), yo, yf) for ds2, yo, yf in zip(d_star_sq, y_obs, y_fit) ], headers="firstrow", ) ) if flex.min(y_obs) > limit: d_min = 1.0 / math.sqrt(flex.max(d_star_sq)) else: try: d_min = 1.0 / math.sqrt(interpolate_value(d_star_sq, y_fit, limit)) except RuntimeError as e: logger.debug(f"Error interpolating value: {e}") d_min = None return ResolutionResult(d_star_sq, y_obs, y_fit, d_min) def _get_cc_half_significance(merging_stats, cc_half_method): """Get the CC½ significance values from the input merging_stats object""" if ( cc_half_method == "sigma_tau" and merging_stats.overall.cc_one_half_sigma_tau_significance is not None ): return flex.bool( b.cc_one_half_sigma_tau_significance for b in merging_stats.bins ).reversed() elif merging_stats.overall.cc_one_half_significance is not None: return flex.bool( b.cc_one_half_significance for b in merging_stats.bins ).reversed() def _get_cc_half_critical_values(merging_stats, cc_half_method): """Get the CC½ critical values from the input merging_stats object""" if ( cc_half_method == "sigma_tau" and merging_stats.overall.cc_one_half_sigma_tau_critical_value is not None ): return flex.double( b.cc_one_half_sigma_tau_critical_value for b in merging_stats.bins ).reversed() elif merging_stats.overall.cc_one_half_critical_value is not None: critical = [ b.cc_one_half_critical_value if b.cc_one_half_critical_value is not None else 0.0 for b in merging_stats.bins ] return flex.double(critical).reversed() def resolution_cc_half( merging_stats, limit, cc_half_method="half_dataset", model=tanh_fit ): """Estimate a resolution limit based on CC½ The function defined by `model` will be fit to the CC½ values that have been pre-calculated by the `merging_stats` object. The estimated resolution limit is chosen as the `d_star_sq` value at which the fitted function equals `limit`. Args: merging_stats (iotbx.merging_statistics.dataset_statistics): Pre-calculated merging statistics object cc_half_method (str): The method for calculating CC½. Either "half_dataset" or "sigma_tau" (See Assmann et al., J. Appl. Cryst. (2016). 49, 1021–1028). model: The function to fit to the selected metric. Must be callable, taking as input x (d_star_sq) and y (the metric to be fitted) values, returning the fitted y(x) values. Default is `tanh_fit`. limit (float): The resolution limit criterion. Returns: The estimated resolution limit in units of Å^-1 """ sel = _get_cc_half_significance(merging_stats, cc_half_method) metric = "cc_one_half_sigma_tau" if cc_half_method == "sigma_tau" else "cc_one_half" result = resolution_fit_from_merging_stats( merging_stats, metric, model, limit, sel=sel ) critical_values = _get_cc_half_critical_values(merging_stats, cc_half_method) if critical_values: result = result._replace(critical_values=critical_values.select(sel)) return result def interpolate_value(x, y, t): """Find the value of x: y(x) = t.""" if t > max(y) or t < min(y): raise RuntimeError(f"t outside of [{min(y):f}, {max(y):f}]") for j in range(1, len(x)): x0 = x[j - 1] y0 = y[j - 1] x1 = x[j] y1 = y[j] if (y0 - t) * (y1 - t) < 0: return x0 + (t - y0) * (x1 - x0) / (y1 - y0) def miller_array_from_mtz(unmerged_mtz, anomalous=False, labels=None): mtz_object = iotbx.mtz.object(file_name=unmerged_mtz) miller_arrays = mtz_object.as_miller_arrays( merge_equivalents=False, anomalous=anomalous ) i_obs = None batches = None all_i_obs = [] for array in miller_arrays: labels = array.info().label_string() if array.is_xray_intensity_array(): all_i_obs.append(array) if labels == "BATCH": assert batches is None batches = array if i_obs is None: if len(all_i_obs) == 0: raise Sorry("No intensities found") elif len(all_i_obs) > 1: if labels is not None: lab_tab = label_table(all_i_obs) i_obs = lab_tab.select_array( label=labels[0], command_line_switch="labels" ) if i_obs is None: raise Sorry( "Multiple intensity arrays - please specify one:\n%s" % "\n".join( [" labels=%s" % a.info().label_string() for a in all_i_obs] ) ) else: i_obs = all_i_obs[0] # need original miller indices otherwise we don't get correct anomalous # merging statistics if "M_ISYM" in mtz_object.column_labels(): indices = mtz_object.extract_original_index_miller_indices() i_obs = i_obs.customized_copy(indices=indices, info=i_obs.info()) return i_obs, batches phil_str = """ rmerge = None .type = float(value_min=0) .help = "Maximum value of Rmerge in the outer resolution shell" .short_caption = "Outer shell Rmerge" .expert_level = 1 completeness = None .type = float(value_min=0) .help = "Minimum completeness in the outer resolution shell" .short_caption = "Outer shell completeness" .expert_level = 1 cc_ref = 0.1 .type = float(value_min=0) .help = "Minimum value of CC vs reference data set in the outer resolution shell" .short_caption = "Outer shell CCref" .expert_level = 1 cc_half = 0.3 .type = float(value_min=0) .help = "Minimum value of CC½ in the outer resolution shell" .short_caption = "Outer shell CC½" .expert_level = 1 cc_half_method = *half_dataset sigma_tau .type = choice .short_caption = "CC½ method" cc_half_significance_level = 0.1 .type = float(value_min=0, value_max=1) .expert_level = 1 .short_caption = "CC½ significance level" cc_half_fit = polynomial *tanh .type = choice .expert_level = 1 .short_caption = "CC½ fit" isigma = None .type = float(value_min=0) .help = "Minimum value of the unmerged <I/sigI> in the outer resolution shell" .short_caption = "Outer shell unmerged <I/sigI>" .expert_level = 1 misigma = None .type = float(value_min=0) .help = "Minimum value of the merged <I/sigI> in the outer resolution shell" .short_caption = "Outer shell merged <I/sigI>" .expert_level = 1 i_mean_over_sigma_mean = None .type = float(value_min=0) .help = "Minimum value of the unmerged <I>/<sigI> in the outer resolution shell" .short_caption = "Outer shell unmerged <I>/<sigI>" .expert_level = 2 nbins = 100 .type = int .help = "Maximum number of resolution bins to use for estimation of resolution limit." .short_caption = "Number of resolution bins." .expert_level = 1 reflections_per_bin = 10 .type = int .help = "Minimum number of reflections per bin." .short_caption = "Minimum number of reflections per bin" binning_method = *counting_sorted volume .type = choice .help = "Use equal-volume bins or bins with approximately equal numbers of reflections per bin." .short_caption = "Equal-volume or equal #ref binning." .expert_level = 1 anomalous = False .type = bool .help = "Keep anomalous pairs separate in merging statistics" .short_caption = "Anomalous" .expert_level = 1 labels = None .type = strings .short_caption = "Labels" space_group = None .type = space_group .expert_level = 1 .short_caption = "Space group" reference = None .type = path .short_caption = "Reference" emax = 4 .type = float(value_min = 0) .help = "Reject reflections with normalised intensities E^2 > emax^2" .short_caption = "Maximum normalised intensity" """ phil_defaults = iotbx.phil.parse( """ resolution { %s batch_range = None .type = ints(size=2, value_min=0) } """ % phil_str ) def plot_result(metric, result): if metric == metrics.CC_HALF: return plots.cc_half_plot( result.d_star_sq, result.y_obs, cc_half_critical_values=result.critical_values, cc_half_fit=result.y_fit, d_min=result.d_min, ) else: d = { metrics.MISIGMA: "Merged <I/σ(I)>", metrics.ISIGMA: "Unmerged <I/σ(I)>", metrics.I_MEAN_OVER_SIGMA_MEAN: "&lt;I&gt;/<σ(I)>", metrics.RMERGE: "R<sub>merge</sub> ", metrics.COMPLETENESS: "Completeness", } d_star_sq_tickvals, d_star_sq_ticktext = plots.d_star_sq_to_d_ticks( result.d_star_sq, 5 ) return { "data": [ { "x": list(result.d_star_sq), # d_star_sq "y": list(result.y_obs), "type": "scatter", "name": "y_obs", }, ( { "x": list(result.d_star_sq), "y": list(result.y_fit), "type": "scatter", "name": "y_fit", "line": {"color": "rgb(47, 79, 79)"}, } if result.y_fit else {} ), ( { "x": [uctbx.d_as_d_star_sq(result.d_min)] * 2, "y": [ 0, max( 1, flex.max(result.y_obs), flex.max(result.y_fit) if result.y_fit else 0, ), ], "type": "scatter", "name": f"d_min = {result.d_min:.2f} Å", "mode": "lines", "line": {"color": "rgb(169, 169, 169)", "dash": "dot"}, } if result.d_min else {} ), ], "layout": { "title": f"{d.get(metric)} vs. resolution", "xaxis": { "title": "Resolution (Å)", "tickvals": d_star_sq_tickvals, "ticktext": d_star_sq_ticktext, }, "yaxis": {"title": d.get(metric), "rangemode": "tozero"}, }, } class ResolutionResult(typing.NamedTuple): d_star_sq: flex.double y_obs: flex.double y_fit: flex.double d_min: float critical_values: flex.double = None class Resolutionizer: """A class to calculate things from merging reflections.""" def __init__(self, i_obs, params, batches=None, reference=None): self._params = params self._reference = reference if self._reference is not None: self._reference = self._reference.merge_equivalents( use_internal_variance=False ).array() i_obs = i_obs.customized_copy( anomalous_flag=params.anomalous, info=i_obs.info() ) if self._params.batch_range is not None and batches is not None: batch_min, batch_max = self._params.batch_range assert batches is not None sel = (batches.data() >= batch_min) & (batches.data() <= batch_max) i_obs = i_obs.select(sel).set_info(i_obs.info()) if self._params.space_group is not None: i_obs = i_obs.customized_copy( space_group_info=self._params.space_group, info=i_obs.info() ) if self._params.emax: normalised = quasi_normalisation(i_obs) e2_cutoff = self._params.emax**2 sel = normalised.data() < e2_cutoff logger.info( f"Removing {sel.count(False)} Wilson outliers with E^2 >= {e2_cutoff}" ) i_obs = i_obs.select(sel) self._intensities = i_obs self._merging_statistics = iotbx.merging_statistics.dataset_statistics( i_obs=i_obs, n_bins=self._params.nbins, reflections_per_bin=self._params.reflections_per_bin, cc_one_half_significance_level=self._params.cc_half_significance_level, cc_one_half_method=self._params.cc_half_method, binning_method=self._params.binning_method, anomalous=params.anomalous, use_internal_variance=False, eliminate_sys_absent=False, assert_is_not_unique_set_under_symmetry=False, ) @classmethod def from_unmerged_mtz(cls, scaled_unmerged, params): """Construct the resolutionizer from an mtz file.""" i_obs, batches = miller_array_from_mtz( scaled_unmerged, anomalous=params.anomalous, labels=params.labels ) if params.reference is not None: reference, _ = miller_array_from_mtz( params.reference, anomalous=params.anomalous, labels=params.labels ) else: reference = None return cls(i_obs, params, batches=batches, reference=reference) @classmethod def from_reflections_and_experiments(cls, reflection_tables, experiments, params): """Construct the resolutionizer from native dials datatypes.""" # add some assertions about data # do batch assignment (same functions as in dials.export) offsets = calculate_batch_offsets(experiments) reflection_tables = assign_batches_to_reflections(reflection_tables, offsets) batches = flex.int() intensities = flex.double() indices = flex.miller_index() variances = flex.double() for table in reflection_tables: if "intensity.scale.value" in table: table = filter_reflection_table( table, ["scale"], partiality_threshold=0.4 ) intensities.extend(table["intensity.scale.value"]) variances.extend(table["intensity.scale.variance"]) else: table = filter_reflection_table( table, ["profile"], partiality_threshold=0.4 ) intensities.extend(table["intensity.prf.value"]) variances.extend(table["intensity.prf.variance"]) indices.extend(table["miller_index"]) batches.extend(table["batch"]) crystal_symmetry = miller.crystal.symmetry( unit_cell=determine_best_unit_cell(experiments), space_group=experiments[0].crystal.get_space_group(), assert_is_compatible_unit_cell=False, ) miller_set = miller.set(crystal_symmetry, indices, anomalous_flag=False) i_obs = miller.array(miller_set, data=intensities, sigmas=flex.sqrt(variances)) i_obs.set_observation_type_xray_intensity() i_obs.set_info(miller.array_info(source="DIALS", source_type="refl")) ms = i_obs.customized_copy() batch_array = miller.array(ms, data=batches) if params.reference is not None: reference, _ = miller_array_from_mtz( params.reference, anomalous=params.anomalous, labels=params.labels ) else: reference = None return cls(i_obs, params, batches=batch_array, reference=reference) def resolution(self, metric, limit=None): if metric == metrics.CC_HALF: return resolution_cc_half( self._merging_statistics, limit, cc_half_method=self._params.cc_half_method, model=tanh_fit if self._params.cc_half_fit == "tanh" else polynomial_fit, ) elif metric == metrics.CC_REF: return self._resolution_cc_ref(limit=self._params.cc_ref) else: model = { metrics.RMERGE: log_inv_fit, metrics.COMPLETENESS: polynomial_fit, metrics.ISIGMA: log_fit, metrics.MISIGMA: log_fit, metrics.I_MEAN_OVER_SIGMA_MEAN: log_fit, }[metric] return resolution_fit_from_merging_stats( self._merging_statistics, metric.value, model, limit ) def resolution_auto(self): """Compute resolution limits based on the current self._params set.""" metric_to_output = { metrics.ISIGMA: "I/sig", metrics.MISIGMA: "Mn(I/sig)", metrics.I_MEAN_OVER_SIGMA_MEAN: "Mn(I)/Mn(sig)", } plot_d = {} for metric in metrics: name = metric.name.lower() limit = getattr(self._params, name) if metric == metrics.CC_REF and not self._reference: limit = None if limit: try: result = self.resolution(metric, limit=limit) except RuntimeError as e: logger.info(f"Resolution fit against {name} failed: {e}") continue pretty_name = metric_to_output.get(metric, name) if result.d_min: logger.info( f"Resolution {pretty_name}:{result.d_min:{18 - len(pretty_name)}.2f}" ) plot_d[name] = plot_result(metric, result) return plot_d def _resolution_cc_ref(self, limit=None): """Compute a resolution limit where cc_ref < 0.5 (limit if set) or the full extent of the data.""" if limit is None: limit = self._params.cc_ref intensities = self._intensities.merge_equivalents( use_internal_variance=False ).array() cc_s = flex.double() for b in self._merging_statistics.bins: cc = intensities.resolution_filter( d_min=b.d_min, d_max=b.d_max ).correlation( self._reference.resolution_filter(d_min=b.d_min, d_max=b.d_max), assert_is_similar_symmetry=False, ) cc_s.append(cc.coefficient()) cc_s = cc_s.reversed() fit = tanh_fit if self._params.cc_half_fit == "tanh" else polynomial_fit d_star_sq = flex.double( 1 / b.d_min**2 for b in self._merging_statistics.bins ).reversed() return resolution_fit(d_star_sq, cc_s, fit, limit)
bsd-3-clause
ff984efb1ae36e06979baa491456eafe
35.198856
100
0.581156
3.57387
false
false
false
false
dials/dials
src/dials/algorithms/symmetry/origin.py
1
3885
""" Analysis of the origin of the diffraction pattern based on indexed and measured intensities. """ from __future__ import annotations def cctbx_crystal_from_dials(crystal): space_group = crystal.get_space_group() unit_cell = crystal.get_unit_cell() from cctbx.crystal import symmetry as crystal_symmetry return crystal_symmetry(unit_cell, space_group.type().lookup_symbol()) def cctbx_i_over_sigi_ms_from_dials_data(reflections, cctbx_crystal_symmetry): from cctbx.miller import set as miller_set from dials.array_family import flex refl = reflections.select(reflections["intensity.sum.variance"] > 0) return miller_set(cctbx_crystal_symmetry, refl["miller_index"]).array( data=refl["intensity.sum.value"], sigmas=flex.sqrt(refl["intensity.sum.variance"]), ) def offset_miller_indices(indices, offset): from dials.array_family import flex return flex.miller_index( *[mi.iround() for mi in (indices.as_vec3_double() + offset).parts()] ) def compute_miller_set_correlation( ms_a, ms_b, map_to_asu=False, merge_equivalents=False ): """Compute correlation between two miller arrays. Args: ms_a (cctbx.miller.array): Input miller.array `a`. ms_b (cctbx.miller.array): Input miller.array `b`. map_to_asu (bool): If ``True``, then map miller indices to the asymmetric unit before matching miller indices between input miller arrays. merge_equivalents (bool): If ``True`` then merge symmetry equivalent reflections before matching miller indices between input miller arrays. Returns: tuple[int, float]: A tuple of the number of observations and the correlation coefficient. """ if map_to_asu: # not obvious that this will help for the reasons stated below ms_a = ms_a.map_to_asu() ms_b = ms_b.map_to_asu() if merge_equivalents: # only want to do this if we have essentially "scaled" the data - if not # then we will get a smooth Wilson plot and about CC=1 (due to general # fall off with resolution) ms_a = ms_a.merge_equivalents().array() ms_b = ms_b.merge_equivalents().array() common_a, common_b = ms_a.common_sets(ms_b) return common_a.size(), common_a.correlation(common_b).coefficient() def get_hkl_offset_correlation_coefficients( dials_reflections, dials_crystal, map_to_asu=False, grid_h=0, grid_k=0, grid_l=0, reference=None, ): # N.B. deliberately ignoring d_min, d_max as these are inconsistent with # changing the miller indices from cctbx import sgtbx from cctbx.miller import set as miller_set from dials.array_family import flex cs = cctbx_crystal_from_dials(dials_crystal) ms = cctbx_i_over_sigi_ms_from_dials_data(dials_reflections, cs) if reference: reference_ms = cctbx_i_over_sigi_ms_from_dials_data(reference, cs) else: reference_ms = None ccs = flex.double() offsets = flex.vec3_int() nref = flex.size_t() if reference: cb_op = sgtbx.change_of_basis_op("x,y,z") else: cb_op = sgtbx.change_of_basis_op("-x,-y,-z") hkl_test = [ (h, k, l) for h in range(-grid_h, grid_h + 1) for k in range(-grid_k, grid_k + 1) for l in range(-grid_l, grid_l + 1) ] for hkl in hkl_test: indices = offset_miller_indices(ms.indices(), hkl) reindexed_indices = cb_op.apply(indices) rms = miller_set(cs, reindexed_indices).array(ms.data()) if reference_ms: _ms = reference_ms else: _ms = miller_set(cs, indices).array(ms.data()) n, cc = compute_miller_set_correlation(_ms, rms, map_to_asu=map_to_asu) ccs.append(cc) offsets.append(hkl) nref.append(n) return offsets, ccs, nref
bsd-3-clause
d4b0d982fdf6cfe0ab0d0de5baa7101d
29.590551
82
0.645302
3.176615
false
false
false
false
dials/dials
src/dials/algorithms/background/modeller.py
1
11518
from __future__ import annotations import logging from dials_algorithms_background_modeller_ext import ( BackgroundStatistics, MultiPanelBackgroundStatistics, ) __all__ = [ "BackgroundModeller", "BackgroundModellerExecutor", "BackgroundModellerResult", "BackgroundStatistics", "FinalizeModel", "MultiPanelBackgroundStatistics", ] logger = logging.getLogger(__name__) class FinalizeModel: """ A class to finalize the background model """ def __init__(self, experiments, filter_type="median", kernel_size=10, niter=100): """ Initialize the finalizer :param experiments: The experiment list :param kernel_size: The median filter kernel size :param niter: The number of iterations for filling holes """ from dials.algorithms.background.gmodel import PolarTransform # Set some parameters self.filter_type = filter_type self.kernel_size = kernel_size self.niter = niter # Check the input assert len(experiments) == 1 experiment = experiments[0] assert len(experiment.detector) == 1 # Save the experiment self.experiment = experiment # Create the transform object self.transform = PolarTransform( experiment.beam, experiment.detector[0], experiment.goniometer ) def finalize(self, data, mask): """ Finalize the model :param data: The data array :param mask: The mask array """ from dials.algorithms.image.fill_holes import diffusion_fill, simple_fill from dials.algorithms.image.filter import mean_filter, median_filter from dials.array_family import flex # Print some image properties sub_data = data.as_1d().select(mask.as_1d()) logger.info("Raw image statistics:") logger.info(" min: %d", int(flex.min(sub_data))) logger.info(" max: %d", int(flex.max(sub_data))) logger.info(" mean: %d", int(flex.mean(sub_data))) logger.info("") # Transform to polar logger.info("Transforming image data to polar grid") result = self.transform.to_polar(data, mask) data = result.data() mask = result.mask() sub_data = data.as_1d().select(mask.as_1d()) logger.info("Polar image statistics:") logger.info(" min: %d", int(flex.min(sub_data))) logger.info(" max: %d", int(flex.max(sub_data))) logger.info(" mean: %d", int(flex.mean(sub_data))) logger.info("") # Filter the image to remove noise if self.kernel_size > 0: if self.filter_type == "median": logger.info("Applying median filter") data = median_filter(data, mask, (self.kernel_size, 0), periodic=True) sub_data = data.as_1d().select(mask.as_1d()) logger.info("Median polar image statistics:") logger.info(" min: %d", int(flex.min(sub_data))) logger.info(" max: %d", int(flex.max(sub_data))) logger.info(" mean: %d", int(flex.mean(sub_data))) logger.info("") elif self.filter_type == "mean": logger.info("Applying mean filter") mask_as_int = mask.as_1d().as_int() mask_as_int.reshape(mask.accessor()) data = mean_filter(data, mask_as_int, (self.kernel_size, 0), 1) sub_data = data.as_1d().select(mask.as_1d()) logger.info("Mean polar image statistics:") logger.info(" min: %d", int(flex.min(sub_data))) logger.info(" max: %d", int(flex.max(sub_data))) logger.info(" mean: %d", int(flex.mean(sub_data))) logger.info("") else: raise RuntimeError(f"Unknown filter_type: {self.filter_type}") # Fill any remaining holes logger.info("Filling holes") data = simple_fill(data, mask) data = diffusion_fill(data, mask, self.niter) mask = flex.bool(data.accessor(), True) sub_data = data.as_1d().select(mask.as_1d()) logger.info("Filled polar image statistics:") logger.info(" min: %d", int(flex.min(sub_data))) logger.info(" max: %d", int(flex.max(sub_data))) logger.info(" mean: %d", int(flex.mean(sub_data))) logger.info("") # Transform back logger.info("Transforming image data from polar grid") result = self.transform.from_polar(data, mask) data = result.data() mask = result.mask() sub_data = data.as_1d().select(mask.as_1d()) logger.info("Final image statistics:") logger.info(" min: %d", int(flex.min(sub_data))) logger.info(" max: %d", int(flex.max(sub_data))) logger.info(" mean: %d", int(flex.mean(sub_data))) logger.info("") # Fill in any discontinuities mask = ~self.transform.discontinuity()[:-1, :-1] data = diffusion_fill(data, mask, self.niter) # Get and apply the mask mask = self.experiment.imageset.get_mask(0)[0] mask = mask.as_1d().as_int().as_double() mask.reshape(data.accessor()) data *= mask # Return the result return data class BackgroundModellerResult: """ A class to contain the modelling result """ def __init__( self, mean=None, variance=None, dispersion=None, mask=None, min_image=None, max_image=None, model=None, polar_model=None, ): """ Init the result """ self.mean = mean self.variance = variance self.dispersion = dispersion self.mask = mask self.min_image = min_image self.max_image = max_image self.model = model self.polar_model = polar_model class BackgroundModellerExecutor: def __init__(self, experiments, params): assert len(experiments) == 1 self.min_images = params.modeller.min_images if self.min_images > len(experiments[0].imageset): self.min_images = len(experiments[0].imageset) self.image_type = params.modeller.image_type self.finalizer = FinalizeModel( experiments=experiments, filter_type=params.modeller.filter_type, kernel_size=params.modeller.kernel_size, niter=params.modeller.niter, ) self.result = None def process(self, image_volume, experiments, reflections): from dials.algorithms.integration.processor import job # Write some output logger.info( " Background modelling; job: %d; frames: %d -> %d; # Reflections: %d", job.index, image_volume.frame0(), image_volume.frame1(), len(reflections), ) # Compute the shoebox mask reflections.compute_mask(experiments=experiments, image_volume=image_volume) # Compute the sum, sum^2 and the number of contributing pixels return MultiPanelBackgroundStatistics(image_volume) def accumulate(self, index, data): if self.result is None: self.result = data else: self.result += data def finalize_model(self): logger.info("") logger.info("=" * 80) logger.info("Finalizing model") logger.info("") result = [] for i in range(len(self.result)): # Get the statistics stats = self.result.get(i) mean = stats.mean(self.min_images) variance = stats.variance(self.min_images) dispersion = stats.dispersion(self.min_images) mask = stats.mask(self.min_images) min_image = stats.min() max_image = stats.max() # Create the model if self.image_type == "min": model = self.finalizer.finalize(min_image, mask) elif self.image_type == "mean": model = self.finalizer.finalize(mean, mask) else: raise RuntimeError(f"Unknown image_type: {self.image_type}") # Add to the list result.append( BackgroundModellerResult( mean=mean, variance=variance, dispersion=dispersion, mask=mask, min_image=min_image, max_image=max_image, model=model, ) ) return result class BackgroundModeller: """ A class to help with background modelling """ def __init__(self, experiments, reflections, params): """ Initialize the modeller :param experiments: The experiment list :param reflections: The reflections to process :param params: The parameters to use """ # Check all reflections have same imageset and get it imageset = experiments[0].imageset for expr in experiments: assert expr.imageset == imageset, "All experiments must share and imageset" # Save some stuff self.experiments = experiments self.reflections = reflections self.params = params self.model = None def compute(self): """ Integrate the data """ from dials.algorithms.integration.image_integrator import ProcessorImage from dials.util.command_line import heading # Init the report self.profile_model_report = None self.integration_report = None # Create summary format fmt = ( " Processing the following experiments:\n" "\n" " Experiments: %d\n" " Beams: %d\n" " Detectors: %d\n" " Goniometers: %d\n" " Scans: %d\n" " Crystals: %d\n" " Imagesets: %d\n" ) # Print the summary logger.info( fmt, len(self.experiments), len(self.experiments.beams()), len(self.experiments.detectors()), len(self.experiments.goniometers()), len(self.experiments.scans()), len(self.experiments.crystals()), len(self.experiments.imagesets()), ) # Print a heading logger.info("=" * 80) logger.info("") logger.info(heading("Modelling background")) logger.info("") # Expand n_sigma for expt in self.experiments: expt.profile._n_sigma += 2 # Compute some reflection properties self.reflections.compute_zeta_multi(self.experiments) self.reflections.compute_d(self.experiments) self.reflections.compute_bbox(self.experiments) # Construvt the image integrator processor processor = ProcessorImage(self.experiments, self.reflections, self.params) processor.executor = BackgroundModellerExecutor(self.experiments, self.params) # Do the processing _, time_info = processor.process() # Compute the model self.model = processor.executor.finalize_model() # Print the time info logger.info(str(time_info)) logger.info("") # Return the reflections return self.model
bsd-3-clause
709b2ad9eace00e1bc936f02907ad899
31.908571
87
0.566765
4.014639
false
false
false
false
dials/dials
src/dials/util/normalisation.py
1
1483
from __future__ import annotations from cctbx import uctbx from scitbx.array_family import flex def quasi_normalisation(intensities): """Quasi-normalisation of the input intensities. Args: intensities (cctbx.miller.array): The intensities to be normalised. Returns: cctbx.miller.array: The normalised intensities. """ # handle negative reflections to minimise effect on mean I values. work = intensities.deep_copy() work.data().set_selected(work.data() < 0.0, 0.0) # set up binning objects if work.size() > 20000: n_refl_shells = 20 elif work.size() > 15000: n_refl_shells = 15 else: n_refl_shells = 10 d_star_sq = work.d_star_sq().data() d_star_sq_max = flex.max(d_star_sq) d_star_sq_min = flex.min(d_star_sq) span = d_star_sq_max - d_star_sq_min d_star_sq_max += span * 1e-6 d_star_sq_min -= span * 1e-6 d_star_sq_step = (d_star_sq_max - d_star_sq_min) / n_refl_shells work.setup_binner_d_star_sq_step( d_min=uctbx.d_star_sq_as_d(d_star_sq_min), # cctbx/cctbx_project#588 d_max=uctbx.d_star_sq_as_d(d_star_sq_max), # cctbx/cctbx_project#588 d_star_sq_step=d_star_sq_step, auto_binning=False, ) normalisations = work.intensity_quasi_normalisations() return intensities.customized_copy( data=(intensities.data() / normalisations.data()), sigmas=(intensities.sigmas() / normalisations.data()), )
bsd-3-clause
591fc044f57ed28c8dc2b76d55aac9a5
32.704545
77
0.636548
2.74122
false
false
false
false
douban/dpark
examples/kmeans.py
1
1581
#!/usr/bin/env python from __future__ import absolute_import from __future__ import print_function import sys, os, os.path from six.moves import map from six.moves import range sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__)))) import random from dpark import DparkContext from vector import Vector dpark = DparkContext() def parseVector(line): return Vector(list(map(float, line.strip().split(' ')))) def closestCenter(p, centers): bestDist = p.squaredDist(centers[0]) bestIndex = 0 for i in range(1, len(centers)): d = p.squaredDist(centers[i]) if d < bestDist: bestDist = d bestIndex = i return bestIndex if __name__ == '__main__': D = 4 K = 3 IT = 10 MIN_DIST = 0.01 centers = [Vector([random.random() for j in range(D)]) for i in range(K)] points = dpark.textFile('kmeans_data.txt').map(parseVector).cache() for it in range(IT): print('iteration', it) mappedPoints = points.map(lambda p: (closestCenter(p, centers), (p, 1))) ncenters = mappedPoints.reduceByKey( lambda (s1, c1), (s2, c2): (s1 + s2, c1 + c2) ).map( lambda id_sum_count: (id_sum_count[0], id_sum_count[1][0] / id_sum_count[1][1]) ).collectAsMap() updated = False for i in ncenters: if centers[i].dist(ncenters[i]) > MIN_DIST: centers[i] = ncenters[i] updated = True if not updated: break print(centers) print('final', centers)
bsd-3-clause
4067864e4f4f76c678f069afc2f4f08c
26.736842
91
0.590765
3.349576
false
false
false
false
dials/dials
src/dials/util/rebin_images.py
1
3341
from __future__ import annotations import binascii import os import random from dxtbx.ext import compress def gz_open(filename, mode): import gzip return gzip.GzipFile(filename, mode) def split_counts(image, split): from scitbx.array_family import flex new_images = [flex.int(flex.grid(image.focus()), 0) for k in range(split)] negative = image.as_1d() < 0 positive = image.as_1d() > 0 for new_image in new_images: new_image.as_1d().set_selected(negative, image.as_1d()) for p in positive.iselection(): counts = image[p] for j in range(counts): new_images[random.randint(0, split - 1)][p] += 1 return new_images def merge_counts(images): from scitbx.array_family import flex image = flex.int(flex.grid(images[0].focus()), 0) negative = images[0].as_1d() < 0 for i in images: image += i image.as_1d().set_selected(negative, images[0].as_1d()) return image def read_image(in_image): from dxtbx import load assert os.path.exists(in_image) start_tag = binascii.unhexlify("0c1a04d5") data = gz_open(in_image, "rb").read() data_offset = data.find(start_tag) cbf_header = data[:data_offset] pixel_values = load(in_image).get_raw_data() return pixel_values, cbf_header def write_image(out_image, pixel_values, header, nn=1): assert not os.path.exists(out_image) start_tag = binascii.unhexlify("0c1a04d5") compressed = compress(pixel_values) fixed_header = "" header = header.decode() for record in header.split("\n")[:-1]: if "X-Binary-Size:" in record: fixed_header += f"X-Binary-Size: {len(compressed)}\r\n" elif "Content-MD5" in record: pass elif "Count_cutoff" in record: cutoff = int(record.split()[2]) * nn fixed_header += "# Count_cutoff %d counts\n" % cutoff else: fixed_header += f"{record}\n" tailer = "\r\n--CIF-BINARY-FORMAT-SECTION----\r\n;\r\n" gz_open(out_image, "wb").write( fixed_header.encode() + start_tag + compressed + tailer.encode() ) def main(in_images, out_images): assert len(in_images) == len(out_images) n = len(in_images) for i in in_images: assert os.path.exists(i) for o in out_images: assert not os.path.exists(o) in_image_data = [] in_image_headers = [] for i in in_images: print(f"Reading {i}") pixel, header = read_image(i) in_image_data.append(pixel) in_image_headers.append(header) sum_image = merge_counts(in_image_data) rebin_images = split_counts(sum_image, n) for o, pixel, header in zip(out_images, rebin_images, in_image_headers): print(f"Writing {o}") write_image(o, pixel, header) def main_sum(in_images, out_image): for i in in_images: assert os.path.exists(i) assert not os.path.exists(out_image) in_image_data = [] in_image_headers = [] for i in in_images: print(f"Reading {i}") pixel, header = read_image(i) in_image_data.append(pixel) in_image_headers.append(header) sum_image = merge_counts(in_image_data) print(f"Writing {out_image}") write_image(out_image, sum_image, in_image_headers[0], nn=len(in_images))
bsd-3-clause
9455e05411d582283550c38a3956b876
24.7
78
0.613289
3.191022
false
false
false
false
dials/dials
tests/command_line/test_ssx_reduction.py
1
2808
# test running data reduction programs on ssx data from __future__ import annotations import procrunner def test_ssx_reduction(dials_data, tmp_path): """ Check that dials.cosym, dials.scale, dials.export and dials.merge run successfully on ssx data. Also test a few smaller analysis programs. """ ssx = dials_data("cunir_serial_processed", pathlib=True) ssx_data = dials_data("cunir_serial", pathlib=True) refls = ssx / "integrated.refl" expts = ssx / "integrated.expt" result = procrunner.run( ["dials.cosym", expts, refls], working_directory=tmp_path, ) assert not result.returncode and not result.stderr cosym_expts = tmp_path / "symmetrized.expt" cosym_refls = tmp_path / "symmetrized.refl" assert cosym_expts.is_file() assert cosym_refls.is_file() assert (tmp_path / "dials.cosym.html").is_file() result = procrunner.run( ["dials.scale", cosym_expts, cosym_refls], working_directory=tmp_path, ) assert not result.returncode and not result.stderr scale_expts = tmp_path / "scaled.expt" scale_refls = tmp_path / "scaled.refl" assert scale_expts.is_file() assert scale_refls.is_file() assert (tmp_path / "dials.scale.html").is_file() # run scaling with reference model / cif for reference in [ ssx_data / "2BW4.pdb", ssx_data / "2bw4.cif", ssx_data / "2bw4-sf.cif", ]: result = procrunner.run( [ "dials.scale", cosym_expts, cosym_refls, f"reference={reference}", "output.experiments=scaled_ref.expt", "output.reflections=scaled_ref.refl", ], working_directory=tmp_path, ) assert not result.returncode and not result.stderr result = procrunner.run( ["dials.export", scale_expts, scale_refls], working_directory=tmp_path, ) assert not result.returncode and not result.stderr assert (tmp_path / "scaled.mtz").is_file() result = procrunner.run( ["dials.merge", scale_expts, scale_refls], working_directory=tmp_path, ) assert not result.returncode and not result.stderr assert (tmp_path / "merged.mtz").is_file() result = procrunner.run( ["dials.damage_analysis", scale_expts, scale_refls], working_directory=tmp_path, ) assert not result.returncode and not result.stderr assert (tmp_path / "dials.damage_analysis.html").is_file() result = procrunner.run( ["dials.compute_delta_cchalf", scale_expts, scale_refls], working_directory=tmp_path, ) assert not result.returncode and not result.stderr assert (tmp_path / "compute_delta_cchalf.html").is_file()
bsd-3-clause
053a0b5d963455746a4a6e33460b2e07
32.035294
73
0.624288
3.395405
false
false
false
false
douban/dpark
dpark/shuffle.py
1
26136
from __future__ import absolute_import from __future__ import print_function import os import os.path import random import six from six.moves import urllib, queue, range, zip, reduce, cPickle as pickle import marshal import struct import time import heapq import itertools from operator import itemgetter from itertools import islice from functools import wraps try: import cStringIO as StringIO except ImportError: from six import BytesIO as StringIO import dpark.conf from dpark.utils import compress, decompress, spawn, atomic_file from dpark.utils.memory import ERROR_TASK_OOM from dpark.utils.log import get_logger from dpark.env import env from dpark.tracker import GetValueMessage, SetValueMessage from dpark.utils.heaponkey import HeapOnKey from dpark.dependency import AggregatorBase from dpark.utils.nested_groupby import GroupByNestedIter, cogroup_no_dup logger = get_logger(__name__) # readable F_MAPPING = { (True, True): b'M', (False, True): b'P', (True, False): b'm', (False, False): b'p' } F_MAPPING_R = dict([(v, k) for k, v in F_MAPPING.items()]) def pack_header(length, is_marshal, is_sorted): flag = F_MAPPING[(is_marshal, is_sorted)] return flag + struct.pack("I", length) def unpack_header(head): l = len(head) if l != 5: raise IOError("fetch bad head length %d" % (l,)) flag = head[:1] is_marshal, is_sorted = F_MAPPING_R[flag] length, = struct.unpack("I", head[1:5]) return length, is_marshal, is_sorted def write_buf(stream, buf, is_marshal): buf = compress(buf) size = len(buf) stream.write(pack_header(size, is_marshal, True)) stream.write(buf) return size + 4 class AutoBatchedSerializer(object): """ Choose the size of batch automatically based on the size of object """ size_loaded = 0 def __init__(self, best_size=1 << 17): self.best_size = best_size self.max_num = 0 self.max_size = 0 self.use_marshal = True self.num_batch = 0 self.file_size = 0 def load_stream(self, stream): while True: head = stream.read(5) if not head: return length, is_marshal, is_sorted = unpack_header(head) assert (is_sorted) buf = stream.read(length) if len(buf) < length: raise IOError("length not match: expected %d, but got %d" % (length, len(buf))) buf = decompress(buf) AutoBatchedSerializer.size_loaded += len(buf) if is_marshal: vs = marshal.loads(buf) else: vs = pickle.loads(buf) for v in vs: yield v def dump_stream(self, iterator, stream): self._dump_stream(iter(iterator), stream) logger.debug("max batch num = %d, max batch size = %d", self.max_num, self.max_size) def _dump_stream(self, iterator, stream): batch_num = 1 while True: vs = list(itertools.islice(iterator, batch_num)) self.num_batch += 1 if not vs: break batch_num = self._dump_batch(stream, vs, batch_num) def _dump_batch(self, stream, vs, batch_num): if self.use_marshal: try: buf = marshal.dumps(vs) except: buf = pickle.dumps(vs, -1) self.use_marshal = False else: buf = pickle.dumps(vs, -1) mem_size = len(buf) self.file_size += write_buf(stream, buf, self.use_marshal) if mem_size < self.best_size: batch_num *= 2 if batch_num > self.max_num: self.max_num = batch_num else: if mem_size > self.best_size * 2 and batch_num > 1: batch_num //= 2 if mem_size > self.max_size: self.max_size = mem_size return batch_num class GroupByAutoBatchedSerializer(AutoBatchedSerializer): def _dump_stream(self, iterator, stream): batch_num = 1 def _batching(): batch = [] num = 0 for k, vs in iterator: n = len(vs) if n + num <= batch_num: batch.append((k, vs)) num += n else: if batch: yield batch batch = [] num = 0 if n >= batch_num: sub_it = iter(vs) while True: sub_vs = list(itertools.islice(sub_it, batch_num)) if not sub_vs: break yield [(k, sub_vs)] else: batch.append((k, vs)) num = n if batch: yield batch for k_vs in _batching(): self.num_batch += 1 batch_num = self._dump_batch(stream, k_vs, batch_num) def get_serializer(rddconf): if rddconf.iter_group and (rddconf.is_groupby or rddconf.is_cogroup): return GroupByAutoBatchedSerializer() else: return AutoBatchedSerializer() def fetch_with_retry(f): MAX_RETRY = 3 RETRY_INTERVALS = [1, 10] @wraps(f) def _(self): self.num_batch_done = 0 while True: try: for items in islice(f(self), self.num_batch_done, None): self.num_batch_done += 1 yield items if self.num_retry > 0: logger.info("Fetch retry %d success for url %s, num_batch %d ", self.num_retry, self.url, self.num_batch_done) break except Exception as e: self.num_retry += 1 msg = "Fetch failed for url %s, tried %d/%d times. Exception: %s. " % ( self.url, self.num_retry, MAX_RETRY, e) fail_fast = False emsg = str(e) if any([emsg.find(s) >= 0 for s in ["404"]]): # "many open file", fail_fast = True msg += "no need to retry." if fail_fast or self.num_retry >= MAX_RETRY: logger.warning(msg) from dpark.task import FetchFailed raise FetchFailed(self.uri, self.sid, self.mid, self.rid) else: sleep_time = RETRY_INTERVALS[self.num_retry - 1] msg += "sleep %d secs" % (sleep_time,) logger.debug(msg) time.sleep(sleep_time) return _ class RemoteFile(object): num_open = 0 def __init__(self, uri, shuffle_id, map_id, reduce_id): self.uri = uri self.sid = shuffle_id self.mid = map_id self.rid = reduce_id self.url = ShuffleWorkDir(shuffle_id, map_id, reduce_id).restore(uri) logger.debug("fetch %s", self.url) self.num_retry = 0 self.num_batch_done = 0 def open(self): f = urllib.request.urlopen(self.url) if f.code == 404: f.close() raise IOError("not found") exp_size = int(f.headers['content-length']) return f, exp_size @fetch_with_retry def unsorted_batches(self): f = None # TEST_RETRY = True try: f, exp_size = self.open() total_size = 0 while True: head = f.read(5) if len(head) == 0: break length, is_marshal, is_sorted = unpack_header(head) assert (not is_sorted) total_size += length + 5 d = f.read(length) if length != len(d): raise IOError( "length not match: expected %d, but got %d" % (length, len(d))) d = decompress(d) if is_marshal: items = marshal.loads(d) else: try: items = pickle.loads(d) except: time.sleep(1) items = pickle.loads(d) yield items # if TEST_RETRY and self.num_retry == 0: # raise Exception("test_retry") if total_size != exp_size: raise IOError( "fetch size not match: expected %d, but got %d" % (exp_size, total_size)) env.task_stats.bytes_fetch += exp_size finally: if f: f.close() @fetch_with_retry def sorted_items(self): f = None try: serializer = AutoBatchedSerializer() self.num_open += 1 f, exp_size = self.open() for obj in serializer.load_stream(f): yield obj env.task_stats.bytes_fetch += exp_size finally: # rely on GC to close if generator not exhausted # so Fetcher must not be an attr of RDD if f: f.close() self.num_open -= 1 class ShuffleFetcher(object): @classmethod def _get_uris(cls, shuffle_id): uris = MapOutputTracker.get_locs(shuffle_id) mapid_uris = list(zip(list(range(len(uris))), uris)) random.shuffle(mapid_uris) return mapid_uris @classmethod def get_remote_files(cls, shuffle_id, reduce_id): uris = cls._get_uris(shuffle_id) return [RemoteFile(uri, shuffle_id, map_id, reduce_id) for map_id, uri in uris] def fetch(self, shuffle_id, reduce_id, merge_func): raise NotImplementedError def stop(self): pass class SimpleShuffleFetcher(ShuffleFetcher): def fetch(self, shuffle_id, reduce_id, merge_func): logger.debug( "Fetching outputs for shuffle %d, reduce %d", shuffle_id, reduce_id) for f in self.get_remote_files(): for items in f.unsorted_batches(): merge_func(items) class ParallelShuffleFetcher(SimpleShuffleFetcher): def __init__(self, nthreads): self.nthreads = nthreads self._started = False def start(self): if self._started: return self._started = True self.requests = queue.Queue() self.results = queue.Queue(self.nthreads) self.threads = [spawn(self._fetch_thread) for i in range(self.nthreads)] def _fetch_thread(self): from dpark.task import FetchFailed while True: f = self.requests.get() if f is None: break try: for items in f.unsorted_batches(): if not self._started: break self.results.put((items, f.mid)) if not self._started: break self.results.put(1) except FetchFailed as e: if not self._started: break self.results.put(e) break def fetch(self, shuffle_id, reduce_id, merge_func): self.start() files = self.get_remote_files(shuffle_id, reduce_id) for f in files: self.requests.put(f) t = time.time() from dpark.task import FetchFailed num_done = 0 while num_done < len(files): r = self.results.get() if r == 1: num_done += 1 elif isinstance(r, FetchFailed): self.stop() raise r else: items, map_id = r merge_func(items, map_id) env.task_stats.secs_fetch = time.time() - t def stop(self): if not self._started: return logger.debug("stop parallel shuffle fetcher ...") self._started = False while not self.requests.empty(): self.requests.get_nowait() for i in range(self.nthreads): self.requests.put(None) N = 5 for _ in range(N): while not self.results.empty(): self.results.get_nowait() for t in self.threads: t.join(1) if all([not t.isAlive() for t in self.threads]): return else: logger.info("FIXME: fail to join fetcher threads") class SortShuffleFetcher(ShuffleFetcher): def get_iters(self, shuffle_id, reduce_id): return [f.sorted_items() for f in self.get_remote_files(shuffle_id, reduce_id)] def fetch(self, shuffle_id, reduce_id, merge_func): merge_func(self.get_iters(shuffle_id, reduce_id)) def heap_merged(items_lists, combiner): heap = [] def pushback(_it, _i): try: _k, _v = next(_it) # put i before value, so do not compare the value heapq.heappush(heap, (_k, i, _v)) except StopIteration: pass for i, it in enumerate(items_lists): pushback(it, i) if not heap: return last_key, i, last_value = heapq.heappop(heap) pushback(items_lists[i], i) while heap: k, i, v = heapq.heappop(heap) if k != last_key: yield last_key, last_value last_key, last_value = k, v else: last_value = combiner(last_value, v) pushback(items_lists[i], i) yield last_key, last_value class SortedItemsOnDisk(object): def __init__(self, items, rddconf): self.path = path = env.workdir.alloc_tmp("sorted_items") with atomic_file(path, bufsize=4096) as f: if not isinstance(items, list): items = list(items) items.sort(key=itemgetter(0)) serializer = get_serializer(rddconf) serializer.dump_stream(items, f) self.size = f.tell() self.num_batch = serializer.num_batch def __iter__(self): serializer = AutoBatchedSerializer() with open(self.path, 'rb') as f: for obj in serializer.load_stream(f): yield obj def __dealloc__(self): try: if os.path.exists(self.path): os.remove(self.path) except Exception: pass class Merger(object): def __init__(self, rddconf, aggregator=None, size=None, api_callsite=None): self.rddconf = rddconf self.aggregator = aggregator self.size = size self.api_callsite = api_callsite @classmethod def get(cls, rddconf, aggregator=None, size=0, api_callsite=None): if rddconf.sort_merge: # all mergers keep order c = SortMerger if rddconf.is_cogroup: if rddconf.iter_group: c = IterCoGroupSortMerger elif rddconf.is_groupby: if rddconf.iter_group: c = IterGroupBySortMerger else: c = DiskHashMerger if rddconf.is_groupby: if rddconf.ordered_group: c = OrderedGroupByDiskHashMerger elif rddconf.is_cogroup: if rddconf.ordered_group: c = OrderedCoGroupDiskHashMerger else: c = CoGroupDiskHashMerger logger.debug("%s %s", c, rddconf) return c(rddconf, aggregator, size, api_callsite) class DiskHashMerger(Merger): def __init__(self, rddconf, aggregator=None, size=None, api_callsite=None): Merger.__init__(self, rddconf, aggregator, size, api_callsite) self.combined = {} self.use_disk = rddconf.disk_merge if self.use_disk: env.meminfo.ratio = rddconf.dump_mem_ratio self.archives = [] self.rotate_time = 0 self.last_rotate_ts = time.time() self.rotate_num = 0 self.total_size = 0 def _rotate(self): total_size = self.total_size t0 = time.time() time_since_last = t0 - self.last_rotate_ts dict_size = len(self.combined) rss_before = env.meminfo.rss_rt size = self._dump() self.total_size += size rss_after = env.meminfo.rss_rt t1 = time.time() rotate_time = t1 - t0 self.last_rotate_ts = t1 self.rotate_time += rotate_time self.rotate_num += 1 max_rotate = 1000 if self.rotate_num > max_rotate: logger.warnging('more than %d rotation. exit!', max_rotate) os._exit(ERROR_TASK_OOM) env.meminfo.after_rotate() _log = logger.info if dpark.conf.LOG_ROTATE else logger.debug _log('rotate %d: use %.2f sec, since last %.2f secs, dict_size 0x%x,' 'mem %d -> %d MB, disk size +%d = %d MB', self.rotate_num, rotate_time, time_since_last, dict_size, rss_before >> 20, rss_after >> 20, size >> 20, total_size >> 20) return env.meminfo.mem_limit_soft def disk_size(self): return sum([a.size for a in self.archives]) def _dump(self): import gc items = self.combined.items() f = SortedItemsOnDisk(items, self.rddconf) self.archives.append(f) del items if self.rddconf.is_groupby: for v in self.combined.itervalues(): del v[:] self.combined.clear() gc.collect() return f.size def merge(self, items, map_id, dep_id=0): mem_limit = env.meminfo.mem_limit_soft use_disk = self.use_disk try: env.meminfo.check = use_disk self._merge(items, map_id, dep_id, use_disk, env.meminfo, mem_limit) finally: env.meminfo.check = True def _get_merge_function(self): return self.aggregator.mergeCombiners def _merge(self, items, map_id, dep_id, use_disk, meminfo, mem_limit): combined = self.combined merge_combiner = self.aggregator.mergeCombiners for k, v in items: o = combined.get(k) combined[k] = merge_combiner(o, v) if o is not None else v if use_disk and meminfo.rss > mem_limit: mem_limit = self._rotate() def __iter__(self): if not self.archives: return six.iteritems(self.combined) items = self.combined.items() items.sort(key=itemgetter(0)) combined = items self.archives.append(iter(combined)) iters = list(map(iter, self.archives)) if self.rddconf.is_groupby and self.rddconf.iter_group: heap = HeapOnKey(key=lambda x: x[0], min_heap=True) it = GroupByNestedIter(heap.merge(iters), "") else: it = heap_merged(iters, self._get_merge_function()) return it class OrderedGroupByDiskHashMerger(DiskHashMerger): def _merge(self, items, map_id, dep_id, use_disk, meminfo, mem_limit): combined = self.combined for k, v in items: o = combined.get(k) iv = (map_id, v) if o is None: combined[k] = [iv] else: o.append(iv) if use_disk and meminfo.rss > mem_limit: mem_limit = self._rotate() def __iter__(self): it = DiskHashMerger.__iter__(self) merge_combiner = self.aggregator.mergeCombiners for k, ivs in it: ivs.sort(key=itemgetter(0)) cb = reduce(merge_combiner, (v for _, v in ivs)) yield k, cb class CoGroupDiskHashMerger(DiskHashMerger): def __init__(self, rddconf, aggregator=None, size=None, api_callsite=None): DiskHashMerger.__init__(self, rddconf, aggregator, size, api_callsite) self.direct_upstreams = [] def _get_merge_function(self): def _merge(x, y): for i in range(self.size): x[i].extend(y[i]) return x return _merge def _merge(self, items, map_id, dep_id, use_disk, meminfo, mem_limit): combined = self.combined if map_id < 0: for k, v in items: t = combined.get(k) if t is None: combined[k] = t = tuple([[] for _ in range(self.size)]) t[dep_id].append(v) if use_disk and meminfo.rss > mem_limit: mem_limit = self._rotate() else: for k, vs in items: t = combined.get(k) if t is None: combined[k] = t = tuple([[] for _ in range(self.size)]) t[dep_id].extend(vs) class OrderedCoGroupDiskHashMerger(CoGroupDiskHashMerger): def _merge(self, items, map_id, dep_id, use_disk, meminfo, mem_limit): combined = self.combined if map_id < 0: self.upstreams.append(dep_id) for k, v in items: t = combined.get(k) if t is None: combined[k] = t = tuple([[] for _ in range(self.size)]) t[dep_id].append(v) if use_disk and meminfo.rss > mem_limit: mem_limit = self._rotate() else: for k, vs in items: t = combined.get(k) if t is None: combined[k] = t = tuple([[] for _ in range(self.size)]) t[dep_id].append((map_id, vs)) if use_disk and meminfo.rss > mem_limit: mem_limit = self._rotate() def __iter__(self): it = DiskHashMerger.__iter__(self) direct_upstreams = self.direct_upstreams for k, groups in it: t = list([[] for _ in range(self.size)]) for i, g in enumerate(groups): if g: if i in direct_upstreams: t[i] = g else: g.sort(key=itemgetter(0)) g1 = [] for _, vs in g: g1.extend(vs) t[i] = g1 yield k, tuple(t) class SortMergeAggregator(AggregatorBase): def __init__(self, mergeCombiners): # each item is a combiner self.mergeValue = mergeCombiners def createCombiner(self, v): return v class CoGroupSortMergeAggregator(AggregatorBase): def __init__(self, size): self.size = size def createCombiner(self, v): # v = (rdd_index, value) values = tuple([[] for _ in range(self.size)]) values[v[0]].extend(v[1]) return values def mergeValue(self, c, v): c[v[0]].extend(v[1]) return c class SortMerger(Merger): def __init__(self, rddconf, aggregator=None, size=None, api_callsite=None): Merger.__init__(self, rddconf, aggregator, size, api_callsite) if aggregator: self.aggregator = SortMergeAggregator(self.aggregator.mergeCombiners) else: self.aggregator = CoGroupSortMergeAggregator(size) self.combined = iter([]) self.paths = [] def _merge_sorted(self, iters): heap = HeapOnKey(key=lambda x: x[0], min_heap=True) merged = heap.merge(iters) return self.aggregator.aggregate_sorted(merged) def _disk_merge_sorted(self, iters): t = time.time() s = AutoBatchedSerializer() iters = iter(iters) while True: batch = list(islice(iters, 100)) if not batch: break path = env.workdir.alloc_tmp_file("sort_merger") with open(path, 'wb') as f: s.dump_stream(self._merge_sorted(batch), f) self.paths.append(path) env.task_stats.num_fetch_rotate += 1 files = [s.load_stream(open(p)) for p in self.paths] env.task_stats.secs_fetch = time.time() - t return self._merge_sorted(files) def merge(self, iters): if self.rddconf.disk_merge or len(iters) > dpark.conf.MAX_OPEN_FILE: merged = self._disk_merge_sorted(iters) else: merged = self._merge_sorted(iters) self.combined = merged def __iter__(self): return self.combined class IterGroupBySortMerger(SortMerger): def _merge_sorted(self, iters): heap = HeapOnKey(key=lambda x: x[0], min_heap=True) return GroupByNestedIter(heap.merge(iters), self.api_callsite) class IterCoGroupSortMerger(SortMerger): def _merge_sorted(self, iters): # each item like <key, values> return cogroup_no_dup(list(map(iter, iters))) class MapOutputTracker(object): @classmethod def get_key(cls, shuffle_id): return 'shuffle:{}'.format(shuffle_id) @classmethod def set_locs(cls, shuffle_id, locs): key = cls.get_key(shuffle_id) env.trackerServer.set(key, locs) @classmethod def get_locs(cls, shuffle_id): key = cls.get_key(shuffle_id) if env.trackerServer: return env.trackerServer.get(key) else: return env.trackerClient.call(GetValueMessage(key)) class ShuffleWorkDir(object): def __init__(self, shuffle_id, input_id, output_id): self.subpath = os.path.join(str(shuffle_id), str(input_id), str(output_id)) def get(self): return env.workdir.get_path(self.subpath) @classmethod def alloc_tmp(cls, mem_first=True, datasize=0): return env.workdir.alloc_tmp_file("shuffle", mem_first, datasize) def export(self, tmppath): return env.workdir.export(tmppath, self.subpath) def restore(self, uri): if uri == env.server_uri: # urllib can open local file url = 'file://' + self.get() else: url = "%s/%s" % (uri, self.subpath) return url
bsd-3-clause
dc5560e712c048c7104ed1948d03914c
29.748235
109
0.529232
3.831135
false
false
false
false
douban/dpark
dpark/utils/frame.py
1
5431
import os import inspect from collections import defaultdict import linecache def get_path(p): return os.path.realpath(os.path.abspath(p)) src_dir = os.path.dirname(os.path.dirname(get_path(__file__))) class Frame(object): def __init__(self, f): """working in func_name, exec code at pos""" self.path = get_path(f.f_code.co_filename) self.lineno = f.f_lineno self.lasti = f.f_lasti self.func_name = f.f_code.co_name @property def pos(self): return self.path, self.lineno, self.lasti def frame_tuple(f): return f.f_code.co_filename, f.f_lineno, f.f_lasti def func_info(f): co = getattr(f, "__code__", None) if co: return "{}@{}:{}".format(co.co_name, co.co_filename, co.co_firstlineno) else: return "{}".format(f) # builtin_function_or_method def summary_stack(frames): result = [] for f in frames: co = f.f_code pos = '{}:{}, in {}'.format(co.co_filename, f.f_lineno, co.co_name) line = linecache.getline(co.co_filename, f.f_lineno).strip() if line: line = line.strip() # if f.f_locals: # for name, value in sorted(f.f_locals.items()): # row.append(' {name} = {value}\n'.format(name=name, value=value)) result.append({"pos": pos, "line": line}) return result class Scope(object): scopes_by_id = {} scopes_by_stackhash = {} scopes_by_api_callsite_id = {} api_callsites = {} calls_in_oneline = defaultdict(dict) # (path, line_no, fname) -> [lasti...] gid = 0 def __init__(self, name_fmt, stack, stackhash, api, api_callsite, stack_above_api): self.id = Scope.gid Scope.gid += 1 self.name = name_fmt.format(api=api) self.stack = stack self.stackhash = stackhash self.api = api self.api_callsite = api_callsite self.key = "{}@{}".format(api, self.api_callsite) self.api_callsite_id = self.api_callsites.get(api_callsite) self.stack_above_api = stack_above_api if self.api_callsite_id is None: self.api_callsite_id = self.api_callsites[api_callsite] = len(self.api_callsites) self.scopes_by_api_callsite_id[self.api_callsite_id] = [self] else: self.scopes_by_api_callsite_id[self.api_callsite_id].append(self) # print(self.id, self.api_callsite_id, api_callsite, self.name) @classmethod def reset(cls): cls.scopes_by_id = {} cls.scopes_by_stackhash = {} cls.scopes_by_api_callsite_id = {} cls.api_callsites = {} cls.calls_in_oneline = defaultdict(dict) cls.gid = 0 @classmethod def get_callsite(cls, caller, callee): """ Deal with usage like "rdd.map(_).map(_)", distinguish same dpark api called in one line by lasti. To be comprehensible, replace lasti with order of calling of same api in this line , starts with 0. """ callee = Frame(callee) # the dpark api called by user, DparkContext.xxx() or RDD.xxx() caller = Frame(caller) # the first callsite out of dpark package, where user call dpark api key = caller.path, caller.lineno, callee.func_name calls = cls.calls_in_oneline.setdefault(key, []) i = -1 for i, lasti in enumerate(calls): if lasti == caller.lasti: seq = i break else: seq = i + 1 calls.append(caller.lasti) api = callee.func_name api_callsite = "{}:{}@{}:{}".format(callee.func_name, seq, caller.path, caller.lineno) return api, api_callsite @classmethod def get(cls, name_fmt): callee = inspect.currentframe() caller = callee.f_back stack = [] stack_above_api = [] api_caller = None api_callee = None while caller is not None: stack.append(frame_tuple(caller)) if api_callee is None: if src_dir != os.path.dirname(get_path(caller.f_code.co_filename)): api_callee = callee # the dpark api called by user, DparkContext.xxx() or RDD.xxx() api_caller = caller # the first callsite out of dpark package, where user call dpark api stack_above_api.append(caller) else: stack_above_api.append(caller) callee = caller caller = caller.f_back stack = tuple(stack) stackhash = hash(stack) scope = cls.scopes_by_stackhash.get(stackhash) if scope is None: stack_above_api = summary_stack(stack_above_api) api, api_callsite = cls.get_callsite(api_caller, api_callee) scope = Scope(name_fmt, stack, stackhash, api, api_callsite, stack_above_api) cls.scopes_by_stackhash[stackhash] = scope cls.scopes_by_id[scope.id] = scope return scope def get_stacks_of_threads(): import threading, sys, traceback threads = {} for t in threading.enumerate(): f = sys._current_frames()[t.ident] k = t.name stack = traceback.format_stack() v = { "stack": stack, "f_locals": "{}".format(f.f_locals), "f_back.f_locals": "{}".format(f.f_back.f_locals) } threads[k] = v return threads
bsd-3-clause
e39d29b6b3316c096f6b63477778bb7e
31.915152
109
0.573559
3.428662
false
false
false
false
dials/dials
src/dials/util/export_shelx.py
1
6338
from __future__ import annotations import logging from math import isclose from cctbx import crystal, miller from iotbx.shelx.write_ins import LATT_SYMM from dials.algorithms.scaling.scaling_library import determine_best_unit_cell from dials.array_family import flex from dials.util import Sorry from dials.util.filter_reflections import filter_reflection_table logger = logging.getLogger(__name__) def export_shelx(scaled_data, experiment_list, params): """Export scaled data corresponding to experiment_list to a SHELX HKL formatted text file.""" # Handle requesting profile intensities (default via auto) but no column if "profile" in params.intensity and "intensity.prf.value" not in scaled_data: raise Sorry( "Requested profile intensity data but only summed present. Use intensity=sum." ) # use supplied best unit cell or that determined from experiment list to define d in reflection table. best_unit_cell = params.mtz.best_unit_cell if best_unit_cell is None: best_unit_cell = determine_best_unit_cell(experiment_list) else: logger.info("Using supplied unit cell across experiments : %s", best_unit_cell) scaled_data["d"] = best_unit_cell.d(scaled_data["miller_index"]) # Clean up reflection table with mtz defaults (as in export_xds_ascii) scaled_data = filter_reflection_table( scaled_data, intensity_choice=params.intensity, partiality_threshold=params.mtz.partiality_threshold, combine_partials=params.mtz.combine_partials, min_isigi=params.mtz.min_isigi, filter_ice_rings=params.mtz.filter_ice_rings, d_min=params.mtz.d_min, ) # Check that all experiments have the same space group if len({x.crystal.get_space_group().make_tidy() for x in experiment_list}) != 1: raise ValueError("Experiments do not have a unique space group") # Create miller set with space group from 1st crystal in experiment list and best unit cell miller_set = miller.set( crystal_symmetry=crystal.symmetry( unit_cell=best_unit_cell, space_group=experiment_list[0].crystal.get_space_group(), ), indices=scaled_data["miller_index"], anomalous_flag=False, ) intensity_choice = ( params.intensity[0] if params.intensity[0] != "profile" else "prf" ) intensities, variances = ( scaled_data["intensity." + intensity_choice + ".value"], scaled_data["intensity." + intensity_choice + ".variance"], ) assert variances.all_gt(0) i_obs = miller.array(miller_set, data=intensities) i_obs.set_observation_type_xray_intensity() i_obs.set_sigmas(flex.sqrt(variances)) # If requested scale up data to maximise use of scale_range if params.shelx.scale: max_val = max( i_obs.data().min_max_mean().max, i_obs.sigmas().min_max_mean().max ) min_val = min( i_obs.data().min_max_mean().min, i_obs.sigmas().min_max_mean().min ) min_scale = params.shelx.scale_range[0] / min_val max_scale = params.shelx.scale_range[1] / max_val scale = min(min_scale, max_scale) i_obs = i_obs.apply_scaling(factor=scale) # write the SHELX HKL file hkl_file = params.shelx.hklout with open(hkl_file, "w") as hkl_file_object: i_obs.export_as_shelx_hklf( file_object=hkl_file_object, scale_range=params.shelx.scale_range, normalise_if_format_overflow=True, ) logger.info(f"Written {i_obs.size()} reflections to {hkl_file}") # and a stub of an .ins file with information from the .expt file _write_ins( experiment_list, best_unit_cell=params.mtz.best_unit_cell, ins_file=params.shelx.ins, ) logger.info(f"Written {params.shelx.ins}") def _write_ins(experiment_list, best_unit_cell, ins_file): sg = experiment_list[0].crystal.get_space_group() unit_cells = [] wavelengths = [] # Check for single wavelength for exp in experiment_list: wl = exp.beam.get_wavelength() if not any([isclose(wl, w, abs_tol=1e-4) for w in wavelengths]): wavelengths.append(wl) if len(wavelengths) > 1: raise ValueError("Experiments have more than one wavelength") else: wl = wavelengths[0] # if user has supplied best_unit_cell use it if best_unit_cell is not None: uc = best_unit_cell uc_sd = None else: for exp in experiment_list: unit_cells.append( exp.crystal.get_recalculated_unit_cell() or exp.crystal.get_unit_cell() ) if len(unit_cells) > 1: if ( len({uc.parameters() for uc in unit_cells}) > 1 ): # have different cells so no esds uc = determine_best_unit_cell(experiment_list) uc_sd = None else: # identical (recalculated?) unit cell with esds uc = ( experiment_list[0].crystal.get_recalculated_unit_cell() or experiment_list[0].crystal.get_unit_cell() ) uc_sd = ( experiment_list[0].crystal.get_recalculated_cell_parameter_sd() or experiment_list[0].crystal.get_cell_parameter_sd() ) else: # single unit cell uc = ( experiment_list[0].crystal.get_recalculated_unit_cell() or experiment_list[0].crystal.get_unit_cell() ) uc_sd = ( experiment_list[0].crystal.get_recalculated_cell_parameter_sd() or experiment_list[0].crystal.get_cell_parameter_sd() ) with open(ins_file, "w") as f: f.write( f"TITL {sg.type().number()} in {sg.type().lookup_symbol().replace(' ','')}\n" ) f.write( "CELL {:8.5f} {:8.4f} {:8.4f} {:8.4f} {:8.3f} {:8.3f} {:8.3f}\n".format( wl, *uc.parameters() ) ) if uc_sd: f.write( "ZERR {:8.3f} {:8.4f} {:8.4f} {:8.4f} {:8.3f} {:8.3f} {:8.3f}\n".format( sg.order_z(), *uc_sd ) ) LATT_SYMM(f, sg)
bsd-3-clause
1d422dbf01a60dd156ce6f54117cb4e8
36.502959
106
0.5994
3.40204
false
false
false
false
dials/dials
tests/algorithms/refinement/test_restraints_parameterisation.py
1
14559
"""Tests for RestraintsParameterisation and associated classes used in refinement""" from __future__ import annotations import os import random import pytest from dxtbx.model.experiment_list import ExperimentListFactory from libtbx.phil import parse from dials.algorithms.refinement import RefinerFactory from dials.algorithms.refinement.restraints import RestraintsParameterisation from dials.array_family import flex def test_single_crystal_restraints_gradients(): """Simple test with a single triclinic crystal restrained to a target unit cell""" from dxtbx.model.experiment_list import Experiment, ExperimentList from dials.algorithms.refinement.parameterisation.beam_parameters import ( BeamParameterisation, ) from dials.algorithms.refinement.parameterisation.crystal_parameters import ( CrystalOrientationParameterisation, CrystalUnitCellParameterisation, ) from dials.algorithms.refinement.parameterisation.detector_parameters import ( DetectorParameterisationSinglePanel, ) from dials.algorithms.refinement.parameterisation.prediction_parameters import ( XYPhiPredictionParameterisation, ) from . import geometry_phil from .setup_geometry import Extract overrides = """geometry.parameters.crystal.a.length.range = 10 50 geometry.parameters.crystal.b.length.range = 10 50 geometry.parameters.crystal.c.length.range = 10 50""" master_phil = parse(geometry_phil) models = Extract(master_phil, overrides) mydetector = models.detector mygonio = models.goniometer mycrystal = models.crystal mybeam = models.beam # Build a mock scan for a 72 degree sequence from dxtbx.model import ScanFactory sf = ScanFactory() myscan = sf.make_scan( image_range=(1, 720), exposure_times=0.1, oscillation=(0, 0.1), epochs=list(range(720)), deg=True, ) # Create parameterisations of these models det_param = DetectorParameterisationSinglePanel(mydetector) s0_param = BeamParameterisation(mybeam, mygonio) xlo_param = CrystalOrientationParameterisation(mycrystal) xluc_param = CrystalUnitCellParameterisation(mycrystal) # Create an ExperimentList experiments = ExperimentList() experiments.append( Experiment( beam=mybeam, detector=mydetector, goniometer=mygonio, scan=myscan, crystal=mycrystal, imageset=None, ) ) # Build a prediction parameterisation pred_param = XYPhiPredictionParameterisation( experiments, detector_parameterisations=[det_param], beam_parameterisations=[s0_param], xl_orientation_parameterisations=[xlo_param], xl_unit_cell_parameterisations=[xluc_param], ) # Build a restraints parameterisation rp = RestraintsParameterisation( detector_parameterisations=[det_param], beam_parameterisations=[s0_param], xl_orientation_parameterisations=[xlo_param], xl_unit_cell_parameterisations=[xluc_param], ) # make a unit cell target sigma = 1.0 uc = mycrystal.get_unit_cell().parameters() target_uc = [random.gauss(e, sigma) for e in uc] rp.add_restraints_to_target_xl_unit_cell( experiment_id=0, values=target_uc, sigma=[sigma] * 6 ) # get analytical values and gradients vals, grads, weights = rp.get_residuals_gradients_and_weights() assert len(vals) == rp.num_residuals # get finite difference gradients p_vals = pred_param.get_param_vals() deltas = [1.0e-7] * len(p_vals) fd_grad = [] for i, delta in enumerate(deltas): val = p_vals[i] p_vals[i] -= delta / 2.0 pred_param.set_param_vals(p_vals) rev_state, foo, bar = rp.get_residuals_gradients_and_weights() rev_state = flex.double(rev_state) p_vals[i] += delta pred_param.set_param_vals(p_vals) fwd_state, foo, bar = rp.get_residuals_gradients_and_weights() fwd_state = flex.double(fwd_state) p_vals[i] = val fd = (fwd_state - rev_state) / delta fd_grad.append(fd) # for comparison, fd_grad is a list of flex.doubles, each of which corresponds # to a column of the sparse matrix grads. for i, fd in enumerate(fd_grad): # extract dense column from the sparse matrix an = grads.col(i).as_dense_vector() assert an == pytest.approx(fd, abs=1e-5) def test_two_triclinic_crystals(): """Simple test with two triclinic crystals restrained to a target unit cell""" from dxtbx.model.experiment_list import Experiment, ExperimentList from dials.algorithms.refinement.parameterisation.beam_parameters import ( BeamParameterisation, ) from dials.algorithms.refinement.parameterisation.crystal_parameters import ( CrystalOrientationParameterisation, CrystalUnitCellParameterisation, ) from dials.algorithms.refinement.parameterisation.detector_parameters import ( DetectorParameterisationSinglePanel, ) #### Import model parameterisations from dials.algorithms.refinement.parameterisation.prediction_parameters import ( XYPhiPredictionParameterisation, ) from . import geometry_phil from .setup_geometry import Extract overrides = """geometry.parameters.crystal.a.length.range = 10 50 geometry.parameters.crystal.b.length.range = 10 50 geometry.parameters.crystal.c.length.range = 10 50""" master_phil = parse(geometry_phil) models = Extract(master_phil, overrides) mydetector = models.detector mygonio = models.goniometer mycrystal = models.crystal # duplicate the crystal from copy import deepcopy mycrystal2 = deepcopy(mycrystal) mybeam = models.beam # Build a mock scan for a 72 degree sequence from dxtbx.model import ScanFactory sf = ScanFactory() myscan = sf.make_scan( image_range=(1, 720), exposure_times=0.1, oscillation=(0, 0.1), epochs=list(range(720)), deg=True, ) # Create parameterisations of these models det_param = DetectorParameterisationSinglePanel(mydetector) s0_param = BeamParameterisation(mybeam, mygonio) xlo_param = CrystalOrientationParameterisation(mycrystal) xluc_param = CrystalUnitCellParameterisation(mycrystal) xluc_param2 = CrystalUnitCellParameterisation(mycrystal2, experiment_ids=[1]) # Create an ExperimentList with the crystal duplicated experiments = ExperimentList() experiments.append( Experiment( beam=mybeam, detector=mydetector, goniometer=mygonio, scan=myscan, crystal=mycrystal, imageset=None, ) ) experiments.append( Experiment( beam=mybeam, detector=mydetector, goniometer=mygonio, scan=myscan, crystal=mycrystal2, imageset=None, ) ) # Build a prediction parameterisation pred_param = XYPhiPredictionParameterisation( experiments, detector_parameterisations=[det_param], beam_parameterisations=[s0_param], xl_orientation_parameterisations=[xlo_param], xl_unit_cell_parameterisations=[xluc_param, xluc_param2], ) # Build a restraints parameterisation rp = RestraintsParameterisation( detector_parameterisations=[det_param], beam_parameterisations=[s0_param], xl_orientation_parameterisations=[xlo_param], xl_unit_cell_parameterisations=[xluc_param, xluc_param2], ) # make a unit cell target sigma = 1.0 uc = mycrystal.get_unit_cell().parameters() target_uc = [random.gauss(e, sigma) for e in uc] rp.add_restraints_to_target_xl_unit_cell( experiment_id=0, values=target_uc, sigma=[sigma] * 6 ) rp.add_restraints_to_target_xl_unit_cell( experiment_id=1, values=target_uc, sigma=[sigma] * 6 ) # get analytical values and gradients vals, grads, weights = rp.get_residuals_gradients_and_weights() assert len(vals) == rp.num_residuals # get finite difference gradients p_vals = pred_param.get_param_vals() deltas = [1.0e-7] * len(p_vals) fd_grad = [] for i, delta in enumerate(deltas): val = p_vals[i] p_vals[i] -= delta / 2.0 pred_param.set_param_vals(p_vals) rev_state, foo, bar = rp.get_residuals_gradients_and_weights() rev_state = flex.double(rev_state) p_vals[i] += delta pred_param.set_param_vals(p_vals) fwd_state, foo, bar = rp.get_residuals_gradients_and_weights() fwd_state = flex.double(fwd_state) p_vals[i] = val fd = (fwd_state - rev_state) / delta fd_grad.append(fd) # for comparison, fd_grad is a list of flex.doubles, each of which corresponds # to a column of the sparse matrix grads. for i, fd in enumerate(fd_grad): # extract dense column from the sparse matrix an = grads.col(i).as_dense_vector() assert an == pytest.approx(fd, abs=1e-5) def test_10_crystals_with_stills_parameterisation(dials_regression): """Test with multiple crystals, and a stills refiner""" # The phil scope from dials.algorithms.refinement.refiner import phil_scope user_phil = parse( """ refinement { parameterisation { crystal { unit_cell { restraints { tie_to_target { values=95,95,132,90,90,120 sigmas=1,1,1,0,0,0 id=0,1,2,3,4,5,6,7,8,9 } } } } } } """ ) working_phil = phil_scope.fetch(source=user_phil) working_params = working_phil.extract() # use the multi stills test data data_dir = os.path.join(dials_regression, "refinement_test_data", "multi_stills") experiments_path = os.path.join(data_dir, "combined_experiments.json") pickle_path = os.path.join(data_dir, "combined_reflections.pickle") experiments = ExperimentListFactory.from_json_file( experiments_path, check_format=False ) reflections = flex.reflection_table.from_file(pickle_path) refiner = RefinerFactory.from_parameters_data_experiments( working_params, reflections, experiments ) # hack to extract the objects needed from the Refiner rp = refiner._target._restraints_parameterisation pred_param = refiner._pred_param # get analytical values and gradients vals, grads, weights = rp.get_residuals_gradients_and_weights() assert len(vals) == rp.num_residuals # get finite difference gradients p_vals = pred_param.get_param_vals() deltas = [1.0e-7] * len(p_vals) fd_grad = [] for i, delta in enumerate(deltas): val = p_vals[i] p_vals[i] -= delta / 2.0 pred_param.set_param_vals(p_vals) rev_state, foo, bar = rp.get_residuals_gradients_and_weights() rev_state = flex.double(rev_state) p_vals[i] += delta pred_param.set_param_vals(p_vals) fwd_state, foo, bar = rp.get_residuals_gradients_and_weights() fwd_state = flex.double(fwd_state) p_vals[i] = val fd = (fwd_state - rev_state) / delta fd_grad.append(fd) # for comparison, fd_grad is a list of flex.doubles, each of which corresponds # to a column of the sparse matrix grads. pnames = pred_param.get_param_names() for i, (pname, fd) in enumerate(zip(pnames, fd_grad)): # extract dense column from the sparse matrix an = grads.col(i).as_dense_vector() # print pname # print list(an.round(6)) # print list(fd.round(6)) # print assert an == pytest.approx(fd, abs=1e-5) def test_group_restraint_with_multiple_crystals_and_a_stills_refiner(dials_regression): # The phil scope from dials.algorithms.refinement.refiner import phil_scope user_phil = parse( """ refinement { parameterisation { crystal { unit_cell { restraints { tie_to_group { sigmas=1,0,2,0,0,0 } } } } } } """ ) working_phil = phil_scope.fetch(source=user_phil) working_params = working_phil.extract() # use the multi stills test data data_dir = os.path.join(dials_regression, "refinement_test_data", "multi_stills") experiments_path = os.path.join(data_dir, "combined_experiments.json") pickle_path = os.path.join(data_dir, "combined_reflections.pickle") experiments = ExperimentListFactory.from_json_file( experiments_path, check_format=False ) reflections = flex.reflection_table.from_file(pickle_path) refiner = RefinerFactory.from_parameters_data_experiments( working_params, reflections, experiments ) # hack to extract the objects needed from the Refiner rp = refiner._target._restraints_parameterisation pred_param = refiner._pred_param # get analytical values and gradients vals, grads, weights = rp.get_residuals_gradients_and_weights() assert len(vals) == rp.num_residuals # get finite difference gradients p_vals = pred_param.get_param_vals() deltas = [1.0e-7] * len(p_vals) fd_grad = [] for i, delta in enumerate(deltas): val = p_vals[i] p_vals[i] -= delta / 2.0 pred_param.set_param_vals(p_vals) rev_state, foo, bar = rp.get_residuals_gradients_and_weights() rev_state = flex.double(rev_state) p_vals[i] += delta pred_param.set_param_vals(p_vals) fwd_state, foo, bar = rp.get_residuals_gradients_and_weights() fwd_state = flex.double(fwd_state) p_vals[i] = val fd = (fwd_state - rev_state) / delta fd_grad.append(fd) # for comparison, fd_grad is a list of flex.doubles, each of which corresponds # to the gradients of the residuals wrt to a single parameter. pnames = pred_param.get_param_names() for i, (pname, fd) in enumerate(zip(pnames, fd_grad)): # extract dense column from the sparse matrix an = grads.col(i).as_dense_vector() # print pname # print list(an.round(6)) # print list(fd.round(6)) # print assert an == pytest.approx(fd, abs=1e-5)
bsd-3-clause
82f8544ad45ee0fd5c28325c33bdd28f
28.895277
87
0.647984
3.462307
false
false
false
false
dials/dials
tests/algorithms/scaling/test_scaling_restraints.py
1
7745
""" Tests for the scaling restraints module. """ from __future__ import annotations from unittest.mock import Mock import pytest from scitbx import sparse from dials.algorithms.scaling.scaling_restraints import ( ScalingRestraintsCalculator, SingleScalingRestraintsCalculator, ) from dials.array_family import flex @pytest.fixture def mock_restrained_component(): """Mock a component with restraints.""" component = Mock() component.n_params = 3 component.calculate_restraints.return_value = [ flex.double([1.0, 2.0, 3.0]), flex.double([0.1, 0.2, 0.3]), ] jacobian_restr = sparse.matrix(component.n_params, component.n_params) jacobian_restr[0, 0] = 1.0 component.calculate_jacobian_restraints.return_value = [ flex.double([1.0, 1.1, 1.2]), jacobian_restr, flex.double([1.0, 2.0, 3.0]), ] return component @pytest.fixture def mock_unrestrained_component(): """Mock a component without restraints.""" component = Mock() component.n_params = 5 component.calculate_restraints.return_value = None component.calculate_jacobian_restraints.return_value = None return component @pytest.fixture def mock_parameter_manager(mock_restrained_component, mock_unrestrained_component): """Mock a parameter manager to handle the components required for the ScalingRestraints class.""" apm = Mock() apm.components = { "restrained": { "object": mock_restrained_component, "n_params": mock_restrained_component.n_params, "start_idx": 0, }, "unrestrained": { "object": mock_unrestrained_component, "n_params": mock_unrestrained_component.n_params, "start_idx": mock_restrained_component.n_params, }, } apm.n_active_params = ( mock_restrained_component.n_params + mock_unrestrained_component.n_params ) return apm @pytest.fixture def mock_unrestrained_apm(mock_unrestrained_component): """Mock a parameter manager to handle no restrained components.""" apm = Mock() apm.components = { "unrestrained": { "object": mock_unrestrained_component, "n_params": mock_unrestrained_component.n_params, "start_idx": 0, } } apm.n_active_params = mock_unrestrained_component.n_params return apm @pytest.fixture def mock_multi_apm(mock_parameter_manager): """Mock a multi-dataset parameter manager.""" multi_apm = Mock() multi_apm.apm_list = [mock_parameter_manager, mock_parameter_manager] n_params = mock_parameter_manager.n_active_params multi_apm.n_active_params = n_params * 2 multi_apm.apm_data = { 0: {"start_idx": 0, "end_idx": n_params}, 1: {"start_idx": n_params, "end_idx": n_params * 2}, } return multi_apm @pytest.fixture def mock_multi_unrestrained_apm(mock_unrestrained_apm): """Mock a parameter manager to handle no restrained components.""" multi_apm = Mock() multi_apm.apm_list = [mock_unrestrained_apm, mock_unrestrained_apm] return multi_apm def test_unrestrained_ScalingRestraints( mock_unrestrained_apm, mock_multi_unrestrained_apm ): """Test the case of unrestrained components. None should be returned in each case.""" assert ( SingleScalingRestraintsCalculator.calculate_restraints(mock_unrestrained_apm) is None ) assert ( SingleScalingRestraintsCalculator.calculate_jacobian_restraints( mock_unrestrained_apm ) is None ) assert ( ScalingRestraintsCalculator.calculate_restraints(mock_multi_unrestrained_apm) is None ) assert ( ScalingRestraintsCalculator.calculate_jacobian_restraints( mock_multi_unrestrained_apm ) is None ) def test_ScalingRestraints( mock_parameter_manager, mock_restrained_component, mock_unrestrained_component ): """Test for the single scaling restraints manager.""" # Test the call to calculate restraints. This should return a residual # vector of the same length as the restraints of the restrained component, # and a gradient vector of the total length of all parameters. restraints = SingleScalingRestraintsCalculator.calculate_restraints( mock_parameter_manager ) abs_restraints = mock_restrained_component.calculate_restraints() assert list(restraints[0]) == list(abs_restraints[0]) assert list(restraints[1]) == ( list(abs_restraints[1]) + [0.0] * mock_unrestrained_component.n_params ) # Test the call to calculate jacobian restraints. This should return a # restraints vector, jacobian and weights vector from the components. The # restraints and weights should be the length of the restrained component. # The jacobian has n_rows equal to the number of restrainted parameters, # n_cols equal to the total number of parameters. Check that these are # correctly composed. jacobian_restraints = ( SingleScalingRestraintsCalculator.calculate_jacobian_restraints( mock_parameter_manager ) ) abs_restraints = mock_restrained_component.calculate_jacobian_restraints() assert list(jacobian_restraints[0]) == list(abs_restraints[0]) assert jacobian_restraints[1].n_cols == mock_parameter_manager.n_active_params assert jacobian_restraints[1].n_rows == mock_restrained_component.n_params for i in range(mock_restrained_component.n_params): for j in range(mock_restrained_component.n_params): assert jacobian_restraints[1][i, j] == abs_restraints[1][i, j] # All other elements should be zero assert abs_restraints[1].non_zeroes == jacobian_restraints[1].non_zeroes def test_MultiScalingRestraints( mock_multi_apm, mock_restrained_component, mock_unrestrained_component ): """Test for the multi-dataset scaling restraints manager.""" # Test the call to calculate restraints. Expected return is the individual # dataset vectors joined together. restraints = ScalingRestraintsCalculator.calculate_restraints(mock_multi_apm) abs_restraints = mock_restrained_component.calculate_restraints() assert list(restraints[0]) == (list(abs_restraints[0]) + list(abs_restraints[0])) assert list(restraints[1]) == ( list(abs_restraints[1]) + [0.0] * mock_unrestrained_component.n_params + list(abs_restraints[1]) + [0.0] * mock_unrestrained_component.n_params ) # Test the call to calculate jacobian restraints. Again, the expected return # is the individual dataset vectors joined together. jacobian_restraints = ScalingRestraintsCalculator.calculate_jacobian_restraints( mock_multi_apm ) abs_restraints = mock_restrained_component.calculate_jacobian_restraints() assert list(jacobian_restraints[0]) == ( list(abs_restraints[0]) + list(abs_restraints[0]) ) n_abs_params = mock_restrained_component.n_params n_total_params = mock_multi_apm.apm_data[0]["end_idx"] assert jacobian_restraints[1].n_cols == mock_multi_apm.n_active_params assert jacobian_restraints[1].n_rows == n_abs_params * 2 # Check that both restraints jacobians were set in correct location. for i in range(mock_restrained_component.n_params): for j in range(mock_restrained_component.n_params): assert jacobian_restraints[1][i, j] == abs_restraints[1][i, j] assert jacobian_restraints[1][i + n_abs_params, j + n_total_params] == ( abs_restraints[1][i, j] ) assert abs_restraints[1].non_zeroes * 2 == jacobian_restraints[1].non_zeroes
bsd-3-clause
ae08ee0232d71403a2ef04b2e7f4f72f
34.856481
85
0.680568
3.477773
false
true
false
false
dials/dials
src/dials/command_line/export_bitmaps.py
1
10013
from __future__ import annotations import os import sys from PIL import Image import iotbx.phil from dxtbx.model.detector_helpers import get_detector_projection_2d_axes from dials.algorithms.image.threshold import DispersionThresholdDebug from dials.array_family import flex from dials.util import Sorry, show_mail_handle_errors from dials.util.image_viewer.slip_viewer.flex_image import ( get_flex_image, get_flex_image_multipanel, ) from dials.util.options import ArgumentParser, flatten_experiments help_message = """ Export raw diffraction image files as bitmap images, optionally exporting images from intermediate spot-finding steps (local mean and variance maps, or sigma_b, sigma_s or threshold-filtered images). Appearance of the images can be altered via the brightness and colour_scheme parameters, and optionally binning of pixels can be used to reduce image sizes. Examples:: dials.export_bitmaps image.cbf dials.export_bitmaps models.expt dials.export_bitmaps image.cbf display=variance colour_scheme=inverse_greyscale """ phil_scope = iotbx.phil.parse( """ binning = 1 .type = int(value_min=1) brightness = 100 .type = float(value_min=0.0) colour_scheme = *greyscale rainbow heatmap inverse_greyscale .type = choice projection = lab *image .type = choice padding = 4 .type = int(value_min=0) imageset_index = None .type = int .multiple = True .help = "The index/indices from an imageset to export. The first image of " "the set is 1." .expert_level=2 display = *image mean variance dispersion sigma_b \ sigma_s threshold global_threshold .type = choice nsigma_b = 6 .type = float(value_min=0) nsigma_s = 3 .type = float(value_min=0) global_threshold = 0 .type = float(value_min=0) kernel_size = 3,3 .type = ints(size=2, value_min=1) min_local = 2 .type = int gain = 1 .type = float(value_min=0) saturation = 0 .type = int show_mask = False .type = bool png { compress_level = 1 .type = int(value_min=0, value_max=9) .help = "ZLIB compression level, a number between 0 and 9: 1 gives best " "speed, 9 gives best compression, 0 gives no compression at all." } jpeg { quality = 75 .type = int(value_min=1, value_max=95) .help = "The image quality, on a scale from 1 (worst) to 95 (best)" } include scope dials.util.options.format_phil_scope output { prefix = "image" .type = str directory = None .type = path file = None .type = str .help = "Full name of the output file. Overrides 'prefix' and the default " "file extension. Only makes sense if a single file is written." format = jpeg *png tiff .type = choice }""", process_includes=True, ) colour_schemes = {"greyscale": 0, "rainbow": 1, "heatmap": 2, "inverse_greyscale": 3} @show_mail_handle_errors() def run(args=None): usage = "dials.export_bitmaps [options] models.expt | image.cbf" parser = ArgumentParser( usage=usage, phil=phil_scope, read_experiments=True, read_experiments_from_images=True, check_format=True, epilog=help_message, ) params, options = parser.parse_args(args, show_diff_phil=True) experiments = flatten_experiments(params.input.experiments) if len(experiments) == 0: parser.print_help() exit(0) imagesets = experiments.imagesets() for imageset in imagesets: imageset_as_bitmaps(imageset, params) def imageset_as_bitmaps(imageset, params): brightness = params.brightness / 100 vendortype = "made up" # check that binning is a power of 2 binning = params.binning if not (binning > 0 and ((binning & (binning - 1)) == 0)): raise Sorry("binning must be a power of 2") output_dir = params.output.directory if output_dir is None: output_dir = "." elif not os.path.exists(output_dir): os.makedirs(output_dir) output_files = [] detector = imageset.get_detector() # Furnish detector with 2D projection axes detector.projected_2d = get_detector_projection_2d_axes(detector) detector.projection = params.projection panel = detector[0] scan = imageset.get_scan() # XXX is this inclusive or exclusive? saturation = panel.get_trusted_range()[1] if params.saturation: saturation = params.saturation if scan is not None and not scan.is_still() and not params.imageset_index: start, end = scan.get_image_range() else: start, end = 1, len(imageset) # If the user specified an image range index, only export those image_range = [ i for i in range(start, end + 1) if not params.imageset_index or i in params.imageset_index ] if params.output.file and len(image_range) != 1: sys.exit("output.file can only be specified if a single image is exported") for i_image in image_range: image = imageset.get_raw_data(i_image - start) mask = imageset.get_mask(i_image - start) if mask is None: mask = [p.get_trusted_range_mask(im) for im, p in zip(image, detector)] if params.show_mask: for rd, m in zip(image, mask): rd.set_selected(~m, -2) image = image_filter( image, mask, display=params.display, gain_value=params.gain, nsigma_b=params.nsigma_b, nsigma_s=params.nsigma_s, global_threshold=params.global_threshold, min_local=params.min_local, kernel_size=params.kernel_size, ) show_untrusted = params.show_mask if len(detector) > 1: # FIXME This doesn't work properly, as flex_image.size2() is incorrect # also binning doesn't work flex_image = get_flex_image_multipanel( brightness=brightness, detector=detector, image_data=image, binning=binning, beam=imageset.get_beam(), show_untrusted=show_untrusted, ) else: flex_image = get_flex_image( brightness=brightness, data=image[0], binning=binning, saturation=saturation, vendortype=vendortype, show_untrusted=show_untrusted, ) flex_image.setWindow(0, 0, 1) flex_image.adjust(color_scheme=colour_schemes.get(params.colour_scheme)) # now export as a bitmap flex_image.prep_string() # XXX is size//binning safe here? pil_img = Image.frombytes( "RGB", (flex_image.ex_size2(), flex_image.ex_size1()), flex_image.as_bytes() ) if params.output.file: path = os.path.join(output_dir, params.output.file) else: path = os.path.join( output_dir, "{prefix}{image:0{padding}}.{format}".format( image=i_image, prefix=params.output.prefix, padding=params.padding, format=params.output.format, ), ) print(f"Exporting {path}") output_files.append(path) with open(path, "wb") as tmp_stream: pil_img.save( tmp_stream, format=params.output.format, compress_level=params.png.compress_level, quality=params.jpeg.quality, ) return output_files def image_filter( raw_data, mask, display, gain_value, nsigma_b, nsigma_s, global_threshold, min_local, kernel_size, ): if display == "image": return raw_data assert gain_value > 0 gain_map = [flex.double(rd.accessor(), gain_value) for rd in raw_data] kabsch_debug_list = [ DispersionThresholdDebug( data.as_double(), mask[i_panel], gain_map[i_panel], kernel_size, nsigma_b, nsigma_s, global_threshold, min_local, ) for i_panel, data in enumerate(raw_data) ] if display == "mean": display_data = [kabsch.mean() for kabsch in kabsch_debug_list] elif display == "variance": display_data = [kabsch.variance() for kabsch in kabsch_debug_list] elif display == "dispersion": display_data = [kabsch.index_of_dispersion() for kabsch in kabsch_debug_list] elif display == "sigma_b": cv = [kabsch.index_of_dispersion() for kabsch in kabsch_debug_list] display_data = (kabsch.cv_mask() for kabsch in kabsch_debug_list) display_data = [_mask.as_1d().as_double() for _mask in display_data] for i, _mask in enumerate(display_data): _mask.reshape(cv[i].accessor()) elif display == "sigma_s": cv = [kabsch.index_of_dispersion() for kabsch in kabsch_debug_list] display_data = (kabsch.value_mask() for kabsch in kabsch_debug_list) display_data = [_mask.as_1d().as_double() for _mask in display_data] for i, _mask in enumerate(display_data): _mask.reshape(cv[i].accessor()) elif display == "global_threshold": cv = [kabsch.index_of_dispersion() for kabsch in kabsch_debug_list] display_data = (kabsch.global_mask() for kabsch in kabsch_debug_list) display_data = [_mask.as_1d().as_double() for _mask in display_data] for i, _mask in enumerate(display_data): _mask.reshape(cv[i].accessor()) elif display == "threshold": cv = [kabsch.index_of_dispersion() for kabsch in kabsch_debug_list] display_data = (kabsch.final_mask() for kabsch in kabsch_debug_list) display_data = [_mask.as_1d().as_double() for _mask in display_data] for i, _mask in enumerate(display_data): _mask.reshape(cv[i].accessor()) return display_data if __name__ == "__main__": run()
bsd-3-clause
d906256aaa0051ae98491e7035c95672
30.487421
88
0.613602
3.508409
false
false
false
false
dials/dials
src/dials/model/experiment/profile.py
1
2922
from __future__ import annotations class ProfileModelExt: """ The definition for a profile model. """ @staticmethod def create( params, reflections, crystal, beam, detector, goniometer=None, scan=None ): """ Create the profile model from data. :param params: The phil parameters :param reflections: The reflections :param crystal: The crystal model :param beam: The beam model :param detector: The detector model :param goniometer: The goniometer model :param scan: The scan model :return: An instance of the profile model """ return None def predict_reflections( self, imageset, crystal, beam, detector, goniometer=None, scan=None, **kwargs ): """ Given an experiment, predict the reflections. :param imageset: The imageset :param crystal: The crystal model :param beam: The beam model :param detector: The detector model :param goniometer: The goniometer model :param scan: The scan model """ pass def compute_partiality( self, reflections, crystal, beam, detector, goniometer=None, scan=None, **kwargs ): """ Given an experiment and list of reflections, compute the partiality of the reflections :param reflections: The reflection table :param crystal: The crystal model :param beam: The beam model :param detector: The detector model :param goniometer: The goniometer model :param scan: The scan model """ pass def compute_bbox( self, reflections, crystal, beam, detector, goniometer=None, scan=None, **kwargs ): """Given an experiment and list of reflections, compute the bounding box of the reflections on the detector (and image frames). :param reflections: The reflection table :param crystal: The crystal model :param beam: The beam model :param detector: The detector model :param goniometer: The goniometer model :param scan: The scan model """ pass def compute_mask( self, reflections, crystal, beam, detector, goniometer=None, scan=None, **kwargs ): """ Given an experiment and list of reflections, compute the foreground/background mask of the reflections. :param reflections: The reflection table :param crystal: The crystal model :param beam: The beam model :param detector: The detector model :param goniometer: The goniometer model :param scan: The scan model """ pass def fitting_class(self): """ Get the profile fitting algorithm associated with this profile model :return: The profile fitting class """ return None
bsd-3-clause
82caa9878e814e0762e4bfac9d61ac63
29.757895
88
0.619781
4.297059
false
false
false
false
dials/dials
src/dials/algorithms/scaling/scale_and_filter.py
1
19957
"""Definitions of functions and classes for scaling and filtering algorithm.""" from __future__ import annotations import math from libtbx import phil from scitbx.array_family import flex from dials.report.plots import d_star_sq_to_d_ticks phil_scope = phil.parse( """ filtering { method = None deltacchalf .type = choice .help = "Choice of whether to do any filtering cycles, default None." deltacchalf { max_cycles = 6 .type = int(value_min=1) max_percent_removed = 10 .type = float min_completeness = None .type = float(value_min=0, value_max=100) .help = "Desired minimum completeness, as a percentage (0 - 100)." mode = *dataset image_group .type = choice .help = "Perform analysis on whole datasets or batch groups" group_size = 10 .type = int(value_min=1) .help = "The number of images to group together when calculating delta" "cchalf in image_group mode" stdcutoff = 4.0 .type = float .help = "Datasets with a ΔCC½ below (mean - stdcutoff*std) are removed" } output { scale_and_filter_results = "scale_and_filter_results.json" .type = str .help = "Filename for output json of scale and filter results." } } """ ) def ordinal(n): return "%d%s" % (n, "tsnrhtdd"[(n / 10 % 10 != 1) * (n % 10 < 4) * n % 10 :: 4]) def log_cycle_results(results, scaling_script, filter_script): """Log results from the scripts for this cycle and add to the results.""" cycle_results = {"merging_stats": scaling_script.merging_statistics_result} if not results.get_cycle_results(): results.initial_n_reflections = scaling_script.scaled_miller_array.size() cycle_results["delta_cc_half_values"] = filter_script.results_summary[ "per_dataset_delta_cc_half_values" ]["delta_cc_half_values"] cycle_results["mean_cc_half"] = filter_script.results_summary["mean_cc_half"] removal_summary = filter_script.results_summary["dataset_removal"] if removal_summary["mode"] == "image_group": cycle_results["image_ranges_removed"] = removal_summary["image_ranges_removed"] cycle_results["removed_datasets"] = removal_summary["experiments_fully_removed"] cycle_results["removed_ids"] = removal_summary["experiment_ids_fully_removed"] cycle_results["n_removed"] = filter_script.results_summary["dataset_removal"][ "n_reflections_removed" ] n_removed = ( sum(res["n_removed"] for res in results.get_cycle_results()) + cycle_results["n_removed"] ) percent_removed = n_removed / results.initial_n_reflections * 100 cycle_results["cumul_percent_removed"] = percent_removed results.add_cycle_results(cycle_results) return results class AnalysisResults: """Class to store results from scaling and filtering.""" def __init__(self): self.termination_reason = None self.cycle_results = [] self.initial_n_reflections = None self.initial_expids_and_image_ranges = None self.expids_and_image_ranges = None self.final_stats = None def add_cycle_results(self, results_dict): """Add the results dict from a scale and filter cycle.""" merging_stats_dict = self._parse_merging_stats(results_dict["merging_stats"]) results_dict["merging_stats"] = merging_stats_dict self.cycle_results.append(results_dict) @staticmethod def _parse_merging_stats(merging_stats_obj): merging_stats = {} overall = merging_stats_obj.overall merging_stats["ccs"] = [b.cc_one_half for b in merging_stats_obj.bins] merging_stats["rmerge"] = [b.r_merge for b in merging_stats_obj.bins] merging_stats["rpim"] = [b.r_pim for b in merging_stats_obj.bins] merging_stats["d_min"] = [b.d_min for b in merging_stats_obj.bins] merging_stats["overall"] = { "cc_one_half": overall.cc_one_half, "r_merge": overall.r_merge, "r_pim": overall.r_pim, "i_over_sigma_mean": overall.i_over_sigma_mean, "completeness": 100 * overall.completeness, "n_obs": overall.n_obs, } return merging_stats def get_cycle_results(self): """Get the results from all cycles.""" return self.cycle_results def get_last_cycle_results(self): """Get the results from the latest recorded cycle.""" return self.cycle_results[-1] def add_final_stats(self, final_stats): """Add additional final merging stats from final rescale.""" self.final_stats = self._parse_merging_stats(final_stats) def get_merging_stats(self): """Get all merging stats, including additional final stats if present.""" stats = [res["merging_stats"] for res in self.cycle_results] if self.final_stats: stats += [self.final_stats] return stats def finish(self, termination_reason): """Set the termination reason/""" assert termination_reason in [ "no_more_removed", "max_cycles", "max_percent_removed", "below_completeness_limit", ] self.termination_reason = termination_reason def to_dict(self): """Return the stored data as a dictionary.""" return { "termination_reason": self.termination_reason, "initial_n_reflections": self.initial_n_reflections, "initial_expids_and_image_ranges": self.initial_expids_and_image_ranges, "expids_and_image_ranges": self.expids_and_image_ranges, "cycle_results": {i + 1: val for i, val in enumerate(self.cycle_results)}, "final_stats": self.final_stats, } @staticmethod def from_dict(dictionary): """Configure the class from its dictionary form.""" results = AnalysisResults() results.termination_reason = dictionary["termination_reason"] results.initial_expids_and_image_ranges = dictionary[ "initial_expids_and_image_ranges" ] results.expids_and_image_ranges = dictionary["expids_and_image_ranges"] results.cycle_results = [ dictionary["cycle_results"][str(key)] for key in sorted(int(k) for k in dictionary["cycle_results"].keys()) ] results.initial_n_reflections = dictionary["initial_n_reflections"] results.final_stats = dictionary["final_stats"] return results def __str__(self): """Make summary of results.""" msg = "\nSummary of data removed:\n" for i, res in enumerate(self.get_cycle_results()): msg += f"Cycle number: {i + 1}\n" if "image_ranges_removed" in res: if res["image_ranges_removed"]: removed = "\n ".join( str(t[0]) + ", dataset " + str(t[1]) for t in res["image_ranges_removed"] ) msg += f" Removed image ranges: \n {removed}" else: if res["removed_ids"]: msg += f" Removed datasets: {res['removed_ids']}\n" msg += ( " cumulative %% of reflections removed: %.3f\n" % res["cumul_percent_removed"] ) return msg color_list = [ "#F44336", "#FFC107", "#FFEB3B", "#8BC34A", "#03A9F4", "#3F51B5", "#607D8B", ] def make_scaling_filtering_plots(data): """Make the plots for scaling and filtering.""" d = make_filtering_merging_stats_plots(data["merging_stats"]) d.update(make_histogram_plots(data["cycle_results"])) if data["mode"] == "image_group": d.update( make_reduction_plots( data["initial_expids_and_image_ranges"], data["expids_and_image_ranges"] ) ) return d def make_filtering_merging_stats_plots(merging_stats): """Generate plotting dicts for merging statistics.""" n_datasets = len(merging_stats) colors = [ (color_list * int(math.ceil(n_datasets / len(color_list))))[i] for i in range(n_datasets) ] colors[-1] = "k" legends = ["initial scale"] if len(merging_stats) > 2: legends += [ ordinal(i) + " rescale" for i in range(1, len(merging_stats) - 1) ] + ["final rescale"] elif len(merging_stats) == 2: legends += ["final rescale"] overall_ccs = [m["overall"]["cc_one_half"] for m in merging_stats] overall_rpim = [m["overall"]["r_pim"] for m in merging_stats] overall_ioversigma = [m["overall"]["i_over_sigma_mean"] for m in merging_stats] overall_completeness = [m["overall"]["completeness"] for m in merging_stats] # first make overall plots d = { "cc_one_half_vs_cycle": { "data": [ { "y": overall_ccs, "x": list(range(1, n_datasets + 1)), "type": "scatter", "mode": "lines", } ], "layout": { "title": "CC<sub>½</sub> vs cycle", "xaxis": {"title": "Cycle number"}, "yaxis": {"title": "CC<sub>½</sub>"}, }, }, "r_pim_vs_cycle": { "data": [ { "y": overall_rpim, "x": list(range(1, n_datasets + 1)), "type": "scatter", "mode": "lines", } ], "layout": { "title": "R-pim vs cycle", "xaxis": {"title": "Cycle number"}, "yaxis": {"title": "R-pim"}, }, }, "i_over_sigma_vs_cycle": { "data": [ { "y": overall_ioversigma, "x": list(range(1, n_datasets + 1)), "type": "scatter", "mode": "lines", } ], "layout": { "title": "<I/σ(I)> vs cycle", "xaxis": {"title": "Cycle number"}, "yaxis": {"title": "<I/σ(I)>"}, }, }, "completeness_vs_cycle": { "data": [ { "y": overall_completeness, "x": list(range(1, n_datasets + 1)), "type": "scatter", "mode": "lines", } ], "layout": { "title": "Completeness vs cycle", "xaxis": {"title": "Cycle number"}, "yaxis": {"title": "Completeness"}, }, }, } cc_one_half_bins = merging_stats[0]["ccs"] r_pim_bins = merging_stats[0]["rpim"] r_merge_bins = merging_stats[0]["rmerge"] resolution = [1.0 / x**2 for x in merging_stats[0]["d_min"]] vals, txt = d_star_sq_to_d_ticks(resolution, 5) d.update( { "cc_one_half_filter": { "data": [ { "x": resolution, # d_star_sq "y": cc_one_half_bins, "type": "scatter", "name": legends[0], "mode": "lines", "line": {"color": colors[0]}, } ], "layout": { "title": "CC<sub>½</sub> vs resolution", "xaxis": { "title": "Resolution (Å)", "tickvals": vals, "ticktext": txt, }, "yaxis": {"title": "CC<sub>½</sub>", "range": [0, 1]}, }, } } ) d.update( { "r_pim_filter": { "data": [ { "x": resolution, # d_star_sq "y": r_pim_bins, "type": "scatter", "name": legends[0], "mode": "lines", "line": {"color": colors[0]}, } ], "layout": { "title": "R-pim vs resolution", "xaxis": { "title": "Resolution (Å)", "tickvals": vals, "ticktext": txt, }, "yaxis": { "title": "R-pim", "range": [0, min(1.5, max(r_pim_bins))], }, }, } } ) d.update( { "r_merge_filter": { "data": [ { "x": resolution, # d_star_sq "y": r_merge_bins, "type": "scatter", "name": legends[0], "mode": "lines", "line": {"color": colors[0]}, } ], "layout": { "title": "R-merge vs resolution", "xaxis": { "title": "Resolution (Å)", "tickvals": vals, "ticktext": txt, }, "yaxis": { "title": "R-merge", "range": [0, min(1.5, max(r_merge_bins))], }, }, } } ) for c, stats in enumerate(merging_stats[1:]): cc_one_half_bins = stats["ccs"] r_pim_bins = stats["rpim"] r_merge_bins = stats["rmerge"] resolution = [1.0 / x**2 for x in stats["d_min"]] d["cc_one_half_filter"]["data"].append( { "x": resolution, # d_star_sq "y": cc_one_half_bins, "type": "scatter", "name": legends[c + 1], "line": {"color": colors[c + 1]}, } ) d["r_pim_filter"]["data"].append( { "x": resolution, # d_star_sq "y": r_pim_bins, "type": "scatter", "name": legends[c + 1], "line": {"color": colors[c + 1]}, } ) d["r_merge_filter"]["data"].append( { "x": resolution, # d_star_sq "y": r_merge_bins, "type": "scatter", "name": legends[c + 1], "line": {"color": colors[c + 1]}, } ) return d def make_histogram_plots(cycle_results): """Make the histogram plots.""" delta_cc_half_lists = [res["delta_cc_half_values"] for res in cycle_results] if not delta_cc_half_lists: return {} n = len(delta_cc_half_lists) overall_mean_ccs = [res["mean_cc_half"] for res in cycle_results] d = { "mean_cc_one_half_vs_cycle": { "data": [ { "y": overall_mean_ccs, "x": list(range(1, n + 1)), "type": "scatter", "mode": "lines", } ], "layout": { "title": "Resolution-averaged CC<sub>½</sub> (σ-τ) vs cycle", "xaxis": {"title": "Cycle number"}, "yaxis": {"title": "Resolution-averaged CC<sub>½</sub> (σ-τ)"}, }, } } colors = [(color_list * int(math.ceil(n / len(color_list))))[i] for i in range(n)] if n == 1: legends = ["ΔCC<sub>½</sub> analysis"] else: legends = [ordinal(i) + " ΔCC<sub>½</sub> analysis" for i in range(1, n + 1)] if "image_ranges_removed" in cycle_results[0]: n_rej = [len(res["image_ranges_removed"]) for res in cycle_results] else: n_rej = [len(res["removed_datasets"]) for res in cycle_results] def _color_bar_charts(counts, index): n = 0 bar_colors = [] n_rej_this = n_rej[index] for count in counts: if n >= n_rej_this: bar_colors.append(colors[index]) else: bar_colors.append("k") n += count return bar_colors def _add_new_histogram(d, hist, index): d.update( { f"scale_filter_histograms_{index}": { "data": [ { "x": list(hist.slot_centers()), "y": list(hist.slots()), "type": "bar", "name": legends[index], "marker": {"color": _color_bar_charts(hist.slots(), index)}, } ], "layout": { "title": f"{legends[index]}", "xaxis": {"title": "ΔCC<sub>½</sub>"}, "yaxis": { "title": "Number of datasets/groups", "range": [0, min(max(hist.slots()), 50)], }, "bargap": 0, }, } } ) for c, deltas in enumerate(delta_cc_half_lists): flex_deltas = flex.double(deltas) # convert list to flex array if flex_deltas.all_eq(flex_deltas[0]): continue hist = flex.histogram( flex_deltas * 100, flex.min(flex_deltas) * 100, flex.max(flex_deltas) * 100, n_slots=40, ) _add_new_histogram(d, hist, c) return d def make_per_dataset_plot(delta_cchalf_i): """Make a line plot of ΔCC½ per group.""" return { "per_dataset_plot": { "data": [ { "y": [i * 100 for i in list(delta_cchalf_i.values())], "x": list(delta_cchalf_i.keys()), "type": "scatter", "mode": "lines", } ], "layout": { "title": "ΔCC<sub>½</sub> vs group", "xaxis": {"title": "Group number"}, "yaxis": {"title": "ΔCC<sub>½</sub>"}, }, } } def make_reduction_plots(initial_expids_and_image_ranges, expids_and_image_ranges): """Make a chart showing excluded image ranges.""" x = list(range(len(initial_expids_and_image_ranges))) initial_n_images = [] initial_expids = [] for expid_and_img in initial_expids_and_image_ranges: initial_n_images.append(expid_and_img[1][1] - expid_and_img[1][0] + 1) initial_expids.append(expid_and_img[0]) final_n_images = [0] * len(x) for expid, valid in expids_and_image_ranges: loc = initial_expids.index(expid) for t in valid: final_n_images[loc] = t[1] - t[0] + 1 n_removed = [i - j for (i, j) in zip(initial_n_images, final_n_images)] d = { "reduction_plot": { "data": [ { "x": x, "y": final_n_images, "type": "bar", "name": "final image ranges", }, {"x": x, "y": n_removed, "type": "bar", "name": "removed image ranges"}, ], "layout": { "title": "Image range plot", "xaxis": {"title": "Experiment number"}, "yaxis": {"title": "Image number"}, "bargap": 0, "barmode": "stack", }, } } return d
bsd-3-clause
a971f491670eb3f2691d62390bafa03f
33.961404
88
0.464723
3.877797
false
false
false
false
douban/dpark
dpark/utils/hotcounter.py
1
1593
from __future__ import absolute_import from __future__ import print_function import operator import six from six.moves import range class HotCounter(object): def __init__(self, vs=None, limit=20): if vs is None: vs = [] self.limit = limit self.total = {} self.updates = {} self._max = 0 for v in vs: self.add(v) def add(self, v): c = self.updates.get(v, 0) + 1 self.updates[v] = c if c > self._max: self._max = c if len(self.updates) > self.limit * 5 and self._max > 5: self._merge() def _merge(self): for k, c in six.iteritems(self.updates): if c > 1: self.total[k] = self.total.get(k, 0) + c self._max = 0 self.updates = {} if len(self.total) > self.limit * 5: self.total = dict(self.top(self.limit * 3)) def update(self, o): self._merge() if isinstance(o, HotCounter): o._merge() for k, c in six.iteritems(o.total): self.total[k] = self.total.get(k, 0) + c def top(self, limit): return sorted(list(self.total.items()), key=operator.itemgetter(1), reverse=True)[:limit] def test(): import random import math t = HotCounter() for j in range(10): c = HotCounter() for i in range(10000): v = int(math.sqrt(random.randint(0, 1000000))) c.add(v) t.update(c) for k, v in t.top(20): print(k, v) if __name__ == '__main__': test()
bsd-3-clause
508b659df54d9596249061097ade8617
23.507692
97
0.511613
3.425806
false
false
false
false
dials/dials
src/dials/algorithms/refinement/parameterisation/crystal_parameters.py
1
7778
from __future__ import annotations import logging logger = logging.getLogger(__name__) from rstbx.symmetry.constraints.parameter_reduction import symmetrize_reduce_enlarge from scitbx import matrix from dials.algorithms.refinement.parameterisation.model_parameters import ( ModelParameterisation, Parameter, ) from dials.algorithms.refinement.refinement_helpers import CrystalOrientationCompose class CrystalOrientationMixin: """Mix-in class defining some functionality unique to crystal orientation parameterisations that can be shared by static and scan-varying versions""" @staticmethod def _build_p_list(parameter_type=Parameter): """Build the list of parameters, using the parameter_type callback to select between versions of the Parameter class""" # set up the parameters phi1 = parameter_type(0.0, matrix.col((1, 0, 0)), "angle (mrad)", "Phi1") phi2 = parameter_type(0.0, matrix.col((0, 1, 0)), "angle (mrad)", "Phi2") phi3 = parameter_type(0.0, matrix.col((0, 0, 1)), "angle (mrad)", "Phi3") # build the parameter list in a specific, maintained order p_list = [phi1, phi2, phi3] return p_list class CrystalOrientationParameterisation( ModelParameterisation, CrystalOrientationMixin ): """A parameterisation of the orientation of a Crystal model. The Crystal orientation matrix U is parameterised by three Tait-Bryan angles expressed in mrad""" def __init__(self, crystal, experiment_ids=None): """Initialise the CrystalOrientationParameterisation object Args: crystal: A dxtbx Crystal object to be parameterised. experiment_ids (list): The experiment IDs affected by this parameterisation. Defaults to None, which is replaced by [0]. """ # The state of a crystal orientation parameterisation is an orientation # matrix '[U]'. The initial state is a snapshot of the crystal # orientation at the time of initialisation '[U0]'. Future states are # composed by rotations around axes of the phi-axis frame by Tait-Bryan # angles. # # [U] = [Phi3][Phi2][Phi1][U0] # set up the initial state if experiment_ids is None: experiment_ids = [0] istate = matrix.sqr(crystal.get_U()) # build the parameter list p_list = self._build_p_list() # set up the base class ModelParameterisation.__init__( self, crystal, istate, p_list, experiment_ids=experiment_ids ) # call compose to calculate all the derivatives self.compose() return def compose(self): # Extract orientation from the initial state U0 = self._initial_state # extract parameters from the internal list phi1, phi2, phi3 = self._param # calculate using the helper class coc = CrystalOrientationCompose( U0, phi1.value, phi1.axis, phi2.value, phi2.axis, phi3.value, phi3.axis ) # compose new state self._model.set_U(coc.U()) # store derivatives self._dstate_dp = [coc.dU_dphi1(), coc.dU_dphi2(), coc.dU_dphi3()] return def get_state(self): # only a single crystal is parameterised here, so no multi_state_elt # argument is allowed return matrix.sqr(self._model.get_U()) class CrystalUnitCellMixin: """Mix-in class defining some functionality unique to crystal unit cell parameterisations that can be shared by static and scan-varying versions""" def _build_p_list(self, crystal, parameter_type=Parameter): """Build the list of parameters, using the parameter_type callback to select between versions of the Parameter class""" # Set up symmetrizing object S = symmetrize_reduce_enlarge(crystal.get_space_group()) S.set_orientation(orientation=crystal.get_B()) X = S.forward_independent_parameters() # Set up the independent parameters, with a change of scale p_list = [ parameter_type(e * 1.0e5, name="g_param_%d" % i) for i, e in enumerate(X) ] return p_list def _compose_core(self, raw_vals): # obtain metrical matrix parameters on natural scale vals = [v * 1.0e-5 for v in raw_vals] # set parameter values in the symmetrizing object and obtain new B S = symmetrize_reduce_enlarge(self._model.get_space_group()) S.set_orientation(orientation=self._model.get_B()) S.forward_independent_parameters() # Set Bconverter as side-effect try: newB = matrix.sqr(S.backward_orientation(vals).reciprocal_matrix()) except RuntimeError as e: # write original error to debug log logger.debug("Unable to compose the crystal model") logger.debug("Original error message: %s", str(e)) logger.debug("Failing now.") raise RuntimeError( "Unable to compose the crystal model. Please check that the " "experiments match the indexing of the reflections." ) # returns the independent parameters given the set_orientation() B matrix # used here for side effects S.forward_independent_parameters() # get the derivatives of state wrt metrical matrix parameters on the # adjusted sale dB_dval = [matrix.sqr(g) * 1.0e-5 for g in S.forward_gradients()] return newB, dB_dval class CrystalUnitCellParameterisation(ModelParameterisation, CrystalUnitCellMixin): """A parameterisation for the unit cell of a Crystal model. The Crystal reciprocal space orthogonalisation matrix B is parameterised using up to 6 metrical matrix elements, rescaled by a multiplicative factor. """ def __init__(self, crystal, experiment_ids=None): """Initialise the CrystalUnitCellParameterisation object Args: crystal: A dxtbx Crystal object to be parameterised. experiment_ids (list): The experiment IDs affected by this parameterisation. Defaults to None, which is replaced by [0]. """ # The state of the unit cell parameterisation is the reciprocal space # orthogonalisation matrix 'B'. The initial state is irrelevant for # this model, as composition of a new B matrix and derivatives can be # done with just the values of 6 unit cell parameters, without # defining axial directions (which are selected by choice of the PDB # convention). For this reason also, the axes of the # parameters are irrelevant and are set here to None. ### Set up the initial state if experiment_ids is None: experiment_ids = [0] istate = None # build the parameter list p_list = self._build_p_list(crystal) # set up the base class ModelParameterisation.__init__( self, crystal, istate, p_list, experiment_ids=experiment_ids ) # call compose to calculate all the derivatives self.compose() return def compose(self): # calculate new B and derivatives newB, self._dstate_dp = self._compose_core([p.value for p in self._param]) # Now pass new B to the crystal model self._model.set_B(newB) return def get_state(self): # only a single crystal is parameterised here, so no multi_state_elt # argument is allowed return matrix.sqr(self._model.get_B()) def set_state_uncertainties(self, var_cov, multi_state_elt=None): self._model.set_B_covariance(var_cov) return
bsd-3-clause
580b013adde02d67bc6f6329af531187
33.878924
85
0.647339
3.968367
false
false
false
false
dials/dials
tests/algorithms/image/connected_components/test_connected_components.py
1
6312
from __future__ import annotations class Test2d: def setup_class(self): from dials.algorithms.image.connected_components import LabelImageStack2d self.size = (500, 500) self.label_images = LabelImageStack2d(self.size) from scitbx.array_family import flex self.data_list = [] self.mask_list = [] for i in range(10): data = flex.random_int_gaussian_distribution( self.size[0] * self.size[1], 100, 10 ) data.reshape(flex.grid(self.size)) mask = flex.random_bool(self.size[0] * self.size[1], 0.1) mask.reshape(flex.grid(self.size)) self.data_list.append(data) self.mask_list.append(mask) for i in range(10): self.label_images.add_image(self.data_list[i], self.mask_list[i]) self.labels = self.label_images.labels() self.coords = self.label_images.coords() self.values = list(self.label_images.values()) assert len(self.labels) > 0 assert len(self.labels) == len(self.coords) assert len(self.labels) == len(self.values) def test_coords_are_valid(self): # Ensure that the values are all ok and in the right order vi = 0 for k in range(10): ind = 0 for j in range(self.size[0]): for i in range(self.size[1]): m = self.mask_list[k][ind] if m: c1 = (k, j, i) c2 = self.coords[vi] vi += 1 assert c1 == c2 ind += 1 def test_values_are_valid(self): # Ensure that the values are all ok and in the right order vi = 0 for k in range(10): for d, m in zip(self.data_list[k], self.mask_list[k]): if m: v1 = d v2 = self.values[vi] vi += 1 assert v1 == v2 def test_labels_are_valid(self): from scitbx.array_family import flex # Create a map of labels label_map = flex.int(flex.grid(10, self.size[0], self.size[1])) for c, l in zip(self.coords, self.labels): assert c[0] >= 0 and c[0] < 10 assert c[1] >= 0 and c[1] < self.size[0] assert c[2] >= 0 and c[2] < self.size[1] label_map[c] = l # Ensure all labels are correct vi = 0 for k in range(10): for j in range(self.size[0]): for i in range(self.size[1]): if self.mask_list[k][j, i]: l1 = self.labels[vi] if k > 0 and self.mask_list[k - 1][j, i]: l2 = label_map[k - 1, j, i] assert l2 != l1 if j > 0 and self.mask_list[k][j - 1, i]: l2 = label_map[k, j - 1, i] assert l2 == l1 if i > 0 and self.mask_list[k][j, i - 1]: l2 = label_map[k, j, i - 1] assert l2 == l1 vi += 1 class Test3d: def setup_class(self): from dials.algorithms.image.connected_components import LabelImageStack3d self.size = (500, 500) self.label_images = LabelImageStack3d(self.size) from scitbx.array_family import flex self.data_list = [] self.mask_list = [] for i in range(10): data = flex.random_int_gaussian_distribution( self.size[0] * self.size[1], 100, 10 ) data.reshape(flex.grid(self.size)) mask = flex.random_bool(self.size[0] * self.size[1], 0.1) mask.reshape(flex.grid(self.size)) self.data_list.append(data) self.mask_list.append(mask) for i in range(10): self.label_images.add_image(self.data_list[i], self.mask_list[i]) self.labels = self.label_images.labels() self.coords = self.label_images.coords() self.values = list(self.label_images.values()) assert len(self.labels) > 0 assert len(self.labels) == len(self.coords) assert len(self.labels) == len(self.values) def test_coords_are_valid(self): # Ensure that the values are all ok and in the right order vi = 0 for k in range(10): ind = 0 for j in range(self.size[0]): for i in range(self.size[1]): m = self.mask_list[k][ind] if m: c1 = (k, j, i) c2 = self.coords[vi] vi += 1 assert c1 == c2 ind += 1 def test_values_are_valid(self): # Ensure that the values are all ok and in the right order vi = 0 for k in range(10): for d, m in zip(self.data_list[k], self.mask_list[k]): if m: v1 = d v2 = self.values[vi] vi += 1 assert v1 == v2 def test_labels_are_valid(self): from scitbx.array_family import flex # Create a map of labels label_map = flex.int(flex.grid(10, self.size[0], self.size[1])) for c, l in zip(self.coords, self.labels): label_map[c] = l # Ensure all labels are correct vi = 0 for k in range(10): for j in range(self.size[0]): for i in range(self.size[1]): if self.mask_list[k][j, i]: l1 = self.labels[vi] if k > 0 and self.mask_list[k - 1][j, i]: l2 = label_map[k - 1, j, i] assert l2 == l1 if j > 0 and self.mask_list[k][j - 1, i]: l2 = label_map[k, j - 1, i] assert l2 == l1 if i > 0 and self.mask_list[k][j, i - 1]: l2 = label_map[k, j, i - 1] assert l2 == l1 vi += 1
bsd-3-clause
1179d2d7a1bf33ffd0120ac8e9d71f1a
34.863636
81
0.460551
3.752675
false
false
false
false
dials/dials
src/dials/command_line/slice_sequence.py
1
8936
from __future__ import annotations from os.path import basename, splitext from dxtbx.model.experiment_list import ExperimentList import dials.util from dials.algorithms.refinement.refinement_helpers import calculate_frame_numbers from dials.array_family import flex from dials.util import Sorry from dials.util.multi_dataset_handling import generate_experiment_identifiers from dials.util.slice import slice_experiments, slice_reflections help_message = """ Slice a sequence to produce a smaller sequence within the bounds of the original. If experiments or experiments are provided, modify the scan objects within these. If reflections are provided, remove reflections outside the provided image ranges. Each image_range parameter refers to a single experiment ID, counting up from zero. Any reflections with experiment ID not matched by a image_range parameter are removed. Examples:: dials.slice_sequence models.expt observations.refl "image_range=1 20" dials.slice_sequence models.expt "image_range=1 20" # two experiments and reflections with IDs '0' and '1' dials.slice_sequence models.expt observations.refl \ "image_range=1 20" "image_range=5 30" """ from libtbx.phil import parse phil_scope = parse( """ output { reflections_filename = None .type = str .help = "The filename for output reflections sliced to remove those" "outside the reduced image range. By default generated" "automatically from the input name" experiments_filename = None .type = str .help = "The filename for the output experiments with sliced scans. By default generated automatically from the input name" } image_range = None .help = "Range in images to slice a sequence. The number of arguments" "must be a factor of two. Each pair of arguments gives a range" "that follows C conventions (e.g. j0 <= j < j1) when slicing the" "reflections by observed centroid." .type = ints(size=2) .multiple = True block_size = None .type = float .help = "Overrides image_range if present. This option splits each sequence" "into the nearest integer number of equal size blocks close to" "block_size degrees in width" """ ) def calculate_block_ranges(scan, block_size): """ :param scans :type a scan object :param block_size: :type block_size: target block size in degrees""" image_ranges = [] nimages = scan.get_num_images() osc_range = scan.get_oscillation_range(deg=True) osc_width = abs(osc_range[1] - osc_range[0]) nblocks = max(int(round(osc_width / block_size)), 1) nblocks = min(nblocks, nimages) # equal sized blocks except the last one that may contain extra images # to make up the remainder nimages_per_block = [nimages // nblocks] * (nblocks - 1) + [ nimages // nblocks + nimages % nblocks ] start = scan.get_image_range()[0] for nim in nimages_per_block: image_ranges.append((start, start + nim - 1)) start += nim return image_ranges class Script: """A class for running the script.""" def __init__(self): """Initialise the script.""" from dials.util.options import ArgumentParser usage = ( "usage: dials.slice_sequence [options] [param.phil] " "models.expt observations.refl" ) # Create the parser self.parser = ArgumentParser( usage=usage, phil=phil_scope, read_reflections=True, read_experiments=True, check_format=False, epilog=help_message, ) def run(self, args=None): """Execute the script.""" from dials.util.options import reflections_and_experiments_from_files # Parse the command line params, options = self.parser.parse_args(args, show_diff_phil=True) reflections, experiments = reflections_and_experiments_from_files( params.input.reflections, params.input.experiments ) # Try to load the models and data slice_exps = len(experiments) > 0 slice_refs = len(reflections) > 0 # Catch case of nothing to do if not slice_exps and not slice_refs: print("No suitable input provided") self.parser.print_help() return if reflections: if len(reflections) > 1: raise Sorry("Only one reflections list can be imported at present") reflections = reflections[0] # calculate frame numbers if needed if experiments: reflections = calculate_frame_numbers(reflections, experiments) # if we still don't have the right column give up if "xyzobs.px.value" not in reflections: raise Sorry( "These reflections do not have frame numbers set, and " "there are no experiments provided to calculate these." ) # set trivial case where no scan range is provided at all if not params.image_range: params.image_range = [None] # check if slicing into blocks if params.block_size is not None: if not slice_exps: raise Sorry( "For slicing into blocks, an experiment file must be provided" ) if len(experiments) > 1: raise Sorry("For slicing into blocks please provide a single scan only") scan = experiments[0].scan # Having extracted the scan, calculate the blocks params.image_range = calculate_block_ranges(scan, params.block_size) # Do the slicing then recombine sliced = [ slice_experiments(experiments, [sr])[0] for sr in params.image_range ] generate_experiment_identifiers(sliced) sliced_experiments = ExperimentList(sliced) # slice reflections if present if slice_refs: sliced = [ slice_reflections(reflections, [sr]) for sr in params.image_range ] sliced_reflections = flex.reflection_table() identifiers = sliced_experiments.identifiers() # resetting experiment identifiers for i, rt in enumerate(sliced): for k in rt.experiment_identifiers().keys(): del rt.experiment_identifiers()[k] rt["id"] = flex.int(rt.size(), i) # set id rt.experiment_identifiers()[i] = identifiers[i] sliced_reflections.extend(rt) else: # slice each dataset into the requested subset if slice_exps: sliced_experiments = slice_experiments(experiments, params.image_range) if slice_refs: sliced_reflections = slice_reflections(reflections, params.image_range) # Save sliced experiments if slice_exps: output_experiments_filename = params.output.experiments_filename if output_experiments_filename is None: # take first filename as template bname = basename(params.input.experiments[0].filename) bname = splitext(bname)[0] if not bname: bname = "experiments" if len(params.image_range) == 1 and params.image_range[0] is not None: ext = "_{}_{}.expt".format(*params.image_range[0]) else: ext = "_sliced.expt" output_experiments_filename = bname + ext print(f"Saving sliced experiments to {output_experiments_filename}") sliced_experiments.as_file(output_experiments_filename) # Save sliced reflections if slice_refs: output_reflections_filename = params.output.reflections_filename if output_reflections_filename is None: # take first filename as template bname = basename(params.input.reflections[0].filename) bname = splitext(bname)[0] if not bname: bname = "reflections" if len(params.image_range) == 1 and params.image_range[0] is not None: ext = "_{}_{}.refl".format(*params.image_range[0]) else: ext = "_sliced.refl" output_reflections_filename = bname + ext print(f"Saving sliced reflections to {output_reflections_filename}") sliced_reflections.as_file(output_reflections_filename) return @dials.util.show_mail_handle_errors() def run(args=None): script = Script() script.run(args) if __name__ == "__main__": run()
bsd-3-clause
517252d1a51a026bcfef991487f9d1f6
34.744
88
0.606647
4.211122
false
false
false
false
dials/dials
src/dials/algorithms/profile_model/ellipsoid/indexer.py
1
7256
from __future__ import annotations import logging from math import floor, sqrt import numpy as np from numpy.linalg import inv, norm from cctbx.array_family import flex from dxtbx import flumpy from scitbx import matrix from dials.algorithms.profile_model.ellipsoid import chisq_quantile from dials.algorithms.statistics.fast_mcd import FastMCD, maha_dist_sq logger = logging.getLogger("dials") def _index(reflection_table, experiment, fail_on_bad_index=False): """Index the strong spots""" # Get some stuff from experiment A = np.array(experiment.crystal.get_A(), dtype=np.float64).reshape(3, 3) s0 = np.array([experiment.beam.get_s0()], dtype=np.float64).reshape(3, 1) s0_length = norm(s0) detector = experiment.detector # Create array if necessary if "miller_index" not in reflection_table: reflection_table["miller_index"] = flex.miller_index(len(reflection_table)) # Index all the reflections miller_index = reflection_table["miller_index"] selection = flex.size_t() num_reindexed = 0 for i, xyz in enumerate(reflection_table["xyzobs.px.value"]): # Get the observed pixel coordinate x, y, _ = xyz # Get the lab coord s1 = np.array( detector[0].get_pixel_lab_coord((x, y)), dtype=np.float64 ).reshape(3, 1) s1_norm = norm(s1) s1 *= s0_length / s1_norm # Get the reciprocal lattice vector r = s1 - s0 # Compute the fractional miller index hf = np.matmul(inv(A), r) # Compute the integer miller index h = np.array([int(floor(j + 0.5)) for j in hf[:, 0]], dtype=int).reshape(3, 1) # Print warning if reindexing if tuple(h) != miller_index[i]: logger.warn( "Reindexing (% 3d, % 3d, % 3d) -> (% 3d, % 3d, % 3d)" % (miller_index[i] + tuple(h)) ) num_reindexed += 1 miller_index[i] = matrix.col(flumpy.from_numpy(h)) if fail_on_bad_index: raise RuntimeError("Bad index") # If its not indexed as 0, 0, 0 then append if h.any() and norm(h - hf) < 0.3: selection.append(i) # Print some info logger.info( "Reindexed %d/%d input reflections" % (num_reindexed, len(reflection_table)) ) logger.info( "Selected %d/%d input reflections" % (len(selection), len(reflection_table)) ) # Select all the indexed reflections reflection_table.set_flags(selection, reflection_table.flags.indexed) reflection_table = reflection_table.select(selection) return reflection_table def _predict(reflection_table, experiment): """ Predict the position of the spots """ # Get some stuff from experiment A = np.array(experiment.crystal.get_A(), dtype=np.float64).reshape((3, 3)) s0 = np.array([experiment.beam.get_s0()], dtype=np.float64).reshape(3, 1) s0_length = norm(s0) # Compute the vector to the reciprocal lattice point # since this is not on the ewald sphere, lets call it s2 s1 = flex.vec3_double(reflection_table.size()) s2 = flex.vec3_double(reflection_table.size()) for i, h in enumerate(reflection_table["miller_index"]): r = np.matmul(A, np.array([h], dtype=np.float64).reshape(3, 1)) s2_i = r + s0 s2[i] = matrix.col(flumpy.from_numpy(s2_i)) s1[i] = matrix.col(flumpy.from_numpy(s2_i * s0_length / norm(s2_i))) reflection_table["s1"] = s1 reflection_table["s2"] = s2 reflection_table["entering"] = flex.bool(reflection_table.size(), False) # Compute the ray intersections xyzpx = flex.vec3_double() xyzmm = flex.vec3_double() for ss in s1: mm = experiment.detector[0].get_ray_intersection(ss) px = experiment.detector[0].millimeter_to_pixel(mm) xyzpx.append(px + (0,)) xyzmm.append(mm + (0,)) reflection_table["xyzcal.mm"] = xyzmm reflection_table["xyzcal.px"] = xyzpx logger.info("Do prediction for %d reflections" % len(reflection_table)) return reflection_table def _filter_reflections_based_on_centroid_distance( reflection_table, experiment, outlier_probability=0.975, max_separation=2, ): """ Filter reflections too far from predicted position """ # Compute the x and y residuals Xobs, Yobs, _ = reflection_table["xyzobs.px.value"].parts() Xcal, Ycal, _ = reflection_table["xyzcal.px"].parts() Xres = Xobs - Xcal Yres = Yobs - Ycal # Compute the epsilon residual s0_length = 1.0 / experiment.beam.get_wavelength() s1x, s1y, s1z = reflection_table["s2"].parts() s1_length = flex.sqrt(s1x**2 + s1y**2 + s1z**2) Eres = s1_length - s0_length # Initialise the fast_mcd outlier algorithm # fast_mcd = FastMCD((Xres, Yres, Eres)) try: fast_mcd = FastMCD((Xres, Yres)) except AssertionError as e: raise RuntimeError(e) # get location and MCD scatter estimate T, S = fast_mcd.get_corrected_T_and_S() # get squared Mahalanobis distances # d2s = maha_dist_sq((Xres, Yres, Eres), T, S) d2s = maha_dist_sq((Xres, Yres), T, S) # Compute the cutoff mahasq_cutoff = chisq_quantile(2, outlier_probability) # compare to the threshold and select reflections selection1 = d2s < mahasq_cutoff selection2 = flex.sqrt(Xres**2 + Yres**2) < max_separation selection = selection1 & selection2 reflection_table = reflection_table.select(selection) n_refl = reflection_table.size() # Print some stuff logger.info("-" * 80) logger.info("Centroid outlier rejection") logger.info(f" Using MCD algorithm with probability = {outlier_probability}") logger.info(" Max X residual: %f" % flex.max(flex.abs(Xres))) logger.info(" Max Y residual: %f" % flex.max(flex.abs(Yres))) logger.info(" Max E residual: %f" % flex.max(flex.abs(Eres))) logger.info(" Mean X RMSD: %f" % (sqrt(flex.sum(Xres**2) / len(Xres)))) logger.info(" Mean Y RMSD: %f" % (sqrt(flex.sum(Yres**2) / len(Yres)))) logger.info(" Mean E RMSD: %f" % (sqrt(flex.sum(Eres**2) / len(Eres)))) logger.info(" MCD location estimate: %.4f, %.4f" % tuple(T)) logger.info( """ MCD scatter estimate: %.7f, %.7f, %.7f, %.7f""" % tuple(S) ) logger.info(" Number of outliers: %d" % selection1.count(False)) logger.info( " Number of reflections with residual > %0.2f pixels: %d" % (max_separation, selection2.count(False)) ) logger.info(f"Number of reflections selection for refinement: {n_refl}") logger.info("-" * 80) return reflection_table def reindex( reflection_table, experiment, outlier_probability=0.975, max_separation=2, fail_on_bad_index=False, ): """Reindex strong spots and perform filtering""" reflection_table = _index(reflection_table, experiment, fail_on_bad_index) reflection_table = _predict(reflection_table, experiment) reflection_table = _filter_reflections_based_on_centroid_distance( reflection_table, experiment, outlier_probability=outlier_probability, max_separation=max_separation, ) return reflection_table
bsd-3-clause
e0047b382072ee9e0a86a285f21fb547
33.226415
86
0.633958
3.26259
false
false
false
false
dials/dials
src/dials/util/image_viewer/slip_viewer/uc_frame.py
1
14583
from __future__ import annotations import math import wx from wx.lib.agw.floatspin import EVT_FLOATSPIN, FloatSpin import cctbx.miller from cctbx.crystal import symmetry from scitbx.matrix import col class UCSettingsFrame(wx.MiniFrame): def __init__(self, *args, **kwds): super().__init__(*args, **kwds) szr = wx.BoxSizer(wx.VERTICAL) self.phil_params = args[0].params panel = UCSettingsPanel(self) self.SetSizer(szr) szr.Add(panel, 1, wx.EXPAND) szr.Fit(panel) self.panel = panel self.sizer = szr self.Fit() self.Bind(wx.EVT_CLOSE, lambda evt: self.Destroy(), self) class UCSettingsPanel(wx.Panel): def __init__(self, *args, **kwds): super().__init__(*args, **kwds) self.phil_params = args[0].phil_params # Needed to draw and delete the rings. XXX Applies to # calibration_frame as well? self._pyslip = self.GetParent().GetParent().pyslip sizer = wx.BoxSizer(wx.VERTICAL) self.SetSizer(sizer) # Number of decimal digits for distances. self.digits = 2 # Wavelength control. beam = self._pyslip.tiles.raw_image.get_beam() self._wavelength = beam.get_wavelength() # Unit cell controls. if self.phil_params.calibrate_unit_cell.unit_cell is not None: self._cell = list( self.phil_params.calibrate_unit_cell.unit_cell.parameters() ) else: self._cell = [4.18, 4.72, 58.38, 89.44, 89.63, 75.85] if self.phil_params.calibrate_unit_cell.space_group is not None: self._spacegroup = self.phil_params.calibrate_unit_cell.space_group else: self._spacegroup = "P1" self._show_hkl = self.phil_params.calibrate_unit_cell.show_hkl self._cell_control_names = [ "uc_a_ctrl", "uc_b_ctrl", "uc_c_ctrl", "uc_alpha_ctrl", "uc_beta_ctrl", "uc_gamma_ctrl", ] box = wx.BoxSizer(wx.HORIZONTAL) self.uc_a = FloatSpin( self, digits=self.digits, name=self._cell_control_names[0], value=self._cell[0], ) box.Add( self.uc_a, 0, wx.RIGHT | wx.TOP | wx.BOTTOM | wx.ALIGN_CENTER_VERTICAL, 5 ) box.Add(wx.StaticText(self, label="a"), 0, wx.ALL | wx.ALIGN_CENTER_VERTICAL, 5) self.Bind(EVT_FLOATSPIN, self.OnSpinCell, self.uc_a) self.uc_alpha = FloatSpin( self, digits=self.digits, name=self._cell_control_names[3], value=self._cell[3], ) box.Add( self.uc_alpha, 0, wx.RIGHT | wx.TOP | wx.BOTTOM | wx.ALIGN_CENTER_VERTICAL, 5, ) box.Add( wx.StaticText(self, label="alpha"), 0, wx.ALL | wx.ALIGN_CENTER_VERTICAL, 5 ) self.Bind(EVT_FLOATSPIN, self.OnSpinCell, self.uc_alpha) sizer.Add(box) box = wx.BoxSizer(wx.HORIZONTAL) self.uc_b = FloatSpin( self, digits=self.digits, name=self._cell_control_names[1], value=self._cell[1], ) box.Add( self.uc_b, 0, wx.RIGHT | wx.TOP | wx.BOTTOM | wx.ALIGN_CENTER_VERTICAL, 5 ) box.Add(wx.StaticText(self, label="b"), 0, wx.ALL | wx.ALIGN_CENTER_VERTICAL, 5) self.Bind(EVT_FLOATSPIN, self.OnSpinCell, self.uc_b) self.uc_beta = FloatSpin( self, digits=self.digits, name=self._cell_control_names[4], value=self._cell[4], ) box.Add( self.uc_beta, 0, wx.RIGHT | wx.TOP | wx.BOTTOM | wx.ALIGN_CENTER_VERTICAL, 5 ) box.Add( wx.StaticText(self, label="beta"), 0, wx.ALL | wx.ALIGN_CENTER_VERTICAL, 5 ) self.Bind(EVT_FLOATSPIN, self.OnSpinCell, self.uc_beta) sizer.Add(box) box = wx.BoxSizer(wx.HORIZONTAL) self.uc_c = FloatSpin( self, digits=self.digits, name=self._cell_control_names[2], value=self._cell[2], ) box.Add( self.uc_c, 0, wx.RIGHT | wx.TOP | wx.BOTTOM | wx.ALIGN_CENTER_VERTICAL, 5 ) box.Add(wx.StaticText(self, label="c"), 0, wx.ALL | wx.ALIGN_CENTER_VERTICAL, 5) self.Bind(EVT_FLOATSPIN, self.OnSpinCell, self.uc_c) self.uc_gamma = FloatSpin( self, digits=self.digits, name=self._cell_control_names[5], value=self._cell[5], ) box.Add( self.uc_gamma, 0, wx.RIGHT | wx.TOP | wx.BOTTOM | wx.ALIGN_CENTER_VERTICAL, 5, ) box.Add( wx.StaticText(self, label="gamma"), 0, wx.ALL | wx.ALIGN_CENTER_VERTICAL, 5 ) self.Bind(EVT_FLOATSPIN, self.OnSpinCell, self.uc_gamma) sizer.Add(box) # Space group control box = wx.BoxSizer(wx.HORIZONTAL) self.space_group_ctrl = wx.TextCtrl( self, name="space group", value=self._spacegroup ) box.Add( self.space_group_ctrl, 0, wx.RIGHT | wx.TOP | wx.BOTTOM | wx.ALIGN_CENTER_VERTICAL, 5, ) box.Add( wx.StaticText(self, label="Space group"), 0, wx.ALL | wx.ALIGN_CENTER_VERTICAL, 5, ) self.Bind(wx.EVT_TEXT, self.OnSpaceGroup, self.space_group_ctrl) sizer.Add(box) # Distance control img = self.GetParent().GetParent()._img box = wx.BoxSizer(wx.HORIZONTAL) self.distance_ctrl = FloatSpin( self, digits=self.digits, name="Detector Distance", value=img.get_detector_distance(), ) self.distance_ctrl.SetIncrement(0.5) box.Add( self.distance_ctrl, 0, wx.RIGHT | wx.TOP | wx.BOTTOM | wx.ALIGN_CENTER_VERTICAL, 5, ) txtd = wx.StaticText(self, label="Detector Distance") box.Add(txtd, 0, wx.ALL | wx.ALIGN_CENTER_VERTICAL, 5) self.Bind(EVT_FLOATSPIN, self.OnSpin, self.distance_ctrl) sizer.Add(box) # Wavelength control img = self.GetParent().GetParent()._img box = wx.BoxSizer(wx.HORIZONTAL) self.wavelength_ctrl = FloatSpin( self, digits=4, name="Wavelength", value=img.get_wavelength() ) self.wavelength_ctrl.SetIncrement(0.05) box.Add( self.wavelength_ctrl, 0, wx.RIGHT | wx.TOP | wx.BOTTOM | wx.ALIGN_CENTER_VERTICAL, 5, ) txtw = wx.StaticText(self, label="Wavelength") box.Add(txtw, 0, wx.ALL | wx.ALIGN_CENTER_VERTICAL, 5) self.Bind(EVT_FLOATSPIN, self.OnSpin, self.wavelength_ctrl) sizer.Add(box) # d_min control if self.phil_params.calibrate_unit_cell.d_min is not None: self.d_min = self.phil_params.calibrate_unit_cell.d_min else: self.d_min = 3.5 box = wx.BoxSizer(wx.HORIZONTAL) self.d_min_ctrl = FloatSpin( self, digits=self.digits, name="d_min", value=self.d_min ) box.Add( self.d_min_ctrl, 0, wx.RIGHT | wx.TOP | wx.BOTTOM | wx.ALIGN_CENTER_VERTICAL, 5, ) txtd = wx.StaticText(self, label="Highest resolution for ring display") box.Add(txtd, 0, wx.ALL | wx.ALIGN_CENTER_VERTICAL, 5) self.Bind(EVT_FLOATSPIN, self.OnSpin, self.d_min_ctrl) sizer.Add(box) # Centering controls. self._center = [0, 0] box = wx.BoxSizer(wx.HORIZONTAL) self.spinner_fast = FloatSpin( self, digits=self.digits, name="fast_ctrl", value=self._center[0] ) box.Add( self.spinner_fast, 0, wx.RIGHT | wx.TOP | wx.BOTTOM | wx.ALIGN_CENTER_VERTICAL, 5, ) box.Add( wx.StaticText(self, label="Center fast"), 0, wx.ALL | wx.ALIGN_CENTER_VERTICAL, 5, ) self.Bind(EVT_FLOATSPIN, self.OnSpinCenter, self.spinner_fast) self.spinner_slow = FloatSpin( self, digits=self.digits, name="slow_ctrl", value=self._center[1] ) box.Add( self.spinner_slow, 0, wx.RIGHT | wx.TOP | wx.BOTTOM | wx.ALIGN_CENTER_VERTICAL, 5, ) box.Add( wx.StaticText(self, label="Center slow"), 0, wx.ALL | wx.ALIGN_CENTER_VERTICAL, 5, ) sizer.Add(box) box = wx.BoxSizer(wx.HORIZONTAL) self.clear_button = wx.Button(self, -1, "Clear") box.Add(self.clear_button, 0, wx.ALL | wx.ALIGN_CENTER_VERTICAL, 5) self.Bind(wx.EVT_BUTTON, self.OnClear, self.clear_button) sizer.Add(box) origin_box = wx.BoxSizer(wx.HORIZONTAL) self.origin = wx.StaticText(self, label="") origin_box.Add(self.origin, 0, wx.ALL | wx.ALIGN_CENTER_VERTICAL, 5) self.Bind(EVT_FLOATSPIN, self.OnSpinCenter, self.spinner_slow) sizer.Add(origin_box) self.DrawRings() def __del__(self): if hasattr(self, "_ring_layer") and self._ring_layer is not None: self._pyslip.DeleteLayer(self._ring_layer) self._ring_layer = None def OnSpinCenter(self, event): obj = event.EventObject name = obj.GetName() if name == "fast_ctrl": self._center[0] = obj.GetValue() elif name == "slow_ctrl": self._center[1] = obj.GetValue() self.DrawRings() def OnSpinCell(self, event): obj = event.EventObject name = obj.GetName() self._cell[self._cell_control_names.index(name)] = obj.GetValue() self.DrawRings() def OnSpin(self, event): self.DrawRings() def OnSpaceGroup(self, event): obj = event.EventObject self._spacegroup = obj.GetValue() self.DrawRings() def OnClear(self, event): self.__del__() def _draw_rings_layer(self, dc, data, map_rel): """Draw a points layer. dc the device context to draw on data an iterable of point tuples: (x, y, place, radius, colour, x_off, y_off, pdata) map_rel points relative to map if True, MUST BE TRUE for lightweight Assumes all points are the same colour, saving 100's of ms. """ assert map_rel is True if len(data) == 0: return (lon, lat, place, radius, colour, x_off, y_off, pdata) = data[0] scale = 2**self._pyslip.tiles.zoom_level # Draw points on map/view, using transparency if implemented. try: dc = wx.GCDC(dc) except NotImplementedError: pass dc.SetPen(wx.Pen(colour)) dc.SetBrush(wx.Brush(colour, wx.TRANSPARENT)) for (lon, lat, place, radius, colour, x_off, y_off, pdata) in data: (x, y) = self._pyslip.ConvertGeo2View((lon, lat)) dc.DrawCircle(x, y, radius * scale) def DrawRings(self): frame = self.GetParent().GetParent() try: uc = symmetry( unit_cell=self._cell, space_group_symbol=str(self._spacegroup) ) hkl_list = cctbx.miller.build_set( uc, False, d_min=self.d_min_ctrl.GetValue() ) except Exception as e: frame.update_statusbar(str(e)) return if self._show_hkl: hkl_list = hkl_list.common_set( cctbx.miller.set( crystal_symmetry=uc, indices=self._show_hkl, anomalous_flag=False ) ) frame.update_statusbar( "%d %d %d %d %d %d, " % tuple(self._cell) + f"number of indices: {len(hkl_list.indices())}" ) spacings = sorted(hkl_list.d_spacings(), key=lambda s: s[1], reverse=True) print(f"Printing spacings, len: {len(spacings)}") for d in spacings: print(d) detector = self._pyslip.tiles.raw_image.get_detector() wavelength = float(self.wavelength_ctrl.GetValue()) distance = float(self.distance_ctrl.GetValue()) pixel_size = detector[0].get_pixel_size()[ 0 ] # FIXME assumes square pixels, and that all panels use same pixel size twotheta = hkl_list.two_theta(wavelength=wavelength) L_mm = [] L_pixels = [] for tt in twotheta: L_mm.append(distance * math.tan(tt[1])) for lmm in L_mm: L_pixels.append(lmm / pixel_size) xrayframe = self.GetParent().GetParent() panel_id, beam_pixel_fast, beam_pixel_slow = xrayframe.get_beam_center_px() ( beam_pixel_slow, beam_pixel_fast, ) = xrayframe.pyslip.tiles.flex_image.tile_readout_to_picture( panel_id, beam_pixel_slow - 0.5, beam_pixel_fast - 0.5 ) center = self._pyslip.tiles.picture_fast_slow_to_map_relative( beam_pixel_fast + self._center[0], beam_pixel_slow + self._center[1] ) # XXX Transparency? ring_data = [ (center[0], center[1], {"colour": "red", "radius": pxl}) for pxl in L_pixels ] # Remove the old ring layer, and draw a new one. if hasattr(self, "_ring_layer") and self._ring_layer is not None: self._pyslip.DeleteLayer(self._ring_layer) self._ring_layer = None self._ring_layer = self._pyslip.AddPointLayer( ring_data, map_rel=True, visible=True, show_levels=[-3, -2, -1, 0, 1, 2, 3, 4, 5], renderer=self._draw_rings_layer, name="<ring_layer>", ) panel = detector[0] fast = col(panel.get_fast_axis()) slow = col(panel.get_slow_axis()) norm = col(panel.get_normal()) x = -panel.pixel_to_millimeter(self._center)[0] y = -panel.pixel_to_millimeter(self._center)[1] z = -(panel.get_distance() - distance) origin = (fast * x + slow * y + norm * z) + col(panel.get_origin()) self.origin.SetLabel("Panel 0 origin: %f, %f, %f" % origin.elems)
bsd-3-clause
2cfba83399d8fda8687408520438cd98
30.980263
88
0.539807
3.486254
false
false
false
false
dials/dials
tests/algorithms/shoebox/test_mask_overlapping.py
1
5935
from __future__ import annotations import numpy as np from dxtbx.model.experiment_list import Experiment, ExperimentList from dxtbx.serialize import load from scitbx import matrix from dials.algorithms import shoebox from dials.algorithms.profile_model.gaussian_rs import Model from dials.array_family import flex def predict_reflections(sequence, crystal): # Get models from the sequence beam = sequence.get_beam() detector = sequence.get_detector() gonio = sequence.get_goniometer() scan = sequence.get_scan() sigma_b = beam.get_sigma_divergence(deg=True) sigma_m = crystal.get_mosaicity(deg=True) exlist = ExperimentList() exlist.append( Experiment( imageset=sequence, beam=beam, detector=detector, goniometer=gonio, scan=scan, crystal=crystal, profile=Model(None, 3, sigma_b, sigma_m, deg=True), ) ) predicted = flex.reflection_table.from_predictions(exlist[0]) predicted["id"] = flex.int(len(predicted), 0) predicted.compute_bbox(exlist) # Find overlapping reflections overlaps = shoebox.find_overlapping(predicted["bbox"]) # Return the reflections and overlaps return predicted, overlaps def test(dials_data): # Load the sequence and crystal sequence = load.imageset( dials_data("centroid_test_data", pathlib=True) / "sweep.json" ) crystal = load.crystal( str(dials_data("centroid_test_data", pathlib=True) / "crystal.json") ) # Get models from the sequence detector = sequence.get_detector() # Get the reflections and overlaps reflections, adjacency_list = predict_reflections(sequence, crystal) reflections["shoebox"] = flex.shoebox(reflections["panel"], reflections["bbox"]) reflections["shoebox"].allocate_with_value(shoebox.MaskCode.Valid) # If the adjacency list is given, then create the reflection mask assert len(detector) == 1 image_size = detector[0].get_image_size() shoeboxes = reflections["shoebox"] coords = reflections["xyzcal.px"] shoebox_masker = shoebox.MaskOverlapping() shoebox_masker(shoeboxes, coords, adjacency_list) # Loop through all edges overlapping = [] for e in adjacency_list.edges(): v1, v2 = adjacency_list.source(e), adjacency_list.target(e) overlapping.append(v1) overlapping.append(v2) # Ensure elements are unique overlapping = set(overlapping) # Ensure we have some overlaps assert len(overlapping) > 0 # Get all non-overlapping reflections all_r = set(range(len(reflections))) non_overlapping = all_r.difference(overlapping) # Run the tests tst_non_overlapping(reflections, non_overlapping, detector[0].get_image_size()) tst_overlapping(reflections, overlapping, adjacency_list, image_size) def tst_non_overlapping(reflections, non_overlapping, image_size): """Ensure non-overlapping reflections have all their values 1.""" # Check that all elements in non_overlapping masks are 1 shoeboxes = reflections["shoebox"] for i in non_overlapping: mask = shoeboxes[i].mask assert mask.all_eq(shoebox.MaskCode.Valid) def tst_overlapping(reflections, overlapping, adjacency_list, image_size): """Ensure masks for overlapping reflections are set properly.""" # Loop through all overlaps shoeboxes = reflections["shoebox"] coord = reflections["xyzcal.px"] for i in overlapping: r1 = shoeboxes[i] bbox_1 = r1.bbox r1_coord = matrix.col(coord[i]) # Create a mask that we expect r1_size = (bbox_1[5] - bbox_1[4], bbox_1[3] - bbox_1[2], bbox_1[1] - bbox_1[0]) expected_mask = np.zeros(shape=r1_size, dtype=np.int32) expected_mask[:, :, :] = shoebox.MaskCode.Valid # Loop through all reflections which this reflection overlaps for j in adjacency_list.adjacent_vertices(i): r2 = shoeboxes[j] bbox_2 = r2.bbox r2_coord = matrix.col(coord[j]) # Get bounding box of intersection bbox_3 = ( max(bbox_1[0], bbox_2[0]), min(bbox_1[1], bbox_2[1]), max(bbox_1[2], bbox_2[2]), min(bbox_1[3], bbox_2[3]), max(bbox_1[4], bbox_2[4]), min(bbox_1[5], bbox_2[5]), ) # Check intersection is valid assert bbox_3[0] < bbox_3[1] assert bbox_3[2] < bbox_3[3] assert bbox_3[4] < bbox_3[5] # Get the coordinates are all mask values mask_coord = [] for k in range(bbox_3[4], bbox_3[5]): for j in range(bbox_3[2], bbox_3[3]): for i in range(bbox_3[0], bbox_3[1]): mask_coord.append(matrix.col((i + 0.5, j + 0.5, k + 0.5))) def dist(a, m): return np.array([(a - b).length() for b in m]) # Find the indices in the intersection area where r2 is closer to # the point than r1 ind = np.where(dist(r1_coord, mask_coord) > dist(r2_coord, mask_coord))[0] # Set the mask values for r1 where r2 is closer to 0 k0, k1 = bbox_3[4] - bbox_1[4], bbox_3[5] - bbox_1[4] j0, j1 = bbox_3[2] - bbox_1[2], bbox_3[3] - bbox_1[2] i0, i1 = bbox_3[0] - bbox_1[0], bbox_3[1] - bbox_1[0] intersect_mask = expected_mask[k0:k1, j0:j1, i0:i1] intersect_mask_1d = intersect_mask.reshape(-1) intersect_mask_1d[ind] = 0 intersect_mask[:, :] = intersect_mask_1d.reshape(intersect_mask.shape) expected_mask[k0:k1, j0:j1, i0:i1] = intersect_mask # Check the masks are the same calculated_mask = r1.mask.as_numpy_array() assert np.all(calculated_mask == expected_mask)
bsd-3-clause
38498a53652ebc9721ed979f91f37c86
34.753012
87
0.613311
3.428654
false
false
false
false
douban/dpark
dpark/schedule.py
1
53619
from __future__ import absolute_import import marshal import multiprocessing import os import socket import sys import time import six from six.moves import map, range, urllib, queue, cPickle import weakref import threading import json from collections import Counter import zmq from addict import Dict from pymesos import MesosSchedulerDriver, encode_data, decode_data import dpark.conf as conf from dpark.accumulator import Accumulator from dpark.dependency import ShuffleDependency from dpark.env import env from dpark.taskset import TaskSet, TaskCounter from dpark.mutable_dict import MutableDict from dpark.task import ResultTask, ShuffleMapTask, TTID, TaskState, TaskEndReason from dpark.hostatus import TaskHostManager from dpark.shuffle import MapOutputTracker from dpark.utils import ( compress, decompress, spawn, getuser, sec2nanosec) from dpark.utils.log import get_logger from dpark.utils.frame import Scope from dpark.utils import dag logger = get_logger(__name__) EXECUTOR_CPUS = 0.01 EXECUTOR_MEMORY = 128 # cache POLL_TIMEOUT = 0.1 RESUBMIT_TIMEOUT = 60 MAX_IDLE_TIME = 60 * 30 class Stage(object): def __init__(self, rdd, shuffleDep, parents, pipelines, pipeline_edges, rdd_pipelines): """ :param rdd: output rdd of this stage :param shuffleDep: for mapOutputStage, determine how computing result will be aggregated, partitioned """ self.id = self.new_id() self.num_try = 0 self.rdd = rdd self.shuffleDep = shuffleDep self.is_final = (shuffleDep is None) self.parents = parents self.numPartitions = len(rdd) self.num_finished = 0 # for final stage self.outputLocs = [[] for _ in range(self.numPartitions)] self.task_stats = [[] for _ in range(self.numPartitions)] self.taskcounters = [] # a TaskCounter object for each run/retry self.submit_time = 0 self.finish_time = 0 self.pipelines = pipelines # each pipeline is a list of rdds self.pipeline_edges = pipeline_edges # ((src_stage_id, src_pipeline_id), (dst_stage_id, dst_pipeline_id)): N self.rdd_pipelines = rdd_pipelines # rdd_id: pipeline_id def __str__(self): return '<Stage(%d) for %s>' % (self.id, self.rdd) def __getstate__(self): raise Exception('should not pickle stage') def __len__(self): return self.numPartitions nextId = 0 @classmethod def new_id(cls): cls.nextId += 1 return cls.nextId @property def try_id(self): return TTID.make_taskset_id(self.id, self.num_try + 1) # incr num_try After create TaskSet @property def isAvailable(self): if not self.parents and self.shuffleDep is None: return True return all(self.outputLocs) @property def num_task_finished(self): if self.is_final: return self.num_finished else: return len([i for i in self.outputLocs if i]) @property def num_task_running(self): if self.taskcounters: return self.taskcounters[-1].running else: return 0 def addOutputLoc(self, partition, host): self.outputLocs[partition].append(host) # def removeOutput(self, partition, host): # prev = self.outputLocs[partition] # self.outputLocs[partition] = [h for h in prev if h != host] def removeHost(self, host): becameUnavailable = False for ls in self.outputLocs: if host in ls: ls.remove(host) becameUnavailable = True if becameUnavailable: msg = ("%s is now unavailable on host %s, " "postpone resubmit until %d secs later " "to wait for futher fetch failure") logger.info(msg, self, host, RESUBMIT_TIMEOUT) def finish(self): if not self.finish_time: self.finish_time = time.time() def _summary_stats(self): stats = [x[-1] for x in self.task_stats if x] d = {} def _summary(lst): lst.sort() r = {'max': max(lst), 'min': min(lst), 'sum': sum(lst) } return r if stats: for attr in dir(stats[0]): if not attr.startswith('_'): d[attr] = _summary(list([getattr(s, attr) for s in stats])) return d def get_node_id(self, stage_id, pipeline_id): if stage_id == -1: stage_id = self.id return "PIPELINE_{}.{}".format(stage_id, pipeline_id) def _fmt_node(self, stage_id, pipeline_id): if stage_id == -1: stage_id = self.id rdds = [{"rdd_name": rdd.ui_label, "rdd_id": rdd.id, "api_callsite_id": rdd.scope.api_callsite_id, "params": rdd.params} for rdd in self.pipelines[pipeline_id]] n = { dag.KW_TYPE: "stage", dag.KW_ID: self.get_node_id(stage_id, pipeline_id), dag.KW_LABEL: str(stage_id), "rdds": rdds } return n def _fmt_edge(self, e): e, nrdd = e src, dst = [self.get_node_id(*n) for n in e] info = {} if nrdd > 1: info['#rdd'] = nrdd return { # dag.KW_ID: "{}_{}".format(src, dst), dag.KW_SRC: src, dag.KW_DST: dst, "info": info } def get_pipeline_graph(self): nodes = [self._fmt_node(self.id, pipeline_id) for pipeline_id in self.pipelines.keys()] edges = [self._fmt_edge(e) for e in six.iteritems(self.pipeline_edges)] g = {dag.KW_NODES: nodes, dag.KW_EDGES: edges} return g def fmt_stats(self): n = self.numPartitions stats = self._summary_stats() msg = "[metric = min/avg/max]: " for k, d in six.iteritems(stats): sm = d['sum'] unit = k.split('_')[0] if unit == 'num': fmt = " = %d/%d/%d" else: fmt = " = %.2f/%.2f/%.2f" if sm > 0: msg += k vs = d['min'], sm / n, d['max'] vs = tuple(map(int, vs)) unit_s = " " if unit == "bytes": vs = tuple([v >> 20 for v in vs]) fmt = " = %.2f/%.2f/%.2f" unit_s = " MB " msg += (fmt % vs) msg += unit_s return msg def _summary_counters(self): def _sum(attr): return sum([getattr(counter, attr) for counter in self.taskcounters]) counters = { "task": { "all": len(self), "running": self.num_task_running, "finished": self.num_task_finished, }, "fail": dict([(attr[5:], _sum(attr)) for attr in TaskCounter(0).get_fail_types()]) } return counters def get_prof(self): stats = self._summary_stats() counters = self._summary_counters() graph = self.get_pipeline_graph() info = { 'id': self.id, 'parents': [p.id for p in self.parents], 'output_rdd': self.rdd.__class__.__name__, 'output_pipeline': self.get_node_id(self.id, self.rdd.id), 'api_callsite': self.rdd.scope.api_callsite, 'start_time': self.submit_time, 'finish_time': self.finish_time, 'num_partition': self.numPartitions, 'mem': self.rdd.mem, } res = { "info": info, "stats": stats, "counters": counters, 'graph': graph } return res def get_tree_stages(self): stages = [] to_visit = [self] seen = set() while to_visit: s = to_visit.pop() stages.append(s) for ss in s.parents: if ss.id not in seen: to_visit.append(ss) seen.add(ss.id) return stages class Scheduler: def start(self): pass def runJob(self, rdd, func, partitions, allowLocal): pass def clear(self): pass def stop(self): pass def defaultParallelism(self): return 2 class CompletionEvent: def __init__(self, task, reason, result, accumUpdates, stats): self.task = task self.reason = reason self.result = result self.accumUpdates = accumUpdates self.stats = stats def walk_dependencies(rdd, edge_func=lambda r, d: True, node_func=lambda r: True): visited = set() to_visit = [rdd] while to_visit: r = to_visit.pop(0) if r.id in visited: continue visited.add(r.id) if node_func(r): for dep in r.dependencies: if edge_func(r, dep): to_visit.append(dep.rdd) class DAGScheduler(Scheduler): def __init__(self): self.id = self.new_id() self.completionEvents = queue.Queue() self.idToStage = weakref.WeakValueDictionary() self.shuffleToMapStage = {} self.cacheLocs = {} self.idToRunJob = {} self.runJobTimes = 0 self.frameworkId = None self.loghub_dir = None self.jobstats = [] self.is_dstream = False self.current_scope = None self.final_lock = threading.RLock() self.final_stage = None self.final_rdd = None nextId = 0 @classmethod def new_id(cls): cls.nextId += 1 return cls.nextId def clear(self): self.idToStage.clear() self.shuffleToMapStage.clear() self.cacheLocs.clear() def submitTasks(self, tasks): raise NotImplementedError def taskEnded(self, task, reason, result, accumUpdates, stats=None): self.completionEvents.put( CompletionEvent( task, reason, result, accumUpdates, stats)) def abort(self): self.completionEvents.put(None) def getCacheLocs(self, rdd): return self.cacheLocs.get(rdd.id, [[] for _ in range(len(rdd))]) def updateCacheLocs(self): self.cacheLocs = env.cacheTrackerServer.getLocationsSnapshot() def newStage(self, output_rdd, shuffleDep): """ A stage may contain multi data pipeline, which form a tree with one final output pipline as root. Zip, CartesianRDD, and Union may commine diff data sources, so lead to a split of the tree. The leaves of the tree may be one of: 1. a pipeline start from a source RDD (TextFileRDD, Collection) 2. a root pipeline of a parent stage . Unioned rdds with same lineage keep only one by add it to dep_rdds and assign a pipeline_id. --- Be careful: - On one hand, logic for ui should not risk mixing newStage, the latter is much more important. - On the other hand, input pipeline need to link to parent stages. """ parent_stages = set() pipelines = {output_rdd.id: [output_rdd]} pipeline_edges = Counter() rdd_pipelines = {output_rdd.id: output_rdd.id} to_visit = [output_rdd] visited = set() dep_filter = set() while to_visit: r = to_visit.pop(0) if r.id in visited: continue visited.add(r.id) my_pipeline_id = rdd_pipelines.get(r.id) if my_pipeline_id is not None: # not all rdd have my_pipeline_id my_pipeline = pipelines.get(my_pipeline_id) if my_pipeline is None: logger.warning("miss pipeline: {} ".format(r.scope.key)) if r.shouldCache: env.cacheTrackerServer.registerRDD(r.id, len(r)) dep_rdds = [] dep_stages = [] for dep in r.dependencies: if isinstance(dep, ShuffleDependency): stage = self.getShuffleMapStage(dep) parent_stages.add(stage) dep_stages.append(stage) if my_pipeline_id is not None: pipeline_edges[(stage.id, stage.rdd.id), (-1, my_pipeline_id)] += 1 # -1 : current_stage else: logger.warning("miss pipeline: {} {}".format(r.scope.key, dep.rdd.scope.key)) else: to_visit.append(dep.rdd) if r.id not in dep_filter and dep.rdd.id in r.dep_lineage_counts: dep_rdds.append(dep.rdd) else: dep_filter.add(dep.rdd.id) if my_pipeline is None: continue ns, nr = len(dep_stages), len(dep_rdds) if ns + nr <= 1: if nr == 1: dep_rdd = dep_rdds[0] my_pipeline.append(dep_rdd) rdd_pipelines[dep_rdd.id] = my_pipeline_id else: for dep_rdd in dep_rdds: did = dep_rdd.id nrdd = r.dep_lineage_counts[did] pipelines[did] = [dep_rdd] # create a new pipeline/branch rdd_pipelines[did] = did pipeline_edges[((-1, did), (-1, my_pipeline_id))] = nrdd # -1 : current_stage stage = Stage(output_rdd, shuffleDep, list(parent_stages), pipelines, dict(pipeline_edges), rdd_pipelines) self.idToStage[stage.id] = stage logger.debug('new stage: %s', stage) return stage def getShuffleMapStage(self, dep): stage = self.shuffleToMapStage.get(dep.shuffleId, None) if stage is None: stage = self.newStage(dep.rdd, dep) self.shuffleToMapStage[dep.shuffleId] = stage return stage def getMissingParentStages(self, stage): missing = set() def _(r, dep): if r.shouldCache and all(self.getCacheLocs(r)): return False if isinstance(dep, ShuffleDependency): stage = self.getShuffleMapStage(dep) if not stage.isAvailable: missing.add(stage) return False return True walk_dependencies(stage.rdd, _) return list(missing) def get_call_graph(self, final_rdd): edges = Counter() # <parent, child > : count visited = set() to_visit = [final_rdd] while to_visit: r = to_visit.pop(0) if r.id in visited: continue visited.add(r.id) for dep in r.dependencies: to_visit.append(dep.rdd) if dep.rdd.scope.api_callsite_id != r.scope.api_callsite_id: edges[(dep.rdd.scope.api_callsite_id, r.scope.api_callsite_id)] += 1 nodes = set() run_scope = self.current_scope edges[(final_rdd.scope.api_callsite_id, run_scope.api_callsite_id)] = 1 for s, d in edges.keys(): nodes.add(s) nodes.add(d) return sorted(list(nodes)), dict(edges) @classmethod def fmt_call_graph(cls, g0): nodes0, edges0 = g0 nodes = [] edges = [{dag.KW_ID: "{}_{}".format(parent, child), dag.KW_SRC: parent, dag.KW_DST: child, "count": count} for ((parent, child), count) in edges0.items()] for n in nodes0: scope = Scope.scopes_by_api_callsite_id[n][0] nodes.append({dag.KW_ID: n, dag.KW_LABEL: scope.api, dag.KW_DETAIL: [scope.api_callsite, scope.stack_above_api]}) return {dag.KW_NODES: nodes, dag.KW_EDGES: edges} def get_profs(self): res = [marshal.loads(j) for j in self.jobstats] running = self.get_running_prof() if running: res.append(marshal.loads(marshal.dumps(running))) return res def get_running_prof(self): if self.final_stage: with self.final_lock: return self._get_stats(self.final_rdd, self.final_stage) def runJob(self, finalRdd, func, partitions, allowLocal): self.runJobTimes += 1 self.current_scope = Scope.get("Job %d:{api}" % (self.runJobTimes, )) outputParts = list(partitions) numOutputParts = len(partitions) finalStage = self.newStage(finalRdd, None) with self.final_lock: self.final_rdd = finalRdd self.final_stage = finalStage try: from dpark.web.ui.views.rddopgraph import StageInfo stage_info = StageInfo() stage_info.create_stage_info(finalStage) def create_stage_info_recur(cur_stage, is_final=False): if not cur_stage or cur_stage.id in self.idToRunJob: return for par_stage in cur_stage.parents: create_stage_info_recur(par_stage) if cur_stage.id not in self.idToRunJob: self.idToRunJob[cur_stage.id] = StageInfo.idToStageInfo[cur_stage.id] self.idToRunJob[cur_stage.id].is_final = is_final create_stage_info_recur(finalStage, is_final=True) except ImportError: pass results = [None] * numOutputParts finished = [None] * numOutputParts last_finished = 0 finalStage.num_finished = 0 waiting = set() running = set() failed = set() pendingTasks = {} # stage -> set([task_id..]) lastFetchFailureTime = 0 self.updateCacheLocs() logger.debug('Final stage: %s, %d', finalStage, numOutputParts) logger.debug('Parents of final stage: %s', finalStage.parents) logger.debug( 'Missing parents: %s', self.getMissingParentStages(finalStage)) def onStageFinished(stage): def _(r, dep): return r._do_checkpoint() MutableDict.merge() walk_dependencies(stage.rdd, _) logger.info("stage %d finish %s", stage.id, stage.fmt_stats()) if (allowLocal and ( not finalStage.parents or not self.getMissingParentStages(finalStage) ) and numOutputParts == 1): split = finalRdd.splits[outputParts[0]] yield func(finalRdd.iterator(split)) onStageFinished(finalStage) return def submitStage(stage): if not stage.submit_time: stage.submit_time = time.time() logger.debug('submit stage %s', stage) if stage not in waiting and stage not in running: missing = self.getMissingParentStages(stage) if not missing: submitMissingTasks(stage) running.add(stage) else: for parent in missing: submitStage(parent) waiting.add(stage) def submitMissingTasks(stage): myPending = pendingTasks.setdefault(stage, set()) tasks = [] have_prefer = True if stage == finalStage: for i in range(numOutputParts): if not finished[i]: part = outputParts[i] if have_prefer: locs = self.getPreferredLocs(finalRdd, part) if not locs: have_prefer = False else: locs = [] tasks.append(ResultTask(finalStage.id, finalStage.try_id, part, finalRdd, func, locs, i)) else: for part in range(stage.numPartitions): if not stage.outputLocs[part]: if have_prefer: locs = self.getPreferredLocs(stage.rdd, part) if not locs: have_prefer = False else: locs = [] tasks.append(ShuffleMapTask(stage.id, stage.try_id, part, stage.rdd, stage.shuffleDep, locs)) logger.debug('add to pending %s tasks', len(tasks)) myPending |= set(t.id for t in tasks) self.submitTasks(tasks) submitStage(finalStage) while finalStage.num_finished != numOutputParts: try: evt = self.completionEvents.get(False) except queue.Empty: if (failed and time.time() > lastFetchFailureTime + RESUBMIT_TIMEOUT): self.updateCacheLocs() for stage in failed: logger.info('Resubmitting failed stages: %s', stage) submitStage(stage) failed.clear() else: time.sleep(0.1) continue if evt is None: # aborted for taskset in self.active_tasksets.values(): self.tasksetFinished(taskset) if not self.is_dstream: self._keep_stats(finalRdd, finalStage) raise RuntimeError('TaskSet aborted!') task, reason = evt.task, evt.reason stage = self.idToStage[task.stage_id] if stage not in pendingTasks: # stage from other taskset continue logger.debug('remove from pending %s from %s', task, stage) pendingTasks[stage].remove(task.id) if reason == TaskEndReason.success: Accumulator.merge(evt.accumUpdates) stage.task_stats[task.partition].append(evt.stats) if isinstance(task, ResultTask): finished[task.outputId] = True finalStage.num_finished += 1 results[task.outputId] = evt.result while last_finished < numOutputParts and finished[last_finished]: yield results[last_finished] results[last_finished] = None last_finished += 1 stage.finish() elif isinstance(task, ShuffleMapTask): stage = self.idToStage[task.stage_id] stage.addOutputLoc(task.partition, evt.result) if all(stage.outputLocs): stage.finish() logger.debug( '%s finished; looking for newly runnable stages', stage ) if pendingTasks[stage]: logger.warn('dirty stage %d with %d tasks' '(select at most 10 tasks:%s) not clean', stage.id, len(pendingTasks[stage]), str(list(pendingTasks[stage])[:10])) del pendingTasks[stage] onStageFinished(stage) running.remove(stage) if stage.shuffleDep is not None: MapOutputTracker.set_locs( stage.shuffleDep.shuffleId, [l[-1] for l in stage.outputLocs]) self.updateCacheLocs() newlyRunnable = set( stage for stage in waiting if not self.getMissingParentStages(stage) ) waiting -= newlyRunnable running |= newlyRunnable logger.debug( 'newly runnable: %s, %s', waiting, newlyRunnable) for stage in newlyRunnable: submitMissingTasks(stage) elif reason == TaskEndReason.fetch_failed: exception = evt.result if stage in running: waiting.add(stage) running.remove(stage) mapStage = self.shuffleToMapStage[exception.shuffleId] mapStage.removeHost(exception.serverUri) failed.add(mapStage) lastFetchFailureTime = time.time() else: logger.error( 'task %s failed: %s %s %s', task, reason, type(reason), reason.message) raise Exception(reason.message) onStageFinished(finalStage) if not self.is_dstream: self._keep_stats(finalRdd, finalStage) assert all(finished) with self.final_lock: self.final_stage = None self.final_rdd = None return def getPreferredLocs(self, rdd, partition): return rdd.preferredLocations(rdd.splits[partition]) def _keep_stats(self, final_rdd, final_stage): try: stats = self._get_stats(final_rdd, final_stage) self.jobstats.append(marshal.dumps(stats)) if self.loghub_dir: self._dump_stats(stats) except Exception as e: logger.exception("Fail to dump job stats: %s.", e) def _dump_stats(self, stats): name = "_".join(map(str, ['sched', self.id, "job", self.runJobTimes])) + ".json" path = os.path.join(self.loghub_dir, name) logger.info("writing profile to %s", path) with open(path, 'w') as f: json.dump(stats, f, indent=4) def _get_stats(self, final_rdd, final_stage): call_graph = self.fmt_call_graph(self.get_call_graph(final_rdd)) cmd = '[dpark] ' + \ os.path.abspath(sys.argv[0]) + ' ' + ' '.join(sys.argv[1:]) stages = sorted([s.get_prof() for s in final_stage.get_tree_stages()], key=lambda x: x['info']['start_time']) sink_scope = self.current_scope sink_id = "SINK_{}_{}".format(self.id, self.runJobTimes) sink_node = { dag.KW_TYPE: "sink", dag.KW_ID: sink_id, dag.KW_LABEL: sink_scope.name, "call_id": sink_scope.api_callsite_id } sink_edge = { "source": final_stage.get_node_id(final_stage.id, final_stage.rdd_pipelines[final_rdd.id]), "target": sink_id, } run = {'framework': self.frameworkId, 'scheduler': self.id, "run": self.runJobTimes, 'sink': { "call_site": sink_scope.api_callsite, "node": sink_node, "edges": sink_edge, }, 'stages': stages, "call_graph": call_graph, } ret = { 'script': { 'cmd': cmd, 'env': {'PWD': os.getcwd()} }, 'run': run } return ret def run_task(task, aid): logger.debug('Running task %r', task) try: Accumulator.clear() result = task.run(aid) accumUpdates = Accumulator.values() MutableDict.flush() return task.id, result, accumUpdates except Exception as e: logger.error('error in task %s', task) import traceback traceback.print_exc() e.task_id = task.id raise e class LocalScheduler(DAGScheduler): attemptId = 0 def nextAttempId(self): self.attemptId += 1 return self.attemptId def submitTasks(self, tasks): logger.debug('submit tasks %s in LocalScheduler', tasks) for task in tasks: task_copy = cPickle.loads(cPickle.dumps(task, -1)) try: _, result, update = run_task(task_copy, self.nextAttempId()) self.taskEnded(task, TaskEndReason.success, result, update) except Exception: self.taskEnded(task, TaskEndReason.other_failure, None, None) def run_task_in_process(task, tid, environ): try: return TaskEndReason.success, run_task(task, tid) except KeyboardInterrupt: sys.exit(0) except Exception as e: return TaskEndReason.other_failure, e class MultiProcessScheduler(LocalScheduler): def __init__(self, threads): LocalScheduler.__init__(self) self.threads = threads self.tasks = {} self.pool = None def submitTasks(self, tasks): if not tasks: return logger.info('Got a taskset with %d tasks: %s', len(tasks), tasks[0].rdd) total, self.finished, start = len(tasks), 0, time.time() def initializer(): # when called on subprocess of multiprocessing's Pool, # default sighandler of SIGTERM will be called to quit gracefully, # and we ignore other signals to prevent dead lock. import signal from .context import _signals for sig in _signals: if sig == signal.SIGTERM: signal.signal(sig, signal.SIG_DFL) else: signal.signal(sig, signal.SIG_IGN) def callback(args): state, data = args logger.debug('task end: %s', state) if state == TaskEndReason.other_failure: logger.warning('task failed: %s', data) self.taskEnded(data.task_id, TaskEndReason.other_failure, result=None, accumUpdates=None) return tid, result, update = data task = self.tasks.pop(tid) self.finished += 1 logger.info('Task %s finished (%d/%d) \x1b[1A', tid, self.finished, total) if self.finished == total: logger.info( 'TaskSet finished in %.1f seconds' + ' ' * 20, time.time() - start) self.taskEnded(task, TaskEndReason.success, result, update) for task in tasks: logger.debug('put task async: %s', task) self.tasks[task.id] = task if not self.pool: # daemonic processes are not allowed to have children from dpark.broadcast import start_download_manager start_download_manager() self.pool = multiprocessing.Pool( self.threads or 2, initializer=initializer ) self.pool.apply_async(run_task_in_process, [task, self.nextAttempId(), env.environ], callback=callback) def stop(self): if self.pool: self.pool.terminate() self.pool.join() logger.debug('process pool stopped') def safe(f): def _(self, *a, **kw): with self.lock: r = f(self, *a, **kw) return r return _ class LogReceiver(object): def __init__(self, output): self.output = output self._started = False self.addr = None def start(self): ctx = zmq.Context() sock = ctx.socket(zmq.PULL) port = sock.bind_to_random_port('tcp://0.0.0.0') self._started = True def collect_log(): while self._started: if sock.poll(1000, zmq.POLLIN): line = sock.recv() self.output.write(line) sock.close() ctx.destroy() spawn(collect_log) host = socket.gethostname() self.addr = 'tcp://%s:%d' % (host, port) logger.debug('log collecter start at %s', self.addr) def stop(self): self._started = False class MesosScheduler(DAGScheduler): def __init__(self, master, options, webui_url=None): DAGScheduler.__init__(self) self.master = master self.cpus = options.cpus self.mem = options.mem self.task_per_node = options.parallel or 8 self.group = options.group self.logLevel = options.logLevel self.options = options self.role = options.role self.color = options.color self.webui_url = webui_url self.started = False self.last_finish_time = 0 self.last_task_launch_time = None self.is_suppressed = False self.isRegistered = False self.executor = None self.driver = None self.out_logger = LogReceiver(sys.stdout) self.err_logger = LogReceiver(sys.stderr) self.lock = threading.RLock() self.task_host_manager = TaskHostManager() self.init_tasksets() def init_tasksets(self): self.active_tasksets = {} self.ttid_to_agent_id = {} self.agent_id_to_ttids = {} def clear(self): DAGScheduler.clear(self) self.init_tasksets() def processHeartBeat(self): # no need in dpark now, just for compatibility with pymesos pass def start(self): self.out_logger.start() self.err_logger.start() def start_driver(self): name = '[dpark] ' + \ os.path.abspath(sys.argv[0]) + ' ' + ' '.join(sys.argv[1:]) if len(name) > 256: name = name[:256] + '...' framework = Dict() framework.user = getuser() if framework.user == 'root': raise Exception('dpark is not allowed to run as \'root\'') framework.name = name if self.role: framework.role = self.role framework.hostname = socket.gethostname() if self.webui_url: framework.webui_url = self.webui_url self.driver = MesosSchedulerDriver( self, framework, self.master, use_addict=True ) self.driver.start() logger.debug('Mesos Scheudler driver started') self.started = True self.last_finish_time = time.time() def check(): while self.started: with self.lock: now = time.time() if (not self.active_tasksets and now - self.last_finish_time > MAX_IDLE_TIME): logger.info('stop mesos scheduler after %d seconds idle', now - self.last_finish_time) self.stop() break for taskset in self.active_tasksets.values(): if taskset.check_task_timeout(): self.requestMoreResources() time.sleep(1) spawn(check) @safe def registered(self, driver, frameworkId, masterInfo): self.isRegistered = True self.frameworkId = frameworkId.value logger.debug('connect to master %s:%s, registered as %s', masterInfo.hostname, masterInfo.port, frameworkId.value) self.executor = self.getExecutorInfo(str(frameworkId.value)) from dpark.utils.log import add_loghub _, self.loghub_dir = add_loghub(self.frameworkId) @safe def reregistered(self, driver, masterInfo): logger.warning('re-connect to mesos master %s:%s', masterInfo.hostname, masterInfo.port) def disconnected(self, driver): logger.debug('framework is disconnected') def _get_container_image(self): return self.options.image @safe def getExecutorInfo(self, framework_id): info = Dict() info.framework_id.value = framework_id info.command.value = '%s %s' % ( sys.executable, os.path.abspath( os.path.join( os.path.dirname(__file__), 'executor.py')) ) info.executor_id.value = env.get('DPARK_ID', 'default') info.command.environment.variables = variables = [] info.shutdown_grace_period.nanoseconds = 5 * (10 ** 9) v = Dict() variables.append(v) v.name = 'UID' v.value = str(os.getuid()) v = Dict() variables.append(v) v.name = 'GID' v.value = str(os.getgid()) container_image = self._get_container_image() if container_image: info.container.type = 'DOCKER' info.container.docker.image = container_image info.container.docker.parameters = parameters = [] p = Dict() p.key = 'memory-swap' p.value = '-1' parameters.append(p) info.container.volumes = volumes = [] for path in ['/etc/passwd', '/etc/group']: v = Dict() volumes.append(v) v.host_path = v.container_path = path v.mode = 'RO' for path in conf.MOOSEFS_MOUNT_POINTS: v = Dict() volumes.append(v) v.host_path = v.container_path = path v.mode = 'RW' for path in conf.DPARK_WORK_DIR.split(','): v = Dict() volumes.append(v) v.host_path = v.container_path = path v.mode = 'RW' def _mount_volume(_volumes, _host_path, _container_path, _mode): _v = Dict() _volumes.append(_v) _v.container_path = _container_path _v.mode = _mode if _host_path: _v.host_path = _host_path if self.options.volumes: for volume in self.options.volumes.split(','): fields = volume.split(':') if len(fields) == 3: host_path, container_path, mode = fields mode = mode.upper() assert mode in ('RO', 'RW') elif len(fields) == 2: host_path, container_path = fields mode = 'RW' elif len(fields) == 1: container_path, = fields host_path = '' mode = 'RW' else: raise Exception('cannot parse volume %s', volume) _mount_volume(volumes, host_path, container_path, mode) info.resources = resources = [] mem = Dict() resources.append(mem) mem.name = 'mem' mem.type = 'SCALAR' mem.scalar.value = EXECUTOR_MEMORY cpus = Dict() resources.append(cpus) cpus.name = 'cpus' cpus.type = 'SCALAR' cpus.scalar.value = EXECUTOR_CPUS Script = os.path.realpath(sys.argv[0]) info.name = Script info.data = encode_data(marshal.dumps( ( Script, os.getcwd(), sys.path, dict(os.environ), self.task_per_node, self.out_logger.addr, self.err_logger.addr, self.logLevel, self.color, env.environ ) )) assert len(info.data) < (50 << 20), \ 'Info data too large: %s' % (len(info.data),) return info @safe def submitTasks(self, tasks): if not tasks: return rdd = tasks[0].rdd assert all(t.rdd is rdd for t in tasks) taskset = TaskSet(self, tasks, rdd.cpus or self.cpus, rdd.mem or self.mem, rdd.gpus, self.task_host_manager) self.active_tasksets[taskset.id] = taskset stage_scope = '' try: from dpark.web.ui.views.rddopgraph import StageInfo stage_scope = StageInfo.idToRDDNode[tasks[0].rdd.id].scope.api_callsite except: pass stage = self.idToStage[tasks[0].stage_id] stage.num_try += 1 stage.taskcounters.append(taskset.counter) logger.info( 'Got taskset %s with %d tasks for stage: %d ' 'at scope[%s] and rdd:%s', taskset.id, len(tasks), tasks[0].stage_id, stage_scope, tasks[0].rdd) need_revive = self.started if not self.started: self.start_driver() while not self.isRegistered: self.lock.release() time.sleep(0.01) self.lock.acquire() if need_revive: self.requestMoreResources() def requestMoreResources(self): logger.debug('reviveOffers') self.driver.reviveOffers() self.is_suppressed = False @safe def resourceOffers(self, driver, offers): rf = Dict() now = time.time() if not self.active_tasksets or (all(taskset.counter.launched == taskset.counter.n for taskset in self.active_tasksets.values()) and self.last_task_launch_time is not None and self.last_task_launch_time + conf.TIME_TO_SUPPRESS < now): logger.debug('suppressOffers') driver.suppressOffers() self.is_suppressed = True rf.refuse_seconds = 60 * 5 for o in offers: driver.declineOffer(o.id, rf) return start = time.time() filter_offer = [] for o in offers: try: if conf.ban(o.hostname): logger.debug("skip offer on banned node: %s", o.hostname) driver.declineOffer(o.id, filters=Dict(refuse_seconds=0xFFFFFFFF)) continue except: logger.exception("bad ban() func in dpark.conf") group = ( self.getAttribute( o.attributes, 'group') or 'None') if (self.group or group.startswith( '_')) and group not in self.group: driver.declineOffer(o.id, filters=Dict(refuse_seconds=0xFFFFFFFF)) continue unavailability = o.get('unavailability') if (unavailability is not None and sec2nanosec(time.time() + conf.DEFAULT_TASK_TIME) >= unavailability['start']['nanoseconds']): logger.debug('the host %s plan to maintain, so skip it', o.hostname) driver.declineOffer(o.id, filters=Dict(refuse_seconds=600)) continue if self.task_host_manager.is_unhealthy_host(o.hostname): logger.warning('the host %s is unhealthy so skip it', o.hostname) driver.declineOffer(o.id, filters=Dict(refuse_seconds=1800)) continue self.task_host_manager.register_host(o.hostname) filter_offer.append(o) offers = filter_offer cpus = [self.getResource(o.resources, 'cpus') for o in offers] gpus = [self.getResource(o.resources, 'gpus') for o in offers] mems = [self.getResource(o.resources, 'mem') - (o.agent_id.value not in self.agent_id_to_ttids and EXECUTOR_MEMORY or 0) for o in offers] # logger.debug('get %d offers (%s cpus, %s mem, %s gpus), %d tasksets', # len(offers), sum(cpus), sum(mems), sum(gpus), len(self.active_tasksets)) mesos_tasks = {} tasks = {} max_create_time = 0 for taskset in self.active_tasksets.values(): while True: host_offers = {} for i, o in enumerate(offers): if self.agent_id_to_ttids.get(o.agent_id.value, 0) >= self.task_per_node: logger.debug('the task limit exceeded at host %s', o.hostname) continue if (mems[i] < self.mem + EXECUTOR_MEMORY or cpus[i] < self.cpus + EXECUTOR_CPUS): continue host_offers[o.hostname] = (i, o) assigned_list = taskset.taskOffer(host_offers, cpus, mems, gpus) if not assigned_list: break for i, o, t in assigned_list: t0 = time.time() mesos_task = self.createTask(o, t) max_create_time = max(max_create_time, time.time() - t0) mesos_tasks.setdefault(o.id.value, []).append(mesos_task) tasks.setdefault(o.id.value, []).append(t) logger.debug('dispatch %s into %s', t, o.hostname) ttid = mesos_task.task_id.value agent_id = o.agent_id.value taskset.ttids.add(ttid) self.ttid_to_agent_id[ttid] = agent_id self.agent_id_to_ttids[agent_id] = self.agent_id_to_ttids.get(agent_id, 0) + 1 cpus[i] -= min(cpus[i], t.cpus) mems[i] -= t.mem gpus[i] -= t.gpus used = time.time() - start if used > 10: logger.warning('use too much time in resourceOffers: %.2fs, %d offers,' 'assigned %d tasks, max_create_time = %ds', used, len(offers), len(mesos_tasks), max_create_time) for o in offers: oid = o.id.value if oid in mesos_tasks: driver.launchTasks(o.id, mesos_tasks[oid]) for task in tasks[oid]: task.stage_time = time.time() else: driver.declineOffer(o.id) if tasks: self.last_task_launch_time = time.time() # logger.debug('reply with %d tasks, %s cpus %s mem %s gpus left', # sum(len(ts) for ts in tasks.values()), # sum(cpus), sum(mems), sum(gpus)) @safe def inverseOffers(self, driver, offers): for o in offers: driver.acceptInverseOffers(o.id) @safe def offerRescinded(self, driver, offer_id): logger.debug('rescinded offer: %s', offer_id) if self.active_tasksets: self.requestMoreResources() def getResource(self, res, name): for r in res: if r.name == name: return r.scalar.value return 0.0 def getAttribute(self, attrs, name): for r in attrs: if r.name == name: return r.text.value def createTask(self, o, t): task = Dict() tid = t.try_id task.name = 'task %s' % tid task.task_id.value = tid task.agent_id.value = o.agent_id.value task.data = encode_data( compress(cPickle.dumps((t, tid), -1)) ) task.executor = self.executor if len(task.data) > 1000 * 1024: logger.warning('task too large: %s %d', t, len(task.data)) assert len(task.data) < (50 << 20), \ 'Task data too large: %s' % (len(task.data),) resources = task.resources = [] cpu = Dict() resources.append(cpu) cpu.name = 'cpus' cpu.type = 'SCALAR' cpu.scalar.value = t.cpus mem = Dict() resources.append(mem) mem.name = 'mem' mem.type = 'SCALAR' mem.scalar.value = t.mem gpu = Dict() resources.append(gpu) gpu.name = 'gpus' gpu.type = 'SCALAR' gpu.scalar.value = t.gpus return task @safe def statusUpdate(self, driver, status): def plot_progresses(): if self.color: total = len(self.active_tasksets) logger.info('\x1b[2K\x1b[J\x1b[1A') for i, taskset_id in enumerate(self.active_tasksets): if i == total - 1: ending = '\x1b[%sA' % total else: ending = '' tasksets = self.active_tasksets[taskset_id] tasksets.progress(ending) mesos_task_id = status.task_id.value state = status.state source = status.source reason = status.get('reason') msg = status.get('message') # type: str if source == 'SOURCE_EXECUTOR' and msg: reason, msg = msg.split(':', 1) data = status.get('data') if data is not None: data = cPickle.loads(decode_data(data)) logger.debug('status update: %s %s %s %s', mesos_task_id, state, reason, msg) ttid = TTID(mesos_task_id) taskset = self.active_tasksets.get(ttid.taskset_id) # type: TaskSet if taskset is None: if state == TaskState.running: logger.debug('kill task %s as its taskset has gone', mesos_task_id) self.driver.killTask(Dict(value=mesos_task_id)) else: logger.debug('ignore task %s as its taskset has gone', mesos_task_id) return if mesos_task_id not in taskset.ttids: logger.debug('ignore task %s as it has finished or failed, new msg: %s', mesos_task_id, (state, reason)) return if state == TaskState.running: taskset.statusUpdate(ttid.task_id, ttid.task_try, state) if taskset.counter.finished == 0: plot_progresses() return # terminal state taskset.ttids.discard(mesos_task_id) if mesos_task_id in self.ttid_to_agent_id: agent_id = self.ttid_to_agent_id[mesos_task_id] if agent_id in self.agent_id_to_ttids: self.agent_id_to_ttids[agent_id] -= 1 del self.ttid_to_agent_id[mesos_task_id] if state == TaskState.finished: try: result, accUpdate, task_stats = data flag, data = result if flag >= 2: try: data = urllib.request.urlopen(data).read() except IOError: # try again data = urllib.request.urlopen(data).read() flag -= 2 data = decompress(data) if flag == 0: result = marshal.loads(data) else: result = cPickle.loads(data) taskset.statusUpdate(ttid.task_id, ttid.task_try, state, result=result, update=accUpdate, stats=task_stats) plot_progresses() except Exception as e: logger.warning('error when cPickle.loads(): %s, data:%s', e, len(data)) state = TaskState.failed taskset.statusUpdate(ttid.task_id, ttid.task_try, state, reason=TaskEndReason.load_failed, message='load failed: %s' % e) else: exception = data if source == 'SOURCE_EXECUTOR' else None # type: Optional[Exception] taskset.statusUpdate(ttid.task_id, ttid.task_try, state, reason, msg, result=exception) @safe def tasksetFinished(self, taskset): logger.debug('taskset %s finished', taskset.id) if taskset.id in self.active_tasksets: self.last_finish_time = time.time() for mesos_task_id in taskset.ttids: self.driver.killTask(Dict(value=mesos_task_id)) del self.active_tasksets[taskset.id] if not self.active_tasksets: self.agent_id_to_ttids.clear() @safe def error(self, driver, message): logger.error('Mesos error message: %s', message) raise RuntimeError(message) # @safe def stop(self): if not self.started: return logger.debug('stop scheduler') self.started = False self.isRegistered = False self.driver.stop(False) self.driver.join() self.driver = None self.out_logger.stop() self.err_logger.stop() def defaultParallelism(self): return 16 def frameworkMessage(self, driver, executor_id, agent_id, data): logger.warning('[agent %s] %s', agent_id.value, data) def executorLost(self, driver, executor_id, agent_id, status): logger.warning( 'executor at %s %s lost: %s', agent_id.value, executor_id.value, status) self.agent_id_to_ttids.pop(agent_id.value, None) def slaveLost(self, driver, agent_id): logger.warning('agent %s lost', agent_id.value) self.agent_id_to_ttids.pop(agent_id.value, None) def killTask(self, task_id, num_try): tid = Dict() tid.value = TTID.make_ttid(task_id, num_try) self.driver.killTask(tid)
bsd-3-clause
fed4a412a9cddbc2fb7686b148d07adb
33.930945
117
0.516254
4.057435
false
false
false
false
dials/dials
tests/algorithms/indexing/lattice_search/test_lattice_search.py
1
4073
from __future__ import annotations import py.path import pytest from cctbx import sgtbx, uctbx from dxtbx.imageset import ImageSet from dxtbx.serialize import load from dials.algorithms.indexing import lattice_search, stills_indexer from dials.array_family import flex from dials.command_line.index import phil_scope from dials.command_line.slice_sequence import slice_experiments, slice_reflections @pytest.fixture def i04_weak_data(dials_regression): data_dir = py.path.local(dials_regression).join( "indexing_test_data", "i04_weak_data" ) reflections_path = data_dir.join("full.pickle").strpath experiments_path = data_dir.join("experiments_import.json").strpath reflections = flex.reflection_table.from_file(reflections_path) experiments = load.experiment_list(experiments_path, check_format=False) return { "reflections": slice_reflections(reflections, [(1, 20)]), "experiments": slice_experiments(experiments, [(1, 20)]), } @pytest.mark.parametrize( "indexing_method,space_group,unit_cell", ( ("fft1d", None, None), ("fft3d", None, None), ("real_space_grid_search", "P422", (57.8, 57.8, 150.0, 90, 90, 90)), ), ) def test_BasisVectorSearch_i04_weak_data( i04_weak_data, indexing_method, space_group, unit_cell ): reflections = i04_weak_data["reflections"] experiments = i04_weak_data["experiments"] params = phil_scope.fetch().extract() params.indexing.refinement_protocol.n_macro_cycles = 2 params.indexing.basis_vector_combinations.max_refine = 5 params.indexing.method = indexing_method if unit_cell is not None: params.indexing.known_symmetry.unit_cell = uctbx.unit_cell(unit_cell) if space_group is not None: params.indexing.known_symmetry.space_group = sgtbx.space_group_info(space_group) idxr = lattice_search.BasisVectorSearch(reflections, experiments, params) idxr.index() indexed_experiments = idxr.refined_experiments assert len(indexed_experiments) == 1 assert indexed_experiments[0].crystal.get_unit_cell().parameters() == pytest.approx( (57.752, 57.776, 150.013, 90.0101, 89.976, 90.008), rel=1e-3 ) @pytest.mark.xfel @pytest.mark.parametrize( "indexing_method,space_group,unit_cell", ( ("fft3d", "P422", (57.8, 57.8, 150.0, 90, 90, 90)), ("fft1d", "P422", (57.8, 57.8, 150.0, 90, 90, 90)), ("real_space_grid_search", "P422", (57.8, 57.8, 150.0, 90, 90, 90)), ("low_res_spot_match", "P422", (57.8, 57.8, 150.0, 90, 90, 90)), ), ) def test_stills_indexer_methods_i04_weak_data( i04_weak_data, indexing_method, space_group, unit_cell ): reflections = slice_reflections(i04_weak_data["reflections"], [(1, 2)]) experiments = slice_experiments(i04_weak_data["experiments"], [(1, 2)]) for experiment in experiments: experiment.imageset = ImageSet( experiment.imageset.data(), experiment.imageset.indices() ) experiment.imageset.set_scan(None) experiment.imageset.set_goniometer(None) experiment.scan = None experiment.goniometer = None params = phil_scope.fetch().extract() params.indexing.method = indexing_method params.indexing.basis_vector_combinations.max_refine = 5 if unit_cell is not None: params.indexing.known_symmetry.unit_cell = uctbx.unit_cell(unit_cell) if space_group is not None: params.indexing.known_symmetry.space_group = sgtbx.space_group_info(space_group) try: idxr = stills_indexer.StillsIndexerBasisVectorSearch( reflections, experiments, params ) except RuntimeError: idxr = stills_indexer.StillsIndexerLatticeSearch( reflections, experiments, params ) idxr.index() indexed_experiments = idxr.refined_experiments assert len(indexed_experiments) == 1 assert indexed_experiments[0].crystal.get_unit_cell().parameters() == pytest.approx( (57.752, 57.776, 150.013, 90.0101, 89.976, 90.008), rel=1e-2 )
bsd-3-clause
9e8d5f72f075f7e7386043ea208406b1
36.712963
88
0.67616
3.21722
false
true
false
false
dials/dials
src/dials/util/image_viewer/slip_viewer/ring_frame.py
1
11177
from __future__ import annotations import math import wx class RingSettingsFrame(wx.MiniFrame): def __init__(self, *args, **kwds): super().__init__(*args, **kwds) szr = wx.BoxSizer(wx.VERTICAL) panel = RingSettingsPanel(self) self.SetSizer(szr) szr.Add(panel, 1, wx.EXPAND) szr.Fit(panel) self.panel = panel self.sizer = szr self.Fit() self.Bind(wx.EVT_CLOSE, lambda evt: self.Destroy(), self) class RingSettingsPanel(wx.Panel): def __init__(self, *args, **kwds): # XXX Support several rings. Plot radial distribution somewhere # (not here), but maybe distribution along ring. Drop-down menu # for ring center, and button to reset to beam center. super().__init__(*args, **kwds) # Needed to draw and delete the rings. XXX Applies to # calibration_frame as well? self._pyslip = self.GetParent().GetParent().pyslip sizer = wx.BoxSizer(wx.VERTICAL) self.SetSizer(sizer) # Number of decimal digits for distances. self.digits = 2 # Distance control XXX probably does not belong here # print "DISTANCE",self.GetParent().GetParent().viewer._img # box = wx.BoxSizer(wx.HORIZONTAL) # from wxtbx.phil_controls.floatctrl import FloatCtrl # from wxtbx.phil_controls import EVT_PHIL_CONTROL # self.distance_ctrl = FloatCtrl(self, pos=(300,180), size=(80,-1), # value=80.00, # name="Detector Distance") # self.distance_ctrl.SetMax(1000) # self.distance_ctrl.SetMin(5) # self.distance_ctrl.SetOptional(False) # box.Add(self.distance_ctrl, 0, wx.ALL|wx.ALIGN_CENTER_VERTICAL, 5) # txtd = wx.StaticText(self, label="Detector Distance") # box.Add(txtd, 0, wx.ALL|wx.ALIGN_CENTER_VERTICAL, 5) # s.Add(box) from wx.lib.agw.floatspin import EVT_FLOATSPIN, FloatSpin # XXX Should really make value be in Aangstroem resolution, and # have a non-linear slider. self._radius = 100 self._center = [0, 0] radius_max = 2000 radius_min = 10 # Radius controls. box = wx.BoxSizer(wx.HORIZONTAL) self.slider = wx.Slider( self, maxValue=radius_max, minValue=radius_min, size=(250, -1), style=wx.SL_AUTOTICKS | wx.SL_HORIZONTAL, value=self._radius, ) box.Add( self.slider, 0, wx.RIGHT | wx.TOP | wx.BOTTOM | wx.ALIGN_CENTER_VERTICAL, 5 ) self.Bind(wx.EVT_SLIDER, self.OnSlide, self.slider) self.spinner = FloatSpin( self, digits=self.digits, max_val=radius_max, min_val=radius_min, value=self._radius, ) box.Add( self.spinner, 0, wx.RIGHT | wx.TOP | wx.BOTTOM | wx.ALIGN_CENTER_VERTICAL, 5 ) self.Bind(EVT_FLOATSPIN, self.OnSpin, self.spinner) self.auto = wx.Button(self, label="Auto fit") self.Bind(wx.EVT_BUTTON, self.OnAutoFit, self.auto) box.Add( self.auto, 0, wx.RIGHT | wx.TOP | wx.BOTTOM | wx.ALIGN_CENTER_VERTICAL, 5 ) sizer.Add(box) # Centering controls. box = wx.BoxSizer(wx.HORIZONTAL) self.spinner_fast = FloatSpin( self, digits=self.digits, name="fast_ctrl", value=self._center[0] ) box.Add( self.spinner_fast, 0, wx.RIGHT | wx.TOP | wx.BOTTOM | wx.ALIGN_CENTER_VERTICAL, 5, ) box.Add( wx.StaticText(self, label="Center fast"), 0, wx.ALL | wx.ALIGN_CENTER_VERTICAL, 5, ) self.Bind(EVT_FLOATSPIN, self.OnSpinCenter, self.spinner_fast) self.spinner_slow = FloatSpin( self, digits=self.digits, name="slow_ctrl", value=self._center[1] ) box.Add( self.spinner_slow, 0, wx.RIGHT | wx.TOP | wx.BOTTOM | wx.ALIGN_CENTER_VERTICAL, 5, ) box.Add( wx.StaticText(self, label="Center slow"), 0, wx.ALL | wx.ALIGN_CENTER_VERTICAL, 5, ) self.Bind(EVT_FLOATSPIN, self.OnSpinCenter, self.spinner_slow) self.clear_button = wx.Button(self, -1, "Clear") box.Add(self.clear_button, 0, wx.ALL | wx.ALIGN_CENTER_VERTICAL, 5) self.Bind(wx.EVT_BUTTON, self.OnClear, self.clear_button) sizer.Add(box) self.DrawRing() def __del__(self): if hasattr(self, "_ring_layer") and self._ring_layer is not None: self._pyslip.DeleteLayer(self._ring_layer) self._ring_layer = None def OnSlide(self, event): # Keep slider and spinner synchronized. obj = event.EventObject # XXX Is this construct sane? See below # and in calibration_frame, too! self._radius = obj.GetValue() self.spinner.SetValue(self._radius) self.DrawRing() def OnAutoFit(self, event): jitter = 6 detector = self._pyslip.tiles.raw_image.get_detector() beam = self._pyslip.tiles.raw_image.get_beam() # FIXME assumes all detector elements use the same millimeter-to-pixel convention if detector[0].get_distance() > 0: h = detector.hierarchy() if len(h) > 0: beam_pixel_fast, beam_pixel_slow = detector[0].millimeter_to_pixel( detector.hierarchy().get_beam_centre(beam.get_s0()) ) else: beam_pixel_fast, beam_pixel_slow = detector[0].millimeter_to_pixel( detector[0].get_beam_centre(beam.get_s0()) ) avg_distance = -sum(p.get_distance() for p in detector) / len(detector) beam_pixel_fast += self._center[0] beam_pixel_slow += self._center[1] def PointsOnCircle(center, radius, count): for r in range(count): t = (r / count) * 2 * math.pi yield ( center[0] + (radius * math.cos(t)), center[1] + (radius * math.sin(t)), ) best = float("-inf") bestc = [self._center[0], self._center[1]] bestr = self._radius image_data = self._pyslip.tiles.raw_image.get_image_data() if not isinstance(image_data, tuple): image_data = (image_data,) for j in range(-jitter, jitter, 1): j /= 2 for i in range(-jitter, jitter, 1): i /= 2 for r in range(-jitter, jitter, 1): r /= 2 total = 0.0 for point in PointsOnCircle( (beam_pixel_fast + i, beam_pixel_slow + j), self._radius + r, 360, ): mm = detector[0].pixel_to_millimeter(point) mm = (mm[0], mm[1], avg_distance) pid = detector.get_panel_intersection(mm) if pid >= 0: px = detector[pid].get_ray_intersection_px(mm) px = [int(round(px[0])), int(round(px[1]))] data = image_data[pid] if ( px[0] >= 0 and px[0] < data.focus()[1] and px[1] >= 0 and px[1] < data.focus()[0] ): total += data[px[1], px[0]] if total > best: best = total bestc = [self._center[0] + i, self._center[1] + j] bestr = self._radius + r print(f"r: {r: 3.1f}, i: {i: 3.1f}, j: {j: 3.1f}, best: {best:f}") print("DONE", bestc, bestr) self._radius = bestr self._center = bestc self.spinner.SetValue(bestr) self.spinner_fast.SetValue(bestc[0]) self.spinner_slow.SetValue(bestc[1]) self.DrawRing() def OnSpin(self, event): # Keep slider and spinner synchronized. XXX OnSpinRadius()? obj = event.EventObject self._radius = obj.GetValue() self.slider.SetValue(self._radius) self.DrawRing() def OnSpinCenter(self, event): obj = event.EventObject name = obj.GetName() if name == "fast_ctrl": self._center[0] = obj.GetValue() elif name == "slow_ctrl": self._center[1] = obj.GetValue() self.DrawRing() def OnClear(self, event): self.__del__() def _draw_ring_layer(self, dc, data, map_rel): """Draw a points layer. dc the device context to draw on data an iterable of point tuples: (x, y, place, radius, colour, x_off, y_off, pdata) map_rel points relative to map if True, MUST BE TRUE for lightweight Assumes all points are the same colour, saving 100's of ms. """ assert map_rel is True if len(data) == 0: return (lon, lat, place, radius, colour, x_off, y_off, pdata) = data[0] scale = 2**self._pyslip.tiles.zoom_level # Draw points on map/view, using transparency if implemented. try: dc = wx.GCDC(dc) except NotImplementedError: pass dc.SetPen(wx.Pen(colour)) dc.SetBrush(wx.Brush(colour, wx.TRANSPARENT)) for (lon, lat, place, radius, colour, x_off, y_off, pdata) in data: (x, y) = self._pyslip.ConvertGeo2View((lon, lat)) dc.DrawCircle(x, y, radius * scale) def DrawRing(self): xrayframe = self.GetParent().GetParent() panel_id, beam_pixel_fast, beam_pixel_slow = xrayframe.get_beam_center_px() ( beam_pixel_slow, beam_pixel_fast, ) = xrayframe.pyslip.tiles.flex_image.tile_readout_to_picture( panel_id, beam_pixel_slow - 0.5, beam_pixel_fast - 0.5 ) center = self._pyslip.tiles.picture_fast_slow_to_map_relative( beam_pixel_fast + self._center[0], beam_pixel_slow + self._center[1] ) # XXX Transparency? ring_data = [(center[0], center[1], {"colour": "red", "radius": self._radius})] # Remove the old ring layer, and draw a new one. XXX Why # disappears at highest levels? if hasattr(self, "_ring_layer") and self._ring_layer is not None: self._pyslip.DeleteLayer(self._ring_layer) self._ring_layer = None self._ring_layer = self._pyslip.AddPointLayer( ring_data, map_rel=True, visible=True, show_levels=[-3, -2, -1, 0, 1, 2, 3, 4, 5], renderer=self._draw_ring_layer, name="<ring_layer>", )
bsd-3-clause
ca2456609d5c38ed931b835473404f17
33.928125
89
0.521607
3.709592
false
false
false
false
dials/dials
tests/algorithms/profile_model/test_calculator.py
1
1715
from __future__ import annotations from dials.algorithms.profile_model.gaussian_rs.calculator import ( _select_reflections_for_sigma_calc, ) from dials.array_family import flex def test_select_reflections_for_sigma_calc(): """Test the reflection selection helper function.""" # first select all if below threshold reflections = flex.reflection_table() reflections["id"] = flex.int(range(0, 1000)) reflections.set_flags(flex.bool(1000, True), reflections.flags.used_in_refinement) reflections = _select_reflections_for_sigma_calc(reflections, 10000) assert reflections.size() == 1000 # select used_in_refinement if not all used in refinement and above threshold reflections = flex.reflection_table() reflections["id"] = flex.int(range(0, 1000)) good = flex.bool(1000, False) sel = flex.size_t(i for i in range(0, 1000, 2)) good.set_selected(sel, True) reflections.set_flags(good, reflections.flags.used_in_refinement) reflections = _select_reflections_for_sigma_calc( reflections, min_number_of_refl=200 ) assert reflections.size() == 500 assert list(reflections["id"])[0:50] == list(range(0, 100, 2)) # top up if not enough used in refinement compared to threshold reflections = flex.reflection_table() reflections["id"] = flex.int(range(0, 1000)) good = flex.bool(1000, False) sel = flex.size_t(i for i in range(0, 1000, 2)) good.set_selected(sel, True) reflections.set_flags(good, reflections.flags.used_in_refinement) reflections = _select_reflections_for_sigma_calc( reflections, min_number_of_refl=700 ) assert reflections.size() > 700 assert reflections.size() < 1000
bsd-3-clause
74fd6274194088eff914b1c6ff57baff
39.833333
86
0.698542
3.375984
false
false
false
false
dials/dials
src/dials/algorithms/refinement/two_theta_refiner.py
1
9766
"""Versions of refinement classes for two theta refinement of the unit cell""" from __future__ import annotations import logging from math import pi, sqrt from scitbx import matrix from scitbx.math import five_number_summary from dials.algorithms.refinement.parameterisation.prediction_parameters import ( PredictionParameterisation, ) from dials.algorithms.refinement.prediction.managed_predictors import ( ExperimentsPredictor, ) from dials.algorithms.refinement.reflection_manager import ReflectionManager from dials.algorithms.refinement.target import Target from dials.array_family import flex from dials.util import tabulate logger = logging.getLogger(__name__) # constants RAD2DEG = 180.0 / pi DEG2RAD = pi / 180.0 class ConstantTwoThetaWeightingStrategy: def calculate_weights(self, reflections): reflections["2theta.weights"] = flex.double(len(reflections), 1) return reflections def calc_2theta(reflections, experiments): """Calculate and return 2theta angles in radians""" twotheta = flex.double(len(reflections), 0.0) for iexp, exp in enumerate(experiments): isel = (reflections["id"] == iexp).iselection() sub_ref = reflections.select(isel) s0 = matrix.col(exp.beam.get_s0()) for ipanel in range(len(exp.detector)): sel = sub_ref["panel"] == ipanel panel_ref = sub_ref.select(sel) x, y, phi = panel_ref["xyzobs.mm.value"].parts() s1 = exp.detector[ipanel].get_lab_coord(flex.vec2_double(x, y)) s1 = s1 / s1.norms() * s0.length() sub_isel = isel.select(sel) twotheta.set_selected(sub_isel, s1.angle(s0)) return twotheta class TwoThetaReflectionManager(ReflectionManager): _weighting_strategy = ConstantTwoThetaWeightingStrategy() def __init__(self, *args, **kwargs): # call base __init__ super().__init__(*args, **kwargs) # set observed 2theta angles self._reflections["2theta_obs.rad"] = calc_2theta( self._reflections, self._experiments ) # placeholder for calculated 2theta angles self._reflections["2theta_cal.rad"] = flex.double(len(self._reflections), 0.0) return def print_stats_on_matches(self): l = self.get_matches() nref = len(l) if nref == 0: logger.warning( "Unable to calculate summary statistics for zero observations" ) return twotheta_resid = l["2theta_resid"] w_2theta = l["2theta.weights"] msg = ( f"\nSummary statistics for {nref} observations" + " matched to predictions:" ) header = ["", "Min", "Q1", "Med", "Q3", "Max"] rows = [] row_data = five_number_summary(twotheta_resid) rows.append( ["2theta_c - 2theta_o (deg)"] + [f"{e * RAD2DEG:.4g}" for e in row_data] ) row_data = five_number_summary(w_2theta) rows.append(["2theta weights"] + [f"{e * DEG2RAD ** 2:.4g}" for e in row_data]) logger.info(msg) logger.info(tabulate(rows, header) + "\n") class TwoThetaExperimentsPredictor(ExperimentsPredictor): def _predict_one_experiment(self, experiment, reflections): B = flex.mat3_double(len(reflections), experiment.crystal.get_B()) r0 = B * reflections["miller_index"].as_vec3_double() r0len = r0.norms() wl = experiment.beam.get_wavelength() # 2theta = 2 * arcsin( |r0| / (2 * |s0| ) ) reflections["2theta_cal.rad"] = 2.0 * flex.asin(0.5 * r0len * wl) reflections.set_flags( flex.size_t(len(reflections)), reflections.flags.predicted ) class TwoThetaTarget(Target): _grad_names = ["d2theta_dp"] rmsd_names = ["RMSD_2theta"] rmsd_units = ["rad"] def __init__( self, experiments, predictor, reflection_manager, prediction_parameterisation ): Target.__init__( self, experiments, predictor, reflection_manager, prediction_parameterisation, ) # set the single cutoff for 2theta residual to essentially zero self._binsize_cutoffs = [1.0e-6] # predict reflections and finalise reflection manager self.predict() self._reflection_manager.finalise() return def predict(self): """perform reflection prediction for the working reflections and update the reflection manager""" # get the matches reflections = self._reflection_manager.get_obs() # reset the 'use' flag for all observations self._reflection_manager.reset_accepted_reflections() # set twotheta in place self._reflection_predictor(reflections) # calculate residuals reflections["2theta_resid"] = ( reflections["2theta_cal.rad"] - reflections["2theta_obs.rad"] ) reflections["2theta_resid2"] = flex.pow2(reflections["2theta_resid"]) # set used_in_refinement flag to all those that had predictions mask = reflections.get_flags(reflections.flags.predicted) reflections.set_flags(mask, reflections.flags.used_in_refinement) # collect the matches self.update_matches(force=True) return @staticmethod def _extract_residuals_and_weights(matches): # return residuals and weights as 1d flex.double vectors residuals = matches["2theta_resid"] weights = matches["2theta.weights"] return residuals, weights @staticmethod def _extract_squared_residuals(matches): residuals2 = matches["2theta_resid2"] return residuals2 def _rmsds_core(self, reflections): """calculate unweighted RMSDs for the specified reflections""" resid_2theta = flex.sum(reflections["2theta_resid2"]) n = len(reflections) rmsds = (sqrt(resid_2theta / n),) return rmsds def achieved(self): """RMSD criterion for target achieved""" r = self._rmsds if self._rmsds else self.rmsds() # reset cached rmsds to avoid getting out of step self._rmsds = None if r[0] < self._binsize_cutoffs[0]: return True return False class TwoThetaPredictionParameterisation(PredictionParameterisation): _grad_names = ("d2theta_dp",) def __init__(self, *args, **kwargs): super().__init__(*args, **kwargs) # check that only the unit cell is parameterised assert not self._detector_parameterisations assert not self._beam_parameterisations assert not self._xl_orientation_parameterisations assert not self._goniometer_parameterisations return def _local_setup(self, reflections): # we want the wavelength self._wavelength = 1.0 / self._s0.norms() return def _xl_unit_cell_derivatives(self, isel, parameterisation=None, reflections=None): # Get required data h = self._h.select(isel) B = self._B.select(isel) wl = self._wavelength.select(isel) # get derivatives of the B matrix wrt the parameters dB_dxluc_p = [ None if der is None else flex.mat3_double(len(isel), der.elems) for der in parameterisation.get_ds_dp(use_none_as_null=True) ] d2theta_dp = [] # loop through the parameters for der in dB_dxluc_p: if der is None: d2theta_dp.append(None) continue r0 = B * h dr0 = der * h r0len = r0.norms() dr0len = dr0.dot(r0) / r0len # 2theta = 2 * arcsin( |r0| / (2 * |s0| ) ) sintheta = 0.5 * r0len * wl fac = 1.0 / flex.sqrt(flex.double(len(wl), 1.0) - flex.pow2(sintheta)) val = fac * wl * dr0len d2theta_dp.append(val) return d2theta_dp def _grads_xl_unit_cell_loop(self, reflections, results, callback=None): """Loop over all crystal unit cell parameterisations, calculate gradients and extend the results""" # loop over the crystal unit cell parameterisations for xlucp in self._xl_unit_cell_parameterisations: # Determine (sub)set of reflections affected by this parameterisation isel = flex.size_t() for exp_id in xlucp.get_experiment_ids(): isel.extend(self._experiment_to_idx[exp_id]) # Extend derivative vectors for this crystal unit cell parameterisation results = self._extend_gradient_vectors( results, self._nref, xlucp.num_free(), keys=self._grad_names ) if len(isel) == 0: # if no reflections are in this experiment, skip calculation of # gradients, but must still process null gradients by a callback if callback is not None: for iparam in range(xlucp.num_free()): results[self._iparam] = callback(results[self._iparam]) self._iparam += 1 else: self._iparam += xlucp.num_free() continue d2theta_dp = self._xl_unit_cell_derivatives( isel, parameterisation=xlucp, reflections=reflections ) for d2theta in d2theta_dp: if d2theta is not None: results[self._iparam][self._grad_names[0]].set_selected( isel, d2theta ) # increment the parameter index pointer self._iparam += 1 return results
bsd-3-clause
11472e60790228a1c6103527c6b9e595
30.811075
88
0.601577
3.767747
false
false
false
false
dials/dials
src/dials/command_line/stills_process.py
1
75177
from __future__ import annotations import collections import copy import glob import logging import os import pickle import sys import tarfile import time from io import BytesIO from dxtbx.model.experiment_list import ( Experiment, ExperimentList, ExperimentListFactory, ) from libtbx.phil import parse from libtbx.utils import Abort, Sorry import dials.util from dials.array_family import flex from dials.util import log logger = logging.getLogger("dials.command_line.stills_process") help_message = """ DIALS script for processing still images. Import, index, refine, and integrate are all done for each image separately. """ def _control_phil_str(): return """ input { file_list = None .type = path .help = Path to a text file with a list of images glob = None .type = str .help = For large, multi-file datasets, specify the paths using wildcards (e.g. *.cbf) .multiple = True image_tag = None .type = str .multiple = True .help = Only process images with these tag(s). For single-image files (like CBFs or SMVs), the image \ tag for each file is the file name. For multi-image files like HDF5, the image tag is \ filename_imagenumber (including leading zeros). Use show_image_tags=True to see the list \ of image tags that will be used for a dataset. show_image_tags = False .type = bool .help = Show the set of image tags that would be used during processing. To process subsets of image \ files, use these tags with the image_tag parameter. max_images = None .type = int .help = Limit total number of processed images to max_images ignore_gain_mismatch = False .type = bool .expert_level = 3 .help = Detector gain should be set on the detector models loaded from the images or in the \ processing parameters, not both. Override the check that this is true with this flag. \ } dispatch { pre_import = False .type = bool .expert_level = 2 .help = If True, before processing import all the data. Needed only if processing \ multiple multi-image files at once (not a recommended use case) process_percent = None .type = int(value_min=1, value_max=100) .help = Percent of events to process refine = False .expert_level = 2 .type = bool .help = If True, after indexing, refine the experimental models squash_errors = True .expert_level = 2 .type = bool .help = If True, if an image fails to process, continue to the next image. \ otherwise, halt processing and show the error. find_spots = True .expert_level = 2 .type = bool .help = Whether to do spotfinding. Needed for indexing/integration index = True .expert_level = 2 .type = bool .help = Attempt to index images. find_spots also needs to be True for this to work integrate = True .expert_level = 2 .type = bool .help = Integrate indexed images. Ignored if index=False or find_spots=False coset = False .expert_level = 2 .type = bool .help = Within the integrate dispatcher, integrate a sublattice coset intended to represent \ negative control spots with no Bragg diffraction. hit_finder{ enable = True .type = bool .help = Whether to do hitfinding. hit_finder=False: process all images minimum_number_of_reflections = 16 .type = int .help = If the number of strong reflections on an image is less than this, and \ the hitfinder is enabled, discard this image. maximum_number_of_reflections = None .type = int .help = If specified, ignores images with more than this many number of reflections } } output { output_dir = . .type = str .help = Directory output files will be placed composite_output = True .type = bool .help = If True, save one set of experiment/reflection files per process, where each is a \ concatenated list of all the successful events examined by that process. \ If False, output a separate experiment/reflection file per image (generates a \ lot of files). logging_dir = None .type = str .help = Directory output log files will be placed experiments_filename = None .type = str .help = The filename for output experiments. For example, %s_imported.expt strong_filename = None .type = str .help = The filename for strong reflections from spot finder output. For example: \ %s_strong.refl indexed_filename = %s_indexed.refl .type = str .help = The filename for indexed reflections. refined_experiments_filename = %s_refined.expt .type = str .help = The filename for saving refined experimental models integrated_filename = %s_integrated.refl .type = str .help = The filename for final integrated reflections. integrated_experiments_filename = %s_integrated.expt .type = str .help = The filename for saving final experimental models. coset_filename = %s_coset%d.refl .type = str .help = The filename for final coset reflections. coset_experiments_filename = %s_coset%d.expt .type = str .help = The filename for saving final coset experimental models. profile_filename = None .type = str .help = The filename for output reflection profile parameters integration_pickle = int-%d-%s.pickle .type = str .help = Filename for cctbx.xfel-style integration pickle files } mp { method = *multiprocessing sge lsf pbs mpi .type = choice .help = "The multiprocessing method to use" nproc = 1 .type = int(value_min=1) .help = "The number of processes to use." composite_stride = None .type = int .help = For MPI, if using composite mode, specify how many ranks to \ aggregate data from. For example, if you have 100 processes, \ composite mode will output N*100 files, where N is the number \ of file types (expt, refl, etc). If you specify stride = 25, \ then each group of 25 process will send their results to 4 \ processes and only N*4 files will be created. Ideally, match \ stride to the number of processors per node. debug .expert_level = 2 { cProfile = False .type = bool .help = Enable code profiling. Profiling file will be available in \ the debug folder. Use (for example) runsnake to visualize \ processing performance output_debug_logs = True .type = bool .help = Whether to write debugging information for every image \ processed } } """ def _dials_phil_str(): return """ input { reference_geometry = None .type = str .help = Provide an models.expt file with exactly one detector model. Data processing will use \ that geometry instead of the geometry found in the image headers. sync_reference_geom = True .type = bool .help = ensures the reference hierarchy agrees with the image format } output { shoeboxes = True .type = bool .help = Save the raw pixel values inside the reflection shoeboxes during spotfinding. } include scope dials.util.options.geometry_phil_scope include scope dials.algorithms.spot_finding.factory.phil_scope include scope dials.algorithms.indexing.indexer.phil_scope indexing { include scope dials.algorithms.indexing.lattice_search.basis_vector_search_phil_scope } include scope dials.algorithms.refinement.refiner.phil_scope include scope dials.algorithms.integration.integrator.phil_scope include scope dials.algorithms.profile_model.factory.phil_scope include scope dials.algorithms.spot_prediction.reflection_predictor.phil_scope include scope dials.algorithms.integration.stills_significance_filter.phil_scope indexing { stills { method_list = None .type = strings .help = List of indexing methods. If indexing fails with first method, indexing will be \ attempted with the next, and so forth known_orientations = None .type = path .multiple = True .expert_level = 2 .help = Paths to previous processing results including crystal orientations. \ If specified, images will not be re-indexed, but instead the known \ orientations will be used. Provide paths to experiment list files, using \ wildcards as needed. require_known_orientation = False .type = bool .expert_level = 2 .help = If known_orientations are provided, and an orientation for an image is not \ found, this is whether or not to attempt to index the image from scratch \ using indexing.method } } integration { include scope dials.algorithms.integration.kapton_correction.absorption_phil_scope coset { transformation = 6 .type = int(value_min=0, value_max=6) .multiple = False .help = The index number(s) of the modulus=2 sublattice transformation(s) used to produce distince coset results. \ 0=Double a, 1=Double b, 2=Double c, 3=C-face centering, 4=B-face centering, 5=A-face centering, 6=Body centering \ See Sauter and Zwart, Acta D (2009) 65:553 } integration_only_overrides { trusted_range = None .type = floats(size=2) .help = "Override the panel trusted range [min_trusted_value, max_trusted_value] during integration." .short_caption = "Panel trusted range" } } profile { gaussian_rs { parameters { sigma_b_cutoff = 0.1 .type = float .help = Maximum sigma_b before the image is rejected } } } """ def _program_defaults_phil_str(): return """ indexing { method = fft1d } refinement { parameterisation { auto_reduction { min_nref_per_parameter = 1 action = fix } beam.fix = all detector.fix = all } reflections { weighting_strategy.override = stills outlier.algorithm = null } } integration { integrator = stills profile.fitting = False background { algorithm = simple simple { outlier.algorithm = plane model.algorithm = linear2d } } } profile.gaussian_rs.min_spots.overall = 0 """ control_phil_str = _control_phil_str() dials_phil_str = _dials_phil_str() program_defaults_phil_str = _program_defaults_phil_str() phil_scope = parse(control_phil_str + dials_phil_str, process_includes=True).fetch( parse(program_defaults_phil_str) ) def do_import(filename, load_models=True): logger.info("Loading %s", os.path.basename(filename)) experiments = ExperimentListFactory.from_filenames([filename], load_models=False) if len(experiments) == 0: try: experiments = ExperimentListFactory.from_json_file(filename) except ValueError: raise Abort(f"Could not load {filename}") if len(experiments) == 0: raise Abort(f"Could not load {filename}") from dxtbx.imageset import ImageSetFactory all_experiments = ExperimentList() for experiment in experiments: # Convert from ImageSequence to ImageSet, if needed imageset = ImageSetFactory.imageset_from_anyset(experiment.imageset) for i in range(len(imageset)): # Preserve original models if they were available (in the case of an image file # they will not be, but in the case of a previously processed experiment list, # then they may be available expt = Experiment( imageset=imageset[i : i + 1], detector=experiment.detector, beam=experiment.beam, scan=experiment.scan, goniometer=experiment.goniometer, crystal=experiment.crystal, ) if load_models: expt.load_models() all_experiments.append(expt) return all_experiments def sync_geometry(src, dest): dest.set_local_frame( src.get_local_fast_axis(), src.get_local_slow_axis(), src.get_local_origin() ) if not src.is_panel(): for src_child, dest_child in zip(src, dest): sync_geometry(src_child, dest_child) class Script: """A class for running the script.""" def __init__(self): """Initialise the script.""" from dials.util.options import ArgumentParser # The script usage usage = "usage: dials.stills_process [options] [param.phil] filenames" self.tag = None self.reference_detector = None # Create the parser self.parser = ArgumentParser(usage=usage, phil=phil_scope, epilog=help_message) def load_reference_geometry(self): if self.params.input.reference_geometry is None: return try: ref_experiments = ExperimentListFactory.from_json_file( self.params.input.reference_geometry, check_format=False ) except Exception: try: import dxtbx img = dxtbx.load(self.params.input.reference_geometry) except Exception: raise Sorry( "Couldn't load geometry file %s" % self.params.input.reference_geometry ) else: self.reference_detector = img.get_detector() else: assert len(ref_experiments.detectors()) == 1 self.reference_detector = ref_experiments.detectors()[0] def run(self, args=None): """Execute the script.""" from libtbx import easy_mp try: from mpi4py import MPI except ImportError: rank = 0 size = 1 else: comm = MPI.COMM_WORLD rank = comm.Get_rank() # each process in MPI has a unique id, 0-indexed size = comm.Get_size() # size: number of processes running in this job if rank == 0: # Parse the command line params, options, all_paths = self.parser.parse_args( args, show_diff_phil=False, return_unhandled=True, quick_parse=True ) if params.input.glob: all_paths.extend(params.input.glob) globbed = [] for p in all_paths: g = glob.glob(p) if not g: sys.exit(f"Error: Unhandled path or option: {p}") globbed.extend(g) all_paths = globbed if not all_paths and params.input.file_list is not None: all_paths.extend( [path.strip() for path in open(params.input.file_list).readlines()] ) if params.indexing.stills.known_orientations: known_orientations = {} for path in params.indexing.stills.known_orientations: for g in glob.glob(path): ko_expts = ExperimentList.from_file(g, check_format=False) for expt in ko_expts: assert ( len(expt.imageset.indices()) == 1 and len(expt.imageset.paths()) == 1 ) key = ( os.path.basename(expt.imageset.paths()[0]), expt.imageset.indices()[0], ) if key not in known_orientations: known_orientations[key] = [] known_orientations[key].append(expt.crystal) if not known_orientations: raise Sorry( "No known_orientations found at the locations specified: %s" % ", ".join(params.indexing.stills.known_orientations) ) params.indexing.stills.known_orientations = known_orientations if size > 1: if rank == 0: transmitted_info = params, options, all_paths else: transmitted_info = None params, options, all_paths = comm.bcast(transmitted_info, root=0) # Check we have some filenames if not all_paths: self.parser.print_help() return if params.mp.debug.cProfile: import cProfile self.pr = cProfile.Profile() self.pr.enable() print(f"Have {len(all_paths)} files") # Mask validation for mask_path in params.spotfinder.lookup.mask, params.integration.lookup.mask: if mask_path is not None and not os.path.isfile(mask_path): raise Sorry(f"Mask {mask_path} not found") # Save the options self.options = options self.params = params st = time.time() if params.mp.method == "mpi": # Configure the logging if params.output.logging_dir is None: logfile = None else: log_path = os.path.join( params.output.logging_dir, "log_rank%04d.out" % rank ) error_path = os.path.join( params.output.logging_dir, "error_rank%04d.out" % rank ) print(f"Redirecting stdout to {log_path}") print(f"Redirecting stderr to {error_path}") sys.stdout = open(log_path, "a") sys.stderr = open(error_path, "a") print("Should be redirected now") logfile = os.path.join( params.output.logging_dir, "info_rank%04d.out" % rank ) log.config(verbosity=options.verbose, logfile=logfile) else: # Configure logging log.config(verbosity=options.verbose, logfile="dials.process.log") bad_phils = [f for f in all_paths if os.path.splitext(f)[1] == ".phil"] if len(bad_phils) > 0: self.parser.print_help() logger.error( "Error: the following phil files were not understood: %s", ", ".join(bad_phils), ) return # Log the diff phil diff_phil = self.parser.diff_phil.as_str() if diff_phil != "": logger.info("The following parameters have been modified:\n") logger.info(diff_phil) for abs_params in self.params.integration.absorption_correction: if abs_params.apply: if not ( self.params.integration.debug.output and not self.params.integration.debug.separate_files ): raise Sorry( "Shoeboxes must be saved to integration intermediates to apply an absorption correction. " + "Set integration.debug.output=True, integration.debug.separate_files=False and " + "integration.debug.delete_shoeboxes=True to temporarily store shoeboxes." ) self.load_reference_geometry() from dials.command_line.dials_import import ManualGeometryUpdater update_geometry = ManualGeometryUpdater(params) # Import stuff logger.info("Loading files...") pre_import = params.dispatch.pre_import or len(all_paths) == 1 if pre_import: # Handle still imagesets by breaking them apart into multiple experiments # Further handle single file still imagesets (like HDF5) by tagging each # frame using its index experiments = ExperimentList() for path in sorted(all_paths): experiments.extend(do_import(path, load_models=False)) indices = [] basenames = [] basename_counts = {} split_experiments = [] for i, imageset in enumerate(experiments.imagesets()): assert len(imageset) == 1 paths = imageset.paths() indices.append(i) basename = os.path.splitext(os.path.basename(paths[0]))[0] basenames.append(basename) if basename in basename_counts: basename_counts[basename] += 1 else: basename_counts[basename] = 1 split_experiments.append(experiments[i : i + 1]) tags = [] split_experiments2 = [] for i, basename in zip(indices, basenames): if basename_counts[basename] > 1: tag = "%s_%05d" % (basename, i) else: tag = basename if ( not self.params.input.image_tag or tag in self.params.input.image_tag ): tags.append(tag) split_experiments2.append(split_experiments[i]) split_experiments = split_experiments2 # Wrapper function def do_work(i, item_list, processor=None, finalize=True): if not processor: processor = Processor( copy.deepcopy(params), composite_tag="%04d" % i, rank=i ) for item in item_list: tag = item[0] experiments = split_experiments[item[1]] try: assert len(experiments) == 1 experiment = experiments[0] experiment.load_models() imageset = experiment.imageset update_geometry(imageset) experiment.beam = imageset.get_beam() experiment.detector = imageset.get_detector() except RuntimeError as e: logger.warning("Error updating geometry on item %s, %s", tag, e) continue if self.reference_detector is not None: experiment = experiments[0] if self.params.input.sync_reference_geom: imageset = experiment.imageset sync_geometry( self.reference_detector.hierarchy(), imageset.get_detector().hierarchy(), ) experiment.detector = imageset.get_detector() else: experiment.detector = copy.deepcopy(self.reference_detector) processor.process_experiments(tag, experiments) imageset.clear_cache() if finalize: processor.finalize() return processor iterable = list(zip(tags, range(len(split_experiments)))) else: basenames = collections.defaultdict(int) sorted_paths = sorted(all_paths) for filename in sorted_paths: basename = os.path.splitext(os.path.basename(filename))[0] basenames[basename] += 1 tags = [] all_paths2 = [] for i, (basename, count) in enumerate(basenames.items()): if count > 1: tag = "%s_%05d" % (basename, i) else: tag = basename if ( not self.params.input.image_tag or tag in self.params.input.image_tag ): tags.append(tag) all_paths2.append(sorted_paths[i]) all_paths = all_paths2 # Wrapper function def do_work(i, item_list, processor=None, finalize=True): if not processor: processor = Processor( copy.deepcopy(params), composite_tag="%04d" % i, rank=i ) for item in item_list: tag, filename = item experiments = do_import(filename, load_models=True) imagesets = experiments.imagesets() if len(imagesets) == 0 or len(imagesets[0]) == 0: logger.info("Zero length imageset in file: %s", filename) return if len(imagesets) > 1: raise Abort(f"Found more than one imageset in file: {filename}") if len(imagesets[0]) > 1: raise Abort( "Found a multi-image file. Run again with pre_import=True" ) try: update_geometry(imagesets[0]) experiment = experiments[0] experiment.beam = imagesets[0].get_beam() experiment.detector = imagesets[0].get_detector() except RuntimeError as e: logger.warning("Error updating geometry on item %s, %s", tag, e) continue if self.reference_detector is not None: if self.params.input.sync_reference_geom: imageset = experiments[0].imageset sync_geometry( self.reference_detector.hierarchy(), imageset.get_detector().hierarchy(), ) experiments[0].detector = imageset.get_detector() else: experiments[0].detector = copy.deepcopy( self.reference_detector ) processor.process_experiments(tag, experiments) if finalize: processor.finalize() return processor iterable = list(zip(tags, all_paths)) if params.input.max_images: iterable = iterable[: params.input.max_images] if params.input.show_image_tags: print("Showing image tags for this dataset and exiting") for tag, item in iterable: print(tag) return # prepare fractions of process_percent, if given process_fractions = None if params.dispatch.process_percent: import fractions percent = params.dispatch.process_percent / 100 process_fractions = fractions.Fraction(percent).limit_denominator(100) def process_this_event(nevent): # nevent modulo the denominator gives us which fraction we're in n_mod_denom = nevent % process_fractions.denominator # compare the 0-indexed modulo against the 1-indexed numerator (intentionally not <=) n_accept = n_mod_denom < process_fractions.numerator return n_accept # Process the data if params.mp.method == "mpi": if size <= 2: # client/server only makes sense for n>2 subset = [ item for i, item in enumerate(iterable) if (i + rank) % size == 0 ] do_work(rank, subset) else: processor = Processor( copy.deepcopy(params), composite_tag="%04d" % rank, rank=rank ) if rank == 0: # server process for item_num, item in enumerate(iterable): if process_fractions and not process_this_event(item_num): continue print("Getting next available process") rankreq = comm.recv(source=MPI.ANY_SOURCE) print(f"Process {rankreq} is ready, sending {item[0]}\n") comm.send(item, dest=rankreq) # send a stop command to each process print("MPI DONE, sending stops\n") for rankreq in range(size - 1): rankreq = comm.recv(source=MPI.ANY_SOURCE) print("Sending stop to %d\n" % rankreq) comm.send("endrun", dest=rankreq) print("All stops sent.") else: # client process while True: # inform the server this process is ready for an event print("Rank %d getting next task" % rank) comm.send(rank, dest=0) print("Rank %d waiting for response" % rank) item = comm.recv(source=0) if item == "endrun": print("Rank %d received endrun" % rank) break print("Rank %d beginning processing" % rank) try: processor = do_work(rank, [item], processor, finalize=False) except Exception as e: print( "Rank %d unhandled exception processing event" % rank, str(e), ) print("Rank %d event processed" % rank) processor.finalize() else: from dxtbx.command_line.image_average import splitit if params.mp.nproc == 1: do_work(0, iterable) else: result = list( easy_mp.multi_core_run( myfunction=do_work, argstuples=list(enumerate(splitit(iterable, params.mp.nproc))), nproc=params.mp.nproc, ) ) error_list = [r[2] for r in result] if error_list.count(None) != len(error_list): print( "Some processes failed execution. Not all images may have processed. Error messages:" ) for error in error_list: if error is None: continue print(error) # Total Time logger.info("") logger.info("Total Time Taken = %f seconds", time.time() - st) if params.mp.debug.cProfile: self.pr.disable() self.pr.dump_stats( os.path.join( self.params.output.output_dir, "debug", "cpu_%d.prof" % comm.rank ) ) class Processor: def __init__(self, params, composite_tag=None, rank=0): self.params = params self.composite_tag = composite_tag # The convention is to put %s in the phil parameter to add a tag to # each output datafile. Save the initial templates here. self.experiments_filename_template = params.output.experiments_filename self.strong_filename_template = params.output.strong_filename self.indexed_filename_template = params.output.indexed_filename self.refined_experiments_filename_template = ( params.output.refined_experiments_filename ) self.integrated_filename_template = params.output.integrated_filename self.integrated_experiments_filename_template = ( params.output.integrated_experiments_filename ) if params.dispatch.coset: self.coset_filename_template = params.output.coset_filename self.coset_experiments_filename_template = ( params.output.coset_experiments_filename ) debug_dir = os.path.join(params.output.output_dir, "debug") if not os.path.exists(debug_dir): try: os.makedirs(debug_dir) except OSError: pass # due to multiprocessing, makedirs can sometimes fail assert os.path.exists(debug_dir) self.debug_file_path = os.path.join(debug_dir, "debug_%d.txt" % rank) write_newline = os.path.exists(self.debug_file_path) if write_newline: # needed if the there was a crash self.debug_write("") if params.output.composite_output: assert composite_tag is not None self.all_imported_experiments = ExperimentList() self.all_strong_reflections = flex.reflection_table() self.all_indexed_experiments = ExperimentList() self.all_indexed_reflections = flex.reflection_table() self.all_integrated_experiments = ExperimentList() self.all_integrated_reflections = flex.reflection_table() self.all_int_pickle_filenames = [] self.all_int_pickles = [] self.all_coset_experiments = ExperimentList() self.all_coset_reflections = flex.reflection_table() self.setup_filenames(composite_tag) def setup_filenames(self, tag): # before processing, set output paths according to the templates if ( self.experiments_filename_template is not None and "%s" in self.experiments_filename_template ): self.params.output.experiments_filename = os.path.join( self.params.output.output_dir, self.experiments_filename_template % ("idx-" + tag), ) if ( self.strong_filename_template is not None and "%s" in self.strong_filename_template ): self.params.output.strong_filename = os.path.join( self.params.output.output_dir, self.strong_filename_template % ("idx-" + tag), ) if ( self.indexed_filename_template is not None and "%s" in self.indexed_filename_template ): self.params.output.indexed_filename = os.path.join( self.params.output.output_dir, self.indexed_filename_template % ("idx-" + tag), ) if ( self.refined_experiments_filename_template is not None and "%s" in self.refined_experiments_filename_template ): self.params.output.refined_experiments_filename = os.path.join( self.params.output.output_dir, self.refined_experiments_filename_template % ("idx-" + tag), ) if ( self.integrated_filename_template is not None and "%s" in self.integrated_filename_template ): self.params.output.integrated_filename = os.path.join( self.params.output.output_dir, self.integrated_filename_template % ("idx-" + tag), ) if ( self.integrated_experiments_filename_template is not None and "%s" in self.integrated_experiments_filename_template ): self.params.output.integrated_experiments_filename = os.path.join( self.params.output.output_dir, self.integrated_experiments_filename_template % ("idx-" + tag), ) if ( self.params.dispatch.coset and self.coset_filename_template is not None and "%s" in self.coset_filename_template ): self.params.output.coset_filename = os.path.join( self.params.output.output_dir, self.coset_filename_template % ("idx-" + tag, self.params.integration.coset.transformation), ) if ( self.params.dispatch.coset and self.coset_experiments_filename_template is not None and "%s" in self.coset_experiments_filename_template ): self.params.output.coset_experiments_filename = os.path.join( self.params.output.output_dir, self.coset_experiments_filename_template % ("idx-" + tag, self.params.integration.coset.transformation), ) def debug_start(self, tag): if not self.params.mp.debug.output_debug_logs: return import socket self.debug_str = f"{socket.gethostname()},{tag}" self.debug_str += ",%s,%s,%s\n" self.debug_write("start") def debug_write(self, string, state=None): if not self.params.mp.debug.output_debug_logs: return from xfel.cxi.cspad_ana import cspad_tbx # XXX move to common timestamp format ts = cspad_tbx.evt_timestamp() # Now debug_file_handle = open(self.debug_file_path, "a") if string == "": debug_file_handle.write("\n") else: if state is None: state = " " debug_file_handle.write(self.debug_str % (ts, state, string)) debug_file_handle.close() def process_experiments(self, tag, experiments): if not self.params.output.composite_output: self.setup_filenames(tag) self.tag = tag self.debug_start(tag) if self.params.output.experiments_filename: if self.params.output.composite_output: self.all_imported_experiments.extend(experiments) else: experiments.as_json(self.params.output.experiments_filename) # Do the processing try: self.pre_process(experiments) except Exception as e: print("Error in pre-process", tag, str(e)) self.debug_write("preprocess_exception", "fail") if not self.params.dispatch.squash_errors: raise return try: if self.params.dispatch.find_spots: self.debug_write("spotfind_start") observed = self.find_spots(experiments) else: print("Spot Finding turned off. Exiting") self.debug_write("data_loaded", "done") return except Exception as e: print("Error spotfinding", tag, str(e)) self.debug_write("spotfinding_exception", "fail") if not self.params.dispatch.squash_errors: raise return try: if self.params.dispatch.index: if ( self.params.dispatch.hit_finder.enable and len(observed) < self.params.dispatch.hit_finder.minimum_number_of_reflections ): print("Not enough spots to index", tag) self.debug_write(f"not_enough_spots_{len(observed)}", "stop") return if ( self.params.dispatch.hit_finder.maximum_number_of_reflections is not None ): if ( self.params.dispatch.hit_finder.enable and len(observed) > self.params.dispatch.hit_finder.maximum_number_of_reflections ): print("Too many spots to index - Possibly junk", tag) self.debug_write(f"too_many_spots_{len(observed)}", "stop") return self.debug_write("index_start") experiments, indexed = self.index(experiments, observed) else: print("Indexing turned off. Exiting") self.debug_write(f"spotfinding_ok_{len(observed)}", "done") return except Exception as e: print("Couldn't index", tag, str(e)) if not self.params.dispatch.squash_errors: raise self.debug_write(f"indexing_failed_{len(observed)}", "stop") return self.debug_write("refine_start") try: experiments, indexed = self.refine(experiments, indexed) except Exception as e: print("Error refining", tag, str(e)) self.debug_write(f"refine_failed_{len(indexed)}", "fail") if not self.params.dispatch.squash_errors: raise return try: if self.params.dispatch.integrate: self.debug_write("integrate_start") integrated = self.integrate(experiments, indexed) else: print("Integration turned off. Exiting") self.debug_write(f"index_ok_{len(indexed)}", "done") return except Exception as e: print("Error integrating", tag, str(e)) self.debug_write(f"integrate_failed_{len(indexed)}", "fail") if not self.params.dispatch.squash_errors: raise return self.debug_write(f"integrate_ok_{len(integrated)}", "done") def pre_process(self, experiments): """Add any pre-processing steps here""" if ( self.params.indexing.stills.known_orientations and self.params.indexing.stills.require_known_orientation ): for expt in experiments: assert ( len(expt.imageset.indices()) == 1 and len(expt.imageset.paths()) == 1 ) key = ( os.path.basename(expt.imageset.paths()[0]), expt.imageset.indices()[0], ) if key not in self.params.indexing.stills.known_orientations: raise RuntimeError("Image not found in set of known orientations") if not self.params.input.ignore_gain_mismatch: g1 = self.params.spotfinder.threshold.dispersion.gain g2 = self.params.integration.summation.detector_gain gain = g1 if g1 is not None else g2 if gain is not None and gain != 1.0: for detector in experiments.detectors(): for panel in detector: if panel.get_gain() != 1.0 and panel.get_gain() != gain: raise RuntimeError( """ The detector is reporting a gain of %f but you have also supplied a gain of %f. Since the detector gain is not 1.0, your supplied gain will be multiplicatively applied in addition to the detector's gain, which is unlikely to be correct. Please re-run, removing spotfinder.dispersion.gain and integration.summation.detector_gain from your parameters. You can override this exception by setting input.ignore_gain_mismatch=True.""" % (panel.get_gain(), gain) ) def find_spots(self, experiments): st = time.time() logger.info("*" * 80) logger.info("Finding Strong Spots") logger.info("*" * 80) # Find the strong spots observed = flex.reflection_table.from_observations( experiments, self.params, is_stills=True ) # Reset z coordinates for dials.image_viewer; see Issues #226 for details xyzobs = observed["xyzobs.px.value"] for i in range(len(xyzobs)): xyzobs[i] = (xyzobs[i][0], xyzobs[i][1], 0) bbox = observed["bbox"] for i in range(len(bbox)): bbox[i] = (bbox[i][0], bbox[i][1], bbox[i][2], bbox[i][3], 0, 1) if self.params.output.composite_output: n = len(self.all_strong_reflections.experiment_identifiers()) for i, experiment in enumerate(experiments): refls = observed.select(observed["id"] == i) refls["id"] = flex.int(len(refls), n) del refls.experiment_identifiers()[i] refls.experiment_identifiers()[n] = experiment.identifier self.all_strong_reflections.extend(refls) n += 1 else: # Save the reflections to file logger.info("\n" + "-" * 80) if self.params.output.strong_filename: self.save_reflections(observed, self.params.output.strong_filename) logger.info("") logger.info("Time Taken = %f seconds", time.time() - st) return observed def index(self, experiments, reflections): from dials.algorithms.indexing.indexer import Indexer st = time.time() logger.info("*" * 80) logger.info("Indexing Strong Spots") logger.info("*" * 80) params = copy.deepcopy(self.params) # don't do scan-varying refinement during indexing params.refinement.parameterisation.scan_varying = False if hasattr(self, "known_crystal_models"): known_crystal_models = self.known_crystal_models elif self.params.indexing.stills.known_orientations: known_crystal_models = [] extended_experiments = ExperimentList() for expt in experiments: assert ( len(expt.imageset.indices()) == 1 and len(expt.imageset.paths()) == 1 ) key = ( os.path.basename(expt.imageset.paths()[0]), expt.imageset.indices()[0], ) if key not in self.params.indexing.stills.known_orientations: if self.params.indexing.stills.require_known_orientation: raise RuntimeError( "Image not found in set of known orientations" ) else: oris = [None] else: oris = self.params.indexing.stills.known_orientations[key] known_crystal_models.extend(oris) extended_experiments.extend(ExperimentList([expt] * len(oris))) experiments = extended_experiments else: known_crystal_models = None indexing_succeeded = False if known_crystal_models: try: idxr = Indexer.from_parameters( reflections, experiments, known_crystal_models=known_crystal_models, params=params, ) idxr.index() logger.info("indexed from known orientation") indexing_succeeded = True except Exception: if self.params.indexing.stills.require_known_orientation: raise if params.indexing.stills.method_list is None and not indexing_succeeded: idxr = Indexer.from_parameters( reflections, experiments, params=params, ) idxr.index() elif not indexing_succeeded: indexing_error = None for method in params.indexing.stills.method_list: params.indexing.method = method try: idxr = Indexer.from_parameters( reflections, experiments, params=params ) idxr.index() except Exception as e: logger.info("Couldn't index using method %s", method) if indexing_error is None: if e is None: e = Exception(f"Couldn't index using method {method}") indexing_error = e else: indexing_error = None break if indexing_error is not None: raise indexing_error indexed = idxr.refined_reflections experiments = idxr.refined_experiments if known_crystal_models is not None: filtered_sel = flex.bool(len(indexed), True) for expt_id in range(len(experiments)): for idx in set( indexed["miller_index"].select(indexed["id"] == expt_id) ): sel = (indexed["miller_index"] == idx) & (indexed["id"] == expt_id) if sel.count(True) > 1: filtered_sel = filtered_sel & ~sel filtered = indexed.select(filtered_sel) logger.info( "Filtered duplicate reflections, %d out of %d remaining", len(filtered), len(indexed), ) print( "Filtered duplicate reflections, %d out of %d remaining" % (len(filtered), len(indexed)) ) indexed = filtered logger.info("") logger.info("Time Taken = %f seconds", time.time() - st) return experiments, indexed def refine(self, experiments, centroids): if self.params.dispatch.refine: from dials.algorithms.refinement import RefinerFactory st = time.time() logger.info("*" * 80) logger.info("Refining Model") logger.info("*" * 80) refiner = RefinerFactory.from_parameters_data_experiments( self.params, centroids, experiments ) refiner.run() experiments = refiner.get_experiments() predicted = refiner.predict_for_indexed() centroids["xyzcal.mm"] = predicted["xyzcal.mm"] centroids["entering"] = predicted["entering"] centroids = centroids.select(refiner.selection_used_for_refinement()) # Re-estimate mosaic estimates from dials.algorithms.indexing.nave_parameters import NaveParameters nv = NaveParameters( params=self.params, experiments=experiments, reflections=centroids, refinery=refiner, graph_verbose=False, ) nv() acceptance_flags_nv = nv.nv_acceptance_flags centroids = centroids.select(acceptance_flags_nv) if self.params.output.composite_output: if ( self.params.output.refined_experiments_filename or self.params.output.indexed_filename ): assert ( self.params.output.refined_experiments_filename is not None and self.params.output.indexed_filename is not None ) n = len(self.all_indexed_experiments) self.all_indexed_experiments.extend(experiments) for i, experiment in enumerate(experiments): refls = centroids.select(centroids["id"] == i) refls["id"] = flex.int(len(refls), n) del refls.experiment_identifiers()[i] refls.experiment_identifiers()[n] = experiment.identifier self.all_indexed_reflections.extend(refls) n += 1 else: # Dump experiments to disk if self.params.output.refined_experiments_filename: experiments.as_json(self.params.output.refined_experiments_filename) if self.params.output.indexed_filename: self.save_reflections(centroids, self.params.output.indexed_filename) if self.params.dispatch.refine: logger.info("") logger.info("Time Taken = %f seconds", time.time() - st) return experiments, centroids def integrate(self, experiments, indexed): st = time.time() logger.info("*" * 80) logger.info("Integrating Reflections") logger.info("*" * 80) indexed, _ = self.process_reference(indexed) if self.params.integration.integration_only_overrides.trusted_range: for detector in experiments.detectors(): for panel in detector: panel.set_trusted_range( self.params.integration.integration_only_overrides.trusted_range ) if self.params.dispatch.coset: from xfel.util.sublattice_helper import integrate_coset integrate_coset(self, experiments, indexed) # Get the integrator from the input parameters logger.info("Configuring integrator from input parameters") from dials.algorithms.integration.integrator import create_integrator from dials.algorithms.profile_model.factory import ProfileModelFactory # Compute the profile model # Predict the reflections # Match the predictions with the reference # Create the integrator experiments = ProfileModelFactory.create(self.params, experiments, indexed) new_experiments = ExperimentList() new_reflections = flex.reflection_table() for expt_id, expt in enumerate(experiments): if ( self.params.profile.gaussian_rs.parameters.sigma_b_cutoff is None or expt.profile.sigma_b() < self.params.profile.gaussian_rs.parameters.sigma_b_cutoff ): refls = indexed.select(indexed["id"] == expt_id) refls["id"] = flex.int(len(refls), len(new_experiments)) # refls.reset_ids() del refls.experiment_identifiers()[expt_id] refls.experiment_identifiers()[len(new_experiments)] = expt.identifier new_reflections.extend(refls) new_experiments.append(expt) else: logger.info( "Rejected expt %d with sigma_b %f" % (expt_id, expt.profile.sigma_b()) ) experiments = new_experiments indexed = new_reflections if len(experiments) == 0: raise RuntimeError("No experiments after filtering by sigma_b") logger.info("") logger.info("=" * 80) logger.info("") logger.info("Predicting reflections") logger.info("") predicted = flex.reflection_table.from_predictions_multi( experiments, dmin=self.params.prediction.d_min, dmax=self.params.prediction.d_max, margin=self.params.prediction.margin, force_static=self.params.prediction.force_static, ) predicted.match_with_reference(indexed) logger.info("") integrator = create_integrator(self.params, experiments, predicted) # Integrate the reflections integrated = integrator.integrate() # correct integrated intensities for absorption correction, if necessary for abs_params in self.params.integration.absorption_correction: if abs_params.apply: if abs_params.algorithm == "fuller_kapton": from dials.algorithms.integration.kapton_correction import ( multi_kapton_correction, ) elif abs_params.algorithm == "kapton_2019": from dials.algorithms.integration.kapton_2019_correction import ( multi_kapton_correction, ) experiments, integrated = multi_kapton_correction( experiments, integrated, abs_params.fuller_kapton, logger=logger )() if self.params.significance_filter.enable: from dials.algorithms.integration.stills_significance_filter import ( SignificanceFilter, ) sig_filter = SignificanceFilter(self.params) filtered_refls = sig_filter(experiments, integrated) accepted_expts = ExperimentList() accepted_refls = flex.reflection_table() logger.info( "Removed %d reflections out of %d when applying significance filter", len(integrated) - len(filtered_refls), len(integrated), ) for expt_id, expt in enumerate(experiments): refls = filtered_refls.select(filtered_refls["id"] == expt_id) if len(refls) > 0: accepted_expts.append(expt) refls["id"] = flex.int(len(refls), len(accepted_expts) - 1) accepted_refls.extend(refls) else: logger.info( "Removed experiment %d which has no reflections left after applying significance filter", expt_id, ) if len(accepted_refls) == 0: raise Sorry("No reflections left after applying significance filter") experiments = accepted_expts integrated = accepted_refls # Delete the shoeboxes used for intermediate calculations, if requested if self.params.integration.debug.delete_shoeboxes and "shoebox" in integrated: del integrated["shoebox"] if self.params.output.composite_output: if ( self.params.output.integrated_experiments_filename or self.params.output.integrated_filename ): assert ( self.params.output.integrated_experiments_filename is not None and self.params.output.integrated_filename is not None ) n = len(self.all_integrated_experiments) self.all_integrated_experiments.extend(experiments) for i, experiment in enumerate(experiments): refls = integrated.select(integrated["id"] == i) refls["id"] = flex.int(len(refls), n) del refls.experiment_identifiers()[i] refls.experiment_identifiers()[n] = experiment.identifier self.all_integrated_reflections.extend(refls) n += 1 else: # Dump experiments to disk if self.params.output.integrated_experiments_filename: experiments.as_json(self.params.output.integrated_experiments_filename) if self.params.output.integrated_filename: # Save the reflections self.save_reflections( integrated, self.params.output.integrated_filename ) self.write_integration_pickles(integrated, experiments) from dials.algorithms.indexing.stills_indexer import ( calc_2D_rmsd_and_displacements, ) rmsd_indexed, _ = calc_2D_rmsd_and_displacements(indexed) log_str = f"RMSD indexed (px): {rmsd_indexed:f}\n" for i in range(6): bright_integrated = integrated.select( ( integrated["intensity.sum.value"] / flex.sqrt(integrated["intensity.sum.variance"]) ) >= i ) if len(bright_integrated) > 0: rmsd_integrated, _ = calc_2D_rmsd_and_displacements(bright_integrated) else: rmsd_integrated = 0 log_str += ( "N reflections integrated at I/sigI >= %d: % 4d, RMSD (px): %f\n" % (i, len(bright_integrated), rmsd_integrated) ) for crystal_model in experiments.crystals(): if hasattr(crystal_model, "get_domain_size_ang"): log_str += ". Final ML model: domain size angstroms: {:f}, half mosaicity degrees: {:f}".format( crystal_model.get_domain_size_ang(), crystal_model.get_half_mosaicity_deg(), ) logger.info(log_str) logger.info("") logger.info("Time Taken = %f seconds", time.time() - st) return integrated def write_integration_pickles(self, integrated, experiments, callback=None): """ Write a serialized python dictionary with integrated intensities and other information suitible for use by cxi.merge or prime.postrefine. @param integrated Reflection table with integrated intensities @param experiments Experiment list. One integration pickle for each experiment will be created. @param callback Deriving classes can use callback to make further modifications to the dictionary before it is serialized. Callback should be a function with this signature: def functionname(params, outfile, frame), where params is the phil scope, outfile is the path to the pickle that will be saved, and frame is the python dictionary to be serialized. """ if not hasattr(self.params.output, "integration_pickle"): return if self.params.output.integration_pickle is not None: from xfel.command_line.frame_extractor import ConstructFrame # Split everything into separate experiments for pickling for e_number, experiment in enumerate(experiments): e_selection = integrated["id"] == e_number reflections = integrated.select(e_selection) frame = ConstructFrame(reflections, experiment).make_frame() frame["pixel_size"] = experiment.detector[0].get_pixel_size()[0] if not hasattr(self, "tag") or self.tag is None: try: # if the data was a file on disc, get the path event_timestamp = os.path.splitext( experiments[0].imageset.paths()[0] )[0] except NotImplementedError: # if the data is in memory only, check if the reader set a timestamp on the format object event_timestamp = ( experiment.imageset.reader().get_format(0).timestamp ) event_timestamp = os.path.basename(event_timestamp) if event_timestamp.find("shot-") == 0: event_timestamp = os.path.splitext(event_timestamp)[ 0 ] # micromanage the file name else: event_timestamp = self.tag if hasattr(self.params.output, "output_dir"): outfile = os.path.join( self.params.output.output_dir, self.params.output.integration_pickle % (e_number, event_timestamp), ) else: outfile = os.path.join( os.path.dirname(self.params.output.integration_pickle), self.params.output.integration_pickle % (e_number, event_timestamp), ) if callback is not None: callback(self.params, outfile, frame) if self.params.output.composite_output: self.all_int_pickle_filenames.append(os.path.basename(outfile)) self.all_int_pickles.append(frame) else: with open(outfile, "wb") as fh: pickle.dump(frame, fh, protocol=pickle.HIGHEST_PROTOCOL) def process_reference(self, reference): """Load the reference spots.""" if reference is None: return None, None st = time.time() assert "miller_index" in reference assert "id" in reference logger.info("Processing reference reflections") logger.info(" read %d strong spots", len(reference)) mask = reference.get_flags(reference.flags.indexed) rubbish = reference.select(~mask) if mask.count(False) > 0: reference.del_selected(~mask) logger.info(" removing %d unindexed reflections", mask.count(True)) if len(reference) == 0: raise Sorry( """ Invalid input for reference reflections. Expected > %d indexed spots, got %d """ % (0, len(reference)) ) mask = reference["miller_index"] == (0, 0, 0) if mask.count(True) > 0: rubbish.extend(reference.select(mask)) reference.del_selected(mask) logger.info(" removing %d reflections with hkl (0,0,0)", mask.count(True)) mask = reference["id"] < 0 if mask.count(True) > 0: raise Sorry( """ Invalid input for reference reflections. %d reference spots have an invalid experiment id """ % mask.count(True) ) logger.info(" using %d indexed reflections", len(reference)) logger.info(" found %d junk reflections", len(rubbish)) logger.info(" time taken: %g", time.time() - st) return reference, rubbish def save_reflections(self, reflections, filename): """Save the reflections to file.""" st = time.time() logger.info("Saving %d reflections to %s", len(reflections), filename) reflections.as_file(filename) logger.info(" time taken: %g", time.time() - st) def finalize(self): """Perform any final operations""" if self.params.output.composite_output: if self.params.mp.composite_stride is not None: assert self.params.mp.method == "mpi" stride = self.params.mp.composite_stride from mpi4py import MPI comm = MPI.COMM_WORLD rank = comm.Get_rank() # each process in MPI has a unique id, 0-indexed size = comm.Get_size() # size: number of processes running in this job comm.barrier() if rank % stride == 0: subranks = [rank + i for i in range(1, stride) if rank + i < size] for i in range(len(subranks)): logger.info("Rank %d waiting for sender", rank) ( sender, imported_experiments, strong_reflections, indexed_experiments, indexed_reflections, integrated_experiments, integrated_reflections, coset_experiments, coset_reflections, int_pickles, int_pickle_filenames, ) = comm.recv(source=MPI.ANY_SOURCE) logger.info("Rank %d received data from rank %d", rank, sender) def extend_with_bookkeeping( src_expts, src_refls, dest_expts, dest_refls ): n = len(dest_refls.experiment_identifiers()) src_refls["id"] += n idents = src_refls.experiment_identifiers() keys = idents.keys() values = idents.values() for key in keys: del idents[key] for i, key in enumerate(keys): idents[key + n] = values[i] dest_expts.extend(src_expts) dest_refls.extend(src_refls) if len(imported_experiments) > 0: extend_with_bookkeeping( imported_experiments, strong_reflections, self.all_imported_experiments, self.all_strong_reflections, ) if len(indexed_experiments) > 0: extend_with_bookkeeping( indexed_experiments, indexed_reflections, self.all_indexed_experiments, self.all_indexed_reflections, ) if len(integrated_experiments) > 0: extend_with_bookkeeping( integrated_experiments, integrated_reflections, self.all_integrated_experiments, self.all_integrated_reflections, ) if len(coset_experiments) > 0: extend_with_bookkeeping( coset_experiments, coset_reflections, self.all_coset_experiments, self.all_coset_reflections, ) self.all_int_pickles.extend(int_pickles) self.all_int_pickle_filenames.extend(int_pickle_filenames) else: destrank = (rank // stride) * stride logger.info( "Rank %d sending results to rank %d", rank, (rank // stride) * stride, ) comm.send( ( rank, self.all_imported_experiments, self.all_strong_reflections, self.all_indexed_experiments, self.all_indexed_reflections, self.all_integrated_experiments, self.all_integrated_reflections, self.all_coset_experiments, self.all_coset_reflections, self.all_int_pickles, self.all_int_pickle_filenames, ), dest=destrank, ) self.all_imported_experiments = ( self.all_strong_reflections ) = ( self.all_indexed_experiments ) = ( self.all_indexed_reflections ) = ( self.all_integrated_experiments ) = ( self.all_integrated_reflections ) = ( self.all_coset_experiments ) = ( self.all_coset_reflections ) = self.all_int_pickles = self.all_integrated_reflections = [] # Dump composite files to disk if ( len(self.all_imported_experiments) > 0 and self.params.output.experiments_filename ): self.all_imported_experiments.as_json( self.params.output.experiments_filename ) if ( len(self.all_strong_reflections) > 0 and self.params.output.strong_filename ): self.save_reflections( self.all_strong_reflections, self.params.output.strong_filename ) if ( len(self.all_indexed_experiments) > 0 and self.params.output.refined_experiments_filename ): self.all_indexed_experiments.as_json( self.params.output.refined_experiments_filename ) if ( len(self.all_indexed_reflections) > 0 and self.params.output.indexed_filename ): self.save_reflections( self.all_indexed_reflections, self.params.output.indexed_filename ) if ( len(self.all_integrated_experiments) > 0 and self.params.output.integrated_experiments_filename ): self.all_integrated_experiments.as_json( self.params.output.integrated_experiments_filename ) if ( len(self.all_integrated_reflections) > 0 and self.params.output.integrated_filename ): self.save_reflections( self.all_integrated_reflections, self.params.output.integrated_filename, ) if self.params.dispatch.coset: if ( len(self.all_coset_experiments) > 0 and self.params.output.coset_experiments_filename ): self.all_coset_experiments.as_json( self.params.output.coset_experiments_filename ) if ( len(self.all_coset_reflections) > 0 and self.params.output.coset_filename ): self.save_reflections( self.all_coset_reflections, self.params.output.coset_filename ) # Create a tar archive of the integration dictionary pickles if len(self.all_int_pickles) > 0 and self.params.output.integration_pickle: tar_template_integration_pickle = ( self.params.output.integration_pickle.replace("%d", "%s") ) outfile = ( os.path.join( self.params.output.output_dir, tar_template_integration_pickle % ("x", self.composite_tag), ) + ".tar" ) tar = tarfile.TarFile(outfile, "w") for i, (fname, d) in enumerate( zip(self.all_int_pickle_filenames, self.all_int_pickles) ): string = BytesIO(pickle.dumps(d, protocol=2)) info = tarfile.TarInfo(name=fname) info.size = string.getbuffer().nbytes info.mtime = time.time() tar.addfile(tarinfo=info, fileobj=string) tar.close() @dials.util.show_mail_handle_errors() def run(args=None): script = Script() script.run(args) if __name__ == "__main__": run()
bsd-3-clause
386c9e8ae801952b4b2639f854db7290
39.636216
428
0.534166
4.530372
false
false
false
false
dials/dials
src/dials/algorithms/scaling/model/components/scale_components.py
1
23435
""" Classes that define a component of a scaling model. Each class holds the parameters and relevant data, as a list of arrays, from which to calculate inverse scale factors and derivatives. The components are initialised without any data, which is added by setting the data dict. In order to update the internal data lists in order to calculate the scales and derivatives, the update_reflection_data method should be called, which can optionally be provided with selection arrays to split the data for blockwise/parallel calculations. The scaling algorithm makes use of the components in the following way. First, the data for all 'suitable' reflections are added to the components. Then, at different stages of the algorithm, selection lists are provided to select a subset of this data (e.g. a small subset to prepare the component for minimisation calculation, or a large subset for calculating the scales for all reflections). The selection lists typically come from the Ih_table datastructures so that the data in the components is split in the same way as the data in the Ih_table datastructure. """ from __future__ import annotations from scitbx import sparse from dials.array_family import flex from dials_scaling_ext import calculate_harmonic_tables_from_selections class ScaleComponentBase: """ Base scale component class. This defines an interface to access the parameters, the component of the inverse scale factor and it's derivatives with respect to the parameters. Scale components derived from the base class are designed to be instantiated by a ScalingModel class, by supplying an initial array of parameters and optionally the current estimated standard deviations. The relevant data from a reflection table is added later by a Scaler using the update_reflection_data method. This behaviour allows data to easily be added/changed after selecting subsets of the data. """ def __init__(self, initial_values, parameter_esds=None): """Set the initial parameter values, parameter esds and n_params.""" self._parameters = initial_values self._parameter_esds = parameter_esds self._n_params = len(self._parameters) self._var_cov = None self._n_refl = [] # store as a list, to allow holding of data in blocks self._parameter_restraints = None self._data = {} @property def data(self): """ Return a dictionary of reflection data relevant to the particular component. This is designed to be a dict of arrays which can be selected from when updating the component (i.e. selecting subsets). """ return self._data @data.setter def data(self, data): self._data = data @data.deleter def data(self): self._data = {} @property def parameter_restraints(self): """Restraint weights for the component parameters.""" return self._parameter_restraints @parameter_restraints.setter def parameter_restraints(self, restraints): assert restraints.size() == self.parameters.size() self._parameter_restraints = restraints @property def n_params(self): """Get the number of parameters of the component (read-only).""" return self._n_params @property def parameters(self): """Parameters of the component.""" return self._parameters @parameters.setter def parameters(self, new_parameters): assert len(new_parameters) == len( self._parameters ), f""" attempting to set a new set of parameters of different length than previous assignment: was {len(self._parameters)}, attempting {len(new_parameters)}""" self._parameters = new_parameters @property def free_parameters(self): return self._parameters @free_parameters.setter def free_parameters(self, parameters): self._parameters = parameters @property def parameter_esds(self): """Return the estimated standard deviations of the parameters.""" return self._parameter_esds @parameter_esds.setter def parameter_esds(self, esds): assert len(esds) == len(self.parameters) self._parameter_esds = esds @property def free_parameter_esds(self): """Return the estimated standard deviations of the parameters.""" return self._parameter_esds @free_parameter_esds.setter def free_parameter_esds(self, esds): assert len(esds) == len(self.free_parameters) self._parameter_esds = esds def calculate_restraints(self): """Calculate residual and gradient restraints for the component.""" return None def calculate_jacobian_restraints(self): """Calculate residual and jacobian restraints for the component.""" return None @property def var_cov_matrix(self): """Return the variance-covariance matrix of the parameters.""" return self._var_cov @var_cov_matrix.setter def var_cov_matrix(self, var_cov): self._var_cov = var_cov @property def n_refl(self): """Return a list of the number of reflections in each block.""" return self._n_refl def update_reflection_data(self, selection=None, block_selections=None): """ Update the internal data arrays. Use the data stored in self.data, optionally with a selection array or list of selections, to populate a list of internal arrays e.g n_refl, normalised_values etc. to allow scale and derivative calculations. If no selection arrays are provided, the internal arrays will be lists containing one array/value, depending on the data type needed for derivative and scale calculation. Args: selection: A flex.bool selection array to select a subset of the internal data. block_selections (list): A list of flex.size_t arrays to select subsets of the internal data. """ raise NotImplementedError() def calculate_scales_and_derivatives(self, block_id=0): """Calculate and return inverse scales and derivatives for a given block.""" raise NotImplementedError() def calculate_scales(self, block_id=0): """Calculate and return inverse scales for a given block.""" raise NotImplementedError() class SingleScaleFactor(ScaleComponentBase): """ A model component consisting of a single global scale parameter. The inverse scale factor for every reflection is the parameter value itself and the derivatives are therefore all 1.0. """ null_parameter_value = 1.0 def __init__(self, initial_values, parameter_esds=None): """Set the initial parameter values, parameter esds and n_params.""" assert ( len(initial_values) == 1 ), """ This model component can only hold a single parameter.""" super().__init__(initial_values, parameter_esds) @ScaleComponentBase.data.setter def data(self, data): """Set the data dict in the parent class.""" assert set(data.keys()) == {"id"}, set(data.keys()) self._data = data def update_reflection_data(self, selection=None, block_selections=None): """ Update the internal n_refl list. Use the data stored in self.data, optionally with a boolean selection array or list of flex.size_t index selections, to make a list of n_refl (of length 1 or len(block_selections)) by inspecting the size of the selection result, in order to allow scale and derivative calculations. Args: selection: Optional, a flex.bool selection array to select a subset of the internal data. block_selections (list): Optional, a list of flex.size_t arrays to select subsets of the internal data. """ data = self.data["id"] if selection: self._n_refl = [data.select(selection).size()] elif block_selections: self._n_refl = [data.select(sel).size() for sel in block_selections] else: self._n_refl = [data.size()] def calculate_scales_and_derivatives(self, block_id=0): """Calculate and return inverse scales and derivatives for a given block.""" scales = flex.double(self.n_refl[block_id], self._parameters[0]) derivatives = sparse.matrix(self.n_refl[block_id], 1) for i in range(self.n_refl[block_id]): derivatives[i, 0] = 1.0 return scales, derivatives def calculate_scales(self, block_id=0): """Calculate and return inverse scales for a given block.""" return flex.double(self.n_refl[block_id], self._parameters[0]) class SingleBScaleFactor(ScaleComponentBase): """ A model component for a single global B-factor parameter. The inverse scale factor for each reflection is given by S = exp(B/(2 * d^2)), the derivatives are S/(2 * d^2). """ null_parameter_value = 0.0 def __init__(self, initial_values, parameter_esds=None): """Set the initial parameter values, parameter esds and n_params.""" super().__init__(initial_values, parameter_esds) self._d_values = [] @property def d_values(self): """Return a list of arrays of d-values associated with this component.""" return self._d_values @ScaleComponentBase.data.setter def data(self, data): """Set the data dict in the parent class.""" assert set(data.keys()) == {"d"}, set(data.keys()) self._data = data def update_reflection_data(self, selection=None, block_selections=None): """ Update the internal n_refl and d_values lists. Use the data stored in self.data, optionally with a boolean selection array or list of flex.size_t index selections, to make a lists of n_refl and d_value arrays (of length 1 or len(block_selections)), in order to allow scale and derivative calculations. Args: selection: Optional, a flex.bool selection array to select a subset of the internal data. block_selections (list): Optional, a list of flex.size_t arrays to select subsets of the internal data. """ data = self.data["d"] if selection: self._d_values = [data.select(selection)] elif block_selections: self._d_values = [data.select(sel) for sel in block_selections] else: self._d_values = [data] self._n_refl = [dvalues.size() for dvalues in self._d_values] def calculate_scales_and_derivatives(self, block_id=0): """Calculate and return inverse scales and derivatives for a given block.""" d_squared = self._d_values[block_id] * self._d_values[block_id] scales = flex.exp( flex.double(self._n_refl[block_id], self._parameters[0]) / (2.0 * d_squared) ) derivatives = sparse.matrix(self._n_refl[block_id], 1) for i in range(self._n_refl[block_id]): derivatives[i, 0] = scales[i] / (2.0 * d_squared[i]) return scales, derivatives def calculate_scales(self, block_id=0): """Calculate and return inverse scales for a given block.""" scales = flex.exp( flex.double(self._n_refl[block_id], self._parameters[0]) / (2.0 * (self._d_values[block_id] * self._d_values[block_id])) ) return scales class LinearDoseDecay(ScaleComponentBase): """ A model component for a decay that depends linearly on dose (see Holton Acta D 2019 D75 113-122) For the dose dependent, the form is I = I0 exp(-ln(2) D/ Hd). Parameterise this as linear function of rotation with an overall factor to refine. T(r) = exp(Cr/d - i.e. a one parameter model with the overall 'dose' proportional factor C. """ null_parameter_value = 0.0 def __init__(self, initial_values, parameter_esds=None): """Set the initial parameter values, parameter esds and n_params.""" super().__init__(initial_values, parameter_esds) self._d_values = [] self._x = [] # rotation/time @property def d_values(self): """Return a list of arrays of d-values associated with this component.""" return self._d_values @ScaleComponentBase.data.setter def data(self, data): """Set the data dict in the parent class.""" assert set(data.keys()) == {"d", "x"}, set(data.keys()) self._data = data def update_reflection_data(self, selection=None, block_selections=None): """ Update the internal n_refl and d_values lists. Use the data stored in self.data, optionally with a boolean selection array or list of flex.size_t index selections, to make a lists of n_refl and d_value arrays (of length 1 or len(block_selections)), in order to allow scale and derivative calculations. Args: selection: Optional, a flex.bool selection array to select a subset of the internal data. block_selections (list): Optional, a list of flex.size_t arrays to select subsets of the internal data. """ d = self.data["d"] x = self.data["x"] if selection: self._d_values = [d.select(selection)] self._x = [x.select(selection)] elif block_selections: self._d_values = [d.select(sel) for sel in block_selections] self._x = [x.select(sel) for sel in block_selections] else: self._d_values = [d] self._x = [x] self._n_refl = [dvalues.size() for dvalues in self._d_values] def calculate_scales_and_derivatives(self, block_id=0): """Calculate and return inverse scales and derivatives for a given block.""" scales = flex.exp( self._parameters[0] * self._x[block_id] / self._d_values[block_id] ) derivatives = sparse.matrix(self._n_refl[block_id], 1) for i in range(self._n_refl[block_id]): derivatives[i, 0] = scales[i] * ( self._x[block_id][i] / self._d_values[block_id][i] ) return scales, derivatives def calculate_scales(self, block_id=0): """Calculate and return inverse scales for a given block.""" scales = flex.exp( self._parameters[0] * self._x[block_id] / self._d_values[block_id] ) return scales class QuadraticDoseDecay(LinearDoseDecay): """ A model component for a decay that depends quadratically on dose For the dose dependent, the form is I = I0 exp(-C D/ d^2). Parameterise this as linear function of rotation with an overall factor to refine. T(r) = exp(Cr/d^2) - i.e. a one parameter model with the overall 'dose' proportional factor C. """ def calculate_scales_and_derivatives(self, block_id=0): """Calculate and return inverse scales and derivatives for a given block.""" scales = flex.exp( self._parameters[0] * self._x[block_id] / (self._d_values[block_id] ** 2) ) derivatives = sparse.matrix(self._n_refl[block_id], 1) for i in range(self._n_refl[block_id]): derivatives[i, 0] = scales[i] * ( self._x[block_id][i] / (self._d_values[block_id][i] ** 2) ) return scales, derivatives def calculate_scales(self, block_id=0): """Calculate and return inverse scales for a given block.""" scales = flex.exp( self._parameters[0] * self._x[block_id] / (self._d_values[block_id] ** 2) ) return scales class SHScaleComponent(ScaleComponentBase): """ A model component for a spherical harmonic absorption correction. This component uses a set of spherical harmonic functions to define an absorption surface for the crystal. A matrix of spherical harmonic coefficients for the data is stored in self._harmonic_values and is used to calculate the scales and derivatives. The scale is given by S = 1 + (sum_l sum_m Clm * Ylm) where Clm are the model parameters and Ylm are the spherical harmonic coefficients, the derivatives are then simply the coefficients Ylm. """ null_parameter_value = 0.0 coefficients_list = None # shared class variable to reduce memory load def __init__(self, initial_values, parameter_esds=None): """Set the initial parameter values, parameter esds and n_params.""" super().__init__(initial_values, parameter_esds) self._harmonic_values = [] self._matrices = [] @property def harmonic_values(self): """Return the matrix of harmonic coefficients for the internal data.""" return self._harmonic_values @property def sph_harm_table(self): """Return the matrix of the full harmonic coefficient for a reflection table.""" return self._data["sph_harm_table"] @sph_harm_table.setter def sph_harm_table(self, sht): self._data["sph_harm_table"] = sht @ScaleComponentBase.data.setter def data(self, data): """Set the data dict in the parent class.""" try: assert set(data.keys()) == {"s1_lookup", "s0_lookup"}, set(data.keys()) self._mode = "memory" except AssertionError: assert set(data.keys()) == {"sph_harm_table"}, set(data.keys()) self._mode = "speed" # Note: only speedier for small datasets self._data = data def calculate_restraints(self): """Calculate residual and gradient restraints for the component.""" residual = self.parameter_restraints * self._parameters * self._parameters gradient = 2.0 * self.parameter_restraints * self._parameters return residual, gradient def calculate_jacobian_restraints(self): """Calculate residual and jacobian restraints for the component.""" jacobian = sparse.matrix(self.n_params, self.n_params) for i in range(self.n_params): jacobian[i, i] = 1.0 return self._parameters, jacobian, self._parameter_restraints def update_reflection_data(self, selection=None, block_selections=None): """ Update the internal n_refl and harmonic_values lists. Use the harmonic values matrix stored in self.data, optionally with a boolean selection array or list of flex.size_t index selections, to make lists of n_refl and harmonic_value arrays (of length 1 or len(block_selections)), in order to allow scale and derivative calculations. Args: selection: Optional, a flex.bool selection array to select a subset of the internal data. block_selections (list): Optional, a list of flex.size_t arrays to select subsets of the internal data. """ if self._mode == "speed": self._update_reflection_data_speedmode(selection, block_selections) elif self._mode == "memory": self._update_reflection_data_memorymode(selection, block_selections) else: raise ValueError def _update_reflection_data_memorymode(self, selection=None, block_selections=None): if len(self.coefficients_list) != self.n_params: self.coefficients_list = self.coefficients_list[0 : self.n_params] # modify only for this instance, only needs to be done once per instance. if selection: n0 = self.data["s0_lookup"].select(selection) n1 = self.data["s1_lookup"].select(selection) values, matrix = calculate_harmonic_tables_from_selections( n0, n1, self.coefficients_list ) self._harmonic_values = [values] self._matrices = [matrix] elif block_selections: self._harmonic_values = [] self._matrices = [] for sel in block_selections: n0 = self.data["s0_lookup"].select(sel) n1 = self.data["s1_lookup"].select(sel) values, matrix = calculate_harmonic_tables_from_selections( n0, n1, self.coefficients_list ) self._harmonic_values.append(values) self._matrices.append(matrix) else: n0 = self.data["s0_lookup"] n1 = self.data["s1_lookup"] values, matrix = calculate_harmonic_tables_from_selections( n0, n1, self.coefficients_list ) self._harmonic_values = [values] self._matrices = [matrix] self._n_refl = [val[0].size() for val in self._harmonic_values] def _update_reflection_data_speedmode(self, selection=None, block_selections=None): if selection: sel_sph_harm_table = self.data["sph_harm_table"].select_columns( selection.iselection() ) self._harmonic_values = [sel_sph_harm_table.transpose()] elif block_selections: self._harmonic_values = [] for sel in block_selections: block_sph_harm_table = self.data["sph_harm_table"].select_columns(sel) self._harmonic_values.append(block_sph_harm_table.transpose()) else: self._harmonic_values = [self.data["sph_harm_table"].transpose()] self._n_refl = [val.n_rows for val in self._harmonic_values] def calculate_scales(self, block_id=0): """Calculate and return inverse scales for a given block.""" if self._mode == "speed": return self._calculate_scales_and_derivatives_speedmode( block_id, derivatives=False ) elif self._mode == "memory": return self._calculate_scales_and_derivatives_memorymode( block_id, derivatives=False ) def calculate_scales_and_derivatives(self, block_id=0): """Calculate and return inverse scales and derivatives for a given block.""" if self._mode == "speed": return self._calculate_scales_and_derivatives_speedmode(block_id) elif self._mode == "memory": return self._calculate_scales_and_derivatives_memorymode(block_id) def _calculate_scales_and_derivatives_speedmode(self, block_id, derivatives=True): abs_scale = flex.double( self._harmonic_values[block_id].n_rows, 1.0 ) # Unity term for i, col in enumerate(self._harmonic_values[block_id].cols()): abs_scale += flex.double(col.as_dense_vector() * self._parameters[i]) if derivatives: return abs_scale, self._harmonic_values[block_id] return abs_scale def _calculate_scales_and_derivatives_memorymode(self, block_id, derivatives=True): abs_scale = flex.double( self._harmonic_values[block_id][0].size(), 1.0 ) # Unity term for i, arr in enumerate( self._harmonic_values[block_id] ): # iterate over a list of arrays abs_scale += arr * self._parameters[i] if derivatives: return abs_scale, self._matrices[block_id] return abs_scale
bsd-3-clause
342874cb9ccda9895592da2d584d2c75
38.855442
88
0.632174
4.069283
false
false
false
false
douban/dpark
dpark/utils/profile.py
1
1268
from __future__ import absolute_import from __future__ import print_function import socket import getpass import sys from cProfile import Profile from pstats import Stats from tempfile import NamedTemporaryFile from functools import wraps from datetime import datetime def profile(hostname=None, to_stdout=False): def print_stats(stats): stats.strip_dirs() stats.sort_stats('time', 'calls') stats.print_stats(20) stats.sort_stats('cumulative') stats.print_stats(20) def decorator(f): @wraps(f) def _(*args, **kwargs): prof = Profile() try: return prof.runcall(f, *args, **kwargs) finally: if to_stdout: stats = Stats(prof) print_stats(stats) else: with NamedTemporaryFile(prefix='dpark_profile_', delete=False) as fd: print('===\n', datetime.today(), getpass.getuser(), sys.argv[0], file=fd) stats = Stats(prof, stream=fd) print_stats(stats) return _ if hostname is None or socket.gethostname() == hostname: return decorator else: return lambda f: f
bsd-3-clause
7629686e7d14f4b70124c4ac29ec4419
28.488372
97
0.566246
4.464789
false
false
false
false
dials/dials
src/dials/command_line/find_spots_client.py
1
7823
from __future__ import annotations import http.client import json import os import select import socket as pysocket import sys import urllib.error import urllib.parse import urllib.request from multiprocessing.pool import ThreadPool import libtbx.phil from dxtbx.model.crystal import CrystalFactory from dxtbx.util import get_url_scheme from libtbx.introspection import number_of_processors from scitbx.array_family import flex import dials.util def work(host, port, filename, params): conn = http.client.HTTPConnection(host, port) path = filename for param in params: path += f";{param}" conn.request("GET", path) return conn.getresponse().read() def _nproc(): return number_of_processors(return_value_if_unknown=-1) def response_to_xml(d): if "n_spots_total" in d: response = f"""<image>{d['image']}</image> <spot_count>{d['n_spots_total']}</spot_count> <spot_count_no_ice>{d['n_spots_no_ice']}</spot_count_no_ice> <d_min>{d['estimated_d_min']:.2f}</d_min> <d_min_method_1>{d['d_min_distl_method_1']:.2f}</d_min_method_1> <d_min_method_2>{d['d_min_distl_method_2']:.2f}</d_min_method_2> <total_intensity>{d['total_intensity']:.0f}</total_intensity>""" else: assert "error" in d return f"<response>\n{d['error']}\n</response>" if "lattices" in d: for lattice in d["lattices"]: crystal = CrystalFactory.from_dict(lattice["crystal"]) response = "\n".join( [ response, "<unit_cell>%.6g %.6g %.6g %.6g %.6g %.6g</unit_cell>" % (crystal.get_unit_cell().parameters()), ] ) response = "\n".join( [ response, "<n_indexed>%i</n_indexed>" % d["n_indexed"], "<fraction_indexed>%.2f</fraction_indexed>" % d["fraction_indexed"], ] ) if "integrated_intensity" in d: response = "\n".join( [ response, "<integrated_intensity>%.0f</integrated_intensity>" % d["integrated_intensity"], ] ) return f"<response>\n{response}\n</response>" def work_all( host, port, filenames, params, plot=False, table=False, json_file=None, grid=None, nproc=None, ): if nproc is None: nproc = _nproc() with ThreadPool(processes=nproc) as pool: threads = {} for filename in filenames: threads[filename] = pool.apply_async(work, (host, port, filename, params)) results = [] for filename in filenames: response = threads[filename].get() d = json.loads(response) results.append(d) print(response_to_xml(d)) if json_file is not None: with open(json_file, "wb") as f: json.dump(results, f) if plot or table: from dials.algorithms.spot_finding.per_image_analysis import ( StatsMultiImage, plot_stats, ) estimated_d_min = flex.double() d_min_distl_method_1 = flex.double() d_min_distl_method_2 = flex.double() n_spots_total = flex.int() n_spots_no_ice = flex.int() total_intensity = flex.double() for d in results: estimated_d_min.append(d["estimated_d_min"]) d_min_distl_method_1.append(d["d_min_distl_method_1"]) d_min_distl_method_2.append(d["d_min_distl_method_2"]) n_spots_total.append(d["n_spots_total"]) n_spots_no_ice.append(d["n_spots_no_ice"]) total_intensity.append(d["total_intensity"]) stats = StatsMultiImage( n_spots_total=n_spots_total, n_spots_no_ice=n_spots_no_ice, n_spots_4A=None, total_intensity=total_intensity, estimated_d_min=estimated_d_min, d_min_distl_method_1=d_min_distl_method_1, d_min_distl_method_2=d_min_distl_method_2, noisiness_method_1=None, noisiness_method_2=None, ) if plot: plot_stats(stats) if table: print(stats) if grid is not None: from matplotlib import pyplot n_spots_no_ice.reshape(flex.grid(grid)) print(n_spots_no_ice.size()) pyplot.figure() pyplot.pcolormesh(n_spots_no_ice.as_numpy_array(), cmap=pyplot.cm.Reds) pyplot.savefig("spot_count.png") def stop(host, port, nproc): stopped = 0 for j in range(nproc): try: url_request = urllib.request.Request(f"http://{host}:{port}/Ctrl-C") socket = urllib.request.urlopen(url_request, None, 3) if socket.getcode() == "200": stopped = stopped + 1 else: print("socket returned code", socket.getcode()) except (pysocket.timeout, urllib.error.HTTPError) as e: print("error on stopping server:", e) except urllib.error.URLError as e: if e.reason.errno != 111: print("error on stopping server:", e) except pysocket.error: # Assuming this means the server killed itself before the reply left the send buffer. stopped = stopped + 1 except http.client.BadStatusLine: # Regular occurrence. Probably means the server stopped anyway. stopped = stopped + 1 return stopped phil_scope = libtbx.phil.parse( """\ nproc = Auto .type = int(value_min=1) host = localhost .type = str port = 1701 .type = int(value_min=1) plot = False .type = bool table = False .type = bool json = None .type = path grid = None .type = ints(size=2, value_min=1) """ ) @dials.util.show_mail_handle_errors() def run(args=None): mixed_args = args or sys.argv[1:] if os.name != "nt": r, w, x = select.select([sys.stdin], [], [], 0) if len(r) > 0: mixed_args.extend([l.strip() for rr in r for l in rr.readlines()]) filenames = [] args = [] for arg in mixed_args: if get_url_scheme(arg): # Make this look like a path. If you squint. And are looking away. filenames.append("/" + urllib.parse.quote(arg)) else: if os.path.isfile(arg): filenames.append(arg) else: args.append(arg) interp = phil_scope.command_line_argument_interpreter() params, unhandled = interp.process_and_fetch( args, custom_processor="collect_remaining" ) params = params.extract() if params.nproc is libtbx.Auto: nproc = None params.nproc = 1024 else: nproc = params.nproc if len(unhandled) and unhandled[0] == "stop": stopped = stop(params.host, params.port, params.nproc) print("Stopped %d findspots processes" % stopped) elif len(unhandled) and unhandled[0] == "ping": url = "http://%s:%i" % (params.host, params.port) try: _ = urllib.request.urlopen(url).read() print("Success") sys.exit(0) except Exception: print("Failure") sys.exit(1) else: if len(filenames) == 1: response = work(params.host, params.port, filenames[0], unhandled) print(response_to_xml(json.loads(response))) else: work_all( params.host, params.port, filenames, unhandled, plot=params.plot, table=params.table, json_file=params.json, grid=params.grid, nproc=nproc, ) if __name__ == "__main__": run()
bsd-3-clause
90f2bed13999919cf55880ab04c1caaa
28.299625
97
0.555797
3.547846
false
false
false
false
dials/dials
src/dials/command_line/rl_png.py
1
6628
# LIBTBX_PRE_DISPATCHER_INCLUDE_SH export PHENIX_GUI_ENVIRONMENT=1 # DIALS_ENABLE_COMMAND_LINE_COMPLETION from __future__ import annotations import logging import math import libtbx.phil from scitbx import matrix from scitbx.array_family import flex import dials.util from dials.algorithms.indexing.indexer import find_max_cell from dials.command_line.search_beam_position import run_dps from dials.util.reciprocal_lattice import Render3d try: import matplotlib except ImportError: exit() # To pass through the "make" step, for graphics-free HPC build # Offline backend matplotlib.use("Agg") logger = logging.getLogger("dials.command_line.rl_png") help_message = """ Generate a png of the strong spots from spotfinding in reciprocal space. Examples:: dials.rl_png imported.expt strong.refl dials.rl_png indexed.expt indexed.refl """ phil_scope = libtbx.phil.parse( """ include scope dials.util.reciprocal_lattice.phil_scope marker_size = 5 .type = int(value_min=1) basis_vector_search { n_solutions = 3 .type = int } plot { size_inches = 10,10 .type = floats(size=2, value_min=0) } """, process_includes=True, ) class ReciprocalLatticePng(Render3d): def __init__(self, settings=None): Render3d.__init__(self, settings=settings) self.viewer = PngScene(settings=self.settings) class PngScene: def __init__(self, settings): self.settings = settings self.rotation_axis = None self.beam_vector = None self.points = None self.colors = None self.palette = None def set_rotation_axis(self, axis): self.rotation_axis = axis def set_beam_vector(self, beam): self.beam_vector = beam def set_points(self, points): self.points = points def set_points_data(self, reflections): # we do not label reciprocal lattice points here pass def set_colors(self, colors): # convert whites to black (background is white) colors.set_selected((colors.norms() == math.sqrt(3)), (0, 0, 0)) self.colors = colors def set_palette(self, palette): self.palette = palette def set_reciprocal_lattice_vectors(self, *args, **kwargs): # we do not draw reciprocal lattice vectors at this time pass def set_reciprocal_crystal_vectors(self, *args, **kwargs): # we do not draw reciprocal crystal vectors at this time either pass def project_2d(self, n): d = self.points.dot(n.elems) p = d * flex.vec3_double(len(d), n.elems) points2d = self.points - p x = matrix.col((1, 0, 0)) if x.angle(n) == 0 or x.angle(-n) == 0: x = matrix.col((0, 1, 0)) x = (x - x.dot(n) * n).normalize() y = x.cross(n) # assert y.angle(x, deg=True) == 90 # assert y.angle(matrix.col(n), deg=True) == 90 px2d = points2d.dot(x) py2d = points2d.dot(y) return px2d, py2d def plot(self, filename, n=(1, 0, 0)): from matplotlib import pyplot n = matrix.col(n).normalize() x, y = self.project_2d(n) fig = pyplot.figure(figsize=self.settings.plot.size_inches) pyplot.scatter( x.as_numpy_array(), y.as_numpy_array(), marker="+", s=self.settings.marker_size, c=list(self.colors), ) pyplot.title("Plane normal: (%.2g, %.2g, %.2g)" % (n.elems)) fig.savefig(filename) pyplot.close() @dials.util.show_mail_handle_errors() def run(args=None): from dials.util import log from dials.util.options import ( ArgumentParser, reflections_and_experiments_from_files, ) usage = "dials.rl_png [options] experiments.json observations.refl" parser = ArgumentParser( usage=usage, phil=phil_scope, read_experiments=True, read_reflections=True, check_format=False, epilog=help_message, ) params, options = parser.parse_args(args) reflections, experiments = reflections_and_experiments_from_files( params.input.reflections, params.input.experiments ) if len(experiments) == 0 or len(reflections) == 0: parser.print_help() exit(0) # Configure the logging log.config(logfile="dials.rl_png.log") # Log the diff phil diff_phil = parser.diff_phil.as_str() if diff_phil != "": logger.info("The following parameters have been modified:\n") logger.info(diff_phil) reflections = reflections[0] f = ReciprocalLatticePng(settings=params) f.load_models(experiments, reflections) rotation_axis = matrix.col(experiments[0].goniometer.get_rotation_axis()) s0 = matrix.col(experiments[0].beam.get_s0()) e1 = rotation_axis.normalize() e2 = s0.normalize() e3 = e1.cross(e2).normalize() f.viewer.plot("rl_rotation_axis.png", n=e1.elems) f.viewer.plot("rl_beam_vector", n=e2.elems) f.viewer.plot("rl_e3.png", n=e3.elems) n_solutions = params.basis_vector_search.n_solutions if experiments.crystals().count(None) < len(experiments): for i, c in enumerate(experiments.crystals()): A = matrix.sqr(c.get_A()) direct_matrix = A.inverse() a = direct_matrix[:3] b = direct_matrix[3:6] c = direct_matrix[6:9] prefix = "" if len(experiments.crystals()) > 1: prefix = "%i_" % (i + 1) f.viewer.plot(f"rl_{prefix}a.png", n=a) f.viewer.plot(f"rl_{prefix}b.png", n=b) f.viewer.plot(f"rl_{prefix}c.png", n=c) elif n_solutions: if "imageset_id" not in reflections: reflections["imageset_id"] = reflections["id"] reflections.centroid_px_to_mm(experiments) reflections.map_centroids_to_reciprocal_space(experiments) if params.d_min is not None: d_spacings = 1 / reflections["rlp"].norms() sel = d_spacings > params.d_min reflections = reflections.select(sel) # derive a max_cell from mm spots max_cell = find_max_cell( reflections, max_cell_multiplier=1.3, step_size=45 ).max_cell result = run_dps(experiments[0], reflections, max_cell) if result: solutions = [matrix.col(v) for v in result["solutions"]] for i in range(min(n_solutions, len(solutions))): v = solutions[i] f.viewer.plot(f"rl_solution_{i + 1}.png", n=v.elems) if __name__ == "__main__": run()
bsd-3-clause
945a1be18c35e59dcc89810c93978eae
26.616667
77
0.610893
3.323972
false
false
false
false
dials/dials
src/dials/algorithms/indexing/assign_indices.py
1
6043
from __future__ import annotations from cctbx.array_family import flex import dials_algorithms_indexing_ext as ext from dials.algorithms.indexing import DialsIndexError class AssignIndicesStrategy: def __init__(self, d_min=None): self._d_min = d_min def __call__(self, reciprocal_lattice_vectors): raise NotImplementedError() class AssignIndicesGlobal(AssignIndicesStrategy): def __init__(self, tolerance=0.3): super().__init__() self._tolerance = tolerance def __call__(self, reflections, experiments, d_min=None): reciprocal_lattice_points = reflections["rlp"] reflections["miller_index"] = flex.miller_index(len(reflections), (0, 0, 0)) if d_min is not None: d_spacings = 1 / reciprocal_lattice_points.norms() inside_resolution_limit = d_spacings > d_min else: inside_resolution_limit = flex.bool(reciprocal_lattice_points.size(), True) sel = inside_resolution_limit & (reflections["id"] == -1) isel = sel.iselection() rlps = reciprocal_lattice_points.select(isel) refs = reflections.select(isel) phi = refs["xyzobs.mm.value"].parts()[2] UB_matrices = flex.mat3_double([cm.get_A() for cm in experiments.crystals()]) imgset_ids = reflections["imageset_id"].select(sel) for i_imgset, imgset in enumerate(experiments.imagesets()): sel_imgset = imgset_ids == i_imgset result = ext.AssignIndices( rlps.select(sel_imgset), phi.select(sel_imgset), UB_matrices, tolerance=self._tolerance, ) miller_indices = result.miller_indices() crystal_ids = result.crystal_ids() expt_ids = flex.int(crystal_ids.size(), -1) for i_cryst, cryst in enumerate(experiments.crystals()): sel_cryst = crystal_ids == i_cryst for i_expt in experiments.where(crystal=cryst, imageset=imgset): expt_ids.set_selected(sel_cryst, i_expt) if experiments[i_expt].identifier: reflections.experiment_identifiers()[i_expt] = experiments[ i_expt ].identifier reflections["miller_index"].set_selected( isel.select(sel_imgset), miller_indices ) reflections["id"].set_selected(isel.select(sel_imgset), expt_ids) reflections.set_flags( reflections["miller_index"] != (0, 0, 0), reflections.flags.indexed ) reflections["id"].set_selected(reflections["miller_index"] == (0, 0, 0), -1) class AssignIndicesLocal(AssignIndicesStrategy): def __init__( self, d_min=None, epsilon=0.05, delta=8, l_min=0.8, nearest_neighbours=20 ): super().__init__() self._epsilon = epsilon self._delta = delta self._l_min = l_min self._nearest_neighbours = nearest_neighbours def __call__(self, reflections, experiments, d_min=None): from libtbx.math_utils import nearest_integer as nint from scitbx import matrix reciprocal_lattice_points = reflections["rlp"] if "miller_index" not in reflections: reflections["miller_index"] = flex.miller_index(len(reflections)) if d_min is not None: d_spacings = 1 / reciprocal_lattice_points.norms() inside_resolution_limit = d_spacings > d_min else: inside_resolution_limit = flex.bool(reciprocal_lattice_points.size(), True) sel = inside_resolution_limit & (reflections["id"] == -1) isel = sel.iselection() rlps = reciprocal_lattice_points.select(isel) refs = reflections.select(isel) phi = refs["xyzobs.mm.value"].parts()[2] if len(rlps) <= self._nearest_neighbours: raise DialsIndexError( "index_assignment.local.nearest_neighbour must be smaller than the number of accepted reflections (%d)" % len(rlps) ) UB_matrices = flex.mat3_double([cm.get_A() for cm in experiments.crystals()]) result = ext.AssignIndicesLocal( rlps, phi, UB_matrices, epsilon=self._epsilon, delta=self._delta, l_min=self._l_min, nearest_neighbours=self._nearest_neighbours, ) miller_indices = result.miller_indices() crystal_ids = result.crystal_ids() hkl = miller_indices.as_vec3_double().iround() assert miller_indices.select(crystal_ids < 0).all_eq((0, 0, 0)) for i_cryst in set(crystal_ids): if i_cryst < 0: continue A = matrix.sqr(experiments[i_cryst].crystal.get_A()) A_inv = A.inverse() cryst_sel = crystal_ids == i_cryst rlp_sel = rlps.select(cryst_sel) hkl_sel = hkl.select(cryst_sel).as_vec3_double() d_sel = 1 / rlp_sel.norms() d_perm = flex.sort_permutation(d_sel, reverse=True) hf_0 = A_inv * rlp_sel[d_perm[0]] h_0 = matrix.col([nint(j) for j in hf_0.elems]) offset = h_0 - matrix.col(hkl_sel[d_perm[0]]) # print "offset:", offset.elems h = hkl_sel + flex.vec3_double(hkl_sel.size(), offset.elems) refs["miller_index"].set_selected( cryst_sel, flex.miller_index(list(h.iround())) ) refs["id"].set_selected(cryst_sel, i_cryst) crystal_ids.set_selected(crystal_ids < 0, -1) refs["id"] = crystal_ids refs["miller_index"].set_selected(crystal_ids < 0, (0, 0, 0)) reflections["miller_index"].set_selected(isel, refs["miller_index"]) reflections["id"].set_selected(isel, refs["id"]) reflections.set_flags( reflections["miller_index"] != (0, 0, 0), reflections.flags.indexed )
bsd-3-clause
c31ae4af13079c5fb87df0ce3e172651
37.490446
119
0.575542
3.525671
false
false
false
false
dials/dials
src/dials/algorithms/integration/image_integrator.py
1
11125
from __future__ import annotations import logging import platform from time import time import dials.algorithms.integration from dials.algorithms.integration.processor import job from dials.model.data import ImageVolume, MultiPanelImageVolume, make_image from dials.util import log from dials.util.log import rehandle_cached_records from dials.util.mp import multi_node_parallel_map from dials_algorithms_integration_integrator_ext import ReflectionManagerPerImage logger = logging.getLogger(__name__) class ProcessorImage: """Top level processor for per image processing.""" def __init__(self, experiments, reflections, params): """ Initialise the manager and the processor. The processor requires a manager class implementing the Manager interface. This class executes all the workers in separate threads and accumulates the results to expose to the user. :param params: The phil parameters """ # Create the processing manager self.manager = ManagerImage(experiments, reflections, params) @property def executor(self): """ Get the executor :return: The executor """ return self.manager.executor @executor.setter def executor(self, function): """ Set the executor :param function: The executor """ self.manager.executor = function def process(self): """ Do all the processing tasks. :return: The processing results """ start_time = time() self.manager.initialize() mp_method = self.manager.params.integration.mp.method mp_nproc = min(len(self.manager), self.manager.params.integration.mp.nproc) mp_njobs = self.manager.params.integration.mp.njobs if ( mp_nproc > 1 and platform.system() == "Windows" ): # platform.system() forks which is bad for MPI, so don't use it unless nproc > 1 logger.warning( "Multiprocessing is not available on windows. Setting nproc = 1\n" ) mp_nproc = 1 assert mp_nproc > 0, "Invalid number of processors" logger.info(self.manager.summary()) logger.info(" Using %s with %d parallel job(s)\n", mp_method, mp_nproc) if mp_nproc > 1: def process_output(result): rehandle_cached_records(result[1]) self.manager.accumulate(result[0]) result[0].reflections = None result[0].data = None def execute_task(task): log.config_simple_cached() result = task() handlers = logging.getLogger("dials").handlers assert len(handlers) == 1, "Invalid number of logging handlers" return result, handlers[0].records multi_node_parallel_map( func=execute_task, iterable=list(self.manager.tasks()), njobs=mp_njobs, nproc=mp_nproc, callback=process_output, cluster_method=mp_method, preserve_order=True, ) else: for task in self.manager.tasks(): self.manager.accumulate(task()) self.manager.finalize() end_time = time() self.manager.time.user_time = end_time - start_time result = self.manager.result() return result, self.manager.time class Task: """ A class to perform a null task. """ def __init__(self, index, frames, reflections, experiments, params, executor): """ Initialise the task :param index: The index of the processing job :param frames: The frames to process :param experiments: The list of experiments :param reflections: The list of reflections :param params The processing parameters :param executor: The executor class """ self.index = index self.frames = frames self.experiments = experiments self.reflections = reflections self.params = params self.executor = executor def __call__(self): """ Do the processing. :return: The processed data """ # Set the job index job.index = self.index # Get the start time start_time = time() # Check all reflections have same imageset and get it exp_id = list(set(self.reflections["id"])) imageset = self.experiments[exp_id[0]].imageset for i in exp_id[1:]: assert ( self.experiments[i].imageset == imageset ), "Task can only handle 1 imageset" # Get the sub imageset frame00, frame01 = self.frames try: frame10, frame11 = imageset.get_array_range() except Exception: frame10, frame11 = (0, len(imageset)) try: assert frame00 < frame01 assert frame10 < frame11 assert frame00 >= frame10 assert frame01 <= frame11 index0 = frame00 - frame10 index1 = index0 + (frame01 - frame00) assert index0 < index1 assert index0 >= 0 assert index1 <= len(imageset) imageset = imageset[index0:index1] except Exception: raise RuntimeError("Programmer Error: bad array range") try: frame0, frame1 = imageset.get_array_range() except Exception: frame0, frame1 = (0, len(imageset)) # Initialise the dataset image_volume = MultiPanelImageVolume() for panel in self.experiments[0].detector: image_volume.add( ImageVolume( frame0, frame1, panel.get_image_size()[1], panel.get_image_size()[0] ) ) # Read all the images into a block of data read_time = 0.0 for i in range(len(imageset)): st = time() image = imageset.get_corrected_data(i) mask = imageset.get_mask(i) if self.params.integration.lookup.mask is not None: assert len(mask) == len( self.params.lookup.mask ), "Mask/Image are incorrect size %d %d" % ( len(mask), len(self.params.integration.lookup.mask), ) mask = tuple( m1 & m2 for m1, m2 in zip(self.params.integration.lookup.mask, mask) ) image_volume.set_image(frame0 + i, make_image(image, mask)) read_time += time() - st del image del mask # Process the data st = time() data = self.executor.process(image_volume, self.experiments, self.reflections) process_time = time() - st # Set the result values return dials.algorithms.integration.Result( index=self.index, reflections=self.reflections, read_time=read_time, process_time=process_time, total_time=time() - start_time, extract_time=0, data=data, ) class ManagerImage: """ A class to manage processing book-keeping """ def __init__(self, experiments, reflections, params): """ Initialise the manager. :param experiments: The list of experiments :param reflections: The list of reflections :param params: The phil parameters """ # Initialise the callbacks self.executor = None # Save some data self.experiments = experiments self.reflections = reflections # Save some parameters self.params = params # Set the finalized flag to False self.finalized = False # Initialise the timing information self.time = dials.algorithms.integration.TimingInfo() def initialize(self): """ Initialise the processing """ # Get the start time start_time = time() # Ensure the reflections contain bounding boxes assert "bbox" in self.reflections, "Reflections have no bbox" # Split the reflections into partials self._split_reflections() # Create the reflection manager frames = self.experiments[0].scan.get_array_range() self.manager = ReflectionManagerPerImage(frames, self.reflections) # Set the initialization time self.time.initialize = time() - start_time def task(self, index): """ Get a task. """ return Task( index=index, frames=self.manager.frames(index), reflections=self.manager.split(index), experiments=self.experiments, params=self.params, executor=self.executor, ) def tasks(self): """ Iterate through the tasks. """ for i in range(len(self)): yield self.task(i) def accumulate(self, result): """ Accumulate the results. """ self.manager.accumulate(result.index, result.reflections) if result.data is not None: self.executor.accumulate(result.index, result.data) self.time.read += result.read_time self.time.process += result.process_time self.time.total += result.total_time def finalize(self): """ Finalize the processing and finish. """ # Get the start time start_time = time() # Check manager is finished assert self.manager.finished(), "Manager is not finished" # Update the time and finalized flag self.time.finalize = time() - start_time self.finalized = True def result(self): """ Return the result. :return: The result """ assert self.finalized, "Manager is not finalized" return self.reflections def finished(self): """ Return if all tasks have finished. :return: True/False all tasks have finished """ return self.finalized and self.manager.finished() def __len__(self): """ Return the number of tasks. :return: the number of tasks """ return len(self.manager) def summary(self): return "" def _split_reflections(self): """ Split the reflections into partials or over job boundaries """ # Optionally split the reflection table into partials, otherwise, # split over job boundaries num_full = len(self.reflections) self.reflections.split_partials() num_partial = len(self.reflections) assert num_partial >= num_full, "Invalid number of partials" if num_partial > num_full: logger.info( " Split %d reflections into %d partial reflections\n", num_full, num_partial, )
bsd-3-clause
0a29c5e8f681a8ca77ba4f002a001e62
29.902778
92
0.571506
4.457131
false
false
false
false
dials/dials
src/dials/algorithms/indexing/model_evaluation.py
1
12078
from __future__ import annotations import collections import copy import logging import math import libtbx from dxtbx.model import Crystal from scitbx import matrix from scitbx.array_family import flex import dials.util from dials.algorithms.indexing.compare_orientation_matrices import ( difference_rotation_matrix_axis_angle, ) from dials.algorithms.refinement import RefinerFactory from dials.util.log import LoggingContext logger = logging.getLogger(__name__) Result = collections.namedtuple( "Result", ( "model_likelihood", "crystal", "rmsds", "n_indexed", "fraction_indexed", "hkl_offset", ), ) def filter_doubled_cell(solutions): accepted_solutions = [] for i1, s1 in enumerate(solutions): doubled_cell = False for (m1, m2, m3) in ( (2, 1, 1), (1, 2, 1), (1, 1, 2), (2, 2, 1), (2, 1, 2), (1, 2, 2), (2, 2, 2), ): if doubled_cell: break a, b, c = (matrix.col(v) for v in s1.crystal.get_real_space_vectors()) new_cryst = Crystal( real_space_a=1 / m1 * a, real_space_b=1 / m2 * b, real_space_c=1 / m3 * c, space_group=s1.crystal.get_space_group(), ) new_unit_cell = new_cryst.get_unit_cell() for s2 in solutions: if s2 is s1: continue if new_unit_cell.is_similar_to( s2.crystal.get_unit_cell(), relative_length_tolerance=0.05 ): R, axis, angle, cb = difference_rotation_matrix_axis_angle( new_cryst, s2.crystal ) if (angle < 1) and (s1.n_indexed < (1.1 * s2.n_indexed)): doubled_cell = True break if not doubled_cell: accepted_solutions.append(s1) return accepted_solutions class ModelRank: def __init__(self): self.all_solutions = [] def append(self, item): self.all_solutions.append(item) def extend(self, items): self.all_solutions.extend(items) def best_model(self): raise NotImplementedError() def __str__(self): raise NotImplementedError() # Tracker for solutions based on code in rstbx/dps_core/basis_choice.py class ModelRankFilter(ModelRank): def __init__( self, check_doubled_cell=True, likelihood_cutoff=0.8, volume_cutoff=1.25, n_indexed_cutoff=0.9, ): super().__init__() self.check_doubled_cell = check_doubled_cell self.likelihood_cutoff = likelihood_cutoff self.volume_cutoff = volume_cutoff self.n_indexed_cutoff = n_indexed_cutoff self.filtered_solutions = [] def append(self, item): super().append(item) self.update_analysis() def extend(self, items): super().extend(items) self.update_analysis() def __len__(self): return len(self.filtered_solutions) def filter_by_likelihood(self, solutions): best_likelihood = max(s.model_likelihood for s in solutions) offset = 0 while (best_likelihood + offset) <= 0: offset += 1 return [ s for s in solutions if (s.model_likelihood + offset) >= (self.likelihood_cutoff * (best_likelihood + offset)) ] def filter_by_volume(self, solutions): # filter by volume - prefer solutions with a smaller unit cell min_volume = min(s.crystal.get_unit_cell().volume() for s in solutions) return [ s for s in solutions if s.crystal.get_unit_cell().volume() < (self.volume_cutoff * min_volume) ] def filter_by_n_indexed(self, solutions, n_indexed_cutoff=None): if n_indexed_cutoff is None: n_indexed_cutoff = self.n_indexed_cutoff # filter by number of indexed reflections - prefer solutions that # account for more of the diffracted spots max_n_indexed = max(s.n_indexed for s in solutions) return [s for s in solutions if s.n_indexed >= n_indexed_cutoff * max_n_indexed] def update_analysis(self): # pre-filter out solutions that only account for a very small # percentage of the indexed spots relative to the best one self.filtered_solutions = self.filter_by_n_indexed( self.all_solutions, n_indexed_cutoff=0.05 ) # 5 percent if self.check_doubled_cell: self.filtered_solutions = filter_doubled_cell(self.filtered_solutions) self.filtered_solutions = self.filter_by_likelihood(self.filtered_solutions) self.filtered_solutions = self.filter_by_volume(self.filtered_solutions) self.filtered_solutions = self.filter_by_n_indexed(self.filtered_solutions) return def best_model(self): self.best_filtered_liklihood = max( s.model_likelihood for s in self.filtered_solutions ) solutions = [ s for s in self.filtered_solutions if s.model_likelihood == self.best_filtered_liklihood ] return solutions[0] def __str__(self): rows = [] rows.append( ["unit_cell", "volume", "n_indexed", "fraction_indexed", "likelihood"] ) for i, s in enumerate(self.all_solutions): s = self.all_solutions[i] rows.append( [ format( s.crystal.get_unit_cell(), "{:.2f} {:.2f} {:.2f} {:.1f} {:.1f} {:.1f}", ), f"{s.crystal.get_unit_cell().volume():.0f}", str(s.n_indexed), f"{s.fraction_indexed * 100:.0f}", f"{s.model_likelihood:.2f}", ] ) return dials.util.tabulate(rows, headers="firstrow") class ModelRankWeighted(ModelRank): def __init__(self, power=2, volume_weight=1, n_indexed_weight=1, rmsd_weight=1): super().__init__() self.volume_weight = volume_weight self.n_indexed_weight = n_indexed_weight self.rmsd_weight = rmsd_weight self.power = power def __len__(self): return len(self.all_solutions) def score_by_volume(self, reverse=False): # smaller volume = better volumes = flex.double( s.crystal.get_unit_cell().volume() for s in self.all_solutions ) score = flex.log(volumes) / math.log(2) return self.volume_weight * (score - flex.min(score)) def score_by_rmsd_xy(self, reverse=False): # smaller rmsds = better rmsd_x, rmsd_y, rmsd_z = flex.vec3_double( s.rmsds for s in self.all_solutions ).parts() rmsd_xy = flex.sqrt(flex.pow2(rmsd_x) + flex.pow2(rmsd_y)) score = flex.log(rmsd_xy) / math.log(2) return self.rmsd_weight * (score - flex.min(score)) def score_by_fraction_indexed(self, reverse=False): # more indexed reflections = better fraction_indexed = flex.double(s.fraction_indexed for s in self.all_solutions) score = flex.log(fraction_indexed) / math.log(2) return self.n_indexed_weight * (-score + flex.max(score)) def best_model(self): scores = self.combined_scores() perm = flex.sort_permutation(scores) return self.all_solutions[perm[0]] def combined_scores(self): scores = sum( flex.pow(score.as_double(), self.power) for score in ( self.score_by_fraction_indexed(), self.score_by_volume(), self.score_by_rmsd_xy(), ) ) return scores def __str__(self): rows = [] rows.append( [ "unit_cell", "volume", "volume score", "#indexed", "% indexed", "% indexed score", "rmsd_xy", "rmsd_xy score", "overall score", ] ) score_by_fraction_indexed = self.score_by_fraction_indexed() score_by_volume = self.score_by_volume() score_by_rmsd_xy = self.score_by_rmsd_xy() combined_scores = self.combined_scores() perm = flex.sort_permutation(combined_scores) rmsd_x, rmsd_y, rmsd_z = flex.vec3_double( s.rmsds for s in self.all_solutions ).parts() rmsd_xy = flex.sqrt(flex.pow2(rmsd_x) + flex.pow2(rmsd_y)) for i in perm: s = self.all_solutions[i] rows.append( [ format( s.crystal.get_unit_cell(), "{:.2f} {:.2f} {:.2f} {:.1f} {:.1f} {:.1f}", ), f"{s.crystal.get_unit_cell().volume():.0f}", f"{score_by_volume[i]:.2f}", str(s.n_indexed), f"{s.fraction_indexed * 100:.0f}", f"{score_by_fraction_indexed[i]:.2f}", f"{rmsd_xy[i]:.2f}", f"{score_by_rmsd_xy[i]:.2f}", f"{combined_scores[i]:.2f}", ] ) return dials.util.tabulate(rows, headers="firstrow") class Strategy: def evaluate(self, experiments, reflections): raise NotImplementedError() class ModelEvaluation(Strategy): def __init__(self, refinement_params): self._params = copy.deepcopy(refinement_params) # override several parameters, mainly for speed self._params.refinement.parameterisation.auto_reduction.action = "fix" self._params.refinement.parameterisation.scan_varying = False self._params.refinement.refinery.max_iterations = 4 if self._params.refinement.reflections.reflections_per_degree is libtbx.Auto: self._params.refinement.reflections.reflections_per_degree = 20 else: self._params.refinement.reflections.reflections_per_degree = min( self._params.refinement.reflections.reflections_per_degree, 20 ) if self._params.refinement.reflections.outlier.block_width is libtbx.Auto: # auto block_width determination is potentially too expensive to do at # this stage: instead set separate_blocks=False and increase value # of tukey.iqr_multiplier to be more tolerant of outliers self._params.refinement.reflections.outlier.separate_blocks = False self._params.refinement.reflections.outlier.tukey.iqr_multiplier = ( 2 * self._params.refinement.reflections.outlier.tukey.iqr_multiplier ) def evaluate(self, experiments, reflections): with LoggingContext("dials.algorithms.refinement", level=logging.ERROR): indexed_reflections = reflections.select(reflections["id"] > -1) try: refiner = RefinerFactory.from_parameters_data_experiments( self._params, indexed_reflections, experiments ) refiner.run() except (RuntimeError, ValueError): return else: rmsds = refiner.rmsds() xy_rmsds = math.sqrt(rmsds[0] ** 2 + rmsds[1] ** 2) model_likelihood = 1.0 - xy_rmsds result = Result( model_likelihood=model_likelihood, crystal=experiments.crystals()[0], rmsds=rmsds, n_indexed=len(indexed_reflections), fraction_indexed=float(len(indexed_reflections)) / len(reflections), hkl_offset=(0, 0, 0), ) return result
bsd-3-clause
b6135dcaf113c6df2ac3641994879213
33.215297
88
0.548435
3.828209
false
false
false
false
dials/dials
src/dials/command_line/model_background.py
1
9235
"""Calculate a global background model and diagnostic images from integrated experiments. This background model may improve subsequent integration in some cases, such as in the presence of ice rings. The method is described in the publication https://doi.org/10.1107/S2052252517010259. Usage: dials.model_background integrated.expt dials.integrate integrated.expt refined.refl background.algorithm=gmodel gmodel.robust.algorithm=True gmodel.model=background.pickle """ from __future__ import annotations import logging import pickle import sys from libtbx.phil import parse import dials.util import dials.util.log logger = logging.getLogger("dials.command_line.model_background") # Set the phil scope phil_scope = parse( """ output { model = 'background.pickle' .type = str .help = "The output filename" log = 'dials.model_background.log' .type = str .help = "The log filename" mean_image_prefix = 'mean' .type = str .help = "The mean background image" variance_image_prefix = 'variance' .type = str .help = "The variance background image" dispersion_image_prefix = 'dispersion' .type = str .help = "The dispersion background image" mask_image_prefix = 'mask' .type = str .help = "The mask background image" min_image_prefix = 'min' .type = str .help = "The min background image" max_image_prefix = 'max' .type = str .help = "The max background image" model_image_prefix = 'model' .type = str .help = "The model background image" polar_model_image_prefix = 'polar' .type = str .help = "The polar model background image" } modeller { min_images = 10 .type = int(value_min=1) .help = "The minimum number of images per pixel" filter_type = *median mean .type = choice .help = "The filter to use on the polar transformed image" kernel_size = 50 .type = int(value_min=0) .help = "The kernel size for the median filter" niter = 100 .type = int(value_min=1) .help = "The number of iterations for filling holes" image_type = min *mean .type = choice .help = "Which image to use" } include scope dials.algorithms.integration.integrator.phil_scope include scope dials.algorithms.spot_prediction.reflection_predictor.phil_scope """, process_includes=True, ) class ImageGenerator: """ Generate diagnostic images """ def __init__(self, model): """ Init the model """ import matplotlib matplotlib.use("Agg") self.model = model def _save_plot( self, name, filename, extractor_fn, bounded=True, colorbar=True, vmax=None ): """ Save the image """ from matplotlib import pylab for i, model in enumerate(self.model): image = extractor_fn(model) pylab.figure(figsize=(6, 4)) if bounded and vmax is None: boundaries = { "vmin": 0, "vmax": sorted(image)[int(0.99 * len(image))], } elif bounded: boundaries = {"vmin": 0, "vmax": vmax} else: boundaries = {} pylab.imshow(image.as_numpy_array(), interpolation="none", **boundaries) ax1 = pylab.gca() ax1.get_xaxis().set_visible(False) ax1.get_yaxis().set_visible(False) if colorbar: cb = pylab.colorbar() cb.ax.tick_params(labelsize=8) logger.info( "Saving %s image for panel %d to %s_%d.png", name, i, filename, i ) pylab.savefig("%s_%d.png" % (filename, i), dpi=600, bbox_inches="tight") def save_min(self, filename): """ Save the min image """ self._save_plot("min", filename, lambda m: m.min_image) def save_max(self, filename): """ Save the max image """ self._save_plot("max", filename, lambda m: m.max_image) def save_mean(self, filename): """ Save the mean image """ self._save_plot("mean", filename, lambda m: m.mean) def save_variance(self, filename): """ Save the variance image """ self._save_plot("variance", filename, lambda m: m.variance) def save_dispersion(self, filename): """ Save the dispersion image """ self._save_plot("dispersion", filename, lambda m: m.dispersion, vmax=2) def save_mask(self, filename): """ Save the dispersion image """ self._save_plot( "mask", filename, lambda m: m.mask, bounded=False, colorbar=False ) def save_model(self, filename): """ Save the model image """ self._save_plot("model", filename, lambda m: m.model) def save_polar_model(self, filename): """ Save the polar model image """ self._save_plot("polar model", filename, lambda m: m.polar_model, bounded=False) class Script: """A class for running the script.""" def __init__(self): """Initialise the script.""" from dials.util.options import ArgumentParser usage = "dials.model_background [options] [param.phil] models.expt" # Initialise the base class self.parser = ArgumentParser( usage=usage, phil=phil_scope, epilog=__doc__, read_experiments=True ) def run(self, args=None): """Execute the script.""" from dials.algorithms.background.modeller import BackgroundModeller from dials.array_family import flex from dials.util.command_line import heading from dials.util.options import flatten_experiments # Parse the command line params, options = self.parser.parse_args(args, show_diff_phil=False) # Configure the logging dials.util.log.config(verbosity=options.verbose, logfile=params.output.log) if params.integration.mp.nproc != 1 or params.integration.mp.njobs != 1: # https://github.com/dials/dials/issues/1083 logger.warning( "Multiprocessing is currently disabled. " "Setting nproc = njobs = 1" ) params.integration.mp.nproc = 1 params.integration.mp.njobs = 1 from dials.util.version import dials_version logger.info(dials_version()) # Log the diff phil diff_phil = self.parser.diff_phil.as_str() if diff_phil != "": logger.info("The following parameters have been modified:\n") logger.info(diff_phil) # Ensure we have a data block experiments = flatten_experiments(params.input.experiments) if len(experiments) == 0: self.parser.print_help() return if any((e.profile is None for e in experiments)): sys.exit( "Experiments must contain a profile model (for example, after integration)" ) # Only handle a single imageset at once imagesets = {expr.imageset for expr in experiments} if len(imagesets) != 1: sys.exit("Can only process a single imageset at a time") # Predict the reflections logger.info("") logger.info("=" * 80) logger.info("") logger.info(heading("Predicting reflections")) logger.info("") predicted = flex.reflection_table.from_predictions_multi( experiments, dmin=params.prediction.d_min, dmax=params.prediction.d_max, margin=params.prediction.margin, force_static=params.prediction.force_static, ) # Create the modeller modeller = BackgroundModeller(experiments, predicted, params) model = modeller.compute() # Save the background model logger.info("Saving background model to %s", params.output.model) from dials.algorithms.background.gmodel import StaticBackgroundModel static_model = StaticBackgroundModel() for m in model: static_model.add(m.model) with open(params.output.model, "wb") as outfile: pickle.dump(static_model, outfile, protocol=pickle.HIGHEST_PROTOCOL) # Output some diagnostic images image_generator = ImageGenerator(model) image_generator.save_mean(params.output.mean_image_prefix) image_generator.save_variance(params.output.variance_image_prefix) image_generator.save_dispersion(params.output.dispersion_image_prefix) image_generator.save_mask(params.output.mask_image_prefix) image_generator.save_min(params.output.min_image_prefix) image_generator.save_max(params.output.max_image_prefix) image_generator.save_model(params.output.model_image_prefix) # image_generator.save_polar_model(params.output.polar_model_image_prefix) @dials.util.show_mail_handle_errors() def run(args=None): script = Script() script.run(args) if __name__ == "__main__": run()
bsd-3-clause
9c5cad30e9d6c0a50ece0d2178d8d4cf
29.081433
136
0.60314
3.949957
false
false
false
false
dials/dials
src/dials/util/slice.py
1
3419
from __future__ import annotations import copy from scitbx.array_family import flex def slice_experiments(experiments, image_ranges): """ :param experiments :type experiments: dxtbx.model.experiment_list.ExperimentList :param image_range: :type image_range: list of 2-tuples defining scan range for each experiment """ # copy the experiments experiments = copy.deepcopy(experiments) if len(experiments) != len(image_ranges): raise ValueError( "Input experiment list and image_ranges are not of the same length" ) for exp, sr in zip(experiments, image_ranges): if sr is None: continue im_range = exp.scan.get_image_range() if sr[0] < im_range[0] or sr[1] > im_range[1]: raise IndexError("requested slice outside current scan range") # slicing uses the array range, not the image range arr_start = exp.scan.get_array_range()[0] beg = sr[0] - 1 - arr_start end = sr[1] - arr_start exp.scan.swap(exp.scan[beg:end]) if exp.imageset is not None: # Gorilla of temporary workarounds for inconsistent scan and imageset slicing # https://github.com/cctbx/dxtbx/issues/213 offset = exp.scan.get_batch_offset() exp.imageset = exp.imageset[beg + offset : end + offset] # account for scan-varying crystal if exp.crystal and exp.crystal.num_scan_points > 0: exp.crystal = slice_crystal(exp.crystal, (beg, end)) return experiments def slice_reflections(reflections, image_ranges): """ :param reflections: reflection table of input reflections :type reflections: dials.array_family.flex.reflection_table :param image_range: list of 2-tuples defining scan range for each experiment id contained within the reflections :type image_range: list of 2-tuples defining scan range for each experiment """ # copy the reflections reflections = copy.deepcopy(reflections) to_keep = flex.size_t() for iexp, sr in enumerate(image_ranges): if sr is None: continue isel = (reflections["id"] == iexp).iselection() frames = (reflections["xyzobs.px.value"].parts()[2]).select(isel) # reflns on image n have frames in range [n-1, n) in_low_lim = frames >= sr[0] - 1 in_high_lim = frames < sr[1] in_lim = in_low_lim & in_high_lim # which indices to keep? sub_isel = isel.select(in_lim) to_keep.extend(sub_isel) # implicitly also removes any reflections with ID outside the range of the # length of image_ranges return reflections.select(to_keep) def slice_crystal(crystal, array_range): """Slice a scan-varying crystal by the provided array range""" if crystal.num_scan_points == 0: return crystal crystal = copy.deepcopy(crystal) UB_mats = [crystal.get_A_at_scan_point(i) for i in range(crystal.num_scan_points)] beg, end = array_range end += 1 # one more scan-point than images in the range UB_mats = UB_mats[beg:end] B_cov = crystal.get_B_covariance_at_scan_points() crystal.reset_scan_points() crystal.set_A_at_scan_points(UB_mats) if len(B_cov) > 0: B_cov = B_cov[beg:end, 0:9, 0:9] crystal.set_B_covariance_at_scan_points(B_cov) return crystal
bsd-3-clause
d62929153722235eb122cedaba756df0
33.19
89
0.638491
3.488776
false
false
false
false
dials/dials
tests/algorithms/integration/profile/test_empirical_modeller.py
1
5054
from __future__ import annotations import math from dials.algorithms.profile_model.modeller import EmpiricalProfileModeller def evaluate_gaussian(x, a, x0, sx): assert len(x) == len(x0) assert len(x) == len(sx) g = 0.0 for xi, x0i, sxi in zip(x, x0, sx): g += (xi - x0i) ** 2 / (2.0 * sxi**2) return a * math.exp(-g) def gaussian(size, a, x0, sx): from scitbx.array_family import flex result = flex.double(flex.grid(size)) index = [0] * len(size) while True: result[index] = evaluate_gaussian(index, a, x0, sx) for j in range(len(size)): index[j] += 1 if index[j] < size[j]: break index[j] = 0 if j == len(size) - 1: return result class Modeller(EmpiricalProfileModeller): def model(self, reflections, profiles): from dials.array_family import flex indices = flex.size_t(range(len(self))) weights = flex.double([1.0] * len(self)) for profile in profiles: self.add(indices, weights, profile) class Test: def setup_class(self): self.n = 9 self.grid_size = (9, 9, 9) self.threshold = 0.0 def test_with_identical_non_negative_profiles(self): from scitbx.array_family import flex # Generate identical non-negative profiles reflections, profiles, profile = self.generate_identical_non_negative_profiles() # Create the reference learner modeller = Modeller(self.n, self.grid_size, self.threshold) # Do the modelling modeller.model(reflections, profiles) modeller.finalize() # Normalize the profile profile = self.normalize_profile(profile) # Check that all the reference profiles are the same eps = 1e-10 for index in range(len(modeller)): reference = modeller.data(index) for k in range(self.grid_size[2]): for j in range(self.grid_size[1]): for i in range(self.grid_size[0]): assert abs(reference[k, j, i] - profile[k, j, i]) <= eps assert abs(flex.sum(reference) - 1.0) <= eps def test_with_systematically_offset_profiles(self): from scitbx.array_family import flex # Generate identical non-negative profiles reflections, profiles = self.generate_systematically_offset_profiles() # Create the reference learner modeller = Modeller(self.n, self.grid_size, self.threshold) # Do the modelling modeller.model(reflections, profiles) modeller.finalize() # Check that all the reference profiles are the same eps = 1e-10 profile = None for index in range(len(modeller)): reference = modeller.data(index) if profile is not None: for k in range(self.grid_size[2]): for j in range(self.grid_size[1]): for i in range(self.grid_size[0]): assert abs(reference[k, j, i] - profile[k, j, i]) <= eps else: profile = reference assert abs(flex.sum(reference) - 1.0) <= eps def normalize_profile(self, profile): from scitbx.array_family import flex max_profile = flex.max(profile) threshold = self.threshold * max_profile sum_profile = 0.0 for i in range(len(profile)): if profile[i] > threshold: sum_profile += profile[i] else: profile[i] = 0.0 result = flex.double(flex.grid(profile.all())) for i in range(len(profile)): result[i] = profile[i] / sum_profile return result def generate_identical_non_negative_profiles(self): from random import uniform from dials.array_family import flex rlist = flex.reflection_table(1000) profile = gaussian(self.grid_size, 1000, (4, 4, 4), (1.5, 1.5, 1.5)) xyz = flex.vec3_double(1000) profiles = [] for i in range(1000): x = uniform(0, 1000) y = uniform(0, 1000) z = uniform(0, 10) xyz[i] = (x, y, z) profiles.append(profile.deep_copy()) rlist["xyzcal.px"] = xyz return rlist, profiles, profile def generate_systematically_offset_profiles(self): from random import uniform from dials.array_family import flex rlist = flex.reflection_table(1000) xyz = flex.vec3_double(1000) profiles = [] for i in range(1000): x = uniform(0, 1000) y = uniform(0, 1000) z = uniform(0, 10) offset = -4.5 + 9 * x / 1000.0 profile = gaussian( self.grid_size, 1000, (4 + offset, 4, 4), (1.5, 1.5, 1.5) ) xyz[i] = (x, y, z) profiles.append(profile) rlist["xyzcal.px"] = xyz return rlist, profiles
bsd-3-clause
13f2d8c0e6f4b06db28924a61ed1f5f7
29.263473
88
0.557578
3.782934
false
false
false
false
feincms/feincms-elephantblog
tests/testapp/test_contents.py
1
3083
from django.template.loader import render_to_string from django.test import TestCase from elephantblog.contents import BlogCategoryListContent, BlogEntryListContent from .factories import EntryFactory, create_category, create_entries class Request: GET = {"page": 1} class ContentsTestCase(TestCase): def test_contents(self): entries = create_entries(EntryFactory) entries[0].richtextcontent_set.create( region="main", ordering=1, text="Hello world" ) entries[0].is_featured = True entries[0].save() category = create_category(title="Category 1") entries[1].categories.add(category) create_category(title="Empty category") BlogEntryListContent._meta.abstract = False # Hack to allow instantiation content = BlogEntryListContent() content.process(Request) html = render_to_string(*content.render()) self.assertIn( 'h2 class="entry-title"><a href="/multilang/2012/10/12/eintrag-1/', html, ) self.assertIn( 'h2 class="entry-title"><a href="/multilang/2012/08/12/entry-1/"', html, ) content.featured_only = True content.process(Request) html = render_to_string(*content.render()) self.assertEqual( html.count("<h2"), 1, ) content.category = category content.process(Request) html = render_to_string(*content.render()) self.assertEqual( html.count("<h2"), 0, ) content.featured_only = False content.process(Request) html = render_to_string(*content.render()) self.assertEqual( html.count("<h2"), 1, ) content = BlogEntryListContent() content.paginate_by = 1 content.process(Request) html = render_to_string(*content.render()) self.assertEqual( html.count("<h2"), 1, ) Request.GET["page"] = 2 content.process(Request) html = render_to_string(*content.render()) self.assertEqual( html.count("<h2"), 1, ) Request.GET["page"] = 3 content.process(Request) html = render_to_string(*content.render()) self.assertEqual( html.count("<h2"), 1, ) Request.GET["page"] = "abc" content.process(Request) html = render_to_string(*content.render()) self.assertEqual( html.count("<h2"), 1, ) BlogCategoryListContent._meta.abstract = False # Hack to allow instantiation content = BlogCategoryListContent() html = render_to_string(*content.render()) self.assertEqual( html.count("<li>"), 1, ) content.show_empty_categories = True html = render_to_string(*content.render()) self.assertEqual( html.count("<li>"), 2, )
bsd-3-clause
08590c79343193c5674d17609560a08b
27.027273
85
0.557249
4.264177
false
false
false
false
chromium/gyp
test/make_global_settings/ar/gyptest-make_global_settings_ar.py
9
3838
#!/usr/bin/env python # Copyright (c) 2014 Google Inc. All rights reserved. # Use of this source code is governed by a BSD-style license that can be # found in the LICENSE file. """ Verifies 'AR' in make_global_settings. """ import os import sys import TestGyp def resolve_path(test, path): if path is None: return None elif test.format == 'make': return '$(abspath %s)' % path elif test.format in ['ninja', 'xcode-ninja']: return os.path.join('..', '..', path) else: test.fail_test() def verify_ar_target(test, ar=None, rel_path=False): if rel_path: ar_expected = resolve_path(test, ar) else: ar_expected = ar # Resolve default values if ar_expected is None: if test.format == 'make': # Make generator hasn't set the default value for AR. # You can remove the following assertion as long as it doesn't # break existing projects. test.must_not_contain('Makefile', 'AR ?= ') return elif test.format in ['ninja', 'xcode-ninja']: if sys.platform == 'win32': ar_expected = 'lib.exe' else: ar_expected = 'ar' if test.format == 'make': test.must_contain('Makefile', 'AR ?= %s' % ar_expected) elif test.format in ['ninja', 'xcode-ninja']: test.must_contain('out/Default/build.ninja', 'ar = %s' % ar_expected) else: test.fail_test() def verify_ar_host(test, ar=None, rel_path=False): if rel_path: ar_expected = resolve_path(test, ar) else: ar_expected = ar # Resolve default values if ar_expected is None: if sys.platform == 'win32': ar_expected = 'lib.exe' else: ar_expected = 'ar' if test.format == 'make': test.must_contain('Makefile', 'AR.host ?= %s' % ar_expected) elif test.format in ['ninja', 'xcode-ninja']: test.must_contain('out/Default/build.ninja', 'ar_host = %s' % ar_expected) else: test.fail_test() test_format = ['ninja'] if sys.platform.startswith('linux') or sys.platform == 'darwin': test_format += ['make'] test = TestGyp.TestGyp(formats=test_format) # Check default values test.run_gyp('make_global_settings_ar.gyp') verify_ar_target(test) # Check default values with GYP_CROSSCOMPILE enabled. with TestGyp.LocalEnv({'GYP_CROSSCOMPILE': '1'}): test.run_gyp('make_global_settings_ar.gyp') verify_ar_target(test) verify_ar_host(test) # Test 'AR' in 'make_global_settings'. with TestGyp.LocalEnv({'GYP_CROSSCOMPILE': '1'}): test.run_gyp('make_global_settings_ar.gyp', '-Dcustom_ar_target=my_ar') verify_ar_target(test, ar='my_ar', rel_path=True) # Test 'AR'/'AR.host' in 'make_global_settings'. with TestGyp.LocalEnv({'GYP_CROSSCOMPILE': '1'}): test.run_gyp('make_global_settings_ar.gyp', '-Dcustom_ar_target=my_ar_target1', '-Dcustom_ar_host=my_ar_host1') verify_ar_target(test, ar='my_ar_target1', rel_path=True) verify_ar_host(test, ar='my_ar_host1', rel_path=True) # Test $AR and $AR_host environment variables. with TestGyp.LocalEnv({'AR': 'my_ar_target2', 'AR_host': 'my_ar_host2'}): test.run_gyp('make_global_settings_ar.gyp') # Ninja generator resolves $AR in gyp phase. Make generator doesn't. if test.format == 'ninja': if sys.platform == 'win32': # TODO(yukawa): Make sure if this is an expected result or not. verify_ar_target(test, ar='lib.exe', rel_path=False) else: verify_ar_target(test, ar='my_ar_target2', rel_path=False) verify_ar_host(test, ar='my_ar_host2', rel_path=False) # Test 'AR' in 'make_global_settings' with $AR_host environment variable. with TestGyp.LocalEnv({'AR_host': 'my_ar_host3'}): test.run_gyp('make_global_settings_ar.gyp', '-Dcustom_ar_target=my_ar_target3') verify_ar_target(test, ar='my_ar_target3', rel_path=True) verify_ar_host(test, ar='my_ar_host3', rel_path=False) test.pass_test()
bsd-3-clause
50f68689e9ac3da224f839e9d4de0faa
29.460317
78
0.652944
3.055732
false
true
false
false
chromium/gyp
pylib/gyp/MSVSUserFile.py
24
5086
# Copyright (c) 2012 Google Inc. All rights reserved. # Use of this source code is governed by a BSD-style license that can be # found in the LICENSE file. """Visual Studio user preferences file writer.""" import os import re import socket # for gethostname import gyp.common import gyp.easy_xml as easy_xml #------------------------------------------------------------------------------ def _FindCommandInPath(command): """If there are no slashes in the command given, this function searches the PATH env to find the given command, and converts it to an absolute path. We have to do this because MSVS is looking for an actual file to launch a debugger on, not just a command line. Note that this happens at GYP time, so anything needing to be built needs to have a full path.""" if '/' in command or '\\' in command: # If the command already has path elements (either relative or # absolute), then assume it is constructed properly. return command else: # Search through the path list and find an existing file that # we can access. paths = os.environ.get('PATH','').split(os.pathsep) for path in paths: item = os.path.join(path, command) if os.path.isfile(item) and os.access(item, os.X_OK): return item return command def _QuoteWin32CommandLineArgs(args): new_args = [] for arg in args: # Replace all double-quotes with double-double-quotes to escape # them for cmd shell, and then quote the whole thing if there # are any. if arg.find('"') != -1: arg = '""'.join(arg.split('"')) arg = '"%s"' % arg # Otherwise, if there are any spaces, quote the whole arg. elif re.search(r'[ \t\n]', arg): arg = '"%s"' % arg new_args.append(arg) return new_args class Writer(object): """Visual Studio XML user user file writer.""" def __init__(self, user_file_path, version, name): """Initializes the user file. Args: user_file_path: Path to the user file. version: Version info. name: Name of the user file. """ self.user_file_path = user_file_path self.version = version self.name = name self.configurations = {} def AddConfig(self, name): """Adds a configuration to the project. Args: name: Configuration name. """ self.configurations[name] = ['Configuration', {'Name': name}] def AddDebugSettings(self, config_name, command, environment = {}, working_directory=""): """Adds a DebugSettings node to the user file for a particular config. Args: command: command line to run. First element in the list is the executable. All elements of the command will be quoted if necessary. working_directory: other files which may trigger the rule. (optional) """ command = _QuoteWin32CommandLineArgs(command) abs_command = _FindCommandInPath(command[0]) if environment and isinstance(environment, dict): env_list = ['%s="%s"' % (key, val) for (key,val) in environment.items()] environment = ' '.join(env_list) else: environment = '' n_cmd = ['DebugSettings', {'Command': abs_command, 'WorkingDirectory': working_directory, 'CommandArguments': " ".join(command[1:]), 'RemoteMachine': socket.gethostname(), 'Environment': environment, 'EnvironmentMerge': 'true', # Currently these are all "dummy" values that we're just setting # in the default manner that MSVS does it. We could use some of # these to add additional capabilities, I suppose, but they might # not have parity with other platforms then. 'Attach': 'false', 'DebuggerType': '3', # 'auto' debugger 'Remote': '1', 'RemoteCommand': '', 'HttpUrl': '', 'PDBPath': '', 'SQLDebugging': '', 'DebuggerFlavor': '0', 'MPIRunCommand': '', 'MPIRunArguments': '', 'MPIRunWorkingDirectory': '', 'ApplicationCommand': '', 'ApplicationArguments': '', 'ShimCommand': '', 'MPIAcceptMode': '', 'MPIAcceptFilter': '' }] # Find the config, and add it if it doesn't exist. if config_name not in self.configurations: self.AddConfig(config_name) # Add the DebugSettings onto the appropriate config. self.configurations[config_name].append(n_cmd) def WriteIfChanged(self): """Writes the user file.""" configs = ['Configurations'] for config, spec in sorted(self.configurations.items()): configs.append(spec) content = ['VisualStudioUserFile', {'Version': self.version.ProjectVersion(), 'Name': self.name }, configs] easy_xml.WriteXmlIfChanged(content, self.user_file_path, encoding="Windows-1252")
bsd-3-clause
932b976a2117c04b3def4b3d94bf950b
33.598639
79
0.593394
4.328511
false
true
false
false
ipython/ipython
IPython/utils/path.py
1
11937
# encoding: utf-8 """ Utilities for path handling. """ # Copyright (c) IPython Development Team. # Distributed under the terms of the Modified BSD License. import os import sys import errno import shutil import random import glob from IPython.utils.process import system #----------------------------------------------------------------------------- # Code #----------------------------------------------------------------------------- fs_encoding = sys.getfilesystemencoding() def _writable_dir(path): """Whether `path` is a directory, to which the user has write access.""" return os.path.isdir(path) and os.access(path, os.W_OK) if sys.platform == 'win32': def _get_long_path_name(path): """Get a long path name (expand ~) on Windows using ctypes. Examples -------- >>> get_long_path_name('c:\\\\docume~1') 'c:\\\\Documents and Settings' """ try: import ctypes except ImportError as e: raise ImportError('you need to have ctypes installed for this to work') from e _GetLongPathName = ctypes.windll.kernel32.GetLongPathNameW _GetLongPathName.argtypes = [ctypes.c_wchar_p, ctypes.c_wchar_p, ctypes.c_uint ] buf = ctypes.create_unicode_buffer(260) rv = _GetLongPathName(path, buf, 260) if rv == 0 or rv > 260: return path else: return buf.value else: def _get_long_path_name(path): """Dummy no-op.""" return path def get_long_path_name(path): """Expand a path into its long form. On Windows this expands any ~ in the paths. On other platforms, it is a null operation. """ return _get_long_path_name(path) def compress_user(path): """Reverse of :func:`os.path.expanduser` """ home = os.path.expanduser('~') if path.startswith(home): path = "~" + path[len(home):] return path def get_py_filename(name): """Return a valid python filename in the current directory. If the given name is not a file, it adds '.py' and searches again. Raises IOError with an informative message if the file isn't found. """ name = os.path.expanduser(name) if os.path.isfile(name): return name if not name.endswith(".py"): py_name = name + ".py" if os.path.isfile(py_name): return py_name raise IOError("File `%r` not found." % name) def filefind(filename: str, path_dirs=None) -> str: """Find a file by looking through a sequence of paths. This iterates through a sequence of paths looking for a file and returns the full, absolute path of the first occurrence of the file. If no set of path dirs is given, the filename is tested as is, after running through :func:`expandvars` and :func:`expanduser`. Thus a simple call:: filefind('myfile.txt') will find the file in the current working dir, but:: filefind('~/myfile.txt') Will find the file in the users home directory. This function does not automatically try any paths, such as the cwd or the user's home directory. Parameters ---------- filename : str The filename to look for. path_dirs : str, None or sequence of str The sequence of paths to look for the file in. If None, the filename need to be absolute or be in the cwd. If a string, the string is put into a sequence and the searched. If a sequence, walk through each element and join with ``filename``, calling :func:`expandvars` and :func:`expanduser` before testing for existence. Returns ------- path : str returns absolute path to file. Raises ------ IOError """ # If paths are quoted, abspath gets confused, strip them... filename = filename.strip('"').strip("'") # If the input is an absolute path, just check it exists if os.path.isabs(filename) and os.path.isfile(filename): return filename if path_dirs is None: path_dirs = ("",) elif isinstance(path_dirs, str): path_dirs = (path_dirs,) for path in path_dirs: if path == '.': path = os.getcwd() testname = expand_path(os.path.join(path, filename)) if os.path.isfile(testname): return os.path.abspath(testname) raise IOError("File %r does not exist in any of the search paths: %r" % (filename, path_dirs) ) class HomeDirError(Exception): pass def get_home_dir(require_writable=False) -> str: """Return the 'home' directory, as a unicode string. Uses os.path.expanduser('~'), and checks for writability. See stdlib docs for how this is determined. For Python <3.8, $HOME is first priority on *ALL* platforms. For Python >=3.8 on Windows, %HOME% is no longer considered. Parameters ---------- require_writable : bool [default: False] if True: guarantees the return value is a writable directory, otherwise raises HomeDirError if False: The path is resolved, but it is not guaranteed to exist or be writable. """ homedir = os.path.expanduser('~') # Next line will make things work even when /home/ is a symlink to # /usr/home as it is on FreeBSD, for example homedir = os.path.realpath(homedir) if not _writable_dir(homedir) and os.name == 'nt': # expanduser failed, use the registry to get the 'My Documents' folder. try: import winreg as wreg with wreg.OpenKey( wreg.HKEY_CURRENT_USER, r"Software\Microsoft\Windows\CurrentVersion\Explorer\Shell Folders" ) as key: homedir = wreg.QueryValueEx(key,'Personal')[0] except: pass if (not require_writable) or _writable_dir(homedir): assert isinstance(homedir, str), "Homedir should be unicode not bytes" return homedir else: raise HomeDirError('%s is not a writable dir, ' 'set $HOME environment variable to override' % homedir) def get_xdg_dir(): """Return the XDG_CONFIG_HOME, if it is defined and exists, else None. This is only for non-OS X posix (Linux,Unix,etc.) systems. """ env = os.environ if os.name == "posix": # Linux, Unix, AIX, etc. # use ~/.config if empty OR not set xdg = env.get("XDG_CONFIG_HOME", None) or os.path.join(get_home_dir(), '.config') if xdg and _writable_dir(xdg): assert isinstance(xdg, str) return xdg return None def get_xdg_cache_dir(): """Return the XDG_CACHE_HOME, if it is defined and exists, else None. This is only for non-OS X posix (Linux,Unix,etc.) systems. """ env = os.environ if os.name == "posix": # Linux, Unix, AIX, etc. # use ~/.cache if empty OR not set xdg = env.get("XDG_CACHE_HOME", None) or os.path.join(get_home_dir(), '.cache') if xdg and _writable_dir(xdg): assert isinstance(xdg, str) return xdg return None def expand_path(s): """Expand $VARS and ~names in a string, like a shell :Examples: In [2]: os.environ['FOO']='test' In [3]: expand_path('variable FOO is $FOO') Out[3]: 'variable FOO is test' """ # This is a pretty subtle hack. When expand user is given a UNC path # on Windows (\\server\share$\%username%), os.path.expandvars, removes # the $ to get (\\server\share\%username%). I think it considered $ # alone an empty var. But, we need the $ to remains there (it indicates # a hidden share). if os.name=='nt': s = s.replace('$\\', 'IPYTHON_TEMP') s = os.path.expandvars(os.path.expanduser(s)) if os.name=='nt': s = s.replace('IPYTHON_TEMP', '$\\') return s def unescape_glob(string): """Unescape glob pattern in `string`.""" def unescape(s): for pattern in '*[]!?': s = s.replace(r'\{0}'.format(pattern), pattern) return s return '\\'.join(map(unescape, string.split('\\\\'))) def shellglob(args): """ Do glob expansion for each element in `args` and return a flattened list. Unmatched glob pattern will remain as-is in the returned list. """ expanded = [] # Do not unescape backslash in Windows as it is interpreted as # path separator: unescape = unescape_glob if sys.platform != 'win32' else lambda x: x for a in args: expanded.extend(glob.glob(a) or [unescape(a)]) return expanded def target_outdated(target,deps): """Determine whether a target is out of date. target_outdated(target,deps) -> 1/0 deps: list of filenames which MUST exist. target: single filename which may or may not exist. If target doesn't exist or is older than any file listed in deps, return true, otherwise return false. """ try: target_time = os.path.getmtime(target) except os.error: return 1 for dep in deps: dep_time = os.path.getmtime(dep) if dep_time > target_time: #print "For target",target,"Dep failed:",dep # dbg #print "times (dep,tar):",dep_time,target_time # dbg return 1 return 0 def target_update(target,deps,cmd): """Update a target with a given command given a list of dependencies. target_update(target,deps,cmd) -> runs cmd if target is outdated. This is just a wrapper around target_outdated() which calls the given command if target is outdated.""" if target_outdated(target,deps): system(cmd) ENOLINK = 1998 def link(src, dst): """Hard links ``src`` to ``dst``, returning 0 or errno. Note that the special errno ``ENOLINK`` will be returned if ``os.link`` isn't supported by the operating system. """ if not hasattr(os, "link"): return ENOLINK link_errno = 0 try: os.link(src, dst) except OSError as e: link_errno = e.errno return link_errno def link_or_copy(src, dst): """Attempts to hardlink ``src`` to ``dst``, copying if the link fails. Attempts to maintain the semantics of ``shutil.copy``. Because ``os.link`` does not overwrite files, a unique temporary file will be used if the target already exists, then that file will be moved into place. """ if os.path.isdir(dst): dst = os.path.join(dst, os.path.basename(src)) link_errno = link(src, dst) if link_errno == errno.EEXIST: if os.stat(src).st_ino == os.stat(dst).st_ino: # dst is already a hard link to the correct file, so we don't need # to do anything else. If we try to link and rename the file # anyway, we get duplicate files - see http://bugs.python.org/issue21876 return new_dst = dst + "-temp-%04X" %(random.randint(1, 16**4), ) try: link_or_copy(src, new_dst) except: try: os.remove(new_dst) except OSError: pass raise os.rename(new_dst, dst) elif link_errno != 0: # Either link isn't supported, or the filesystem doesn't support # linking, or 'src' and 'dst' are on different filesystems. shutil.copy(src, dst) def ensure_dir_exists(path, mode=0o755): """ensure that a directory exists If it doesn't exist, try to create it and protect against a race condition if another process is doing the same. The default permissions are 755, which differ from os.makedirs default of 777. """ if not os.path.exists(path): try: os.makedirs(path, mode=mode) except OSError as e: if e.errno != errno.EEXIST: raise elif not os.path.isdir(path): raise IOError("%r exists but is not a directory" % path)
bsd-3-clause
51ccb99d6424ad0592e4f10e8d2e680d
29.529412
90
0.604758
3.890808
false
false
false
false
ipython/ipython
IPython/utils/frame.py
1
3048
# encoding: utf-8 """ Utilities for working with stack frames. """ #----------------------------------------------------------------------------- # Copyright (C) 2008-2011 The IPython Development Team # # Distributed under the terms of the BSD License. The full license is in # the file COPYING, distributed as part of this software. #----------------------------------------------------------------------------- #----------------------------------------------------------------------------- # Imports #----------------------------------------------------------------------------- import sys #----------------------------------------------------------------------------- # Code #----------------------------------------------------------------------------- def extract_vars(*names,**kw): """Extract a set of variables by name from another frame. Parameters ---------- *names : str One or more variable names which will be extracted from the caller's frame. **kw : integer, optional How many frames in the stack to walk when looking for your variables. The default is 0, which will use the frame where the call was made. Examples -------- :: In [2]: def func(x): ...: y = 1 ...: print(sorted(extract_vars('x','y').items())) ...: In [3]: func('hello') [('x', 'hello'), ('y', 1)] """ depth = kw.get('depth',0) callerNS = sys._getframe(depth+1).f_locals return dict((k,callerNS[k]) for k in names) def extract_vars_above(*names): """Extract a set of variables by name from another frame. Similar to extractVars(), but with a specified depth of 1, so that names are extracted exactly from above the caller. This is simply a convenience function so that the very common case (for us) of skipping exactly 1 frame doesn't have to construct a special dict for keyword passing.""" callerNS = sys._getframe(2).f_locals return dict((k,callerNS[k]) for k in names) def debugx(expr,pre_msg=''): """Print the value of an expression from the caller's frame. Takes an expression, evaluates it in the caller's frame and prints both the given expression and the resulting value (as well as a debug mark indicating the name of the calling function. The input must be of a form suitable for eval(). An optional message can be passed, which will be prepended to the printed expr->value pair.""" cf = sys._getframe(1) print('[DBG:%s] %s%s -> %r' % (cf.f_code.co_name,pre_msg,expr, eval(expr,cf.f_globals,cf.f_locals))) # deactivate it by uncommenting the following line, which makes it a no-op #def debugx(expr,pre_msg=''): pass def extract_module_locals(depth=0): """Returns (module, locals) of the function `depth` frames away from the caller""" f = sys._getframe(depth + 1) global_ns = f.f_globals module = sys.modules[global_ns['__name__']] return (module, f.f_locals)
bsd-3-clause
140496a59121839d756072620fd9e345
32.130435
86
0.536417
4.508876
false
false
false
false
ipython/ipython
IPython/utils/tz.py
44
1352
# encoding: utf-8 """ Timezone utilities Just UTC-awareness right now """ #----------------------------------------------------------------------------- # Copyright (C) 2013 The IPython Development Team # # Distributed under the terms of the BSD License. The full license is in # the file COPYING, distributed as part of this software. #----------------------------------------------------------------------------- #----------------------------------------------------------------------------- # Imports #----------------------------------------------------------------------------- from datetime import tzinfo, timedelta, datetime #----------------------------------------------------------------------------- # Code #----------------------------------------------------------------------------- # constant for zero offset ZERO = timedelta(0) class tzUTC(tzinfo): """tzinfo object for UTC (zero offset)""" def utcoffset(self, d): return ZERO def dst(self, d): return ZERO UTC = tzUTC() def utc_aware(unaware): """decorator for adding UTC tzinfo to datetime's utcfoo methods""" def utc_method(*args, **kwargs): dt = unaware(*args, **kwargs) return dt.replace(tzinfo=UTC) return utc_method utcfromtimestamp = utc_aware(datetime.utcfromtimestamp) utcnow = utc_aware(datetime.utcnow)
bsd-3-clause
cb720515b7bc78b31161bc40b7bf8001
28.391304
78
0.448225
5.473684
false
false
false
false
ipython/ipython
tools/github_stats.py
1
8425
#!/usr/bin/env python """Simple tools to query github.com and gather stats about issues. To generate a report for IPython 2.0, run: python github_stats.py --milestone 2.0 --since-tag rel-1.0.0 """ #----------------------------------------------------------------------------- # Imports #----------------------------------------------------------------------------- import sys from argparse import ArgumentParser from datetime import datetime, timedelta from subprocess import check_output from gh_api import ( get_paged_request, make_auth_header, get_pull_request, is_pull_request, get_milestone_id, get_issues_list, get_authors, ) #----------------------------------------------------------------------------- # Globals #----------------------------------------------------------------------------- ISO8601 = "%Y-%m-%dT%H:%M:%SZ" PER_PAGE = 100 #----------------------------------------------------------------------------- # Functions #----------------------------------------------------------------------------- def round_hour(dt): return dt.replace(minute=0,second=0,microsecond=0) def _parse_datetime(s): """Parse dates in the format returned by the Github API.""" if s: return datetime.strptime(s, ISO8601) else: return datetime.fromtimestamp(0) def issues2dict(issues): """Convert a list of issues to a dict, keyed by issue number.""" idict = {} for i in issues: idict[i['number']] = i return idict def split_pulls(all_issues, project="ipython/ipython"): """split a list of closed issues into non-PR Issues and Pull Requests""" pulls = [] issues = [] for i in all_issues: if is_pull_request(i): pull = get_pull_request(project, i['number'], auth=True) pulls.append(pull) else: issues.append(i) return issues, pulls def issues_closed_since(period=timedelta(days=365), project="ipython/ipython", pulls=False): """Get all issues closed since a particular point in time. period can either be a datetime object, or a timedelta object. In the latter case, it is used as a time before the present. """ which = 'pulls' if pulls else 'issues' if isinstance(period, timedelta): since = round_hour(datetime.utcnow() - period) else: since = period url = "https://api.github.com/repos/%s/%s?state=closed&sort=updated&since=%s&per_page=%i" % (project, which, since.strftime(ISO8601), PER_PAGE) allclosed = get_paged_request(url, headers=make_auth_header()) filtered = [ i for i in allclosed if _parse_datetime(i['closed_at']) > since ] if pulls: filtered = [ i for i in filtered if _parse_datetime(i['merged_at']) > since ] # filter out PRs not against main (backports) filtered = [i for i in filtered if i["base"]["ref"] == "main"] else: filtered = [ i for i in filtered if not is_pull_request(i) ] return filtered def sorted_by_field(issues, field='closed_at', reverse=False): """Return a list of issues sorted by closing date date.""" return sorted(issues, key = lambda i:i[field], reverse=reverse) def report(issues, show_urls=False): """Summary report about a list of issues, printing number and title.""" if show_urls: for i in issues: role = 'ghpull' if 'merged_at' in i else 'ghissue' print(u'* :%s:`%d`: %s' % (role, i['number'], i['title'].replace(u'`', u'``'))) else: for i in issues: print(u'* %d: %s' % (i['number'], i['title'].replace(u'`', u'``'))) #----------------------------------------------------------------------------- # Main script #----------------------------------------------------------------------------- if __name__ == "__main__": print("DEPRECATE: backport_pr.py is deprecated and it is now recommended" "to install `ghpro` from PyPI.", file=sys.stderr) # Whether to add reST urls for all issues in printout. show_urls = True parser = ArgumentParser() parser.add_argument('--since-tag', type=str, help="The git tag to use for the starting point (typically the last major release)." ) parser.add_argument('--milestone', type=str, help="The GitHub milestone to use for filtering issues [optional]." ) parser.add_argument('--days', type=int, help="The number of days of data to summarize (use this or --since-tag)." ) parser.add_argument('--project', type=str, default="ipython/ipython", help="The project to summarize." ) parser.add_argument('--links', action='store_true', default=False, help="Include links to all closed Issues and PRs in the output." ) opts = parser.parse_args() tag = opts.since_tag # set `since` from days or git tag if opts.days: since = datetime.utcnow() - timedelta(days=opts.days) else: if not tag: tag = check_output(['git', 'describe', '--abbrev=0']).strip().decode('utf8') cmd = ['git', 'log', '-1', '--format=%ai', tag] tagday, tz = check_output(cmd).strip().decode('utf8').rsplit(' ', 1) since = datetime.strptime(tagday, "%Y-%m-%d %H:%M:%S") h = int(tz[1:3]) m = int(tz[3:]) td = timedelta(hours=h, minutes=m) if tz[0] == '-': since += td else: since -= td since = round_hour(since) milestone = opts.milestone project = opts.project print("fetching GitHub stats since %s (tag: %s, milestone: %s)" % (since, tag, milestone), file=sys.stderr) if milestone: milestone_id = get_milestone_id(project=project, milestone=milestone, auth=True) issues_and_pulls = get_issues_list(project=project, milestone=milestone_id, state='closed', auth=True, ) issues, pulls = split_pulls(issues_and_pulls, project=project) else: issues = issues_closed_since(since, project=project, pulls=False) pulls = issues_closed_since(since, project=project, pulls=True) # For regular reports, it's nice to show them in reverse chronological order issues = sorted_by_field(issues, reverse=True) pulls = sorted_by_field(pulls, reverse=True) n_issues, n_pulls = map(len, (issues, pulls)) n_total = n_issues + n_pulls # Print summary report we can directly include into release notes. print() since_day = since.strftime("%Y/%m/%d") today = datetime.today().strftime("%Y/%m/%d") print("GitHub stats for %s - %s (tag: %s)" % (since_day, today, tag)) print() print("These lists are automatically generated, and may be incomplete or contain duplicates.") print() ncommits = 0 all_authors = [] if tag: # print git info, in addition to GitHub info: since_tag = tag+'..' cmd = ['git', 'log', '--oneline', since_tag] ncommits += len(check_output(cmd).splitlines()) author_cmd = ['git', 'log', '--use-mailmap', "--format=* %aN", since_tag] all_authors.extend(check_output(author_cmd).decode('utf-8', 'replace').splitlines()) pr_authors = [] for pr in pulls: pr_authors.extend(get_authors(pr)) ncommits = len(pr_authors) + ncommits - len(pulls) author_cmd = ['git', 'check-mailmap'] + pr_authors with_email = check_output(author_cmd).decode('utf-8', 'replace').splitlines() all_authors.extend([ u'* ' + a.split(' <')[0] for a in with_email ]) unique_authors = sorted(set(all_authors), key=lambda s: s.lower()) print("We closed %d issues and merged %d pull requests." % (n_issues, n_pulls)) if milestone: print("The full list can be seen `on GitHub <https://github.com/{project}/issues?q=milestone%3A{milestone}>`__".format(project=project,milestone=milestone) ) print() print("The following %i authors contributed %i commits." % (len(unique_authors), ncommits)) print() print('\n'.join(unique_authors)) if opts.links: print() print("GitHub issues and pull requests:") print() print('Pull Requests (%d):\n' % n_pulls) report(pulls, show_urls) print() print('Issues (%d):\n' % n_issues) report(issues, show_urls)
bsd-3-clause
16b4ab2229458e7a947e14201d6f5eae
35.630435
163
0.56095
3.850548
false
false
false
false
ipython/ipython
IPython/core/inputtransformer2.py
1
28374
"""Input transformer machinery to support IPython special syntax. This includes the machinery to recognise and transform ``%magic`` commands, ``!system`` commands, ``help?`` querying, prompt stripping, and so forth. Added: IPython 7.0. Replaces inputsplitter and inputtransformer which were deprecated in 7.0. """ # Copyright (c) IPython Development Team. # Distributed under the terms of the Modified BSD License. import ast from codeop import CommandCompiler, Compile import re import tokenize from typing import List, Tuple, Optional, Any import warnings _indent_re = re.compile(r'^[ \t]+') def leading_empty_lines(lines): """Remove leading empty lines If the leading lines are empty or contain only whitespace, they will be removed. """ if not lines: return lines for i, line in enumerate(lines): if line and not line.isspace(): return lines[i:] return lines def leading_indent(lines): """Remove leading indentation. If the first line starts with a spaces or tabs, the same whitespace will be removed from each following line in the cell. """ if not lines: return lines m = _indent_re.match(lines[0]) if not m: return lines space = m.group(0) n = len(space) return [l[n:] if l.startswith(space) else l for l in lines] class PromptStripper: """Remove matching input prompts from a block of input. Parameters ---------- prompt_re : regular expression A regular expression matching any input prompt (including continuation, e.g. ``...``) initial_re : regular expression, optional A regular expression matching only the initial prompt, but not continuation. If no initial expression is given, prompt_re will be used everywhere. Used mainly for plain Python prompts (``>>>``), where the continuation prompt ``...`` is a valid Python expression in Python 3, so shouldn't be stripped. Notes ----- If initial_re and prompt_re differ, only initial_re will be tested against the first line. If any prompt is found on the first two lines, prompts will be stripped from the rest of the block. """ def __init__(self, prompt_re, initial_re=None): self.prompt_re = prompt_re self.initial_re = initial_re or prompt_re def _strip(self, lines): return [self.prompt_re.sub('', l, count=1) for l in lines] def __call__(self, lines): if not lines: return lines if self.initial_re.match(lines[0]) or \ (len(lines) > 1 and self.prompt_re.match(lines[1])): return self._strip(lines) return lines classic_prompt = PromptStripper( prompt_re=re.compile(r'^(>>>|\.\.\.)( |$)'), initial_re=re.compile(r'^>>>( |$)') ) ipython_prompt = PromptStripper( re.compile( r""" ^( # Match from the beginning of a line, either: # 1. First-line prompt: ((\[nav\]|\[ins\])?\ )? # Vi editing mode prompt, if it's there In\ # The 'In' of the prompt, with a space \[\d+\]: # Command index, as displayed in the prompt \ # With a mandatory trailing space | # ... or ... # 2. The three dots of the multiline prompt \s* # All leading whitespace characters \.{3,}: # The three (or more) dots \ ? # With an optional trailing space ) """, re.VERBOSE, ) ) def cell_magic(lines): if not lines or not lines[0].startswith('%%'): return lines if re.match(r'%%\w+\?', lines[0]): # This case will be handled by help_end return lines magic_name, _, first_line = lines[0][2:].rstrip().partition(' ') body = ''.join(lines[1:]) return ['get_ipython().run_cell_magic(%r, %r, %r)\n' % (magic_name, first_line, body)] def _find_assign_op(token_line) -> Optional[int]: """Get the index of the first assignment in the line ('=' not inside brackets) Note: We don't try to support multiple special assignment (a = b = %foo) """ paren_level = 0 for i, ti in enumerate(token_line): s = ti.string if s == '=' and paren_level == 0: return i if s in {'(','[','{'}: paren_level += 1 elif s in {')', ']', '}'}: if paren_level > 0: paren_level -= 1 return None def find_end_of_continued_line(lines, start_line: int): """Find the last line of a line explicitly extended using backslashes. Uses 0-indexed line numbers. """ end_line = start_line while lines[end_line].endswith('\\\n'): end_line += 1 if end_line >= len(lines): break return end_line def assemble_continued_line(lines, start: Tuple[int, int], end_line: int): r"""Assemble a single line from multiple continued line pieces Continued lines are lines ending in ``\``, and the line following the last ``\`` in the block. For example, this code continues over multiple lines:: if (assign_ix is not None) \ and (len(line) >= assign_ix + 2) \ and (line[assign_ix+1].string == '%') \ and (line[assign_ix+2].type == tokenize.NAME): This statement contains four continued line pieces. Assembling these pieces into a single line would give:: if (assign_ix is not None) and (len(line) >= assign_ix + 2) and (line[... This uses 0-indexed line numbers. *start* is (lineno, colno). Used to allow ``%magic`` and ``!system`` commands to be continued over multiple lines. """ parts = [lines[start[0]][start[1]:]] + lines[start[0]+1:end_line+1] return ' '.join([p.rstrip()[:-1] for p in parts[:-1]] # Strip backslash+newline + [parts[-1].rstrip()]) # Strip newline from last line class TokenTransformBase: """Base class for transformations which examine tokens. Special syntax should not be transformed when it occurs inside strings or comments. This is hard to reliably avoid with regexes. The solution is to tokenise the code as Python, and recognise the special syntax in the tokens. IPython's special syntax is not valid Python syntax, so tokenising may go wrong after the special syntax starts. These classes therefore find and transform *one* instance of special syntax at a time into regular Python syntax. After each transformation, tokens are regenerated to find the next piece of special syntax. Subclasses need to implement one class method (find) and one regular method (transform). The priority attribute can select which transformation to apply if multiple transformers match in the same place. Lower numbers have higher priority. This allows "%magic?" to be turned into a help call rather than a magic call. """ # Lower numbers -> higher priority (for matches in the same location) priority = 10 def sortby(self): return self.start_line, self.start_col, self.priority def __init__(self, start): self.start_line = start[0] - 1 # Shift from 1-index to 0-index self.start_col = start[1] @classmethod def find(cls, tokens_by_line): """Find one instance of special syntax in the provided tokens. Tokens are grouped into logical lines for convenience, so it is easy to e.g. look at the first token of each line. *tokens_by_line* is a list of lists of tokenize.TokenInfo objects. This should return an instance of its class, pointing to the start position it has found, or None if it found no match. """ raise NotImplementedError def transform(self, lines: List[str]): """Transform one instance of special syntax found by ``find()`` Takes a list of strings representing physical lines, returns a similar list of transformed lines. """ raise NotImplementedError class MagicAssign(TokenTransformBase): """Transformer for assignments from magics (a = %foo)""" @classmethod def find(cls, tokens_by_line): """Find the first magic assignment (a = %foo) in the cell. """ for line in tokens_by_line: assign_ix = _find_assign_op(line) if (assign_ix is not None) \ and (len(line) >= assign_ix + 2) \ and (line[assign_ix+1].string == '%') \ and (line[assign_ix+2].type == tokenize.NAME): return cls(line[assign_ix+1].start) def transform(self, lines: List[str]): """Transform a magic assignment found by the ``find()`` classmethod. """ start_line, start_col = self.start_line, self.start_col lhs = lines[start_line][:start_col] end_line = find_end_of_continued_line(lines, start_line) rhs = assemble_continued_line(lines, (start_line, start_col), end_line) assert rhs.startswith('%'), rhs magic_name, _, args = rhs[1:].partition(' ') lines_before = lines[:start_line] call = "get_ipython().run_line_magic({!r}, {!r})".format(magic_name, args) new_line = lhs + call + '\n' lines_after = lines[end_line+1:] return lines_before + [new_line] + lines_after class SystemAssign(TokenTransformBase): """Transformer for assignments from system commands (a = !foo)""" @classmethod def find(cls, tokens_by_line): """Find the first system assignment (a = !foo) in the cell. """ for line in tokens_by_line: assign_ix = _find_assign_op(line) if (assign_ix is not None) \ and not line[assign_ix].line.strip().startswith('=') \ and (len(line) >= assign_ix + 2) \ and (line[assign_ix + 1].type == tokenize.ERRORTOKEN): ix = assign_ix + 1 while ix < len(line) and line[ix].type == tokenize.ERRORTOKEN: if line[ix].string == '!': return cls(line[ix].start) elif not line[ix].string.isspace(): break ix += 1 def transform(self, lines: List[str]): """Transform a system assignment found by the ``find()`` classmethod. """ start_line, start_col = self.start_line, self.start_col lhs = lines[start_line][:start_col] end_line = find_end_of_continued_line(lines, start_line) rhs = assemble_continued_line(lines, (start_line, start_col), end_line) assert rhs.startswith('!'), rhs cmd = rhs[1:] lines_before = lines[:start_line] call = "get_ipython().getoutput({!r})".format(cmd) new_line = lhs + call + '\n' lines_after = lines[end_line + 1:] return lines_before + [new_line] + lines_after # The escape sequences that define the syntax transformations IPython will # apply to user input. These can NOT be just changed here: many regular # expressions and other parts of the code may use their hardcoded values, and # for all intents and purposes they constitute the 'IPython syntax', so they # should be considered fixed. ESC_SHELL = '!' # Send line to underlying system shell ESC_SH_CAP = '!!' # Send line to system shell and capture output ESC_HELP = '?' # Find information about object ESC_HELP2 = '??' # Find extra-detailed information about object ESC_MAGIC = '%' # Call magic function ESC_MAGIC2 = '%%' # Call cell-magic function ESC_QUOTE = ',' # Split args on whitespace, quote each as string and call ESC_QUOTE2 = ';' # Quote all args as a single string, call ESC_PAREN = '/' # Call first argument with rest of line as arguments ESCAPE_SINGLES = {'!', '?', '%', ',', ';', '/'} ESCAPE_DOUBLES = {'!!', '??'} # %% (cell magic) is handled separately def _make_help_call(target, esc): """Prepares a pinfo(2)/psearch call from a target name and the escape (i.e. ? or ??)""" method = 'pinfo2' if esc == '??' \ else 'psearch' if '*' in target \ else 'pinfo' arg = " ".join([method, target]) #Prepare arguments for get_ipython().run_line_magic(magic_name, magic_args) t_magic_name, _, t_magic_arg_s = arg.partition(' ') t_magic_name = t_magic_name.lstrip(ESC_MAGIC) return "get_ipython().run_line_magic(%r, %r)" % (t_magic_name, t_magic_arg_s) def _tr_help(content): """Translate lines escaped with: ? A naked help line should fire the intro help screen (shell.show_usage()) """ if not content: return 'get_ipython().show_usage()' return _make_help_call(content, '?') def _tr_help2(content): """Translate lines escaped with: ?? A naked help line should fire the intro help screen (shell.show_usage()) """ if not content: return 'get_ipython().show_usage()' return _make_help_call(content, '??') def _tr_magic(content): "Translate lines escaped with a percent sign: %" name, _, args = content.partition(' ') return 'get_ipython().run_line_magic(%r, %r)' % (name, args) def _tr_quote(content): "Translate lines escaped with a comma: ," name, _, args = content.partition(' ') return '%s("%s")' % (name, '", "'.join(args.split()) ) def _tr_quote2(content): "Translate lines escaped with a semicolon: ;" name, _, args = content.partition(' ') return '%s("%s")' % (name, args) def _tr_paren(content): "Translate lines escaped with a slash: /" name, _, args = content.partition(' ') return '%s(%s)' % (name, ", ".join(args.split())) tr = { ESC_SHELL : 'get_ipython().system({!r})'.format, ESC_SH_CAP : 'get_ipython().getoutput({!r})'.format, ESC_HELP : _tr_help, ESC_HELP2 : _tr_help2, ESC_MAGIC : _tr_magic, ESC_QUOTE : _tr_quote, ESC_QUOTE2 : _tr_quote2, ESC_PAREN : _tr_paren } class EscapedCommand(TokenTransformBase): """Transformer for escaped commands like %foo, !foo, or /foo""" @classmethod def find(cls, tokens_by_line): """Find the first escaped command (%foo, !foo, etc.) in the cell. """ for line in tokens_by_line: if not line: continue ix = 0 ll = len(line) while ll > ix and line[ix].type in {tokenize.INDENT, tokenize.DEDENT}: ix += 1 if ix >= ll: continue if line[ix].string in ESCAPE_SINGLES: return cls(line[ix].start) def transform(self, lines): """Transform an escaped line found by the ``find()`` classmethod. """ start_line, start_col = self.start_line, self.start_col indent = lines[start_line][:start_col] end_line = find_end_of_continued_line(lines, start_line) line = assemble_continued_line(lines, (start_line, start_col), end_line) if len(line) > 1 and line[:2] in ESCAPE_DOUBLES: escape, content = line[:2], line[2:] else: escape, content = line[:1], line[1:] if escape in tr: call = tr[escape](content) else: call = '' lines_before = lines[:start_line] new_line = indent + call + '\n' lines_after = lines[end_line + 1:] return lines_before + [new_line] + lines_after _help_end_re = re.compile( r"""(%{0,2} (?!\d)[\w*]+ # Variable name (\.(?!\d)[\w*]+|\[-?[0-9]+\])* # .etc.etc or [0], we only support literal integers. ) (\?\??)$ # ? or ?? """, re.VERBOSE, ) class HelpEnd(TokenTransformBase): """Transformer for help syntax: obj? and obj??""" # This needs to be higher priority (lower number) than EscapedCommand so # that inspecting magics (%foo?) works. priority = 5 def __init__(self, start, q_locn): super().__init__(start) self.q_line = q_locn[0] - 1 # Shift from 1-indexed to 0-indexed self.q_col = q_locn[1] @classmethod def find(cls, tokens_by_line): """Find the first help command (foo?) in the cell. """ for line in tokens_by_line: # Last token is NEWLINE; look at last but one if len(line) > 2 and line[-2].string == '?': # Find the first token that's not INDENT/DEDENT ix = 0 while line[ix].type in {tokenize.INDENT, tokenize.DEDENT}: ix += 1 return cls(line[ix].start, line[-2].start) def transform(self, lines): """Transform a help command found by the ``find()`` classmethod. """ piece = "".join(lines[self.start_line : self.q_line + 1]) indent, content = piece[: self.start_col], piece[self.start_col :] lines_before = lines[: self.start_line] lines_after = lines[self.q_line + 1 :] m = _help_end_re.search(content) if not m: raise SyntaxError(content) assert m is not None, content target = m.group(1) esc = m.group(3) call = _make_help_call(target, esc) new_line = indent + call + '\n' return lines_before + [new_line] + lines_after def make_tokens_by_line(lines:List[str]): """Tokenize a series of lines and group tokens by line. The tokens for a multiline Python string or expression are grouped as one line. All lines except the last lines should keep their line ending ('\\n', '\\r\\n') for this to properly work. Use `.splitlines(keeplineending=True)` for example when passing block of text to this function. """ # NL tokens are used inside multiline expressions, but also after blank # lines or comments. This is intentional - see https://bugs.python.org/issue17061 # We want to group the former case together but split the latter, so we # track parentheses level, similar to the internals of tokenize. # reexported from token on 3.7+ NEWLINE, NL = tokenize.NEWLINE, tokenize.NL # type: ignore tokens_by_line: List[List[Any]] = [[]] if len(lines) > 1 and not lines[0].endswith(("\n", "\r", "\r\n", "\x0b", "\x0c")): warnings.warn( "`make_tokens_by_line` received a list of lines which do not have lineending markers ('\\n', '\\r', '\\r\\n', '\\x0b', '\\x0c'), behavior will be unspecified", stacklevel=2, ) parenlev = 0 try: for token in tokenize.generate_tokens(iter(lines).__next__): tokens_by_line[-1].append(token) if (token.type == NEWLINE) \ or ((token.type == NL) and (parenlev <= 0)): tokens_by_line.append([]) elif token.string in {'(', '[', '{'}: parenlev += 1 elif token.string in {')', ']', '}'}: if parenlev > 0: parenlev -= 1 except tokenize.TokenError: # Input ended in a multiline string or expression. That's OK for us. pass if not tokens_by_line[-1]: tokens_by_line.pop() return tokens_by_line def has_sunken_brackets(tokens: List[tokenize.TokenInfo]): """Check if the depth of brackets in the list of tokens drops below 0""" parenlev = 0 for token in tokens: if token.string in {"(", "[", "{"}: parenlev += 1 elif token.string in {")", "]", "}"}: parenlev -= 1 if parenlev < 0: return True return False def show_linewise_tokens(s: str): """For investigation and debugging""" warnings.warn( "show_linewise_tokens is deprecated since IPython 8.6", DeprecationWarning, stacklevel=2, ) if not s.endswith("\n"): s += "\n" lines = s.splitlines(keepends=True) for line in make_tokens_by_line(lines): print("Line -------") for tokinfo in line: print(" ", tokinfo) # Arbitrary limit to prevent getting stuck in infinite loops TRANSFORM_LOOP_LIMIT = 500 class TransformerManager: """Applies various transformations to a cell or code block. The key methods for external use are ``transform_cell()`` and ``check_complete()``. """ def __init__(self): self.cleanup_transforms = [ leading_empty_lines, leading_indent, classic_prompt, ipython_prompt, ] self.line_transforms = [ cell_magic, ] self.token_transformers = [ MagicAssign, SystemAssign, EscapedCommand, HelpEnd, ] def do_one_token_transform(self, lines): """Find and run the transform earliest in the code. Returns (changed, lines). This method is called repeatedly until changed is False, indicating that all available transformations are complete. The tokens following IPython special syntax might not be valid, so the transformed code is retokenised every time to identify the next piece of special syntax. Hopefully long code cells are mostly valid Python, not using lots of IPython special syntax, so this shouldn't be a performance issue. """ tokens_by_line = make_tokens_by_line(lines) candidates = [] for transformer_cls in self.token_transformers: transformer = transformer_cls.find(tokens_by_line) if transformer: candidates.append(transformer) if not candidates: # Nothing to transform return False, lines ordered_transformers = sorted(candidates, key=TokenTransformBase.sortby) for transformer in ordered_transformers: try: return True, transformer.transform(lines) except SyntaxError: pass return False, lines def do_token_transforms(self, lines): for _ in range(TRANSFORM_LOOP_LIMIT): changed, lines = self.do_one_token_transform(lines) if not changed: return lines raise RuntimeError("Input transformation still changing after " "%d iterations. Aborting." % TRANSFORM_LOOP_LIMIT) def transform_cell(self, cell: str) -> str: """Transforms a cell of input code""" if not cell.endswith('\n'): cell += '\n' # Ensure the cell has a trailing newline lines = cell.splitlines(keepends=True) for transform in self.cleanup_transforms + self.line_transforms: lines = transform(lines) lines = self.do_token_transforms(lines) return ''.join(lines) def check_complete(self, cell: str): """Return whether a block of code is ready to execute, or should be continued Parameters ---------- cell : string Python input code, which can be multiline. Returns ------- status : str One of 'complete', 'incomplete', or 'invalid' if source is not a prefix of valid code. indent_spaces : int or None The number of spaces by which to indent the next line of code. If status is not 'incomplete', this is None. """ # Remember if the lines ends in a new line. ends_with_newline = False for character in reversed(cell): if character == '\n': ends_with_newline = True break elif character.strip(): break else: continue if not ends_with_newline: # Append an newline for consistent tokenization # See https://bugs.python.org/issue33899 cell += '\n' lines = cell.splitlines(keepends=True) if not lines: return 'complete', None if lines[-1].endswith('\\'): # Explicit backslash continuation return 'incomplete', find_last_indent(lines) try: for transform in self.cleanup_transforms: if not getattr(transform, 'has_side_effects', False): lines = transform(lines) except SyntaxError: return 'invalid', None if lines[0].startswith('%%'): # Special case for cell magics - completion marked by blank line if lines[-1].strip(): return 'incomplete', find_last_indent(lines) else: return 'complete', None try: for transform in self.line_transforms: if not getattr(transform, 'has_side_effects', False): lines = transform(lines) lines = self.do_token_transforms(lines) except SyntaxError: return 'invalid', None tokens_by_line = make_tokens_by_line(lines) # Bail if we got one line and there are more closing parentheses than # the opening ones if ( len(lines) == 1 and tokens_by_line and has_sunken_brackets(tokens_by_line[0]) ): return "invalid", None if not tokens_by_line: return 'incomplete', find_last_indent(lines) if tokens_by_line[-1][-1].type != tokenize.ENDMARKER: # We're in a multiline string or expression return 'incomplete', find_last_indent(lines) newline_types = {tokenize.NEWLINE, tokenize.COMMENT, tokenize.ENDMARKER} # type: ignore # Pop the last line which only contains DEDENTs and ENDMARKER last_token_line = None if {t.type for t in tokens_by_line[-1]} in [ {tokenize.DEDENT, tokenize.ENDMARKER}, {tokenize.ENDMARKER} ] and len(tokens_by_line) > 1: last_token_line = tokens_by_line.pop() while tokens_by_line[-1] and tokens_by_line[-1][-1].type in newline_types: tokens_by_line[-1].pop() if not tokens_by_line[-1]: return 'incomplete', find_last_indent(lines) if tokens_by_line[-1][-1].string == ':': # The last line starts a block (e.g. 'if foo:') ix = 0 while tokens_by_line[-1][ix].type in {tokenize.INDENT, tokenize.DEDENT}: ix += 1 indent = tokens_by_line[-1][ix].start[1] return 'incomplete', indent + 4 if tokens_by_line[-1][0].line.endswith('\\'): return 'incomplete', None # At this point, our checks think the code is complete (or invalid). # We'll use codeop.compile_command to check this with the real parser try: with warnings.catch_warnings(): warnings.simplefilter('error', SyntaxWarning) res = compile_command(''.join(lines), symbol='exec') except (SyntaxError, OverflowError, ValueError, TypeError, MemoryError, SyntaxWarning): return 'invalid', None else: if res is None: return 'incomplete', find_last_indent(lines) if last_token_line and last_token_line[0].type == tokenize.DEDENT: if ends_with_newline: return 'complete', None return 'incomplete', find_last_indent(lines) # If there's a blank line at the end, assume we're ready to execute if not lines[-1].strip(): return 'complete', None return 'complete', None def find_last_indent(lines): m = _indent_re.match(lines[-1]) if not m: return 0 return len(m.group(0).replace('\t', ' '*4)) class MaybeAsyncCompile(Compile): def __init__(self, extra_flags=0): super().__init__() self.flags |= extra_flags class MaybeAsyncCommandCompiler(CommandCompiler): def __init__(self, extra_flags=0): self.compiler = MaybeAsyncCompile(extra_flags=extra_flags) _extra_flags = ast.PyCF_ALLOW_TOP_LEVEL_AWAIT compile_command = MaybeAsyncCommandCompiler(extra_flags=_extra_flags)
bsd-3-clause
8051f4f3e27c481bbbe29a6471eb0560
34.601004
171
0.581448
4.014999
false
false
false
false
ipython/ipython
examples/IPython Kernel/ipython-get-history.py
1
1136
#!/usr/bin/env python """Extract a session from the IPython input history. Usage: ipython-get-history.py sessionnumber [outputfile] If outputfile is not given, the relevant history is written to stdout. If outputfile has a .py extension, the translated history (without IPython's special syntax) will be extracted. Example: ./ipython-get-history.py 57 record.ipy This script is a simple demonstration of HistoryAccessor. It should be possible to build much more flexible and powerful tools to browse and pull from the history database. """ import sys from pathlib import Path from IPython.core.history import HistoryAccessor session_number = int(sys.argv[1]) if len(sys.argv) > 2: filepath = Path(sys.argv[2]) dest = open(filepath, "w", encoding="utf-8") raw = not filepath.name.endswith(".py") else: dest = sys.stdout raw = True with dest: dest.write("# coding: utf-8\n") # Profiles other than 'default' can be specified here with a profile= argument: hist = HistoryAccessor() for session, lineno, cell in hist.get_range(session=session_number, raw=raw): dest.write(cell + '\n')
bsd-3-clause
c529b5345687c19586f41fa4407cb6ac
27.4
83
0.725352
3.664516
false
false
false
false
ipython/ipython
IPython/core/tests/test_inputtransformer2_line.py
1
2971
"""Tests for the line-based transformers in IPython.core.inputtransformer2 Line-based transformers are the simpler ones; token-based transformers are more complex. See test_inputtransformer2 for tests for token-based transformers. """ from IPython.core import inputtransformer2 as ipt2 CELL_MAGIC = ("""\ %%foo arg body 1 body 2 """, """\ get_ipython().run_cell_magic('foo', 'arg', 'body 1\\nbody 2\\n') """) def test_cell_magic(): for sample, expected in [CELL_MAGIC]: assert ipt2.cell_magic(sample.splitlines(keepends=True)) == expected.splitlines( keepends=True ) CLASSIC_PROMPT = ("""\ >>> for a in range(5): ... print(a) """, """\ for a in range(5): print(a) """) CLASSIC_PROMPT_L2 = ("""\ for a in range(5): ... print(a) ... print(a ** 2) """, """\ for a in range(5): print(a) print(a ** 2) """) def test_classic_prompt(): for sample, expected in [CLASSIC_PROMPT, CLASSIC_PROMPT_L2]: assert ipt2.classic_prompt( sample.splitlines(keepends=True) ) == expected.splitlines(keepends=True) IPYTHON_PROMPT = ("""\ In [1]: for a in range(5): ...: print(a) """, """\ for a in range(5): print(a) """) IPYTHON_PROMPT_L2 = ("""\ for a in range(5): ...: print(a) ...: print(a ** 2) """, """\ for a in range(5): print(a) print(a ** 2) """) IPYTHON_PROMPT_VI_INS = ( """\ [ins] In [11]: def a(): ...: 123 ...: ...: 123 """, """\ def a(): 123 123 """, ) IPYTHON_PROMPT_VI_NAV = ( """\ [nav] In [11]: def a(): ...: 123 ...: ...: 123 """, """\ def a(): 123 123 """, ) def test_ipython_prompt(): for sample, expected in [ IPYTHON_PROMPT, IPYTHON_PROMPT_L2, IPYTHON_PROMPT_VI_INS, IPYTHON_PROMPT_VI_NAV, ]: assert ipt2.ipython_prompt( sample.splitlines(keepends=True) ) == expected.splitlines(keepends=True) INDENT_SPACES = ("""\ if True: a = 3 """, """\ if True: a = 3 """) INDENT_TABS = ("""\ \tif True: \t\tb = 4 """, """\ if True: \tb = 4 """) def test_leading_indent(): for sample, expected in [INDENT_SPACES, INDENT_TABS]: assert ipt2.leading_indent( sample.splitlines(keepends=True) ) == expected.splitlines(keepends=True) LEADING_EMPTY_LINES = ("""\ \t if True: a = 3 b = 4 """, """\ if True: a = 3 b = 4 """) ONLY_EMPTY_LINES = ("""\ \t """, """\ \t """) def test_leading_empty_lines(): for sample, expected in [LEADING_EMPTY_LINES, ONLY_EMPTY_LINES]: assert ipt2.leading_empty_lines( sample.splitlines(keepends=True) ) == expected.splitlines(keepends=True) CRLF_MAGIC = ([ "%%ls\r\n" ], [ "get_ipython().run_cell_magic('ls', '', '')\n" ]) def test_crlf_magic(): for sample, expected in [CRLF_MAGIC]: assert ipt2.cell_magic(sample) == expected
bsd-3-clause
b549096b76bb1eb869bfc221305e9018
16.790419
88
0.539886
2.99496
false
true
false
false
ipython/ipython
IPython/core/release.py
1
2179
# -*- coding: utf-8 -*- """Release data for the IPython project.""" #----------------------------------------------------------------------------- # Copyright (c) 2008, IPython Development Team. # Copyright (c) 2001, Fernando Perez <fernando.perez@colorado.edu> # Copyright (c) 2001, Janko Hauser <jhauser@zscout.de> # Copyright (c) 2001, Nathaniel Gray <n8gray@caltech.edu> # # Distributed under the terms of the Modified BSD License. # # The full license is in the file COPYING.txt, distributed with this software. #----------------------------------------------------------------------------- # IPython version information. An empty _version_extra corresponds to a full # release. 'dev' as a _version_extra string means this is a development # version _version_major = 8 _version_minor = 7 _version_patch = 0 _version_extra = ".dev" # _version_extra = "rc1" # _version_extra = "" # Uncomment this for full releases # Construct full version string from these. _ver = [_version_major, _version_minor, _version_patch] __version__ = '.'.join(map(str, _ver)) if _version_extra: __version__ = __version__ + _version_extra version = __version__ # backwards compatibility name version_info = (_version_major, _version_minor, _version_patch, _version_extra) # Change this when incrementing the kernel protocol version kernel_protocol_version_info = (5, 0) kernel_protocol_version = "%i.%i" % kernel_protocol_version_info license = "BSD-3-Clause" authors = {'Fernando' : ('Fernando Perez','fperez.net@gmail.com'), 'Janko' : ('Janko Hauser','jhauser@zscout.de'), 'Nathan' : ('Nathaniel Gray','n8gray@caltech.edu'), 'Ville' : ('Ville Vainio','vivainio@gmail.com'), 'Brian' : ('Brian E Granger', 'ellisonbg@gmail.com'), 'Min' : ('Min Ragan-Kelley', 'benjaminrk@gmail.com'), 'Thomas' : ('Thomas A. Kluyver', 'takowl@gmail.com'), 'Jorgen' : ('Jorgen Stenarson', 'jorgen.stenarson@bostream.nu'), 'Matthias' : ('Matthias Bussonnier', 'bussonniermatthias@gmail.com'), } author = 'The IPython Development Team' author_email = 'ipython-dev@python.org'
bsd-3-clause
0a9488f0a18f1cdf92f15ac3c1618c1e
39.351852
80
0.611749
3.213864
false
false
false
false
ipython/ipython
IPython/terminal/pt_inputhooks/osx.py
1
4789
"""Inputhook for OS X Calls NSApp / CoreFoundation APIs via ctypes. """ # obj-c boilerplate from appnope, used under BSD 2-clause import ctypes import ctypes.util from threading import Event objc = ctypes.cdll.LoadLibrary(ctypes.util.find_library("objc")) # type: ignore void_p = ctypes.c_void_p objc.objc_getClass.restype = void_p objc.sel_registerName.restype = void_p objc.objc_msgSend.restype = void_p objc.objc_msgSend.argtypes = [void_p, void_p] msg = objc.objc_msgSend def _utf8(s): """ensure utf8 bytes""" if not isinstance(s, bytes): s = s.encode('utf8') return s def n(name): """create a selector name (for ObjC methods)""" return objc.sel_registerName(_utf8(name)) def C(classname): """get an ObjC Class by name""" return objc.objc_getClass(_utf8(classname)) # end obj-c boilerplate from appnope # CoreFoundation C-API calls we will use: CoreFoundation = ctypes.cdll.LoadLibrary(ctypes.util.find_library("CoreFoundation")) # type: ignore CFFileDescriptorCreate = CoreFoundation.CFFileDescriptorCreate CFFileDescriptorCreate.restype = void_p CFFileDescriptorCreate.argtypes = [void_p, ctypes.c_int, ctypes.c_bool, void_p, void_p] CFFileDescriptorGetNativeDescriptor = CoreFoundation.CFFileDescriptorGetNativeDescriptor CFFileDescriptorGetNativeDescriptor.restype = ctypes.c_int CFFileDescriptorGetNativeDescriptor.argtypes = [void_p] CFFileDescriptorEnableCallBacks = CoreFoundation.CFFileDescriptorEnableCallBacks CFFileDescriptorEnableCallBacks.restype = None CFFileDescriptorEnableCallBacks.argtypes = [void_p, ctypes.c_ulong] CFFileDescriptorCreateRunLoopSource = CoreFoundation.CFFileDescriptorCreateRunLoopSource CFFileDescriptorCreateRunLoopSource.restype = void_p CFFileDescriptorCreateRunLoopSource.argtypes = [void_p, void_p, void_p] CFRunLoopGetCurrent = CoreFoundation.CFRunLoopGetCurrent CFRunLoopGetCurrent.restype = void_p CFRunLoopAddSource = CoreFoundation.CFRunLoopAddSource CFRunLoopAddSource.restype = None CFRunLoopAddSource.argtypes = [void_p, void_p, void_p] CFRelease = CoreFoundation.CFRelease CFRelease.restype = None CFRelease.argtypes = [void_p] CFFileDescriptorInvalidate = CoreFoundation.CFFileDescriptorInvalidate CFFileDescriptorInvalidate.restype = None CFFileDescriptorInvalidate.argtypes = [void_p] # From CFFileDescriptor.h kCFFileDescriptorReadCallBack = 1 kCFRunLoopCommonModes = void_p.in_dll(CoreFoundation, 'kCFRunLoopCommonModes') def _NSApp(): """Return the global NSApplication instance (NSApp)""" objc.objc_msgSend.argtypes = [void_p, void_p] return msg(C('NSApplication'), n('sharedApplication')) def _wake(NSApp): """Wake the Application""" objc.objc_msgSend.argtypes = [ void_p, void_p, void_p, void_p, void_p, void_p, void_p, void_p, void_p, void_p, void_p, ] event = msg( C("NSEvent"), n( "otherEventWithType:location:modifierFlags:" "timestamp:windowNumber:context:subtype:data1:data2:" ), 15, # Type 0, # location 0, # flags 0, # timestamp 0, # window None, # context 0, # subtype 0, # data1 0, # data2 ) objc.objc_msgSend.argtypes = [void_p, void_p, void_p, void_p] msg(NSApp, n('postEvent:atStart:'), void_p(event), True) _triggered = Event() def _input_callback(fdref, flags, info): """Callback to fire when there's input to be read""" _triggered.set() CFFileDescriptorInvalidate(fdref) CFRelease(fdref) NSApp = _NSApp() objc.objc_msgSend.argtypes = [void_p, void_p, void_p] msg(NSApp, n('stop:'), NSApp) _wake(NSApp) _c_callback_func_type = ctypes.CFUNCTYPE(None, void_p, void_p, void_p) _c_input_callback = _c_callback_func_type(_input_callback) def _stop_on_read(fd): """Register callback to stop eventloop when there's data on fd""" _triggered.clear() fdref = CFFileDescriptorCreate(None, fd, False, _c_input_callback, None) CFFileDescriptorEnableCallBacks(fdref, kCFFileDescriptorReadCallBack) source = CFFileDescriptorCreateRunLoopSource(None, fdref, 0) loop = CFRunLoopGetCurrent() CFRunLoopAddSource(loop, source, kCFRunLoopCommonModes) CFRelease(source) def inputhook(context): """Inputhook for Cocoa (NSApp)""" NSApp = _NSApp() _stop_on_read(context.fileno()) objc.objc_msgSend.argtypes = [void_p, void_p] msg(NSApp, n('run')) if not _triggered.is_set(): # app closed without firing callback, # probably due to last window being closed. # Run the loop manually in this case, # since there may be events still to process (#9734) CoreFoundation.CFRunLoopRun()
bsd-3-clause
9f6bc4d3772d385a4ab63898a2861a51
29.503185
100
0.704322
3.377292
false
false
false
false
ipython/ipython
IPython/lib/guisupport.py
3
6305
# coding: utf-8 """ Support for creating GUI apps and starting event loops. IPython's GUI integration allows interactive plotting and GUI usage in IPython session. IPython has two different types of GUI integration: 1. The terminal based IPython supports GUI event loops through Python's PyOS_InputHook. PyOS_InputHook is a hook that Python calls periodically whenever raw_input is waiting for a user to type code. We implement GUI support in the terminal by setting PyOS_InputHook to a function that iterates the event loop for a short while. It is important to note that in this situation, the real GUI event loop is NOT run in the normal manner, so you can't use the normal means to detect that it is running. 2. In the two process IPython kernel/frontend, the GUI event loop is run in the kernel. In this case, the event loop is run in the normal manner by calling the function or method of the GUI toolkit that starts the event loop. In addition to starting the GUI event loops in one of these two ways, IPython will *always* create an appropriate GUI application object when GUi integration is enabled. If you want your GUI apps to run in IPython you need to do two things: 1. Test to see if there is already an existing main application object. If there is, you should use it. If there is not an existing application object you should create one. 2. Test to see if the GUI event loop is running. If it is, you should not start it. If the event loop is not running you may start it. This module contains functions for each toolkit that perform these things in a consistent manner. Because of how PyOS_InputHook runs the event loop you cannot detect if the event loop is running using the traditional calls (such as ``wx.GetApp.IsMainLoopRunning()`` in wxPython). If PyOS_InputHook is set These methods will return a false negative. That is, they will say the event loop is not running, when is actually is. To work around this limitation we proposed the following informal protocol: * Whenever someone starts the event loop, they *must* set the ``_in_event_loop`` attribute of the main application object to ``True``. This should be done regardless of how the event loop is actually run. * Whenever someone stops the event loop, they *must* set the ``_in_event_loop`` attribute of the main application object to ``False``. * If you want to see if the event loop is running, you *must* use ``hasattr`` to see if ``_in_event_loop`` attribute has been set. If it is set, you *must* use its value. If it has not been set, you can query the toolkit in the normal manner. * If you want GUI support and no one else has created an application or started the event loop you *must* do this. We don't want projects to attempt to defer these things to someone else if they themselves need it. The functions below implement this logic for each GUI toolkit. If you need to create custom application subclasses, you will likely have to modify this code for your own purposes. This code can be copied into your own project so you don't have to depend on IPython. """ # Copyright (c) IPython Development Team. # Distributed under the terms of the Modified BSD License. from IPython.core.getipython import get_ipython #----------------------------------------------------------------------------- # wx #----------------------------------------------------------------------------- def get_app_wx(*args, **kwargs): """Create a new wx app or return an exiting one.""" import wx app = wx.GetApp() if app is None: if 'redirect' not in kwargs: kwargs['redirect'] = False app = wx.PySimpleApp(*args, **kwargs) return app def is_event_loop_running_wx(app=None): """Is the wx event loop running.""" # New way: check attribute on shell instance ip = get_ipython() if ip is not None: if ip.active_eventloop and ip.active_eventloop == 'wx': return True # Fall through to checking the application, because Wx has a native way # to check if the event loop is running, unlike Qt. # Old way: check Wx application if app is None: app = get_app_wx() if hasattr(app, '_in_event_loop'): return app._in_event_loop else: return app.IsMainLoopRunning() def start_event_loop_wx(app=None): """Start the wx event loop in a consistent manner.""" if app is None: app = get_app_wx() if not is_event_loop_running_wx(app): app._in_event_loop = True app.MainLoop() app._in_event_loop = False else: app._in_event_loop = True #----------------------------------------------------------------------------- # qt4 #----------------------------------------------------------------------------- def get_app_qt4(*args, **kwargs): """Create a new qt4 app or return an existing one.""" from IPython.external.qt_for_kernel import QtGui app = QtGui.QApplication.instance() if app is None: if not args: args = ([''],) app = QtGui.QApplication(*args, **kwargs) return app def is_event_loop_running_qt4(app=None): """Is the qt4 event loop running.""" # New way: check attribute on shell instance ip = get_ipython() if ip is not None: return ip.active_eventloop and ip.active_eventloop.startswith('qt') # Old way: check attribute on QApplication singleton if app is None: app = get_app_qt4(['']) if hasattr(app, '_in_event_loop'): return app._in_event_loop else: # Does qt4 provide a other way to detect this? return False def start_event_loop_qt4(app=None): """Start the qt4 event loop in a consistent manner.""" if app is None: app = get_app_qt4(['']) if not is_event_loop_running_qt4(app): app._in_event_loop = True app.exec_() app._in_event_loop = False else: app._in_event_loop = True #----------------------------------------------------------------------------- # Tk #----------------------------------------------------------------------------- #----------------------------------------------------------------------------- # gtk #-----------------------------------------------------------------------------
bsd-3-clause
efafd2f4a2d0dfce963fd5951df77706
39.677419
80
0.625852
4.306694
false
false
false
false
ipython/ipython
IPython/external/qt_loaders.py
1
11158
""" This module contains factory functions that attempt to return Qt submodules from the various python Qt bindings. It also protects against double-importing Qt with different bindings, which is unstable and likely to crash This is used primarily by qt and qt_for_kernel, and shouldn't be accessed directly from the outside """ import importlib.abc import sys import types from functools import partial, lru_cache import operator # ### Available APIs. # Qt6 QT_API_PYQT6 = "pyqt6" QT_API_PYSIDE6 = "pyside6" # Qt5 QT_API_PYQT5 = 'pyqt5' QT_API_PYSIDE2 = 'pyside2' # Qt4 QT_API_PYQT = "pyqt" # Force version 2 QT_API_PYQTv1 = "pyqtv1" # Force version 2 QT_API_PYSIDE = "pyside" QT_API_PYQT_DEFAULT = "pyqtdefault" # use system default for version 1 vs. 2 api_to_module = { # Qt6 QT_API_PYQT6: "PyQt6", QT_API_PYSIDE6: "PySide6", # Qt5 QT_API_PYQT5: "PyQt5", QT_API_PYSIDE2: "PySide2", # Qt4 QT_API_PYSIDE: "PySide", QT_API_PYQT: "PyQt4", QT_API_PYQTv1: "PyQt4", # default QT_API_PYQT_DEFAULT: "PyQt6", } class ImportDenier(importlib.abc.MetaPathFinder): """Import Hook that will guard against bad Qt imports once IPython commits to a specific binding """ def __init__(self): self.__forbidden = set() def forbid(self, module_name): sys.modules.pop(module_name, None) self.__forbidden.add(module_name) def find_spec(self, fullname, path, target=None): if path: return if fullname in self.__forbidden: raise ImportError( """ Importing %s disabled by IPython, which has already imported an Incompatible QT Binding: %s """ % (fullname, loaded_api()) ) ID = ImportDenier() sys.meta_path.insert(0, ID) def commit_api(api): """Commit to a particular API, and trigger ImportErrors on subsequent dangerous imports""" modules = set(api_to_module.values()) modules.remove(api_to_module[api]) for mod in modules: ID.forbid(mod) def loaded_api(): """Return which API is loaded, if any If this returns anything besides None, importing any other Qt binding is unsafe. Returns ------- None, 'pyside6', 'pyqt6', 'pyside2', 'pyside', 'pyqt', 'pyqt5', 'pyqtv1' """ if sys.modules.get("PyQt6.QtCore"): return QT_API_PYQT6 elif sys.modules.get("PySide6.QtCore"): return QT_API_PYSIDE6 elif sys.modules.get("PyQt5.QtCore"): return QT_API_PYQT5 elif sys.modules.get("PySide2.QtCore"): return QT_API_PYSIDE2 elif sys.modules.get("PyQt4.QtCore"): if qtapi_version() == 2: return QT_API_PYQT else: return QT_API_PYQTv1 elif sys.modules.get("PySide.QtCore"): return QT_API_PYSIDE return None def has_binding(api): """Safely check for PyQt4/5, PySide or PySide2, without importing submodules Parameters ---------- api : str [ 'pyqtv1' | 'pyqt' | 'pyqt5' | 'pyside' | 'pyside2' | 'pyqtdefault'] Which module to check for Returns ------- True if the relevant module appears to be importable """ module_name = api_to_module[api] from importlib.util import find_spec required = ['QtCore', 'QtGui', 'QtSvg'] if api in (QT_API_PYQT5, QT_API_PYSIDE2, QT_API_PYQT6, QT_API_PYSIDE6): # QT5 requires QtWidgets too required.append('QtWidgets') for submod in required: try: spec = find_spec('%s.%s' % (module_name, submod)) except ImportError: # Package (e.g. PyQt5) not found return False else: if spec is None: # Submodule (e.g. PyQt5.QtCore) not found return False if api == QT_API_PYSIDE: # We can also safely check PySide version import PySide return PySide.__version_info__ >= (1, 0, 3) return True def qtapi_version(): """Return which QString API has been set, if any Returns ------- The QString API version (1 or 2), or None if not set """ try: import sip except ImportError: # as of PyQt5 5.11, sip is no longer available as a top-level # module and needs to be imported from the PyQt5 namespace try: from PyQt5 import sip except ImportError: return try: return sip.getapi('QString') except ValueError: return def can_import(api): """Safely query whether an API is importable, without importing it""" if not has_binding(api): return False current = loaded_api() if api == QT_API_PYQT_DEFAULT: return current in [QT_API_PYQT6, None] else: return current in [api, None] def import_pyqt4(version=2): """ Import PyQt4 Parameters ---------- version : 1, 2, or None Which QString/QVariant API to use. Set to None to use the system default ImportErrors raised within this function are non-recoverable """ # The new-style string API (version=2) automatically # converts QStrings to Unicode Python strings. Also, automatically unpacks # QVariants to their underlying objects. import sip if version is not None: sip.setapi('QString', version) sip.setapi('QVariant', version) from PyQt4 import QtGui, QtCore, QtSvg if QtCore.PYQT_VERSION < 0x040700: raise ImportError("IPython requires PyQt4 >= 4.7, found %s" % QtCore.PYQT_VERSION_STR) # Alias PyQt-specific functions for PySide compatibility. QtCore.Signal = QtCore.pyqtSignal QtCore.Slot = QtCore.pyqtSlot # query for the API version (in case version == None) version = sip.getapi('QString') api = QT_API_PYQTv1 if version == 1 else QT_API_PYQT return QtCore, QtGui, QtSvg, api def import_pyqt5(): """ Import PyQt5 ImportErrors raised within this function are non-recoverable """ from PyQt5 import QtCore, QtSvg, QtWidgets, QtGui # Alias PyQt-specific functions for PySide compatibility. QtCore.Signal = QtCore.pyqtSignal QtCore.Slot = QtCore.pyqtSlot # Join QtGui and QtWidgets for Qt4 compatibility. QtGuiCompat = types.ModuleType('QtGuiCompat') QtGuiCompat.__dict__.update(QtGui.__dict__) QtGuiCompat.__dict__.update(QtWidgets.__dict__) api = QT_API_PYQT5 return QtCore, QtGuiCompat, QtSvg, api def import_pyqt6(): """ Import PyQt6 ImportErrors raised within this function are non-recoverable """ from PyQt6 import QtCore, QtSvg, QtWidgets, QtGui # Alias PyQt-specific functions for PySide compatibility. QtCore.Signal = QtCore.pyqtSignal QtCore.Slot = QtCore.pyqtSlot # Join QtGui and QtWidgets for Qt4 compatibility. QtGuiCompat = types.ModuleType("QtGuiCompat") QtGuiCompat.__dict__.update(QtGui.__dict__) QtGuiCompat.__dict__.update(QtWidgets.__dict__) api = QT_API_PYQT6 return QtCore, QtGuiCompat, QtSvg, api def import_pyside(): """ Import PySide ImportErrors raised within this function are non-recoverable """ from PySide import QtGui, QtCore, QtSvg return QtCore, QtGui, QtSvg, QT_API_PYSIDE def import_pyside2(): """ Import PySide2 ImportErrors raised within this function are non-recoverable """ from PySide2 import QtGui, QtCore, QtSvg, QtWidgets, QtPrintSupport # Join QtGui and QtWidgets for Qt4 compatibility. QtGuiCompat = types.ModuleType('QtGuiCompat') QtGuiCompat.__dict__.update(QtGui.__dict__) QtGuiCompat.__dict__.update(QtWidgets.__dict__) QtGuiCompat.__dict__.update(QtPrintSupport.__dict__) return QtCore, QtGuiCompat, QtSvg, QT_API_PYSIDE2 def import_pyside6(): """ Import PySide6 ImportErrors raised within this function are non-recoverable """ from PySide6 import QtGui, QtCore, QtSvg, QtWidgets, QtPrintSupport # Join QtGui and QtWidgets for Qt4 compatibility. QtGuiCompat = types.ModuleType("QtGuiCompat") QtGuiCompat.__dict__.update(QtGui.__dict__) QtGuiCompat.__dict__.update(QtWidgets.__dict__) QtGuiCompat.__dict__.update(QtPrintSupport.__dict__) return QtCore, QtGuiCompat, QtSvg, QT_API_PYSIDE6 def load_qt(api_options): """ Attempt to import Qt, given a preference list of permissible bindings It is safe to call this function multiple times. Parameters ---------- api_options : List of strings The order of APIs to try. Valid items are 'pyside', 'pyside2', 'pyqt', 'pyqt5', 'pyqtv1' and 'pyqtdefault' Returns ------- A tuple of QtCore, QtGui, QtSvg, QT_API The first three are the Qt modules. The last is the string indicating which module was loaded. Raises ------ ImportError, if it isn't possible to import any requested bindings (either because they aren't installed, or because an incompatible library has already been installed) """ loaders = { # Qt6 QT_API_PYQT6: import_pyqt6, QT_API_PYSIDE6: import_pyside6, # Qt5 QT_API_PYQT5: import_pyqt5, QT_API_PYSIDE2: import_pyside2, # Qt4 QT_API_PYSIDE: import_pyside, QT_API_PYQT: import_pyqt4, QT_API_PYQTv1: partial(import_pyqt4, version=1), # default QT_API_PYQT_DEFAULT: import_pyqt6, } for api in api_options: if api not in loaders: raise RuntimeError( "Invalid Qt API %r, valid values are: %s" % (api, ", ".join(["%r" % k for k in loaders.keys()]))) if not can_import(api): continue #cannot safely recover from an ImportError during this result = loaders[api]() api = result[-1] # changed if api = QT_API_PYQT_DEFAULT commit_api(api) return result else: raise ImportError(""" Could not load requested Qt binding. Please ensure that PyQt4 >= 4.7, PyQt5, PySide >= 1.0.3 or PySide2 is available, and only one is imported per session. Currently-imported Qt library: %r PyQt4 available (requires QtCore, QtGui, QtSvg): %s PyQt5 available (requires QtCore, QtGui, QtSvg, QtWidgets): %s PySide >= 1.0.3 installed: %s PySide2 installed: %s Tried to load: %r """ % (loaded_api(), has_binding(QT_API_PYQT), has_binding(QT_API_PYQT5), has_binding(QT_API_PYSIDE), has_binding(QT_API_PYSIDE2), api_options)) def enum_factory(QT_API, QtCore): """Construct an enum helper to account for PyQt5 <-> PyQt6 changes.""" @lru_cache(None) def _enum(name): # foo.bar.Enum.Entry (PyQt6) <=> foo.bar.Entry (non-PyQt6). return operator.attrgetter( name if QT_API == QT_API_PYQT6 else name.rpartition(".")[0] )(sys.modules[QtCore.__package__]) return _enum
bsd-3-clause
7548a9e8544ff5e72ca3f95595175ee8
26.964912
83
0.6209
3.67402
false
false
false
false
ipython/ipython
IPython/terminal/pt_inputhooks/wx.py
2
7132
"""Enable wxPython to be used interactively in prompt_toolkit """ import sys import signal import time from timeit import default_timer as clock import wx def ignore_keyboardinterrupts(func): """Decorator which causes KeyboardInterrupt exceptions to be ignored during execution of the decorated function. This is used by the inputhook functions to handle the event where the user presses CTRL+C while IPython is idle, and the inputhook loop is running. In this case, we want to ignore interrupts. """ def wrapper(*args, **kwargs): try: func(*args, **kwargs) except KeyboardInterrupt: pass return wrapper @ignore_keyboardinterrupts def inputhook_wx1(context): """Run the wx event loop by processing pending events only. This approach seems to work, but its performance is not great as it relies on having PyOS_InputHook called regularly. """ app = wx.GetApp() if app is not None: assert wx.Thread_IsMain() # Make a temporary event loop and process system events until # there are no more waiting, then allow idle events (which # will also deal with pending or posted wx events.) evtloop = wx.EventLoop() ea = wx.EventLoopActivator(evtloop) while evtloop.Pending(): evtloop.Dispatch() app.ProcessIdle() del ea return 0 class EventLoopTimer(wx.Timer): def __init__(self, func): self.func = func wx.Timer.__init__(self) def Notify(self): self.func() class EventLoopRunner(object): def Run(self, time, input_is_ready): self.input_is_ready = input_is_ready self.evtloop = wx.EventLoop() self.timer = EventLoopTimer(self.check_stdin) self.timer.Start(time) self.evtloop.Run() def check_stdin(self): if self.input_is_ready(): self.timer.Stop() self.evtloop.Exit() @ignore_keyboardinterrupts def inputhook_wx2(context): """Run the wx event loop, polling for stdin. This version runs the wx eventloop for an undetermined amount of time, during which it periodically checks to see if anything is ready on stdin. If anything is ready on stdin, the event loop exits. The argument to elr.Run controls how often the event loop looks at stdin. This determines the responsiveness at the keyboard. A setting of 1000 enables a user to type at most 1 char per second. I have found that a setting of 10 gives good keyboard response. We can shorten it further, but eventually performance would suffer from calling select/kbhit too often. """ app = wx.GetApp() if app is not None: assert wx.Thread_IsMain() elr = EventLoopRunner() # As this time is made shorter, keyboard response improves, but idle # CPU load goes up. 10 ms seems like a good compromise. elr.Run(time=10, # CHANGE time here to control polling interval input_is_ready=context.input_is_ready) return 0 @ignore_keyboardinterrupts def inputhook_wx3(context): """Run the wx event loop by processing pending events only. This is like inputhook_wx1, but it keeps processing pending events until stdin is ready. After processing all pending events, a call to time.sleep is inserted. This is needed, otherwise, CPU usage is at 100%. This sleep time should be tuned though for best performance. """ app = wx.GetApp() if app is not None: assert wx.Thread_IsMain() # The import of wx on Linux sets the handler for signal.SIGINT # to 0. This is a bug in wx or gtk. We fix by just setting it # back to the Python default. if not callable(signal.getsignal(signal.SIGINT)): signal.signal(signal.SIGINT, signal.default_int_handler) evtloop = wx.EventLoop() ea = wx.EventLoopActivator(evtloop) t = clock() while not context.input_is_ready(): while evtloop.Pending(): t = clock() evtloop.Dispatch() app.ProcessIdle() # We need to sleep at this point to keep the idle CPU load # low. However, if sleep to long, GUI response is poor. As # a compromise, we watch how often GUI events are being processed # and switch between a short and long sleep time. Here are some # stats useful in helping to tune this. # time CPU load # 0.001 13% # 0.005 3% # 0.01 1.5% # 0.05 0.5% used_time = clock() - t if used_time > 10.0: # print 'Sleep for 1 s' # dbg time.sleep(1.0) elif used_time > 0.1: # Few GUI events coming in, so we can sleep longer # print 'Sleep for 0.05 s' # dbg time.sleep(0.05) else: # Many GUI events coming in, so sleep only very little time.sleep(0.001) del ea return 0 @ignore_keyboardinterrupts def inputhook_wxphoenix(context): """Run the wx event loop until the user provides more input. This input hook is suitable for use with wxPython >= 4 (a.k.a. Phoenix). It uses the same approach to that used in ipykernel.eventloops.loop_wx. The wx.MainLoop is executed, and a wx.Timer is used to periodically poll the context for input. As soon as input is ready, the wx.MainLoop is stopped. """ app = wx.GetApp() if app is None: return if context.input_is_ready(): return assert wx.IsMainThread() # Wx uses milliseconds poll_interval = 100 # Use a wx.Timer to periodically check whether input is ready - as soon as # it is, we exit the main loop timer = wx.Timer() def poll(ev): if context.input_is_ready(): timer.Stop() app.ExitMainLoop() timer.Start(poll_interval) timer.Bind(wx.EVT_TIMER, poll) # The import of wx on Linux sets the handler for signal.SIGINT to 0. This # is a bug in wx or gtk. We fix by just setting it back to the Python # default. if not callable(signal.getsignal(signal.SIGINT)): signal.signal(signal.SIGINT, signal.default_int_handler) # The SetExitOnFrameDelete call allows us to run the wx mainloop without # having a frame open. app.SetExitOnFrameDelete(False) app.MainLoop() # Get the major wx version number to figure out what input hook we should use. major_version = 3 try: major_version = int(wx.__version__[0]) except Exception: pass # Use the phoenix hook on all platforms for wxpython >= 4 if major_version >= 4: inputhook = inputhook_wxphoenix # On OSX, evtloop.Pending() always returns True, regardless of there being # any events pending. As such we can't use implementations 1 or 3 of the # inputhook as those depend on a pending/dispatch loop. elif sys.platform == 'darwin': inputhook = inputhook_wx2 else: inputhook = inputhook_wx3
bsd-3-clause
ba413fa5a600fb7546b17ea8b9b05bf2
31.56621
79
0.640213
3.938156
false
false
false
false
ipython/ipython
IPython/core/magics/basic.py
1
22932
"""Implementation of basic magic functions.""" from logging import error import io import os from pprint import pformat import sys from warnings import warn from traitlets.utils.importstring import import_item from IPython.core import magic_arguments, page from IPython.core.error import UsageError from IPython.core.magic import Magics, magics_class, line_magic, magic_escapes from IPython.utils.text import format_screen, dedent, indent from IPython.testing.skipdoctest import skip_doctest from IPython.utils.ipstruct import Struct class MagicsDisplay(object): def __init__(self, magics_manager, ignore=None): self.ignore = ignore if ignore else [] self.magics_manager = magics_manager def _lsmagic(self): """The main implementation of the %lsmagic""" mesc = magic_escapes['line'] cesc = magic_escapes['cell'] mman = self.magics_manager magics = mman.lsmagic() out = ['Available line magics:', mesc + (' '+mesc).join(sorted([m for m,v in magics['line'].items() if (v not in self.ignore)])), '', 'Available cell magics:', cesc + (' '+cesc).join(sorted([m for m,v in magics['cell'].items() if (v not in self.ignore)])), '', mman.auto_status()] return '\n'.join(out) def _repr_pretty_(self, p, cycle): p.text(self._lsmagic()) def __str__(self): return self._lsmagic() def _jsonable(self): """turn magics dict into jsonable dict of the same structure replaces object instances with their class names as strings """ magic_dict = {} mman = self.magics_manager magics = mman.lsmagic() for key, subdict in magics.items(): d = {} magic_dict[key] = d for name, obj in subdict.items(): try: classname = obj.__self__.__class__.__name__ except AttributeError: classname = 'Other' d[name] = classname return magic_dict def _repr_json_(self): return self._jsonable() @magics_class class BasicMagics(Magics): """Magics that provide central IPython functionality. These are various magics that don't fit into specific categories but that are all part of the base 'IPython experience'.""" @skip_doctest @magic_arguments.magic_arguments() @magic_arguments.argument( '-l', '--line', action='store_true', help="""Create a line magic alias.""" ) @magic_arguments.argument( '-c', '--cell', action='store_true', help="""Create a cell magic alias.""" ) @magic_arguments.argument( 'name', help="""Name of the magic to be created.""" ) @magic_arguments.argument( 'target', help="""Name of the existing line or cell magic.""" ) @magic_arguments.argument( '-p', '--params', default=None, help="""Parameters passed to the magic function.""" ) @line_magic def alias_magic(self, line=''): """Create an alias for an existing line or cell magic. Examples -------- :: In [1]: %alias_magic t timeit Created `%t` as an alias for `%timeit`. Created `%%t` as an alias for `%%timeit`. In [2]: %t -n1 pass 1 loops, best of 3: 954 ns per loop In [3]: %%t -n1 ...: pass ...: 1 loops, best of 3: 954 ns per loop In [4]: %alias_magic --cell whereami pwd UsageError: Cell magic function `%%pwd` not found. In [5]: %alias_magic --line whereami pwd Created `%whereami` as an alias for `%pwd`. In [6]: %whereami Out[6]: u'/home/testuser' In [7]: %alias_magic h history "-p -l 30" --line Created `%h` as an alias for `%history -l 30`. """ args = magic_arguments.parse_argstring(self.alias_magic, line) shell = self.shell mman = self.shell.magics_manager escs = ''.join(magic_escapes.values()) target = args.target.lstrip(escs) name = args.name.lstrip(escs) params = args.params if (params and ((params.startswith('"') and params.endswith('"')) or (params.startswith("'") and params.endswith("'")))): params = params[1:-1] # Find the requested magics. m_line = shell.find_magic(target, 'line') m_cell = shell.find_magic(target, 'cell') if args.line and m_line is None: raise UsageError('Line magic function `%s%s` not found.' % (magic_escapes['line'], target)) if args.cell and m_cell is None: raise UsageError('Cell magic function `%s%s` not found.' % (magic_escapes['cell'], target)) # If --line and --cell are not specified, default to the ones # that are available. if not args.line and not args.cell: if not m_line and not m_cell: raise UsageError( 'No line or cell magic with name `%s` found.' % target ) args.line = bool(m_line) args.cell = bool(m_cell) params_str = "" if params is None else " " + params if args.line: mman.register_alias(name, target, 'line', params) print('Created `%s%s` as an alias for `%s%s%s`.' % ( magic_escapes['line'], name, magic_escapes['line'], target, params_str)) if args.cell: mman.register_alias(name, target, 'cell', params) print('Created `%s%s` as an alias for `%s%s%s`.' % ( magic_escapes['cell'], name, magic_escapes['cell'], target, params_str)) @line_magic def lsmagic(self, parameter_s=''): """List currently available magic functions.""" return MagicsDisplay(self.shell.magics_manager, ignore=[]) def _magic_docs(self, brief=False, rest=False): """Return docstrings from magic functions.""" mman = self.shell.magics_manager docs = mman.lsmagic_docs(brief, missing='No documentation') if rest: format_string = '**%s%s**::\n\n%s\n\n' else: format_string = '%s%s:\n%s\n' return ''.join( [format_string % (magic_escapes['line'], fname, indent(dedent(fndoc))) for fname, fndoc in sorted(docs['line'].items())] + [format_string % (magic_escapes['cell'], fname, indent(dedent(fndoc))) for fname, fndoc in sorted(docs['cell'].items())] ) @line_magic def magic(self, parameter_s=''): """Print information about the magic function system. Supported formats: -latex, -brief, -rest """ mode = '' try: mode = parameter_s.split()[0][1:] except IndexError: pass brief = (mode == 'brief') rest = (mode == 'rest') magic_docs = self._magic_docs(brief, rest) if mode == 'latex': print(self.format_latex(magic_docs)) return else: magic_docs = format_screen(magic_docs) out = [""" IPython's 'magic' functions =========================== The magic function system provides a series of functions which allow you to control the behavior of IPython itself, plus a lot of system-type features. There are two kinds of magics, line-oriented and cell-oriented. Line magics are prefixed with the % character and work much like OS command-line calls: they get as an argument the rest of the line, where arguments are passed without parentheses or quotes. For example, this will time the given statement:: %timeit range(1000) Cell magics are prefixed with a double %%, and they are functions that get as an argument not only the rest of the line, but also the lines below it in a separate argument. These magics are called with two arguments: the rest of the call line and the body of the cell, consisting of the lines below the first. For example:: %%timeit x = numpy.random.randn((100, 100)) numpy.linalg.svd(x) will time the execution of the numpy svd routine, running the assignment of x as part of the setup phase, which is not timed. In a line-oriented client (the terminal or Qt console IPython), starting a new input with %% will automatically enter cell mode, and IPython will continue reading input until a blank line is given. In the notebook, simply type the whole cell as one entity, but keep in mind that the %% escape can only be at the very start of the cell. NOTE: If you have 'automagic' enabled (via the command line option or with the %automagic function), you don't need to type in the % explicitly for line magics; cell magics always require an explicit '%%' escape. By default, IPython ships with automagic on, so you should only rarely need the % escape. Example: typing '%cd mydir' (without the quotes) changes your working directory to 'mydir', if it exists. For a list of the available magic functions, use %lsmagic. For a description of any of them, type %magic_name?, e.g. '%cd?'. Currently the magic system has the following functions:""", magic_docs, "Summary of magic functions (from %slsmagic):" % magic_escapes['line'], str(self.lsmagic()), ] page.page('\n'.join(out)) @line_magic def page(self, parameter_s=''): """Pretty print the object and display it through a pager. %page [options] OBJECT If no object is given, use _ (last output). Options: -r: page str(object), don't pretty-print it.""" # After a function contributed by Olivier Aubert, slightly modified. # Process options/args opts, args = self.parse_options(parameter_s, 'r') raw = 'r' in opts oname = args and args or '_' info = self.shell._ofind(oname) if info['found']: if raw: txt = str(info["obj"]) else: txt = pformat(info["obj"]) page.page(txt) else: print('Object `%s` not found' % oname) @line_magic def pprint(self, parameter_s=''): """Toggle pretty printing on/off.""" ptformatter = self.shell.display_formatter.formatters['text/plain'] ptformatter.pprint = bool(1 - ptformatter.pprint) print('Pretty printing has been turned', ['OFF','ON'][ptformatter.pprint]) @line_magic def colors(self, parameter_s=''): """Switch color scheme for prompts, info system and exception handlers. Currently implemented schemes: NoColor, Linux, LightBG. Color scheme names are not case-sensitive. Examples -------- To get a plain black and white terminal:: %colors nocolor """ def color_switch_err(name): warn('Error changing %s color schemes.\n%s' % (name, sys.exc_info()[1]), stacklevel=2) new_scheme = parameter_s.strip() if not new_scheme: raise UsageError( "%colors: you must specify a color scheme. See '%colors?'") # local shortcut shell = self.shell # Set shell colour scheme try: shell.colors = new_scheme shell.refresh_style() except: color_switch_err('shell') # Set exception colors try: shell.InteractiveTB.set_colors(scheme = new_scheme) shell.SyntaxTB.set_colors(scheme = new_scheme) except: color_switch_err('exception') # Set info (for 'object?') colors if shell.color_info: try: shell.inspector.set_active_scheme(new_scheme) except: color_switch_err('object inspector') else: shell.inspector.set_active_scheme('NoColor') @line_magic def xmode(self, parameter_s=''): """Switch modes for the exception handlers. Valid modes: Plain, Context, Verbose, and Minimal. If called without arguments, acts as a toggle. When in verbose mode the value `--show` (and `--hide`) will respectively show (or hide) frames with ``__tracebackhide__ = True`` value set. """ def xmode_switch_err(name): warn('Error changing %s exception modes.\n%s' % (name,sys.exc_info()[1])) shell = self.shell if parameter_s.strip() == "--show": shell.InteractiveTB.skip_hidden = False return if parameter_s.strip() == "--hide": shell.InteractiveTB.skip_hidden = True return new_mode = parameter_s.strip().capitalize() try: shell.InteractiveTB.set_mode(mode=new_mode) print('Exception reporting mode:',shell.InteractiveTB.mode) except: xmode_switch_err('user') @line_magic def quickref(self, arg): """ Show a quick reference sheet """ from IPython.core.usage import quick_reference qr = quick_reference + self._magic_docs(brief=True) page.page(qr) @line_magic def doctest_mode(self, parameter_s=''): """Toggle doctest mode on and off. This mode is intended to make IPython behave as much as possible like a plain Python shell, from the perspective of how its prompts, exceptions and output look. This makes it easy to copy and paste parts of a session into doctests. It does so by: - Changing the prompts to the classic ``>>>`` ones. - Changing the exception reporting mode to 'Plain'. - Disabling pretty-printing of output. Note that IPython also supports the pasting of code snippets that have leading '>>>' and '...' prompts in them. This means that you can paste doctests from files or docstrings (even if they have leading whitespace), and the code will execute correctly. You can then use '%history -t' to see the translated history; this will give you the input after removal of all the leading prompts and whitespace, which can be pasted back into an editor. With these features, you can switch into this mode easily whenever you need to do testing and changes to doctests, without having to leave your existing IPython session. """ # Shorthands shell = self.shell meta = shell.meta disp_formatter = self.shell.display_formatter ptformatter = disp_formatter.formatters['text/plain'] # dstore is a data store kept in the instance metadata bag to track any # changes we make, so we can undo them later. dstore = meta.setdefault('doctest_mode',Struct()) save_dstore = dstore.setdefault # save a few values we'll need to recover later mode = save_dstore('mode',False) save_dstore('rc_pprint',ptformatter.pprint) save_dstore('xmode',shell.InteractiveTB.mode) save_dstore('rc_separate_out',shell.separate_out) save_dstore('rc_separate_out2',shell.separate_out2) save_dstore('rc_separate_in',shell.separate_in) save_dstore('rc_active_types',disp_formatter.active_types) if not mode: # turn on # Prompt separators like plain python shell.separate_in = '' shell.separate_out = '' shell.separate_out2 = '' ptformatter.pprint = False disp_formatter.active_types = ['text/plain'] shell.magic('xmode Plain') else: # turn off shell.separate_in = dstore.rc_separate_in shell.separate_out = dstore.rc_separate_out shell.separate_out2 = dstore.rc_separate_out2 ptformatter.pprint = dstore.rc_pprint disp_formatter.active_types = dstore.rc_active_types shell.magic('xmode ' + dstore.xmode) # mode here is the state before we switch; switch_doctest_mode takes # the mode we're switching to. shell.switch_doctest_mode(not mode) # Store new mode and inform dstore.mode = bool(not mode) mode_label = ['OFF','ON'][dstore.mode] print('Doctest mode is:', mode_label) @line_magic def gui(self, parameter_s=''): """Enable or disable IPython GUI event loop integration. %gui [GUINAME] This magic replaces IPython's threaded shells that were activated using the (pylab/wthread/etc.) command line flags. GUI toolkits can now be enabled at runtime and keyboard interrupts should work without any problems. The following toolkits are supported: wxPython, PyQt4, PyGTK, Tk and Cocoa (OSX):: %gui wx # enable wxPython event loop integration %gui qt4|qt # enable PyQt4 event loop integration %gui qt5 # enable PyQt5 event loop integration %gui gtk # enable PyGTK event loop integration %gui gtk3 # enable Gtk3 event loop integration %gui gtk4 # enable Gtk4 event loop integration %gui tk # enable Tk event loop integration %gui osx # enable Cocoa event loop integration # (requires %matplotlib 1.1) %gui # disable all event loop integration WARNING: after any of these has been called you can simply create an application object, but DO NOT start the event loop yourself, as we have already handled that. """ opts, arg = self.parse_options(parameter_s, '') if arg=='': arg = None try: return self.shell.enable_gui(arg) except Exception as e: # print simple error message, rather than traceback if we can't # hook up the GUI error(str(e)) @skip_doctest @line_magic def precision(self, s=''): """Set floating point precision for pretty printing. Can set either integer precision or a format string. If numpy has been imported and precision is an int, numpy display precision will also be set, via ``numpy.set_printoptions``. If no argument is given, defaults will be restored. Examples -------- :: In [1]: from math import pi In [2]: %precision 3 Out[2]: u'%.3f' In [3]: pi Out[3]: 3.142 In [4]: %precision %i Out[4]: u'%i' In [5]: pi Out[5]: 3 In [6]: %precision %e Out[6]: u'%e' In [7]: pi**10 Out[7]: 9.364805e+04 In [8]: %precision Out[8]: u'%r' In [9]: pi**10 Out[9]: 93648.047476082982 """ ptformatter = self.shell.display_formatter.formatters['text/plain'] ptformatter.float_precision = s return ptformatter.float_format @magic_arguments.magic_arguments() @magic_arguments.argument( 'filename', type=str, help='Notebook name or filename' ) @line_magic def notebook(self, s): """Export and convert IPython notebooks. This function can export the current IPython history to a notebook file. For example, to export the history to "foo.ipynb" do "%notebook foo.ipynb". """ args = magic_arguments.parse_argstring(self.notebook, s) outfname = os.path.expanduser(args.filename) from nbformat import write, v4 cells = [] hist = list(self.shell.history_manager.get_range()) if(len(hist)<=1): raise ValueError('History is empty, cannot export') for session, execution_count, source in hist[:-1]: cells.append(v4.new_code_cell( execution_count=execution_count, source=source )) nb = v4.new_notebook(cells=cells) with io.open(outfname, "w", encoding="utf-8") as f: write(nb, f, version=4) @magics_class class AsyncMagics(BasicMagics): @line_magic def autoawait(self, parameter_s): """ Allow to change the status of the autoawait option. This allow you to set a specific asynchronous code runner. If no value is passed, print the currently used asynchronous integration and whether it is activated. It can take a number of value evaluated in the following order: - False/false/off deactivate autoawait integration - True/true/on activate autoawait integration using configured default loop - asyncio/curio/trio activate autoawait integration and use integration with said library. - `sync` turn on the pseudo-sync integration (mostly used for `IPython.embed()` which does not run IPython with a real eventloop and deactivate running asynchronous code. Turning on Asynchronous code with the pseudo sync loop is undefined behavior and may lead IPython to crash. If the passed parameter does not match any of the above and is a python identifier, get said object from user namespace and set it as the runner, and activate autoawait. If the object is a fully qualified object name, attempt to import it and set it as the runner, and activate autoawait. The exact behavior of autoawait is experimental and subject to change across version of IPython and Python. """ param = parameter_s.strip() d = {True: "on", False: "off"} if not param: print("IPython autoawait is `{}`, and set to use `{}`".format( d[self.shell.autoawait], self.shell.loop_runner )) return None if param.lower() in ('false', 'off'): self.shell.autoawait = False return None if param.lower() in ('true', 'on'): self.shell.autoawait = True return None if param in self.shell.loop_runner_map: self.shell.loop_runner, self.shell.autoawait = self.shell.loop_runner_map[param] return None if param in self.shell.user_ns : self.shell.loop_runner = self.shell.user_ns[param] self.shell.autoawait = True return None runner = import_item(param) self.shell.loop_runner = runner self.shell.autoawait = True
bsd-3-clause
d3913db72d4571593746b848447aba0a
33.69289
112
0.586866
4.199231
false
false
false
false
ipython/ipython
IPython/utils/contexts.py
1
1619
# encoding: utf-8 """Miscellaneous context managers. """ import warnings # Copyright (c) IPython Development Team. # Distributed under the terms of the Modified BSD License. class preserve_keys(object): """Preserve a set of keys in a dictionary. Upon entering the context manager the current values of the keys will be saved. Upon exiting, the dictionary will be updated to restore the original value of the preserved keys. Preserved keys which did not exist when entering the context manager will be deleted. Examples -------- >>> d = {'a': 1, 'b': 2, 'c': 3} >>> with preserve_keys(d, 'b', 'c', 'd'): ... del d['a'] ... del d['b'] # will be reset to 2 ... d['c'] = None # will be reset to 3 ... d['d'] = 4 # will be deleted ... d['e'] = 5 ... print(sorted(d.items())) ... [('c', None), ('d', 4), ('e', 5)] >>> print(sorted(d.items())) [('b', 2), ('c', 3), ('e', 5)] """ def __init__(self, dictionary, *keys): self.dictionary = dictionary self.keys = keys def __enter__(self): # Actions to perform upon exiting. to_delete = [] to_update = {} d = self.dictionary for k in self.keys: if k in d: to_update[k] = d[k] else: to_delete.append(k) self.to_delete = to_delete self.to_update = to_update def __exit__(self, *exc_info): d = self.dictionary for k in self.to_delete: d.pop(k, None) d.update(self.to_update)
bsd-3-clause
2302c4dc6208821d7c7e10068c105d19
25.540984
68
0.525015
3.646396
false
false
false
false
lisa-lab/pylearn2
pylearn2/utils/track_version.py
33
7516
#!/usr/bin/env python """ Script to obtain version of Python modules and basic information on the experiment setup (e.g. cpu, os), e.g. * numpy: 1.6.1 | pylearn: a6e634b83d | pylearn2: 57a156beb0 * CPU: x86_64 * OS: Linux-2.6.35.14-106.fc14.x86_64-x86_64-with-fedora-14-Laughlin You can also define the modules to be tracked with the environment variable `PYLEARN2_TRACK_MODULES`. Use ":" to separate module names between them, e.g. `PYLEARN2_TRACK_MODULES = module1:module2:module3` By default, the following modules are tracked: pylearn2, theano, numpy, scipy """ __authors__ = "Olivier Dellaleau and Raul Chandias Ferrari" __copyright__ = "Copyright 2013, Universite de Montreal" __credits__ = ["Olivier Dellaleau", "Raul Chandias Ferrari"] __license__ = "3-clause BSD" __maintainer__ = "Raul Chandias Ferrari" __email__ = "chandiar@iro" import copy import logging import os import platform import socket import subprocess import sys import warnings from theano.compat import six logger = logging.getLogger(__name__) class MetaLibVersion(type): """ Constructor that will be called everytime another's class constructor is called (if the "__metaclass__ = MetaLibVersion" line is present in the other class definition). Parameters ---------- cls : WRITEME name : WRITEME bases : WRITEME dict : WRITEME """ def __init__(cls, name, bases, dict): type.__init__(cls, name, bases, dict) cls.libv = LibVersion() class LibVersion(object): """ Initialize a LibVersion object that will store the version of python packages in a dictionary (versions). The python packages that are supported are: pylearn, pylearn2, theano, jobman, numpy and scipy. The key for the versions dict is the name of the package and the associated value is the version number. """ def __init__(self): self.versions = {} self.str_versions = '' self.exp_env_info = {} self._get_lib_versions() self._get_exp_env_info() def _get_exp_env_info(self): """ Get information about the experimental environment such as the cpu, os and the hostname of the machine on which the experiment is running. """ self.exp_env_info['host'] = socket.gethostname() self.exp_env_info['cpu'] = platform.processor() self.exp_env_info['os'] = platform.platform() if 'theano' in sys.modules: self.exp_env_info['theano_config'] = sys.modules['theano'].config else: self.exp_env_info['theano_config'] = None def _get_lib_versions(self): """Get version of Python packages.""" repos = os.getenv('PYLEARN2_TRACK_MODULES', '') default_repos = 'pylearn2:theano:numpy:scipy' repos = default_repos + ":" + repos repos = set(repos.split(':')) for repo in repos: try: if repo == '': continue __import__(repo) if hasattr(sys.modules[repo], '__version__'): v = sys.modules[repo].__version__ if v != 'unknown': self.versions[repo] = v continue self.versions[repo] = self._get_git_version( self._get_module_parent_path(sys.modules[repo])) except ImportError: self.versions[repo] = None known = copy.copy(self.versions) # Put together all modules with unknown versions. unknown = [k for k, w in known.items() if not w] known = dict((k, w) for k, w in known.items() if w) # Print versions. self.str_versions = ' | '.join( ['%s:%s' % (k, w) for k, w in sorted(six.iteritems(known))] + ['%s:?' % ','.join(sorted(unknown))]) def __str__(self): """ Return version of the Python packages as a string. e.g. numpy:1.6.1 | pylearn:a6e634b83d | pylearn2:57a156beb0 """ return self.str_versions def _get_git_version(self, root): """ Return the git revision of a repository with the letter 'M' appended to the revision if the repo was modified. e.g. 10d3046e85 M Parameters ---------- root : str Root folder of the repository Returns ------- rval : str or None A string with the revision hash, or None if it could not be retrieved (e.g. if it is not actually a git repository) """ if not os.path.isdir(os.path.join(root, '.git')): return None cwd_backup = os.getcwd() try: os.chdir(root) sub_p = subprocess.Popen(['git', 'rev-parse', 'HEAD'], stdout=subprocess.PIPE, stderr=subprocess.PIPE) version = sub_p.communicate()[0][0:10].strip() sub_p = subprocess.Popen(['git', 'diff', '--name-only'], stdout=subprocess.PIPE, stderr=subprocess.PIPE) modified = sub_p.communicate()[0] if len(modified): version += ' M' return version except Exception: pass finally: try: os.chdir(cwd_backup) except Exception: warnings.warn("Could not chdir back to " + cwd_backup) def _get_hg_version(self, root): """Same as `get_git_version` but for a Mercurial repository.""" if not os.path.isdir(os.path.join(root, '.hg')): return None cwd_backup = os.getcwd() try: os.chdir(root) sub_p = subprocess.Popen(['hg', 'parents'], stdout=subprocess.PIPE, stderr=subprocess.PIPE) sub_p_output = sub_p.communicate()[0] finally: os.chdir(cwd_backup) first_line = sub_p_output.split('\n')[0] # The first line looks like: # changeset: 1517:a6e634b83d88 return first_line.split(':')[2][0:10] def _get_module_path(self, module): """Return path to a given module.""" return os.path.realpath(module.__path__[0]) def _get_module_parent_path(self, module): """Return path to the parent directory of a given module.""" return os.path.dirname(self._get_module_path(module)) def print_versions(self): """ Print version of the Python packages as a string. e.g. numpy:1.6.1 | pylearn:a6e634b83d | pylearn2:57a156beb0 """ logger.info(self.__str__()) def print_exp_env_info(self, print_theano_config=False): """ Return basic information about the experiment setup such as the hostname of the machine the experiment was run on, the operating system installed on the machine. Parameters ---------- print_theano_config : bool, optional If True, information about the theano configuration will be displayed. """ logger.info('HOST: {0}'.format(self.exp_env_info['host'])) logger.info('CPU: {0}'.format(self.exp_env_info['cpu'])) logger.info('OS: {0}'.format(self.exp_env_info['os'])) if print_theano_config: logger.info(self.exp_env_info['theano_config'])
bsd-3-clause
b1220beffd88461e1426ca18e532aab1
33.635945
77
0.56546
3.989384
false
false
false
false
lisa-lab/pylearn2
pylearn2/utils/iteration.py
1
36271
""" Iterators providing indices for different kinds of iteration over datasets. Presets: - sequential: iterates through fixed slices of the dataset in sequence - shuffled_sequential: iterates through a shuffled version of the dataset in sequence - random_slice: on each call to next, returns a slice of the dataset, chosen uniformly at random over contiguous slices. Samples with replacement, but still reports that container is empty after num_examples / batch_size calls - random_uniform: on each call to next, returns a random subset of the dataset. Samples with replacement, but still reports that container is empty after num_examples / batch_size calls """ from __future__ import division import warnings import numpy as np from theano.compat import six from pylearn2.space import CompositeSpace from pylearn2.utils import safe_izip, wraps from pylearn2.utils.data_specs import is_flat_specs from pylearn2.utils.exc import reraise_as from pylearn2.utils.rng import make_np_rng import copy # Make sure that the docstring uses restructured text list format. # If you change the module-level docstring, please re-run # pylearn2/doc/scripts/docgen.py and make sure sphinx doesn't issue any # warnings for this file. # This particular docstring was being frequently broken prior to the # addition of this test. # TODO: have nosetests run docgen.py in warning=error mode, remove # tests for specific conditions assert """Presets: - sequential: iterates through fixed slices of the dataset in sequence - s""" in __doc__ class SubsetIterator(object): """ An iterator that returns slices or lists of indices into a dataset of a given fixed size. Parameters ---------- dataset_size : int The number of examples, total, in the dataset. batch_size : int, optional The (typical/maximum) number of examples per batch. Less may be returned in the very last batch if batch size does not evenly divide `dataset_size`. num_batches : int, optional The number of batches to return. Needn't be specified if `batch_size` is specified. If both `batch_size` and `num_batches` are specified then it must be true that `batch_size * num_batches <= dataset_size`. rng : `np.random.RandomState` or seed, optional A `np.random.RandomState` object or the seed to be used to create one. A deterministic default seed is used otherwise. """ # This breaks the doc generation, so until we figure out why, not in the # docstring. # # Attributes # ---------- # batch_size : int # num_batches : int # num_examples : int # uneven : bool # fancy : bool # `True` if this iterator produces lists of indices, # `False` if it produces slices. # stochastic : bool # `True` if this iterator makes use of the random number # generator, and will therefore produce different sequences # depending on the RNG state. `False` otherwise. def __init__(self, dataset_size, batch_size=None, num_batches=None, rng=None): raise NotImplementedError() def next(self): """ Retrieves description of the next batch of examples. Returns ------- next_batch : `slice` or list of int An object describing the indices in the dataset of a batch of data. Either a `slice` object or a list of integers specifying individual indices of examples. Raises ------ StopIteration When there are no more batches to return. """ raise NotImplementedError() def __next__(self): self.next() def __iter__(self): return self # Does this return subsets that need fancy indexing? (i.e. lists # of indices) fancy = False # Does this class make use of random number generators? stochastic = False # Does it ensure that every batch has the same size? uniform_batch_size = False @property def batch_size(self): """ The (maximum) number of examples in each batch. Returns ------- batch_size : int The (maximum) number of examples in each batch. This is either as specified via the constructor, or inferred from the dataset size and the number of batches requested. """ return self._batch_size @property def num_batches(self): """ The total number of batches that the iterator will ever return. Returns ------- num_batches : int The total number of batches the iterator will ever return. This is either as specified via the constructor, or inferred from the dataset size and the batch size. """ return self._num_batches @property def num_examples(self): """ The total number of examples over which the iterator operates. Returns ------- num_examples : int The total number of examples over which the iterator operates. May be less than the dataset size. """ return self.batch_size * self.num_batches @property def uneven(self): """ Whether every batch will be the same size. Returns ------- uneven : bool `True` if returned batches may be of differing sizes, `False` otherwise. """ raise NotImplementedError() class ForcedEvenIterator(SubsetIterator): """ A class which wraps other iterators to ensure equal batch size. This class needs to be completed using type() metaclass, see Examples section to see how to use it. Parameters ---------- dataset_size : int Total number of examples in the dataset batch_size : int or None The size of the batches. If set to None and num_batches is defined, batch_size will be calculated based on dataset_size. num_batches : int or None The number of batch in the dataset. If set to None and batch_size is defined, num_batches will be calculated based on dataset_size. *args : Variable length argument list for _base_iterator_cls **kwargs : Arbitrary keyword arguments for _base_iterator_cls Notes ----- This class can not be initialized because it needs to be completed using type() metaclass. See Examples section for more details. Batches of size unequal to batch_size will be discarded. Those examples will never be visited. Examples -------- >>> dct = ForcedEvenIterator.__dict__.copy() >>> dct["_base_iterator_cls"] = SequentialSubsetIterator >>> dct["fancy"] = SequentialSubsetIterator.fancy >>> dct["stochastic"] = SequentialSubsetIterator.stochastic >>> >>> NewForcedEvenClass = type("ForcedEvenDummyIterator", ... ForcedEvenIterator.__bases__, dct) >>> >>> even_iterator = NewForcedEvenClass(dataset_size=100, ... batch_size=30, num_batches=None) For a shortcut use function as_even() >>> NewForcedEvenClass = as_even(SequentialSubsetIterator) >>> even_iterator = NewForcedEvenClass(dataset_size=100, ... batch_size=30, num_batches=None) """ def __init__(self, dataset_size, batch_size, num_batches, *args, **kwargs): if self.fancy is None or self.stochastic is None or \ self._base_iterator_cls is None: raise ValueError("You must pre-define fancy, stochastic and " "_base_iterator_cls arguments by creating a new " "class using the metaclass type()." "See function as_even() for an example.") if batch_size is None: if num_batches is not None: batch_size = int(dataset_size / num_batches) else: raise ValueError("need one of batch_size, num_batches " "for sequential batch iteration") elif batch_size is not None: if num_batches is not None: max_num_batches = int(dataset_size / batch_size) if num_batches > max_num_batches: raise ValueError("dataset of %d examples can only provide " "%d batches of equal size with batch_size" " %d, but %d batches were requested" % (dataset_size, max_num_batches, batch_size, num_batches)) else: num_batches = int(dataset_size / batch_size) self._base_iterator = self._base_iterator_cls(dataset_size, batch_size, num_batches, *args, **kwargs) # Does it ensure that every batch has the same size? uniform_batch_size = True # Does this return subsets that need fancy indexing? (i.e. lists # of indices) # Needs to be set before initialization. See Examples section in class docs fancy = None # Does this class make use of random number generators? # Needs to be set before initialization. See Examples section in class docs stochastic = None # base iterator that ForcedEvenIterator class wraps # Needs to be set before initialization. See Examples section in class docs _base_iterator_cls = None @property def _dataset_size(self): return self._base_iterator._dataset_size @property def _batch_size(self): return self._base_iterator._batch_size @property def _num_batches(self): return self._base_iterator._num_batches @property def num_examples(self): """ Number of examples that will be visited by the iterator. (May be lower than dataset_size) """ product = self.batch_size * self.num_batches if product > self._dataset_size: return self.batch_size * (self.num_batches - 1) else: return product def next(self): """ Returns next batch of _base_iterator Raises ------ StopException When _base_iterator reaches the end of the dataset Notes ----- Uneven batches may be discarded and StopException will be raised without having iterated throught every examples. """ length = -1 # check if the batch has wrong length, throw it away while length != self.batch_size: batch = self._base_iterator.next() if isinstance(batch, slice): length = batch.stop-batch.start else: length = len(batch) return batch def __next__(self): return self.next() @property @wraps(SubsetIterator.uneven, assigned=(), updated=()) def uneven(self): return False def as_even(iterator_cls): """ Returns a class wrapping iterator_cls that forces equal batch size. Parameters ---------- iterator_cls : class An iterator class that inherits from SubsetIterator Returns ------- class An iterator class ForcedEven{put the name of iterator_cls here}, based on ForcedEvenIterator, that wraps iterator_cls. """ assert issubclass(iterator_cls, SubsetIterator) dct = ForcedEvenIterator.__dict__.copy() dct["_base_iterator_cls"] = iterator_cls dct["fancy"] = iterator_cls.fancy dct["stochastic"] = iterator_cls.stochastic NewForcedEvenClass = type("ForcedEven%s" % iterator_cls.__name__, ForcedEvenIterator.__bases__, dct) return NewForcedEvenClass class SequentialSubsetIterator(SubsetIterator): """ Returns mini-batches proceeding sequentially through the dataset. Notes ----- Returns slice objects to represent ranges of indices (`fancy = False`). See :py:class:`SubsetIterator` for detailed constructor parameter and attribute documentation. """ def __init__(self, dataset_size, batch_size, num_batches, rng=None): if rng is not None: raise ValueError("non-None rng argument not supported for " "sequential batch iteration") assert num_batches is None or num_batches >= 0 self._dataset_size = dataset_size if batch_size is None: if num_batches is not None: batch_size = int(np.ceil(self._dataset_size / num_batches)) else: raise ValueError("need one of batch_size, num_batches " "for sequential batch iteration") elif batch_size is not None: if num_batches is not None: max_num_batches = np.ceil(self._dataset_size / batch_size) if num_batches > max_num_batches: raise ValueError("dataset of %d examples can only provide " "%d batches with batch_size %d, but %d " "batches were requested" % (self._dataset_size, max_num_batches, batch_size, num_batches)) else: num_batches = np.ceil(self._dataset_size / batch_size) self._batch_size = batch_size self._num_batches = num_batches self._next_batch_no = 0 self._idx = 0 self._batch = 0 @wraps(SubsetIterator.next, assigned=(), updated=()) def next(self): if self._batch >= self.num_batches or self._idx >= self._dataset_size: raise StopIteration() # this fix the problem where dataset_size % batch_size != 0 elif (self._idx + self._batch_size) > self._dataset_size: self._last = slice(self._idx, self._dataset_size) self._idx = self._dataset_size return self._last else: self._last = slice(self._idx, self._idx + self._batch_size) self._idx += self._batch_size self._batch += 1 return self._last def __next__(self): return self.next() fancy = False stochastic = False uniform_batch_size = False @property @wraps(SubsetIterator.num_examples, assigned=(), updated=()) def num_examples(self): product = self.batch_size * self.num_batches return min(product, self._dataset_size) @property @wraps(SubsetIterator.uneven, assigned=(), updated=()) def uneven(self): return self.batch_size * self.num_batches > self._dataset_size class ShuffledSequentialSubsetIterator(SequentialSubsetIterator): """ Randomly shuffles the example indices and then proceeds sequentially through the permutation. Notes ----- Returns lists of indices (`fancy = True`). See :py:class:`SubsetIterator` for detailed constructor parameter and attribute documentation. """ stochastic = True fancy = True uniform_batch_size = False def __init__(self, dataset_size, batch_size, num_batches, rng=None): super(ShuffledSequentialSubsetIterator, self).__init__( dataset_size, batch_size, num_batches, None ) self._rng = make_np_rng(rng, which_method=["random_integers", "shuffle"]) self._shuffled = np.arange(self._dataset_size) self._rng.shuffle(self._shuffled) @wraps(SubsetIterator.next) def next(self): if self._batch >= self.num_batches or self._idx >= self._dataset_size: raise StopIteration() # this fix the problem where dataset_size % batch_size != 0 elif (self._idx + self._batch_size) > self._dataset_size: rval = self._shuffled[self._idx: self._dataset_size] self._idx = self._dataset_size return rval else: rval = self._shuffled[self._idx: self._idx + self._batch_size] self._idx += self._batch_size self._batch += 1 return rval def __next__(self): return self.next() class RandomUniformSubsetIterator(SubsetIterator): """ Selects minibatches of examples by drawing indices uniformly at random, with replacement. Notes ----- Returns lists of indices (`fancy = True`). See :py:class:`SubsetIterator` for detailed constructor parameter and attribute documentation. """ def __init__(self, dataset_size, batch_size, num_batches, rng=None): self._rng = make_np_rng(rng, which_method=["random_integers", "shuffle"]) if batch_size is None: raise ValueError("batch_size cannot be None for random uniform " "iteration") elif num_batches is None: raise ValueError("num_batches cannot be None for random uniform " "iteration") self._dataset_size = dataset_size self._batch_size = batch_size self._num_batches = num_batches self._next_batch_no = 0 @wraps(SubsetIterator.next) def next(self): if self._next_batch_no >= self._num_batches: raise StopIteration() else: self._last = self._rng.random_integers(low=0, high=self._dataset_size - 1, size=(self._batch_size,)) self._next_batch_no += 1 return self._last def __next__(self): return self.next() fancy = True stochastic = True uniform_batch_size = True class RandomSliceSubsetIterator(RandomUniformSubsetIterator): """ Returns minibatches that are randomly selected contiguous slices in index space. Notes ----- Returns slice objects to represent ranges of indices (`fancy = False`). See :py:class:`SubsetIterator` for detailed constructor parameter and attribute documentation. """ def __init__(self, dataset_size, batch_size, num_batches, rng=None): if batch_size is None: raise ValueError("batch_size cannot be None for random slice " "iteration") elif num_batches is None: raise ValueError("num_batches cannot be None for random slice " "iteration") super(RandomSliceSubsetIterator, self).__init__(dataset_size, batch_size, num_batches, rng) self._last_start = self._dataset_size - self._batch_size if self._last_start < 0: raise ValueError("batch_size > dataset_size not supported for " "random slice iteration") @wraps(SubsetIterator.next) def next(self): if self._next_batch_no >= self._num_batches: raise StopIteration() else: start = self._rng.random_integers(low=0, high=self._last_start) self._last = slice(start, start + self._batch_size) self._next_batch_no += 1 return self._last def __next__(self): return self.next() fancy = False stochastic = True uniform_batch_size = True class BatchwiseShuffledSequentialIterator(SequentialSubsetIterator): """ Returns minibatches randomly, but sequential inside each minibatch. Notes ----- Returns slice objects to represent ranges of indices (`fancy = False`). See :py:class:`SubsetIterator` for detailed constructor parameter and attribute documentation. """ def __init__(self, dataset_size, batch_size, num_batches=None, rng=None): self._rng = make_np_rng(rng, which_method=["random_integers", "shuffle"]) assert num_batches is None or num_batches >= 0 self._dataset_size = dataset_size if batch_size is None: if num_batches is not None: batch_size = int(np.ceil(self._dataset_size / num_batches)) else: raise ValueError("need one of batch_size, num_batches " "for sequential batch iteration") elif batch_size is not None: if num_batches is not None: max_num_batches = np.ceil(self._dataset_size / batch_size) if num_batches > max_num_batches: raise ValueError("dataset of %d examples can only provide " "%d batches with batch_size %d, but %d " "batches were requested" % (self._dataset_size, max_num_batches, batch_size, num_batches)) else: num_batches = np.ceil(self._dataset_size / batch_size) self._batch_size = batch_size self._num_batches = int(num_batches) self._next_batch_no = 0 self._idx = 0 self._batch_order = list(range(self._num_batches)) self._rng.shuffle(self._batch_order) @wraps(SubsetIterator.next) def next(self): if self._next_batch_no >= self._num_batches: raise StopIteration() else: start = self._batch_order[self._next_batch_no] * self._batch_size if start + self._batch_size > self._dataset_size: self._last = slice(start, self._dataset_size) else: self._last = slice(start, start + self._batch_size) self._next_batch_no += 1 return self._last def __next__(self): return self.next() fancy = False stochastic = True uniform_batch_size = False class EvenSequencesSubsetIterator(SubsetIterator): """ An iterator for datasets with sequential data (e.g. list of words) which returns a list of indices of sequences in the dataset which have the same length. Within one minibatch all sequences will have the same length, so it might return minibatches with different sizes depending on the distribution of the lengths of sequences in the data. Notes ----- Returns lists of indices (`fancy = True`). Parameters ---------- sequence_data : list of lists or ndarray of objects (ndarrays) The sequential data used to determine indices within the dataset such that within a minibatch all sequences will have same lengths. See :py:class:`SubsetIterator` for detailed constructor parameter and attribute documentation. """ def __init__(self, sequence_data, batch_size, num_batches=None, rng=None): self._rng = make_np_rng(rng, which_method=["random_integers", "shuffle"]) if batch_size is None: raise ValueError("batch_size cannot be None for random uniform " "iteration") if num_batches is not None: raise ValueError("EvenSequencesSubsetIterator doesn't support" " fixed number of batches") if isinstance(sequence_data, list): self._dataset_size = len(sequence_data) elif isinstance(sequence_data, np.ndarray): self._dataset_size = sequence_data.shape[0] else: raise ValueError("sequence_data must be of type list or" " ndarray") self._sequence_data = sequence_data self._batch_size = batch_size self.prepare() self.reset() def prepare(self): # find unique lengths in sequences self.lengths = [len(s) for s in self._sequence_data] self.len_unique = np.unique(self.lengths) # store the indices of sequences for each unique length, # and their counts self.len_indices = dict() self.len_counts = dict() for ll in self.len_unique: self.len_indices[ll] = np.where(self.lengths == ll)[0] self.len_counts[ll] = len(self.len_indices[ll]) def reset(self): # make a copy of the number of sequences that share a specific length self.len_curr_counts = copy.copy(self.len_counts) # permute the array of unique lengths every epoch self.len_unique = self._rng.permutation(self.len_unique) self.len_indices_pos = dict() # save current total counts to decide when to stop iteration self.total_curr_counts = 0 for ll in self.len_unique: # keep a pointer to where we should start picking our minibatch of # same length sequences self.len_indices_pos[ll] = 0 # permute the array of indices of sequences with specific lengths # every epoch self.len_indices[ll] = self._rng.permutation(self.len_indices[ll]) self.total_curr_counts += len(self.len_indices[ll]) self.len_idx = -1 @wraps(SubsetIterator.next) def next(self): # stop when there are no more sequences left if self.total_curr_counts == 0: self.reset() raise StopIteration() # pick a length from the permuted array of lengths while True: self.len_idx = np.mod(self.len_idx+1, len(self.len_unique)) curr_len = self.len_unique[self.len_idx] if self.len_curr_counts[curr_len] > 0: break # find the position and the size of the minibatch of sequences # to be returned curr_batch_size = np.minimum(self._batch_size, self.len_curr_counts[curr_len]) curr_pos = self.len_indices_pos[curr_len] # get the actual indices for the sequences curr_indices = self.len_indices[curr_len][curr_pos:curr_pos + curr_batch_size] # update the pointer and counts of sequences in the chosen length self.len_indices_pos[curr_len] += curr_batch_size self.len_curr_counts[curr_len] -= curr_batch_size self.total_curr_counts -= curr_batch_size return curr_indices def __next__(self): return self.next() @property @wraps(SubsetIterator.num_examples, assigned=(), updated=()) def num_examples(self): return len(self._sequence_data) fancy = True stochastic = True uniform_batch_size = False _iteration_schemes = { 'sequential': SequentialSubsetIterator, 'shuffled_sequential': ShuffledSequentialSubsetIterator, 'random_slice': RandomSliceSubsetIterator, 'random_uniform': RandomUniformSubsetIterator, 'batchwise_shuffled_sequential': BatchwiseShuffledSequentialIterator, 'even_sequential': as_even(SequentialSubsetIterator), 'even_shuffled_sequential': as_even(ShuffledSequentialSubsetIterator), 'even_batchwise_shuffled_sequential': as_even(BatchwiseShuffledSequentialIterator), 'even_sequences': EvenSequencesSubsetIterator, } def has_uniform_batch_size(mode): """ Returns True if the iteration scheme has uniform batch size, False if not Parameters ---------- mode: string A string defining an iteration scheme in _iteration_schemes Returns ------- boolean True if the iteration scheme has uniform batch size, False otherwise """ return resolve_iterator_class(mode).uniform_batch_size def is_stochastic(mode): """ """ return resolve_iterator_class(mode).stochastic def resolve_iterator_class(mode): """ Map textual representations of default iteration modes to classes. Parameters ---------- mode : str or class object If a string, identifier string for the built-in iteration modes. See the module documentation of :py:mod:`pylearn2.utils.iteration` for a list of available modes. If a class, it is expected to be a class that respects the constructor and attribute interface defined in :py:class:`SubsetIterator`. Returns ------- subset_iter_class : class A class instance (i.e., an instance of type `type`) that interface defined in :py:class:`SubsetIterator`. """ if isinstance(mode, six.string_types) and mode not in _iteration_schemes: raise ValueError("unknown iteration mode string: %s" % mode) elif mode in _iteration_schemes: subset_iter_class = _iteration_schemes[mode] else: subset_iter_class = mode return subset_iter_class class FiniteDatasetIterator(object): """ A wrapper around subset iterators that actually retrieves data. Parameters ---------- dataset : `Dataset` object The dataset over which to iterate. data_specs : tuple A `(space, source)` tuple. See :ref:`data_specs` for a full description. Must not contain nested composite spaces. subset_iterator : object An iterator object that returns slice objects or lists of examples, conforming to the interface specified by :py:class:`SubsetIterator`. return_tuple : bool, optional Always return a tuple, even if there is exactly one source of data being returned. Defaults to `False`. convert : list of callables A list of callables, in the same order as the sources in `data_specs`, that will be called on the individual source batches prior to any further processing. Notes ----- See the documentation for :py:class:`SubsetIterator` for attribute documentation. The dataset should provide a `get` method which accepts a tuple of source identifiers and a list or slice of indexes and returns a tuple of batches of examples, one for each source. The old interface using `get_data` is still supported for the moment being. """ def __init__(self, dataset, subset_iterator, data_specs=None, return_tuple=False, convert=None): self._data_specs = data_specs self._dataset = dataset self._subset_iterator = subset_iterator self._return_tuple = return_tuple # Keep only the needed sources in self._raw_data. # Remember what source they correspond to in self._source assert is_flat_specs(data_specs) dataset_space, dataset_source = self._dataset.get_data_specs() assert is_flat_specs((dataset_space, dataset_source)) # the dataset's data spec is either a single (space, source) pair, # or a pair of (non-nested CompositeSpace, non-nested tuple). # We could build a mapping and call flatten(..., return_tuple=True) # but simply putting spaces, sources and data in tuples is simpler. if not isinstance(dataset_source, (tuple, list)): dataset_source = (dataset_source,) if not isinstance(dataset_space, CompositeSpace): dataset_sub_spaces = (dataset_space,) else: dataset_sub_spaces = dataset_space.components assert len(dataset_source) == len(dataset_sub_spaces) space, source = data_specs if not isinstance(source, tuple): source = (source,) if not isinstance(space, CompositeSpace): sub_spaces = (space,) else: sub_spaces = space.components assert len(source) == len(sub_spaces) # If `dataset` is incompatible with the new interface, fall back to the # old interface if not hasattr(self._dataset, 'get'): all_data = self._dataset.get_data() if not isinstance(all_data, tuple): all_data = (all_data,) raw_data = [] for s in source: try: raw_data.append(all_data[dataset_source.index(s)]) except ValueError as e: msg = str(e) + '\nThe dataset does not provide '\ 'a source with name: ' + s + '.' reraise_as(ValueError(msg)) self._raw_data = tuple(raw_data) self._source = source self._space = sub_spaces if convert is None: self._convert = [None for s in source] else: assert len(convert) == len(source) self._convert = convert for i, (so, sp) in enumerate(safe_izip(source, sub_spaces)): try: idx = dataset_source.index(so) except ValueError as e: msg = str(e) + '\nThe dataset does not provide '\ 'a source with name: ' + so + '.' reraise_as(ValueError(msg)) dspace = dataset_sub_spaces[idx] fn = self._convert[i] # If there is a fn, it is supposed to take care of the formatting, # and it should be an error if it does not. If there was no fn, # then the iterator will try to format using the generic # space-formatting functions. if fn is None: # "dspace" and "sp" have to be passed as parameters # to lambda, in order to capture their current value, # otherwise they would change in the next iteration # of the loop. fn = (lambda batch, dspace=dspace, sp=sp: dspace.np_format_as(batch, sp)) self._convert[i] = fn def __iter__(self): return self @wraps(SubsetIterator.next) def next(self): """ Retrieves the next batch of examples. Returns ------- next_batch : object An object representing a mini-batch of data, conforming to the space specified in the `data_specs` constructor argument to this iterator. Will be a tuple if more than one data source was specified or if the constructor parameter `return_tuple` was `True`. Raises ------ StopIteration When there are no more batches to return. """ next_index = self._subset_iterator.next() # If the dataset is incompatible with the new interface, fall back to # the old one if hasattr(self._dataset, 'get'): rval = self._next(next_index) else: rval = self._fallback_next(next_index) if not self._return_tuple and len(rval) == 1: rval, = rval return rval def _next(self, next_index): return tuple( fn(batch) if fn else batch for batch, fn in safe_izip(self._dataset.get(self._source, next_index), self._convert) ) def _fallback_next(self, next_index): # TODO: handle fancy-index copies by allocating a buffer and # using np.take() return tuple( fn(data[next_index]) if fn else data[next_index] for data, fn in safe_izip(self._raw_data, self._convert) ) def __next__(self): return self.next() @property @wraps(SubsetIterator.batch_size, assigned=(), updated=()) def batch_size(self): return self._subset_iterator.batch_size @property @wraps(SubsetIterator.num_batches, assigned=(), updated=()) def num_batches(self): return self._subset_iterator.num_batches @property @wraps(SubsetIterator.num_examples, assigned=(), updated=()) def num_examples(self): return self._subset_iterator.num_examples @property @wraps(SubsetIterator.uneven, assigned=(), updated=()) def uneven(self): return self._subset_iterator.uneven @property @wraps(SubsetIterator.stochastic, assigned=(), updated=()) def stochastic(self): return self._subset_iterator.stochastic
bsd-3-clause
feaf896de09339cf0daa2b0b4c1fcb9f
34.180407
79
0.594966
4.497334
false
false
false
false
lisa-lab/pylearn2
pylearn2/utils/datasets.py
44
9068
""" Several utilities to evaluate an ALC on the dataset, to iterate over minibatches from a dataset, or to merge three data with given proportions """ # Standard library imports import logging import os import functools from itertools import repeat import warnings # Third-party imports import numpy import scipy from theano.compat.six.moves import reduce, xrange import theano try: from matplotlib import pyplot from mpl_toolkits.mplot3d import Axes3D except ImportError: warnings.warn("Could not import some dependencies.") # Local imports from pylearn2.utils.rng import make_np_rng logger = logging.getLogger(__name__) ################################################## # 3D Visualization ################################################## def do_3d_scatter(x, y, z, figno=None, title=None): """ Generate a 3D scatterplot figure and optionally give it a title. Parameters ---------- x : WRITEME y : WRITEME z : WRITEME figno : WRITEME title : WRITEME """ fig = pyplot.figure(figno) ax = Axes3D(fig) ax.scatter(x, y, z) ax.set_xlabel("X") ax.set_ylabel("Y") ax.set_zlabel("Z") pyplot.suptitle(title) def save_plot(repr, path, name="figure.pdf", title="features"): """ .. todo:: WRITEME """ # TODO : Maybe run a PCA if shape[1] > 3 assert repr.get_value(borrow=True).shape[1] == 3 # Take the first 3 columns x, y, z = repr.get_value(borrow=True).T do_3d_scatter(x, y, z) # Save the produces figure filename = os.path.join(path, name) pyplot.savefig(filename, format="pdf") logger.info('... figure saved: {0}'.format(filename)) ################################################## # Features or examples filtering ################################################## def filter_labels(train, label, classes=None): """ Filter examples of train for which we have labels Parameters ---------- train : WRITEME label : WRITEME classes : WRITEME Returns ------- WRITEME """ if isinstance(train, theano.tensor.sharedvar.SharedVariable): train = train.get_value(borrow=True) if isinstance(label, theano.tensor.sharedvar.SharedVariable): label = label.get_value(borrow=True) if not (isinstance(train, numpy.ndarray) or scipy.sparse.issparse(train)): raise TypeError('train must be a numpy array, a scipy sparse matrix,' ' or a theano shared array') # Examples for which any label is set if classes is not None: label = label[:, classes] # Special case for sparse matrices if scipy.sparse.issparse(train): idx = label.sum(axis=1).nonzero()[0] return (train[idx], label[idx]) # Compress train and label arrays according to condition condition = label.any(axis=1) return tuple(var.compress(condition, axis=0) for var in (train, label)) def nonzero_features(data, combine=None): """ Get features for which there are nonzero entries in the data. Parameters ---------- data : list of matrices List of data matrices, either in sparse format or not. They must have the same number of features (column number). combine : function, optional A function to combine elementwise which features to keep. Default keeps the intersection of each non-zero columns. Returns ------- indices : ndarray object Indices of the nonzero features. Notes ----- I would return a mask (bool array) here, but scipy.sparse doesn't appear to fully support advanced indexing. """ if combine is None: combine = functools.partial(reduce, numpy.logical_and) # Assumes all values are >0, which is the case for all sparse datasets. masks = numpy.asarray([subset.sum(axis=0) for subset in data]).squeeze() nz_feats = combine(masks).nonzero()[0] return nz_feats # TODO: Is this a duplicate? def filter_nonzero(data, combine=None): """ Filter non-zero features of data according to a certain combining function Parameters ---------- data : list of matrices List of data matrices, either in sparse format or not. They must have the same number of features (column number). combine : function A function to combine elementwise which features to keep. Default keeps the intersection of each non-zero columns. Returns ------- indices : ndarray object Indices of the nonzero features. """ nz_feats = nonzero_features(data, combine) return [set[:, nz_feats] for set in data] ################################################## # Iterator object for minibatches of datasets ################################################## class BatchIterator(object): """ Builds an iterator object that can be used to go through the minibatches of a dataset, with respect to the given proportions in conf Parameters ---------- dataset : WRITEME set_proba : WRITEME batch_size : WRITEME seed : WRITEME """ def __init__(self, dataset, set_proba, batch_size, seed=300): # Local shortcuts for array operations flo = numpy.floor sub = numpy.subtract mul = numpy.multiply div = numpy.divide mod = numpy.mod # Record external parameters self.batch_size = batch_size if (isinstance(dataset[0], theano.Variable)): self.dataset = [set.get_value(borrow=True) for set in dataset] else: self.dataset = dataset # Compute maximum number of samples for one loop set_sizes = [set.shape[0] for set in self.dataset] set_batch = [float(self.batch_size) for i in xrange(3)] set_range = div(mul(set_proba, set_sizes), set_batch) set_range = map(int, numpy.ceil(set_range)) # Upper bounds for each minibatch indexes set_limit = numpy.ceil(numpy.divide(set_sizes, set_batch)) self.limit = map(int, set_limit) # Number of rows in the resulting union set_tsign = sub(set_limit, flo(div(set_sizes, set_batch))) set_tsize = mul(set_tsign, flo(div(set_range, set_limit))) l_trun = mul(flo(div(set_range, set_limit)), mod(set_sizes, set_batch)) l_full = mul(sub(set_range, set_tsize), set_batch) self.length = sum(l_full) + sum(l_trun) # Random number generation using a permutation index_tab = [] for i in xrange(3): index_tab.extend(repeat(i, set_range[i])) # Use a deterministic seed self.seed = seed rng = make_np_rng(seed, which_method="permutation") self.permut = rng.permutation(index_tab) def __iter__(self): """Generator function to iterate through all minibatches""" counter = [0, 0, 0] for chosen in self.permut: # Retrieve minibatch from chosen set index = counter[chosen] minibatch = self.dataset[chosen][ index * self.batch_size:(index + 1) * self.batch_size ] # Increment the related counter counter[chosen] = (counter[chosen] + 1) % self.limit[chosen] # Return the computed minibatch yield minibatch def __len__(self): """Return length of the weighted union""" return self.length def by_index(self): """Same generator as __iter__, but yield only the chosen indexes""" counter = [0, 0, 0] for chosen in self.permut: index = counter[chosen] counter[chosen] = (counter[chosen] + 1) % self.limit[chosen] yield chosen, index ################################################## # Miscellaneous ################################################## def minibatch_map(fn, batch_size, input_data, output_data=None, output_width=None): """ Apply a function on input_data, one minibatch at a time. Storage for the output can be provided. If it is the case, it should have appropriate size. If output_data is not provided, then output_width should be specified. Parameters ---------- fn : WRITEME batch_size : WRITEME input_data : WRITEME output_data : WRITEME output_width : WRITEME Returns ------- WRITEME """ if output_width is None: if output_data is None: raise ValueError('output_data or output_width should be provided') output_width = output_data.shape[1] output_length = input_data.shape[0] if output_data is None: output_data = numpy.empty((output_length, output_width)) else: assert output_data.shape[0] == input_data.shape[0], ('output_data ' 'should have the same length as input_data', output_data.shape[0], input_data.shape[0]) for i in xrange(0, output_length, batch_size): output_data[i:i+batch_size] = fn(input_data[i:i+batch_size]) return output_data
bsd-3-clause
8d83e4b1375f9974b47a4136c87c2eb8
29.02649
79
0.600794
4.123693
false
false
false
false
lisa-lab/pylearn2
pylearn2/scripts/dbm/dbm_metrics.py
7
24693
#!/usr/bin/env python __authors__ = "Vincent Dumoulin" __copyright__ = "Copyright 2013, Universite de Montreal" __credits__ = ["Guillaume Desjargins", "Vincent Dumoulin"] __license__ = "3-clause BSD" __maintainer__ = "Vincent Dumoulin" """ This script computes both an estimate of the partition function of the provided DBM model and an estimate of the log-likelihood on the given training and test sets. This is guaranteed to work only for DBMs with a BinaryVector visible layer and BinaryVectorMaxPool hidden layers with pool sizes of 1. It uses annealed importance sampling (AIS) to estimate Z, the partition function. TODO: add more details, cite paper usage: dbm_metrics.py [-h] {ais} model_path positional arguments: {ais} the desired metric model_path path to the pickled DBM model optional arguments: -h, --help show the help message and exit """ import argparse import warnings import numpy import logging from theano.compat.six.moves import xrange import theano import theano.tensor as T from theano.sandbox.rng_mrg import MRG_RandomStreams as RandomStreams from theano import scan import pylearn2 from pylearn2.compat import OrderedDict from pylearn2.datasets.mnist import MNIST from pylearn2.utils import serial from pylearn2 import utils floatX = theano.config.floatX logging.basicConfig(level=logging.INFO) rng = numpy.random.RandomState(9873242) theano_rng = RandomStreams(rng.randint(2**30)) def _sample_even_odd(W_list, b_list, samples, beta, odd=True): """ Sample from the even (or odd) layers given a list of previous states. Parameters ---------- W_list : array-like object of theano shared variables Weight matrices of the DBM. Its first element is ignored, since in the Pylearn2 framework a visible layer does not have a weight matrix. b_list : array-like object of theano shared variables Biases of the DBM samples : array-like object of theano shared variables Samples corresponding to the previous states beta : theano.tensor.scalar Inverse temperature parameter odd : boolean Whether to sample from the odd or the even layers (defaults to sampling from odd layers) """ for i in xrange(odd, len(samples), 2): samples[i] = sample_hi_given(samples, i, W_list, b_list, beta) def _activation_even_odd(W_list, b_list, samples, beta, odd=True): """ Compute the activation of the even (or odd) layers given a list of previous states. Parameters ---------- W_list : array-like object of theano shared variables Weight matrices of the DBM. Its first element is ignored, since in the Pylearn2 framework a visible layer does not have a weight matrix. b_list : array-like object of theano shared variables Biases of the DBM samples : array-like object of theano shared variables Samples corresponding to the previous states beta : theano.tensor.scalar Inverse temperature parameter odd : boolean Whether to compute activation for the odd or the even layers (defaults to computing for odd layers) """ for i in xrange(odd, len(samples), 2): samples[i] = hi_given(samples, i, W_list, b_list, beta, apply_sigmoid=False) def neg_sampling(W_list, b_list, nsamples, beta=1.0, pa_bias=None, marginalize_odd=True, theano_rng=None): """ Generate a sample from the intermediate distribution defined at inverse temperature 'beta', starting from state 'nsamples'. See file docstring for equation of p_k(h1). Parameters ---------- W_list : array-like object of theano shared variables Weight matrices of the DBM. Its first element is ignored, since in the Pylearn2 framework a visible layer does not have a weight matrix. b_list : array-like object of theano shared variables Biases of the DBM nsamples : array-like object of theano shared variables Negative samples corresponding to the previous states beta : theano.tensor.scalar Inverse temperature parameter marginalize_odd : boolean Whether to marginalize odd layers theano_rng : theano RandomStreams Random number generator Returns ------- new_nsamples : array-like object of symbolic matrices new_nsamples[i] contains new samples for i-th layer. """ # There's as much layers in the DBM as there are bias vectors depth = len(b_list) new_nsamples = [nsamples[i] for i in xrange(depth)] # Contribution from model B, at temperature beta_k _sample_even_odd(W_list, b_list, new_nsamples, beta, odd=marginalize_odd) _activation_even_odd(W_list, b_list, new_nsamples, beta, odd=not marginalize_odd) # Contribution from model A, at temperature (1 - beta_k) new_nsamples[not marginalize_odd] += pa_bias * (1. - beta) # Loop over all layers (not being marginalized) for i in xrange(not marginalize_odd, depth, 2): new_nsamples[i] = T.nnet.sigmoid(new_nsamples[i]) new_nsamples[i] = theano_rng.binomial( size=nsamples[i].get_value().shape, n=1, p=new_nsamples[i], dtype=floatX ) return new_nsamples def free_energy_at_beta(W_list, b_list, samples, beta, pa_bias=None, marginalize_odd=True): """ Compute the free-energy of the sample 'h1_sample', for model p_k(h1). Parameters ---------- W_list : array-like object of theano shared variables Weight matrices of the DBM. Its first element is ignored, since in the Pylearn2 framework a visible layer does not have a weight matrix. b_list : array-like object of theano shared variables Biases of the DBM samples : array-like object of theano shared variable Samples from which we extract the samples of layer h1 beta : theano.tensor.scalar Inverse temperature beta_k of model p_k(h1) at which to measure the free-energy. pa_bias : array-like object of theano shared variables Biases for the A model marginalize_odd : boolean Whether to marginalize odd layers Returns ------- fe : symbolic variable Free-energy of sample 'h1_sample', at inverse temperature beta """ # There's as much layers in the DBM as there are bias vectors depth = len(b_list) fe = 0. # Contribution of biases keep_idx = numpy.arange(not marginalize_odd, depth, 2) for i in keep_idx: fe -= T.dot(samples[i], b_list[i]) * beta # Contribution of biases marg_idx = numpy.arange(marginalize_odd, depth, 2) for i in marg_idx: from_im1 = T.dot(samples[i-1], W_list[i]) if i >= 1 else 0. from_ip1 = T.dot(samples[i+1], W_list[i+1].T) if i < depth-1 else 0 net_input = (from_im1 + from_ip1 + b_list[i]) * beta fe -= T.sum(T.nnet.softplus(net_input), axis=1) fe -= T.dot(samples[not marginalize_odd], pa_bias) * (1. - beta) return fe def compute_log_ais_weights(batch_size, free_energy_fn, sample_fn, betas): """ Compute log of the AIS weights Parameters ---------- batch_size : scalar Size of a batch of samples free_energy_fn : theano.function Function which, given temperature beta_k, computes the free energy of the samples stored in model.samples. This function should return a symbolic vector. sample_fn : theano.function Function which, given temperature beta_k, generates samples h1 ~ p_k(h1). betas : array-like object of scalars Inverse temperature parameters for which to compute the log_ais weights Returns ------- log_ais_w : theano.tensor.vector Vector containing log ais-weights """ # Initialize log-ais weights log_ais_w = numpy.zeros(batch_size, dtype=floatX) # Iterate from inverse temperature beta_k=0 to beta_k=1... for i in range(len(betas) - 1): bp, bp1 = betas[i], betas[i+1] log_ais_w += free_energy_fn(bp) - free_energy_fn(bp1) sample_fn(bp1) if i % 1e3 == 0: logging.info('Temperature %f ' % bp1) return log_ais_w def estimate_from_weights(log_ais_w): """ Safely compute the log-average of the ais-weights Parameters ---------- log_ais_w : theano.tensor.vector Symbolic vector containing log_ais_w^{(m)}. Returns ------- dlogz : theano.tensor.scalar log(Z_B) - log(Z_A) var_dlogz : theano.tensor.scalar Variance of our estimator """ # Utility function for safely computing log-mean of the ais weights ais_w = T.vector() max_ais_w = T.max(ais_w) dlogz = T.log(T.mean(T.exp(ais_w - max_ais_w))) + max_ais_w log_mean = theano.function([ais_w], dlogz, allow_input_downcast=False) # Estimate the log-mean of the AIS weights dlogz = log_mean(log_ais_w) # Estimate log-variance of the AIS weights # VAR(log(X)) \approx VAR(X) / E(X)^2 = E(X^2)/E(X)^2 - 1 m = numpy.max(log_ais_w) var_dlogz = (log_ais_w.shape[0] * numpy.sum(numpy.exp(2 * (log_ais_w - m))) / numpy.sum(numpy.exp(log_ais_w - m)) ** 2 - 1.) return dlogz, var_dlogz def compute_log_za(b_list, pa_bias, marginalize_odd=True): """ Compute the exact partition function of model p_A(h1) Parameters ---------- b_list : array-like object of theano shared variables Biases of the DBM pa_bias : array-like object of theano shared variables Biases for the A model marginalize_odd : boolean Whether to marginalize odd layers Returns ------- log_za : scalar Partition function of model A """ log_za = 0. for i, b in enumerate(b_list): if i == (not marginalize_odd): log_za += numpy.sum(numpy.log(1 + numpy.exp(pa_bias))) else: log_za += numpy.log(2) * b.get_value().shape[0] return log_za def compute_likelihood_given_logz(nsamples, psamples, batch_size, energy_fn, inference_fn, log_z, test_x): """ Compute test set likelihood as below, where q is the variational approximation to the posterior p(h1,h2|v). ln p(v) \approx \sum_h q(h) E(v,h1,h2) + H(q) - ln Z See section 3.2 of DBM paper for details. Parameters ---------- nsamples : array-like object of theano shared variables Negative samples psamples : array-like object of theano shared variables Positive samples batch_size : scalar Size of a batch of samples energy_fn : theano.function Function which computes the (temperature 1) energy of the samples. This function should return a symbolic vector. inference_fn : theano.function Inference function for DBM. Function takes a T.matrix as input (data) and returns a list of length 'length(b_list)', where the i-th element is an ndarray containing approximate samples of layer i. log_z : scalar Estimate partition function of 'model'. test_x : numpy.ndarray Test set data, in dense design matrix format. Returns ------- likelihood : scalar Negative log-likelihood of test data under the model """ i = 0. likelihood = 0 for i in xrange(0, len(test_x), batch_size): # Recast data as floatX and apply preprocessing if required x = numpy.array(test_x[i:numpy.minimum(test_x.shape[0], i + batch_size), :], dtype=floatX) batch_size0 = len(x) if len(x) < batch_size: # concatenate x to have some dummy entries x = numpy.concatenate((x, numpy.zeros((batch_size-len(x),x.shape[1]), dtype=floatX)), axis=0) # Perform inference inference_fn(x) # Entropy of h(q) adds contribution to variational lower-bound hq = 0 for psample in psamples[1:]: temp = \ - psample.get_value() * numpy.log(1e-5 + psample.get_value()) \ - (1.-psample.get_value()) \ * numpy.log(1. - psample.get_value() + 1e-5) hq += numpy.sum(temp, axis=1) # Copy into negative phase buffers to measure energy nsamples[0].set_value(x) for ii, psample in enumerate(psamples): if ii > 0: nsamples[ii].set_value(psample.get_value()) # Compute sum of likelihood for current buffer x_likelihood = numpy.sum((-energy_fn(1.0) + hq - log_z)[:batch_size0]) # Perform moving average of negative likelihood # Divide by len(x) and not bufsize, since last buffer might be smaller likelihood = (i * likelihood + x_likelihood) / (i + batch_size0) return likelihood def hi_given(samples, i, W_list, b_list, beta=1.0, apply_sigmoid=True): """ Compute the state of hidden layer i given all other layers Parameters ---------- samples : array-like object of theano shared variables For the positive phase, samples[0] points to the input, while samples[i] contains the current state of the i-th layer. In the negative phase, samples[i] contains the persistent chain associated with the i-th layer. i : integer Compute activation of layer i of our DBM W_list : array-like object of theano shared variables Weight matrices of the DBM. Its first element is ignored, since in the Pylearn2 framework a visible layer does not have a weight matrix. b_list : array-like object of theano shared variables Biases of the DBM beta : scalar Inverse temperature parameter used when performing AIS apply_sigmoid : boolean When False, hi_given will not apply the sigmoid. Useful for AIS estimate. Returns ------- hi_mean : symbolic variable Activation of the i-th layer """ # There's as much layers in the DBM as there are bias vectors depth = len(samples) hi_mean = 0. if i < depth-1: # Top-down input wip1 = W_list[i+1] hi_mean += T.dot(samples[i+1], wip1.T) * beta if i > 0: # Bottom-up input wi = W_list[i] hi_mean += T.dot(samples[i-1], wi) * beta hi_mean += b_list[i] * beta if apply_sigmoid: return T.nnet.sigmoid(hi_mean) else: return hi_mean def sample_hi_given(samples, i, W_list, b_list, beta=1.0): """ Given current state of our DBM ('samples'), sample the values taken by the i-th layer. Parameters ---------- samples : array-like object of theano shared variables For the positive phase, samples[0] points to the input, while samples[i] contains the current state of the i-th layer. In the negative phase, samples[i] contains the persistent chain associated with the i-th layer. i : integer Compute activation of layer i of our DBM W_list : array-like object of theano shared variables Weight matrices of the DBM. Its first element is ignored, since in the Pylearn2 framework a visible layer does not have a weight matrix. b_list : array-like object of theano shared variables Biases of the DBM beta : scalar Inverse temperature parameter used when performing AIS Returns ------- hi_sample : symbolic variable State of the i-th layer """ hi_mean = hi_given(samples, i, W_list, b_list, beta) hi_sample = theano_rng.binomial( size=samples[i].get_value().shape, n=1, p=hi_mean, dtype=floatX ) return hi_sample def _e_step(psamples, W_list, b_list, n_steps=100, eps=1e-5): """ Performs 'n_steps' of mean-field inference (used to compute positive phase statistics) Parameters ---------- psamples : array-like object of theano shared variables State of each layer of the DBM (during the inference process). psamples[0] points to the input n_steps : integer Number of iterations of mean-field to perform """ depth = len(psamples) # now alternate mean-field inference for even/odd layers def mf_iteration(*psamples): new_psamples = [p for p in psamples] for i in xrange(1, depth, 2): new_psamples[i] = hi_given(psamples, i, W_list, b_list) for i in xrange(2, depth, 2): new_psamples[i] = hi_given(psamples, i, W_list, b_list) score = 0. for i in xrange(1, depth): score = T.maximum(T.mean(abs(new_psamples[i] - psamples[i])), score) return new_psamples, theano.scan_module.until(score < eps) new_psamples, updates = scan( mf_iteration, outputs_info=psamples, n_steps=n_steps ) return [x[-1] for x in new_psamples] def estimate_likelihood(W_list, b_list, trainset, testset, free_energy_fn=None, batch_size=100, large_ais=False, log_z=None, pos_mf_steps=50, pos_sample_steps=0): """ Compute estimate of log-partition function and likelihood of trainset and testset Parameters ---------- W_list : array-like object of theano shared variables b_list : array-like object of theano shared variables Biases of the DBM trainset : pylearn2.datasets.dataset.Dataset Training set testset : pylearn2.datasets.dataset.Dataset Test set free_energy_fn : theano.function Function which, given temperature beta_k, computes the free energy of the samples stored in model.samples. This function should return a symbolic vector. batch_size : integer Size of a batch of examples large_ais : boolean If True, will use 3e5 chains, instead of 3e4 log_z : log-partition function (if precomputed) pos_mf_steps: the number of fixed-point iterations for approximate inference pos_sample_steps: same thing as pos_mf_steps when both pos_mf_steps > 0 and pos_sample_steps > 0, pos_mf_steps has a priority Returns ------- nll : scalar Negative log-likelihood of data.X under `model`. logz : scalar Estimate of log-partition function of `model`. """ warnings.warn("This is garanteed to work only for DBMs with a " + "BinaryVector visible layer and BinaryVectorMaxPool " + "hidden layers with pool sizes of 1.") # Add a dummy placeholder for visible layer's weights in W_list W_list = [None] + W_list # Depth of the DBM depth = len(b_list) # Initialize samples psamples = [] nsamples = [] for i, b in enumerate(b_list): psamples += [utils.sharedX(rng.rand(batch_size, b.get_value().shape[0]), name='psamples%i' % i)] nsamples += [utils.sharedX(rng.rand(batch_size, b.get_value().shape[0]), name='nsamples%i' % i)] psamples[0] = T.matrix('psamples0') ########################## ## BUILD THEANO FUNCTIONS ########################## beta = T.scalar() # For an even number of layers, we marginalize the odd layers # (and vice-versa) marginalize_odd = (depth % 2) == 0 # Build function to retrieve energy. E = -T.dot(nsamples[0], b_list[0]) * beta for i in xrange(1, depth): E -= T.sum(T.dot(nsamples[i-1], W_list[i] * beta) * nsamples[i], axis=1) E -= T.dot(nsamples[i], b_list[i] * beta) energy_fn = theano.function([beta], E) # Build inference function. assert (pos_mf_steps or pos_sample_steps) pos_steps = pos_mf_steps if pos_mf_steps else pos_sample_steps new_psamples = _e_step(psamples, W_list, b_list, n_steps=pos_steps) ups = OrderedDict() for psample, new_psample in zip(psamples[1:], new_psamples[1:]): ups[psample] = new_psample temp = numpy.asarray(trainset.X, dtype=floatX) mean_train = numpy.mean(temp, axis=0) inference_fn = theano.function(inputs=[psamples[0]], outputs=[], updates=ups) # Configure baserate bias for (h0 if `marginalize_odd` else h1) inference_fn(numpy.tile(mean_train, (batch_size, 1))) numpy_psamples = [mean_train[None, :]] + \ [psample.get_value() for psample in psamples[1:]] mean_pos = numpy.minimum(numpy_psamples[not marginalize_odd], 1-1e-5) mean_pos = numpy.maximum(mean_pos, 1e-5) pa_bias = -numpy.log(1./mean_pos[0] - 1.) # Build Theano function to sample from interpolating distributions. updates = OrderedDict() new_nsamples = neg_sampling(W_list, b_list, nsamples, beta=beta, pa_bias=pa_bias, marginalize_odd=marginalize_odd, theano_rng=theano_rng) for (nsample, new_nsample) in zip(nsamples, new_nsamples): updates[nsample] = new_nsample sample_fn = theano.function([beta], [], updates=updates, name='sample_func') # Build function to compute free-energy of p_k(h1). fe_bp_h1 = free_energy_at_beta(W_list, b_list, nsamples, beta, pa_bias, marginalize_odd=marginalize_odd) free_energy_fn = theano.function([beta], fe_bp_h1) ########### ## RUN AIS ########### # Generate exact sample for the base model. for i, nsample_i in enumerate(nsamples): bias = pa_bias if i == 1 else b_list[i].get_value() hi_mean_vec = 1. / (1. + numpy.exp(-bias)) hi_mean = numpy.tile(hi_mean_vec, (batch_size, 1)) r = rng.random_sample(hi_mean.shape) hi_sample = numpy.array(hi_mean > r, dtype=floatX) nsample_i.set_value(hi_sample) # Default configuration for interpolating distributions if large_ais: betas = numpy.cast[floatX]( numpy.hstack((numpy.linspace(0, 0.5, 1e5+1)[:-1], numpy.linspace(0.5, 0.9, 1e5+1)[:-1], numpy.linspace(0.9, 1.0, 1e5)))) else: betas = numpy.cast[floatX]( numpy.hstack((numpy.linspace(0, 0.5, 1e4+1)[:-1], numpy.linspace(0.5, 0.9, 1e4+1)[:-1], numpy.linspace(0.9, 1.0, 1e4)))) if log_z is None: log_ais_w = compute_log_ais_weights(batch_size, free_energy_fn, sample_fn, betas) dlogz, var_dlogz = estimate_from_weights(log_ais_w) log_za = compute_log_za(b_list, pa_bias, marginalize_odd) log_z = log_za + dlogz logging.info('log_z = %f' % log_z) logging.info('log_za = %f' % log_za) logging.info('dlogz = %f' % dlogz) logging.info('var_dlogz = %f' % var_dlogz) train_ll = compute_likelihood_given_logz(nsamples, psamples, batch_size, energy_fn, inference_fn, log_z, trainset.X) logging.info('Training likelihood = %f' % train_ll) test_ll = compute_likelihood_given_logz(nsamples, psamples, batch_size, energy_fn, inference_fn, log_z, testset.X) logging.info('Test likelihood = %f' % test_ll) return (train_ll, test_ll, log_z) if __name__ == '__main__': # Possible metrics metrics = {'ais': estimate_likelihood} datasets = {'mnist': MNIST} # Argument parsing parser = argparse.ArgumentParser() parser.add_argument("metric", help="the desired metric", choices=metrics.keys()) parser.add_argument("dataset", help="the dataset used for computing the " + "metric", choices=datasets.keys()) parser.add_argument("model_path", help="path to the pickled DBM model") args = parser.parse_args() metric = metrics[args.metric] dataset = datasets[args.dataset] model = serial.load(args.model_path) layers = [model.visible_layer] + model.hidden_layers W_list = [theano.shared(hidden_layer.get_weights()) for hidden_layer in model.hidden_layers] b_list = [theano.shared(layer.get_biases()) for layer in layers] trainset = dataset(which_set='train') testset = dataset(which_set='test') metric(W_list, b_list, trainset, testset, pos_mf_steps=5)
bsd-3-clause
c272811532c5b1779aa68f99827312fe
34.478448
105
0.615357
3.677837
false
false
false
false
lisa-lab/pylearn2
pylearn2/utils/serial.py
33
20145
""" Utilities for serializing and deserializing python objects. """ try: from cPickle import BadPickleGet except ImportError: BadPickleGet = KeyError import pickle import logging import numpy as np from theano.compat import six from theano.compat.six.moves import cPickle, xrange import os import time import warnings import sys from pylearn2.utils.string_utils import preprocess from pylearn2.utils.mem import improve_memory_error_message io = None hdf_reader = None import struct from pylearn2.utils.exc import reraise_as from pylearn2.utils.string_utils import match import shutil logger = logging.getLogger(__name__) def load(filepath, retry=True): """ Loads object(s) from file specified by 'filepath'. Parameters ---------- filepath : str A path to a file to load. Should be a pickle, Matlab, or NumPy file; or a .txt or .amat file that numpy.loadtxt can load. retry : bool, optional If True, will make a handful of attempts to load the file before giving up. This can be useful if you are for example calling show_weights.py on a file that is actively being written to by a training script--sometimes the load attempt might fail if the training script writes at the same time show_weights tries to read, but if you try again after a few seconds you should be able to open the file. Returns ------- loaded_object : object The object that was stored in the file. """ return _load(filepath, recurse_depth=0, retry=True) def save(filepath, obj, on_overwrite='ignore'): """ Serialize `object` to a file denoted by `filepath`. Parameters ---------- filepath : str A filename. If the suffix is `.joblib` and joblib can be imported, `joblib.dump` is used in place of the regular pickling mechanisms; this results in much faster saves by saving arrays as separate .npy files on disk. If the file suffix is `.npy` than `numpy.save` is attempted on `obj`. Otherwise, (c)pickle is used. obj : object A Python object to be serialized. on_overwrite : str, optional A string specifying what to do if the file already exists. Possible values include: - "ignore" : Just overwrite the existing file. - "backup" : Make a backup copy of the file (<filepath>.bak). Save the new copy. Then delete the backup copy. This allows recovery of the old version of the file if saving the new one fails. """ filepath = preprocess(filepath) if os.path.exists(filepath): if on_overwrite == 'backup': backup = filepath + '.bak' shutil.move(filepath, backup) save(filepath, obj) try: os.remove(backup) except Exception as e: warnings.warn("Got an error while trying to remove " + backup + ":" + str(e)) return else: assert on_overwrite == 'ignore' try: _save(filepath, obj) except RuntimeError as e: """ Sometimes for large theano graphs, pickle/cPickle exceed the maximum recursion depth. This seems to me like a fundamental design flaw in pickle/cPickle. The workaround I employ here is the one recommended to someone who had a similar problem on stackexchange: http://stackoverflow.com/questions/2134706/hitting-maximum-recursion-depth-using-pythons-pickle-cpickle Obviously this does not scale and could cause a crash but I don't see another solution short of writing our own implementation of pickle. """ if str(e).find('recursion') != -1: logger.warning('pylearn2.utils.save encountered the following ' 'error: ' + str(e) + '\nAttempting to resolve this error by calling ' + 'sys.setrecusionlimit and retrying') old_limit = sys.getrecursionlimit() try: sys.setrecursionlimit(50000) _save(filepath, obj) finally: sys.setrecursionlimit(old_limit) def get_pickle_protocol(): """ Allow configuration of the pickle protocol on a per-machine basis. This way, if you use multiple platforms with different versions of pickle, you can configure each of them to use the highest protocol supported by all of the machines that you want to be able to communicate. """ try: protocol_str = os.environ['PYLEARN2_PICKLE_PROTOCOL'] except KeyError: # If not defined, we default to 0 because this is the default # protocol used by cPickle.dump (and because it results in # maximum portability) protocol_str = '0' if protocol_str == 'pickle.HIGHEST_PROTOCOL': return pickle.HIGHEST_PROTOCOL return int(protocol_str) def _save(filepath, obj): """ .. todo:: WRITEME """ try: import joblib joblib_available = True except ImportError: joblib_available = False if filepath.endswith('.npy'): np.save(filepath, obj) return # This is dumb # assert filepath.endswith('.pkl') save_dir = os.path.dirname(filepath) # Handle current working directory case. if save_dir == '': save_dir = '.' if not os.path.exists(save_dir): os.makedirs(save_dir) if os.path.exists(save_dir) and not os.path.isdir(save_dir): raise IOError("save path %s exists, not a directory" % save_dir) elif not os.access(save_dir, os.W_OK): raise IOError("permission error creating %s" % filepath) try: if joblib_available and filepath.endswith('.joblib'): joblib.dump(obj, filepath) else: if filepath.endswith('.joblib'): warnings.warn('Warning: .joblib suffix specified but joblib ' 'unavailable. Using ordinary pickle.') with open(filepath, 'wb') as filehandle: cPickle.dump(obj, filehandle, get_pickle_protocol()) except Exception as e: logger.exception("cPickle has failed to write an object to " "{0}".format(filepath)) if str(e).find('maximum recursion depth exceeded') != -1: raise try: logger.info('retrying with pickle') with open(filepath, "wb") as f: pickle.dump(obj, f) except Exception as e2: if str(e) == '' and str(e2) == '': logger.exception('neither cPickle nor pickle could write to ' '{0}'.format(filepath)) logger.exception( 'moreover, neither of them raised an exception that ' 'can be converted to a string' ) logger.exception( 'now re-attempting to write with cPickle outside the ' 'try/catch loop so you can see if it prints anything ' 'when it dies' ) with open(filepath, 'wb') as f: cPickle.dump(obj, f, get_pickle_protocol()) logger.info('Somehow or other, the file write worked once ' 'we quit using the try/catch.') else: if str(e2) == 'env': raise import pdb tb = pdb.traceback.format_exc() reraise_as(IOError(str(obj) + ' could not be written to ' + str(filepath) + ' by cPickle due to ' + str(e) + ' nor by pickle due to ' + str(e2) + '. \nTraceback ' + tb)) logger.warning('{0} was written by pickle instead of cPickle, due to ' '{1} (perhaps your object' ' is really big?)'.format(filepath, e)) def clone_via_serialize(obj): """ Makes a "deep copy" of an object by serializing it and then deserializing it. Parameters ---------- obj : object The object to clone. Returns ------- obj2 : object A copy of the object. """ s = cPickle.dumps(obj, get_pickle_protocol()) return cPickle.loads(s) def to_string(obj): """ Serializes an object to a string. Parameters ---------- obj : object The object to serialize. Returns ------- string : str The object serialized as a string. """ return cPickle.dumps(obj, get_pickle_protocol()) def from_string(s): """ Deserializes an object from a string. Parameters ---------- s : str The object serialized as a string. Returns ------- obj : object The object. """ return cPickle.loads(s) def mkdir(filepath): """ Make a directory. Should succeed even if it needs to make more than one directory and nest subdirectories to do so. Raises an error if the directory can't be made. Does not raise an error if the directory already exists. Parameters ---------- filepath : WRITEME """ try: os.makedirs(filepath) except OSError: if not os.path.isdir(filepath): raise def read_int(fin, n=1): """ Reads n ints from a file. Parameters ---------- fin : file Readable file object n : int Number of ints to read Returns ------- rval : int or list The integer or integers requested """ if n == 1: s = fin.read(4) if len(s) != 4: raise ValueError('fin did not contain 4 bytes') return struct.unpack('i', s)[0] else: rval = [] for i in xrange(n): rval.append(read_int(fin)) return rval # dictionary to convert lush binary matrix magic numbers # to dtypes lush_magic = { 507333717: 'uint8', 507333716: 'int32', 507333713: 'float32', 507333715: 'float64' } def read_bin_lush_matrix(filepath): """ Reads a binary matrix saved by the lush library. Parameters ---------- filepath : str The path to the file. Returns ------- matrix : ndarray A NumPy version of the stored matrix. """ f = open(filepath, 'rb') try: magic = read_int(f) except ValueError: reraise_as("Couldn't read magic number") ndim = read_int(f) if ndim == 0: shape = () else: shape = read_int(f, max(3, ndim)) total_elems = 1 for dim in shape: total_elems *= dim try: dtype = lush_magic[magic] except KeyError: reraise_as(ValueError('Unrecognized lush magic number ' + str(magic))) rval = np.fromfile(file=f, dtype=dtype, count=total_elems) excess = f.read(-1) if excess: raise ValueError(str(len(excess)) + ' extra bytes found at end of file.' ' This indicates mismatch between header ' 'and content') rval = rval.reshape(*shape) f.close() return rval def load_train_file(config_file_path, environ=None): """ Loads and parses a yaml file for a Train object. Publishes the relevant training environment variables Parameters ---------- config_file_path : str Path to a config file containing a YAML string describing a pylearn2.train.Train object environ : dict, optional A dictionary used for ${FOO} substitutions in addition to environment variables when parsing the YAML file. If a key appears both in `os.environ` and this dictionary, the value in this dictionary is used. Returns ------- Object described by the YAML string stored in the config file """ from pylearn2.config import yaml_parse suffix_to_strip = '.yaml' # Publish environment variables related to file name if config_file_path.endswith(suffix_to_strip): config_file_full_stem = config_file_path[0:-len(suffix_to_strip)] else: config_file_full_stem = config_file_path os.environ["PYLEARN2_TRAIN_FILE_FULL_STEM"] = config_file_full_stem directory = config_file_path.split('/')[:-1] directory = '/'.join(directory) if directory != '': directory += '/' os.environ["PYLEARN2_TRAIN_DIR"] = directory os.environ["PYLEARN2_TRAIN_BASE_NAME"] = config_file_path.split('/')[-1] os.environ[ "PYLEARN2_TRAIN_FILE_STEM"] = config_file_full_stem.split('/')[-1] return yaml_parse.load_path(config_file_path, environ=environ) def _load(filepath, recurse_depth=0, retry=True): """ Recursively tries to load a file until success or maximum number of attempts. Parameters ---------- filepath : str A path to a file to load. Should be a pickle, Matlab, or NumPy file; or a .txt or .amat file that numpy.loadtxt can load. recurse_depth : int, optional End users should not use this argument. It is used by the function itself to implement the `retry` option recursively. retry : bool, optional If True, will make a handful of attempts to load the file before giving up. This can be useful if you are for example calling show_weights.py on a file that is actively being written to by a training script--sometimes the load attempt might fail if the training script writes at the same time show_weights tries to read, but if you try again after a few seconds you should be able to open the file. Returns ------- loaded_object : object The object that was stored in the file. """ try: import joblib joblib_available = True except ImportError: joblib_available = False if recurse_depth == 0: filepath = preprocess(filepath) if filepath.endswith('.npy') or filepath.endswith('.npz'): return np.load(filepath) if filepath.endswith('.amat') or filepath.endswith('txt'): try: return np.loadtxt(filepath) except Exception: reraise_as("{0} cannot be loaded by serial.load (trying " "to use np.loadtxt)".format(filepath)) if filepath.endswith('.mat'): global io if io is None: import scipy.io io = scipy.io try: return io.loadmat(filepath) except NotImplementedError as nei: if str(nei).find('HDF reader') != -1: global hdf_reader if hdf_reader is None: import h5py hdf_reader = h5py return hdf_reader.File(filepath, 'r') else: raise # this code should never be reached assert False # for loading PY2 pickle in PY3 encoding = {'encoding': 'latin-1'} if six.PY3 else {} def exponential_backoff(): if recurse_depth > 9: logger.info('Max number of tries exceeded while trying to open ' '{0}'.format(filepath)) logger.info('attempting to open via reading string') with open(filepath, 'rb') as f: content = f.read() return cPickle.loads(content, **encoding) else: nsec = 0.5 * (2.0 ** float(recurse_depth)) logger.info("Waiting {0} seconds and trying again".format(nsec)) time.sleep(nsec) return _load(filepath, recurse_depth + 1, retry) try: if not joblib_available: with open(filepath, 'rb') as f: obj = cPickle.load(f, **encoding) else: try: obj = joblib.load(filepath) except Exception as e: if os.path.exists(filepath) and not os.path.isdir(filepath): raise raise_cannot_open(filepath) except MemoryError as e: # We want to explicitly catch this exception because for MemoryError # __str__ returns the empty string, so some of our default printouts # below don't make a lot of sense. # Also, a lot of users assume any exception is a bug in the library, # so we can cut down on mail to pylearn-users by adding a message # that makes it clear this exception is caused by their machine not # meeting requirements. if os.path.splitext(filepath)[1] == ".pkl": improve_memory_error_message(e, ("You do not have enough memory to " "open %s \n" " + Try using numpy.{save,load} " "(file with extension '.npy') " "to save your file. It uses less " "memory when reading and " "writing files than pickled files.") % filepath) else: improve_memory_error_message(e, "You do not have enough memory to " "open %s" % filepath) except (BadPickleGet, EOFError, KeyError) as e: if not retry: reraise_as(e.__class__('Failed to open {0}'.format(filepath))) obj = exponential_backoff() except ValueError: logger.exception if not retry: reraise_as(ValueError('Failed to open {0}'.format(filepath))) obj = exponential_backoff() except Exception: # assert False reraise_as("Couldn't open {0}".format(filepath)) # if the object has no yaml_src, we give it one that just says it # came from this file. could cause trouble if you save obj again # to a different location if not hasattr(obj, 'yaml_src'): try: obj.yaml_src = '!pkl: "' + os.path.abspath(filepath) + '"' except Exception: pass return obj def raise_cannot_open(path): """ Raise an exception saying we can't open `path`. Parameters ---------- path : str The path we cannot open """ pieces = path.split('/') for i in xrange(1, len(pieces) + 1): so_far = '/'.join(pieces[0:i]) if not os.path.exists(so_far): if i == 1: if so_far == '': continue reraise_as(IOError('Cannot open ' + path + ' (' + so_far + ' does not exist)')) parent = '/'.join(pieces[0:i - 1]) bad = pieces[i - 1] if not os.path.isdir(parent): reraise_as(IOError("Cannot open " + path + " because " + parent + " is not a directory.")) candidates = os.listdir(parent) if len(candidates) == 0: reraise_as(IOError("Cannot open " + path + " because " + parent + " is empty.")) if len(candidates) > 100: # Don't attempt to guess the right name if the directory is # huge reraise_as(IOError("Cannot open " + path + " but can open " + parent + ".")) if os.path.islink(path): reraise_as(IOError(path + " appears to be a symlink to a " "non-existent file")) reraise_as(IOError("Cannot open " + path + " but can open " + parent + ". Did you mean " + match(bad, candidates) + " instead of " + bad + "?")) # end if # end for assert False
bsd-3-clause
9f1635ba770f56492e91b962a0949c0e
31.544426
115
0.558451
4.441138
false
false
false
false
lisa-lab/pylearn2
pylearn2/costs/ebm_estimation.py
34
11066
""" Training costs for unsupervised learning of energy-based models """ from functools import wraps import logging import numpy as np import sys from theano import scan import theano.tensor as T from theano.compat.six.moves import zip as izip from pylearn2.compat import OrderedDict from pylearn2.costs.cost import Cost, DefaultDataSpecsMixin from pylearn2.utils import py_integer_types from pylearn2.utils.rng import make_theano_rng from pylearn2.models.rbm import BlockGibbsSampler logger = logging.getLogger(__name__) logger.debug("Cost changing the recursion limit.") # We need this to be high enough that the big theano graphs we make # when unrolling inference don't cause python to complain. # python intentionally declares stack overflow well before the stack # segment is actually exceeded. But we can't make this value too big # either, or we'll get seg faults when the python interpreter really # does go over the stack segment. # IG encountered seg faults on eos3 (a machine at LISA labo) when using # 50000 so for now it is set to 40000. # I think the actual safe recursion limit can't be predicted in advance # because you don't know how big of a stack frame each function will # make, so there is not really a "correct" way to do this. Really the # python interpreter should provide an option to raise the error # precisely when you're going to exceed the stack segment. sys.setrecursionlimit(40000) class NCE(DefaultDataSpecsMixin, Cost): """ Noise-Contrastive Estimation See "Noise-Contrastive Estimation: A new estimation principle for unnormalized models" by Gutmann and Hyvarinen Parameters ---------- noise : WRITEME A Distribution from which noisy examples are generated noise_per_clean : int Number of noisy examples to generate for each clean example given """ def h(self, X, model): """ Computes `h` from the NCE paper. Parameters ---------- X : Theano matrix Batch of input data model : Model Any model with a `log_prob` method. Returns ------- h : A theano symbol for the `h` function from the paper. """ return - T.nnet.sigmoid(self.G(X, model)) def G(self, X, model): """ Computes `G` from the NCE paper. Parameters ---------- X : Theano matrix Batch of input data model : Model Any model with a `log_prob` method. Returns ------- G : A theano symbol for the `G` function from the paper. """ return model.log_prob(X) - self.noise.log_prob(X) def expr(self, model, data, noisy_data=None): """ Computes the NCE objective. Parameters ---------- model : Model Any Model that implements a `log_probs` method. data : Theano matrix noisy_data : Theano matrix, optional The noise samples used for noise-contrastive estimation. Will be generated internally if not provided. The keyword argument allows FixedVarDescr to provide the same noise across several steps of a line search. """ space, source = self.get_data_specs(model) space.validate(data) X = data if X.name is None: X_name = 'X' else: X_name = X.name m_data = X.shape[0] m_noise = m_data * self.noise_per_clean if noisy_data is not None: space.validate(noisy_data) Y = noisy_data else: Y = self.noise.random_design_matrix(m_noise) log_hx = -T.nnet.softplus(-self.G(X, model)) log_one_minus_hy = -T.nnet.softplus(self.G(Y, model)) # based on equation 3 of the paper # ours is the negative of theirs because # they maximize it and we minimize it rval = -T.mean(log_hx)-T.mean(log_one_minus_hy) rval.name = 'NCE('+X_name+')' return rval def __init__(self, noise, noise_per_clean): self.noise = noise assert isinstance(noise_per_clean, py_integer_types) self.noise_per_clean = noise_per_clean class SM(DefaultDataSpecsMixin, Cost): """ (Regularized) Score Matching See: - "Regularized estimation of image statistics by Score Matching", D. Kingma, Y. LeCun, NIPS 2010 - eqn. 4 of "On Autoencoders and Score Matching for Energy Based Models" Swersky et al 2011 Uses the mean over visible units rather than sum over visible units so that hyperparameters won't depend as much on the # of visible units Parameters ---------- lambd : WRITEME """ def __init__(self, lambd=0): assert lambd >= 0 self.lambd = lambd @wraps(Cost.expr) def expr(self, model, data): self.get_data_specs(model)[0].validate(data) X = data X_name = 'X' if X.name is None else X.name def f(i, _X, _dx): return T.grad(_dx[:, i].sum(), _X)[:, i] dx = model.score(X) ddx, _ = scan(f, sequences=[T.arange(X.shape[1])], non_sequences=[X, dx]) ddx = ddx.T assert len(ddx.type.broadcastable) == 2 rval = T.mean(0.5 * dx**2 + ddx + self.lambd * ddx**2) rval.name = 'sm('+X_name+')' return rval class SMD(DefaultDataSpecsMixin, Cost): """ Denoising Score Matching See eqn. 4.3 of "A Connection Between Score Matching and Denoising Autoencoders" by Pascal Vincent for details Note that instead of using half the squared norm we use the mean squared error, so that hyperparameters don't depend as much on the # of visible units Parameters ---------- corruptor : WRITEME WRITEME """ def __init__(self, corruptor): super(SMD, self).__init__() self.corruptor = corruptor @wraps(Cost.expr) def expr(self, model, data): self.get_data_specs(model)[0].validate(data) X = data X_name = 'X' if X.name is None else X.name corrupted_X = self.corruptor(X) if corrupted_X.name is None: corrupted_X.name = 'corrupt('+X_name+')' model_score = model.score(corrupted_X) assert len(model_score.type.broadcastable) == len(X.type.broadcastable) parzen_score = T.grad( - T.sum(self.corruptor.corruption_free_energy(corrupted_X, X)), corrupted_X) assert \ len(parzen_score.type.broadcastable) == len(X.type.broadcastable) score_diff = model_score - parzen_score score_diff.name = 'smd_score_diff('+X_name+')' assert len(score_diff.type.broadcastable) == len(X.type.broadcastable) # TODO: this could probably be faster as a tensordot, # but we don't have tensordot for gpu yet sq_score_diff = T.sqr(score_diff) # sq_score_diff = Print('sq_score_diff',attrs=['mean'])(sq_score_diff) smd = T.mean(sq_score_diff) smd.name = 'SMD('+X_name+')' return smd class SML(Cost): """ Stochastic Maximum Likelihood See "On the convergence of Markovian stochastic algorithms with rapidly decreasing ergodicity rates" by Laurent Younes (1998) Also known as Persistent Constrastive Divergence (PCD) See "Training restricted boltzmann machines using approximations to the likelihood gradient" by Tijmen Tieleman (2008) The number of particles fits the batch size. Parameters ---------- batch_size: int Batch size of the training algorithm nsteps: int Number of steps made by the block Gibbs sampler between each epoch """ def __init__(self, batch_size, nsteps): super(SML, self).__init__() self.nchains = batch_size self.nsteps = nsteps @wraps(Cost.get_gradients) def get_gradients(self, model, data, **kwargs): cost = self._cost(model, data, **kwargs) params = list(model.get_params()) grads = T.grad(cost, params, disconnected_inputs='ignore', consider_constant=[self.sampler.particles]) gradients = OrderedDict(izip(params, grads)) updates = OrderedDict() sampler_updates = self.sampler.updates() updates.update(sampler_updates) return gradients, updates def _cost(self, model, data): """ A fake cost that we differentiate symbolically to derive the SML update rule. Parameters ---------- model : Model data : Batch in get_data_specs format Returns ------- cost : 0-d Theano tensor The fake cost """ if not hasattr(self, 'sampler'): self.sampler = BlockGibbsSampler( rbm=model, particles=0.5+np.zeros((self.nchains, model.get_input_dim())), rng=model.rng, steps=self.nsteps) # compute negative phase updates sampler_updates = self.sampler.updates() # Compute SML cost pos_v = data neg_v = self.sampler.particles ml_cost = (model.free_energy(pos_v).mean() - model.free_energy(neg_v).mean()) return ml_cost @wraps(Cost.expr) def expr(self, model, data): return None @wraps(Cost.get_data_specs) def get_data_specs(self, model): return (model.get_input_space(), model.get_input_source()) class CDk(Cost): """ Contrastive Divergence. See "Training products of experts by minimizing contrastive divergence" by Geoffrey E. Hinton (2002) Parameters ---------- nsteps : int Number of Markov chain steps for the negative sample seed : int Seed for the random number generator """ def __init__(self, nsteps, seed=42): super(CDk, self).__init__() self.nsteps = nsteps self.rng = make_theano_rng(seed, which_method='binomial') def _cost(self, model, data): pos_v = data neg_v = data for k in range(self.nsteps): [neg_v, _locals] = model.gibbs_step_for_v(neg_v, self.rng) # Compute CD cost ml_cost = (model.free_energy(pos_v).mean() - model.free_energy(neg_v).mean()) return ml_cost, neg_v @wraps(Cost.get_gradients) def get_gradients(self, model, data, **kwargs): cost, neg_v = self._cost(model, data, **kwargs) params = list(model.get_params()) grads = T.grad(cost, params, disconnected_inputs='ignore', consider_constant=[neg_v]) gradients = OrderedDict(izip(params, grads)) updates = OrderedDict() return gradients, updates @wraps(Cost.expr) def expr(self, model, data): return None @wraps(Cost.expr) def get_data_specs(self, model): return (model.get_input_space(), model.get_input_source())
bsd-3-clause
3a5a1649e2403c52c14b5639b41045ed
28.121053
79
0.604735
3.827741
false
false
false
false
lisa-lab/pylearn2
pylearn2/train_extensions/__init__.py
49
5909
"""Plugins for the Train object.""" __authors__ = "Ian Goodfellow" __copyright__ = "Copyright 2010-2012, Universite de Montreal" __credits__ = ["Ian Goodfellow", "David Warde-Farley"] __license__ = "3-clause BSD" __maintainer__ = "LISA Lab" __email__ = "pylearn-dev@googlegroups" import functools import logging import numpy as np logger = logging.getLogger(__name__) class TrainExtension(object): """ An object called by pylearn2.train.Train at various points during learning. Useful for adding custom features to the basic learning procedure. This base class implements all callback methods as no-ops. To add a feature to the Train class, implement a subclass of this base class that overrides any subset of these no-op methods. """ def on_save(self, model, dataset, algorithm): """ Train calls this immediately before it saves the model. Parameters ---------- model : pylearn2.models.Model The model object being trained. dataset : pylearn2.datasets.Dataset The dataset object used for training. algorithm : pylearn2.training_algorithms.TrainingAlgorithm The object representing the training algorithm being used to train the model. """ def on_monitor(self, model, dataset, algorithm): """ Train calls this immediately after each call to the Monitor (i.e., when training begins, and at the end of each epoch). Parameters ---------- model : pylearn2.models.Model The model object being trained dataset : pylearn2.datasets.Dataset The dataset object being trained. algorithm : pylearn2.training_algorithms.TrainingAlgorithm The object representing the training algorithm being used to train the model. """ def setup(self, model, dataset, algorithm): """ Train calls this immediately upon instantiation, before any monitoring is done. Parameters ---------- model : pylearn2.models.Model The model object being trained. dataset : pylearn2.datasets.Dataset The dataset object being trained. algorithm : pylearn2.training_algorithms.TrainingAlgorithm The object representing the training algorithm being used to train the model. """ class SharedSetter(TrainExtension): """ Sets shared variables to take on the specified values after the specified amounts of epochs have taken place. epoch_updates = [ [i, x, y] ] means run x.set_value(cast(y)) after i epochs have passed. Parameters ---------- epoch_updates : WRITEME """ def __init__(self, epoch_updates): self._count = 0 self._epoch_to_updates = {} self._vars = set([]) for update in epoch_updates: epoch, var, val = update self._vars.add(var) if epoch not in self._epoch_to_updates: self._epoch_to_updates[epoch] = [] assert hasattr(var, 'get_value') assert var.name is not None self._epoch_to_updates[epoch].append((var,val)) @functools.wraps(TrainExtension.on_monitor) def on_monitor(self, model, dataset, algorithm): # TODO: write more specific docstring if self._count == 0: monitor = model.monitor # TODO: make Monitor support input-less channels so this hack # isn't necessary hack = monitor.channels.values()[0] for var in self._vars: monitor.add_channel(name=var.name, val=var, ipt=hack.graph_input, dataset=hack.dataset) if self._count in self._epoch_to_updates: for update in self._epoch_to_updates[self._count]: var, val = update var.set_value(np.cast[var.dtype](val)) self._count += 1 class ChannelSmoother(TrainExtension): """ Makes a smoothed version of a monitoring channel by averaging together the k most recent values of that channel. This is a little bit dangerous because if other TrainExtensions depend on the channel being up to date they must appear after this one in the extensions list. A better long term solution would be to make the Monitor support this kind of channel directly instead of hacking it in. Note that the Monitor will print this channel as having a value of -1, and then the extension will print the right value. Parameters ---------- channel_to_smooth : WRITEME channel_to_publish : WRITEME k : WRITEME """ def __init__(self, channel_to_smooth, channel_to_publish, k=5): self.__dict__.update(locals()) del self.self @functools.wraps(TrainExtension.setup) def setup(self, model, dataset, algorithm): # TODO: more specific docstring monitor = model.monitor channels = monitor.channels channel_to_smooth = channels[self.channel_to_smooth] ipt = channel_to_smooth.graph_input dataset = channel_to_smooth.dataset monitor.add_channel(name=self.channel_to_publish, ipt=ipt, val=-1., dataset=dataset) self.in_ch = channel_to_smooth self.out_ch = channels[self.channel_to_publish] @functools.wraps(TrainExtension.on_monitor) def on_monitor(self, model, dataset, algorithm): # TODO: write more specific docstring val_record = self.in_ch.val_record start = max(0, len(val_record) - self.k + 1) values = val_record[start:] mean = sum(values) / float(len(values)) self.out_ch.val_record[-1] = mean logger.info('\t{0}: {1}'.format(self.channel_to_publish, mean))
bsd-3-clause
6c7714a92efc601a428708dfdfea5d67
32.196629
79
0.620917
4.412995
false
false
false
false
lisa-lab/pylearn2
pylearn2/scripts/summarize_model.py
44
2999
#!/usr/bin/env python """ This script summarizes a model by showing some statistics about the parameters and checking whether the model completed training succesfully """ from __future__ import print_function __authors__ = "Ian Goodfellow" __copyright__ = "Copyright 2010-2012, Universite de Montreal" __credits__ = ["Ian Goodfellow"] __license__ = "3-clause BSD" __maintainer__ = "LISA Lab" __email__ = "pylearn-dev@googlegroups" import argparse import numpy as np from pylearn2.compat import first_key from pylearn2.utils import serial def summarize(path): """ Summarize the model Parameters ---------- path : str The path to the pickled model to summarize """ model = serial.load(path) for param in model.get_params(): name = param.name if name is None: name = '<anon>' v = param.get_value() print(name + ': ' + str((v.min(), v.mean(), v.max())), end='') print(str(v.shape)) if np.sign(v.min()) != np.sign(v.max()): v = np.abs(v) print('abs(' + name + '): ' + str((v.min(), v.mean(), v.max()))) if v.ndim == 2: row_norms = np.sqrt(np.square(v).sum(axis=1)) print(name + " row norms:", end='') print((row_norms.min(), row_norms.mean(), row_norms.max())) col_norms = np.sqrt(np.square(v).sum(axis=0)) print(name + " col norms:", end='') print((col_norms.min(), col_norms.mean(), col_norms.max())) if hasattr(model, 'monitor'): print('trained on', model.monitor.get_examples_seen(), 'examples') print('which corresponds to ', end='') print(model.monitor.get_batches_seen(), 'batches') key = first_key(model.monitor.channels) hour = float(model.monitor.channels[key].time_record[-1]) / 3600. print('Trained for {0} hours'.format(hour)) try: print(model.monitor.get_epochs_seen(), 'epochs') except Exception: pass if hasattr(model.monitor, 'training_succeeded'): if model.monitor.training_succeeded: print('Training succeeded') else: print('Training was not yet completed ' + 'at the time of this save.') else: print('This pickle file is damaged, or was made before the ' + 'Monitor tracked whether training completed.') def make_argument_parser(): """ Creates an ArgumentParser to read the options for this script from sys.argv """ parser = argparse.ArgumentParser( description="Print some parameter statistics of a pickled model " "and check if it completed training succesfully." ) parser.add_argument('path', help='The pickled model to summarize') return parser if __name__ == "__main__": parser = make_argument_parser() args = parser.parse_args() summarize(args.path)
bsd-3-clause
a2449dff6375002e8a64399017b2a66e
31.956044
76
0.579193
3.935696
false
false
false
false
lisa-lab/pylearn2
pylearn2/models/mnd.py
49
4148
__authors__ = "Ian Goodfellow" __copyright__ = "Copyright 2010-2012, Universite de Montreal" __credits__ = ["Ian Goodfellow"] __license__ = "3-clause BSD" __maintainer__ = "LISA Lab" __email__ = "pylearn-dev@googlegroups" from pylearn2.models.model import Model from pylearn2.utils import sharedX import numpy as np import theano.tensor as T class DiagonalMND(Model): """ A model based on the multivariate normal distribution. This variant is constrained to have diagonal covariance. Parameters ---------- nvis : WRITEME init_beta : WRITEME init_mu : WRITEME min_beta : WRITEME max_beta : WRITEME """ # TODO: unify this with distribution.mnd def __init__(self, nvis, init_beta, init_mu, min_beta, max_beta): #copy all arguments to the object self.__dict__.update( locals() ) del self.self super(DiagonalMND,self).__init__() #build the object self.redo_everything() def redo_everything(self): """ .. todo:: WRITEME """ self.beta = sharedX(np.ones((self.nvis,))*self.init_beta,'beta') self.mu = sharedX(np.ones((self.nvis,))*self.init_mu,'mu') self.redo_theano() def free_energy(self, X): """ .. todo:: WRITEME """ diff = X-self.mu sq = T.sqr(diff) return 0.5 * T.dot( sq, self.beta ) def log_prob(self, X): """ .. todo:: WRITEME """ return -self.free_energy(X) - self.log_partition_function() def log_partition_function(self): """ .. todo:: WRITEME """ # Z^-1 = (2pi)^{-nvis/2} det( beta^-1 )^{-1/2} # Z = (2pi)^(nvis/2) sqrt( det( beta^-1) ) # log Z = (nvis/2) log 2pi - (1/2) sum(log(beta)) return float(self.nvis)/2. * np.log(2*np.pi) - 0.5 * T.sum(T.log(self.beta)) def redo_theano(self): """ .. todo:: WRITEME """ init_names = dir(self) self.censored_updates = {} for param in self.get_params(): self.censored_updates[param] = set([]) final_names = dir(self) self.register_names_to_del( [name for name in final_names if name not in init_names]) def _modify_updates(self, updates): """ .. todo:: WRITEME """ if self.beta in updates and updates[self.beta] not in self.censored_updates[self.beta]: updates[self.beta] = T.clip(updates[self.beta], self.min_beta, self.max_beta ) params = self.get_params() for param in updates: if param in params: self.censored_updates[param] = self.censored_updates[param].union(set([updates[param]])) def get_params(self): """ .. todo:: WRITEME """ return [self.mu, self.beta ] def kl_divergence(q,p): """ .. todo:: WRITEME """ #KL divergence of two DiagonalMNDs #http://en.wikipedia.org/wiki/Kullback%E2%80%93Leibler_divergence#KL_divergence_for_Normal_Distributions #D_KL(q||p) = 0.5 ( beta_q^T beta_p^-1 + beta_p^T sq(mu_p - mu_q) - log(det Siga_q / det Sigma_p) - k) assert isinstance(q,DiagonalMND) assert isinstance(p,DiagonalMND) assert q.nvis == p.nvis k = q.nvis beta_q = q.beta beta_p = p.beta beta_q_inv = 1./beta_q trace_term = T.dot(beta_q_inv,beta_p) assert trace_term.ndim == 0 mu_p = p.mu mu_q = q.mu quad_term = T.dot(beta_p, T.sqr(mu_p-mu_q)) assert quad_term.ndim == 0 # - log ( det Sigma_q / det Sigma_p) # = log det Sigma_p - log det Sigma_q # = log det Beta_p_inv - log det Beta_q_inv # = sum(log(beta_p_inv)) - sum(log(beta_q_inv)) # = sum(log(beta_q)) - sum(log(beta_p)) log_term = T.sum(T.log(beta_q)) - T.sum(T.log(beta_p)) assert log_term.ndim == 0 inside_parens = trace_term + quad_term + log_term - k assert inside_parens.ndim == 0 rval = 0.5 * inside_parens return rval
bsd-3-clause
c6b9c32aa570dc2d2fa491ffa122bbac
23.544379
108
0.543877
3.222999
false
false
false
false
lisa-lab/pylearn2
pylearn2/utils/tests/test_general.py
45
1096
""" Tests for pylearn2.utils.general functions. """ from pylearn2.utils import contains_nan, contains_inf, isfinite import numpy as np def test_contains_nan(): """ Tests that pylearn2.utils.contains_nan correctly identifies `np.nan` values in an array. """ arr = np.random.random(100) assert not contains_nan(arr) arr[0] = np.nan assert contains_nan(arr) def test_contains_inf(): """ Tests that pylearn2.utils.contains_inf correctly identifies `np.inf` values in an array. """ arr = np.random.random(100) assert not contains_inf(arr) arr[0] = np.nan assert not contains_inf(arr) arr[1] = np.inf assert contains_inf(arr) arr[1] = -np.inf assert contains_inf(arr) def test_isfinite(): """ Tests that pylearn2.utils.isfinite correctly identifies `np.nan` and `np.inf` values in an array. """ arr = np.random.random(100) assert isfinite(arr) arr[0] = np.nan assert not isfinite(arr) arr[0] = np.inf assert not isfinite(arr) arr[0] = -np.inf assert not isfinite(arr)
bsd-3-clause
47c760e1650f3b491f82a8e22503a7de
22.826087
63
0.644161
3.321212
false
true
false
false
lisa-lab/pylearn2
pylearn2/dataset_get/dataset-get.py
44
36792
#!/usr/bin/env python # -*- coding: utf-8 ######################################## # # # This file is intentionally monolithic. # It also intentionally restricts itself # to standard library modules, with no # extra dependencies. # from __future__ import print_function __authors__ = "Steven Pigeon" __copyright__ = "(c) 2012, Université de Montréal" __contact__ = "Steven Pigeon: pigeon@iro.umontreal.ca" __version__ = "dataset-get 0.1" __licence__ = "BSD 3-Clause http://www.opensource.org/licenses/BSD-3-Clause " import logging import re,os,sys,shutil,time import warnings import urllib,urllib2 import tarfile import subprocess from theano.compat.six.moves import input logger = logging.getLogger(__name__) ######################################## class package_info: """ A simple class to structure the package's information """ def __init__(self, cf, name,ts,rs,src,whr): self.configuration_file=cf # in which configuration file was it found? self.name=name # the short name, e.g., "mnist" self.timestamp=int(ts) # a unix ctime self.readable_size=rs # a human-readable size, e.g., "401.3MB" self.source=src # the web source self.where=whr # where on this machine ######################################## # # Global variables for the whole module. # dataset_sources="sources.lst" dataset_web="http://www.stevenpigeon.org/secret" dataset_conf_path="" dataset_data_path="" root_conf_path=None root_data_path=None user_conf_path=None user_data_path=None super_powers=False # both dictionaries for fast search # (but are semantically lists) packages_sources={} installed_packages_list={} ######################################## def local_path_as_url( filename ): """ Takes a local, OS-specific path or filename and transforms it into an url starting with file:// (it simplifies a lot of things). :param filename: a relative or absolute pathname :returns: the urlified absolute path """ return "file://"+urllib.pathname2url(os.path.abspath(filename)) ######################################## def has_super_powers(): """ Determines whether or not the program is run as root. :returns: true if run as root, false otherwise """ return os.geteuid()==0 ######################################## def corename( filename ): """ returns the 'corename' of a file. For example, corename("thingie.tar.bz2") returns "thingie" (a totally correct way of doing this would be to use MIME-approved standard extensions, in order to distinguish from, say a file "thingie.tar.bz2" and another file "my.data.tar.bz2"---for which we would have only "my" as corename) :param filename: a (base) filename :returns: the "core" filename """ f1=None f2=os.path.basename(filename) # repeatedly remove the right-most # extension, until none is found # while f1 != f2: f1=f2 (f2,ext)=os.path.splitext(f1) return f2 ######################################## def get_timestamp_from_url( url ): """ Gets the Last-Modified field from the http header associated with the file pointed to by the url. Raises whatever exception urllib2.urlopen raises. It can't lookup local file, unless they are presented as a file:/// url. :param url: a filename or an url :returns: the last-modified timestamp """ obj = urllib2.urlopen( url ) return time.strptime( obj.info()["Last-Modified"], "%a, %d %b %Y %H:%M:%S GMT") # RFC 2822 date format ######################################## def download_from_url( url, filename=None, progress_hook=None ): """ Downloads a file from an URL in the specificed filename (or, if filename is None, to a temporary location). Returns the location of the downloaded file. :param url: url of the file to download :param filename: filename to download to (None means a temp file is created) :param progress_hook: a download hook to display progress :returns: the filename where the file was downloaded """ (temp_filename, headers)=urllib.urlretrieve( url,filename,progress_hook ) return temp_filename ######################################## def file_access_rights( filename, rights, check_above=False ): """ Determines if a file has given rights. If the file exists, it tests for its rights. If it does not exist, and check_above=True, then it checks for the directory's rights, to test for write/creation rights. :param filename: filename of the file to assess :param rights: rights to be tested :param check_above: Check directory rights if file does not exist. :returns: boolean, whether 'rights' rights are OK """ if os.path.exists(filename): return os.access(filename, rights) else: if check_above: return os.access(os.path.dirname(os.path.abspath(filename)), rights) else: return False ######################################## def atomic_replace( src_filename, dst_filename ): """ Does an "atomic" replace of a file by another. If both files reside on the fame FS device, atomic_replace does a regular move. If not, the source file is first copied to a temporary location on the same FS as the destination, then a regular move is performed. caveat: the destination FS must have enough storage left for the temporary file. :param src_filename: The file to replace from :param dst_filename: The file to be replaced :raises: whatever shutil raises """ #################################### def same_fs( filename_a, filename_b): """ Checks if both files reside on the same FS device """ stats_a = os.stat(filename_a) stats_b = os.stat(filename_b) return stats_a.st_dev == stats_b.st_dev; if os.path.exists(dst_filename) and not same_fs(src_filename,dst_filename): # deals with non-atomic move # dst_path = os.path.dirname(os.path.abspath(dst_filename)) dst_temp_filename=os.tempnam(dst_path); shutil.copy(src_filename, dst_temp_filename) # non-atomic shutil.move(dst_temp_filename,dst_filename) # atomic else: # an atomic move is performed # (both files are on the same device, # or the destination doesn't exist) # shutil.move(src_filename, dst_filename) ######################################## def set_defaults(): """ Detects whether the program is run as an ordinary user or as root, and then sets defauts directories for packages, configurations, and sources. caveat: this is an FreeDesktop-friendly version, and we will need eventually to have Windows- and OSX-friendly versions. See: http://freedesktop.org/wiki/Home and: http://www.linuxfoundation.org/collaborate/workgroups/lsb/fhs """ global dataset_conf_path, \ dataset_data_path, \ root_conf_path, \ root_data_path, \ user_conf_path, \ super_powers # a conspicuously LINUX version # (on windows, if we ever do a # windows version, these would # be different, and we may even # not have 'root' per se.) # root_conf_path="/etc/pylearn/" root_data_path="/usr/share/pylearn/dataset/" user_conf_path=os.path.join(os.environ["HOME"],".local/share/pylearn/") user_data_path=os.path.join(os.environ["HOME"],".local/share/pylearn/dataset/") if has_super_powers(): dataset_conf_path=root_conf_path dataset_data_path=root_data_path super_powers=True else: dataset_conf_path=user_conf_path dataset_data_path=user_data_path super_powers=False # check if directories exist, and if not, # create them, and then fetch source.lst # if not os.path.exists(dataset_conf_path): os.makedirs(dataset_conf_path) if not os.path.exists(os.path.join(dataset_conf_path,dataset_sources)): atomic_update(os.path.join(dataset_web,dataset_sources), os.path.join(dataset_conf_path,dataset_sources), progress_bar) if not os.path.exists(dataset_data_path): os.makedirs(dataset_data_path) read_packages_sources() read_installed_packages_list(); ######################################## def read_packages_sources(): """ Reads the sources.lst file and populates the available packages list. caveat: parsing of the sources.lst is pathetic Assigns: packages_sources :raises: RuntimeError if sources.lst cannot be read """ def read_from_file(config_filename): """ Reads a sources.lst file from a given location :param config_filename: the configuration file to read """ global packages_sources try: f=open(config_filename,"r") except Exception as e: # not a problem if not found in a given location pass else: # file opened for line in f: t=line.rstrip().split(' ') # rstrips strips whitespaces at the end (\n) packages_sources[t[0]]=\ this_package=package_info( config_filename, t[0], # name t[1], # timestamp t[2], # human-readable size urllib.unquote(t[3]), # source on the web None) # None as not installed (from source) (may be overridden later) if super_powers: read_from_file(os.path.join(dataset_conf_path,dataset_sources)) else: # read root, user, then paths. paths=[ os.path.join(root_conf_path,dataset_sources), os.path.join(user_conf_path,dataset_sources) ] try: paths+=[ os.path.join(x,dataset_sources) for x in re.split(":|;",os.environ["PYLEARN2_DATA_PATH"]) ] except Exception: # PYLEARN2_DATA_PATH may or mayn't be defined pass for path in paths: read_from_file(path) if len(packages_sources)==0: raise RuntimeError( "[cf] fatal: could not find/read sources.lst (unexpected!)" ) ######################################## def read_installed_packages_list(): """ Reads the various installed.lst files found on the system. First it searches for the root-installed installed.lst, then the user's, then searches the locations specified by the environment variable PYLEARN2_DATA_PATH (which is a standard :-separated list of locations) Assigns: installed_packages_list """ # note: we add and overwrite rather # than clearing and filling (so we can # read many installed.lst, but the last # ones read overrides the earlier ones) # def read_from_file(config_filename): """ Reads an installed.lst file from a given location :param config_filename: the configuration file to read """ global installed_packages_list try: installed_list_file=open(config_filename) except IOError: # not a problem if not found in a location pass else: # read from file and # create a dictionary # for line in installed_list_file: l=line.rstrip().split(' ') # removes trailing whitespaces (\n) if l: installed_packages_list[l[0]]=\ this_package=package_info( config_filename, l[0], # name l[1], # timestamp l[2], # human-readable size urllib.unquote(l[3]), # source on the web urllib.unquote(l[4])) # where installed else: pass # skip blank lines (there shouldn't be any) if super_powers: # then read only root read_from_file(os.path.join(dataset_conf_path,"installed.lst")) else: # read root, user, then paths. paths=[ os.path.join(root_conf_path,"installed.lst"), os.path.join(user_conf_path,"installed.lst") ] try: paths+=[ os.path.join(x,"installed.lst") for x in re.split(":|;",os.environ["PYLEARN2_DATA_PATH"]) ] except Exception: # PYLEARN2_DATA_PATH may or mayn't be defined pass for path in paths: read_from_file(path) if len(installed_packages_list)==0: logger.warning("[cf] no install.lst found " "(will be created on install/upgrade)") ######################################## def write_installed_packages_list(): """ Saves the installed package list and their location (file over-writen depends on run as root or as a normal user) """ global installed_packages_list try: tmp=open(os.path.join(dataset_conf_path,"installed.lst.2"),"w") except IOError: raise RuntimeError("[cf] fatal: cannot create temp file") else: # ok, probably worked? for package in installed_packages_list.values(): # adds only packages that are readable for # this user (maybe some site-installed datasets # are out of his reach) # if package.where!=None and \ file_access_rights(os.path.join(package.where,package.name), os.F_OK | os.R_OK): print( " ".join(map(str,[ package.name, package.timestamp, package.readable_size, urllib.quote(package.source,"/:~"), urllib.quote(package.where,"/:~") ] )), file=tmp) # replace the installed.lst in # a safe way atomic_replace(os.path.join(dataset_conf_path,"installed.lst.2"), os.path.join(dataset_conf_path,"installed.lst")) ######################################## def atomic_update( remote_src, local_dst, hook=None ): """ Takes a (possibly) remote file an checks if it is newer than a(n obligatoritly) local file. If the source is newer, an "atomic update" is performed. Atomic here means that the source is downloaded in a distinct location, and only if download is successful is the destination file replaced atomically. :param remote_src: Url to a (possibly) remote file :param local_dst: file to update :param hook: download progress hook :raises: various IOErrors """ global hook_download_filename # hook-related try: remote_date = get_timestamp_from_url(remote_src); except IOError as e: raise IOError("[ts] %s %s" % (str(e),remote_src)) else: if os.path.exists(local_dst): # it exists, test for update try: local_date = get_timestamp_from_url(local_path_as_url(local_dst)) except Exception as e: raise IOError("[ts] %s %s" % (str(e),local_dst)) else: if (local_date<remote_date): # OK, the file seems to be out-of-date # let's update it # if file_access_rights(local_dst,os.W_OK,check_above=True): # we have write access to the file, or if it doesn't # exist, to the directory where we want to write it. # try: hook_download_filename=remote_src # hook-related temp_filename=download_from_url(remote_src, filename=None, progress_hook=hook) except Exception as e: raise IOError("[dl] %s %s" % (str(e),remote_src)) else: # download to temporary was successful, # let's (try to) perform the atomic replace # try: atomic_replace(temp_filename,local_dst) except Exception as e: raise IOError("[ac] %s %s --> %s" % (str(e),temp_filename,local_dst)) else: raise IOError("[rw] no write access to %s " % local_dst ) else: # file's up to date, everything's fine # and there's nothing else to do # pass else: # file does not exist, just download! # if file_access_rights(local_dst,os.W_OK,check_above=True): try: hook_download_filename=remote_src # hook-related temp_filename=download_from_url(remote_src, filename=None, progress_hook=hook) except Exception as e: raise IOError("[dl] %s %s" % (str(e),remote_src)) else: # yay! download successful! # try: atomic_replace(temp_filename,local_dst) except Exception as e: raise IOError("[ac] %s %s --> %s" % (str(e),temp_filename,local_dst)) else: raise IOError("[rw] no right access to %s" % local_dst) ######################################## def unpack_tarball( tar_filename, dest_path ): """ Unpacks a (bzipped2) tarball to a destination directory :param tar_filename: the bzipped2 tar file :param dest_path: a path to where expand the tarball :raises: various IOErrors """ if os.path.exists(tar_filename): if file_access_rights(dest_path,os.W_OK,check_above=False): try: # open the tarball as read, bz2 # this_tar_file=tarfile.open(tar_filename,"r:bz2") except Exception as e: raise IOError("[tar] cannot open '%s'" % tar_filename) else: # ok, it's openable, let's expand it # try: this_tar_file.extractall(dest_path) except Exception as e: raise IOError("[tar] error while extracting '%s'" %tar_filename) else: # yay! success! pass else: raise IOError("[tar] no right access to '%s'" % dest_path) else: raise IOError("'%s' not found" % tar_filename) ######################################## def run_scripts( package_location, scripts ): """ Search for installation scripts speficied by the scripts list :param package_location: "root" path for the package :param scripts: list of scripts to look for (and execute) :raises: subprocess exceptions """ path=os.path.join(package_location,"scripts/") cwd=os.getcwd() os.chdir(path) for script in scripts: if os.path.exists( script ): # throws CalledProcessError if return # return code is not zero. # try: subprocess.check_call( script, stdout=sys.stdout, stderr=sys.stderr ) except Exception: os.chdir(cwd) raise # ok, success (or not), let's unstack os.chdir(cwd) ######################################## def install_package( package, src, dst ): """ Unpacks a (bzipped2) tarball and expands it to the given location. If unpacking is successful, installation scripts are run. :param package: package information :param src: the source tarball :param dst: the destination directory :raises: IOErrors and subprocess exceptions """ #FIXME: change creation flags to group-public # readable when invoked with super-powers # unpack_tarball(src,dst) run_scripts(dst+package.name, scripts=["getscript","postinst"] ) ######################################## def remove_package(package,dst): """ Removes a script by running the various removal scripts, then by deleting files and directories. :param package: package information :param dst: packages root (where packages are installed) """ #path=os.path.join(dst,package.name) path=os.path.join(package.where,package.name) #print path run_scripts(path,scripts=["prerm"]) shutil.rmtree(os.path.join(path,"data/")) shutil.rmtree(os.path.join(path,"docs/")) run_scripts(os.path.join(dst,package.name),scripts=["postrm"]) shutil.rmtree(os.path.join(path,"scripts/")) shutil.rmtree(path) update_installed_list("r",package) ######################################## def update_installed_list( op, package ): """ Updates the internal list of installed packages. The operation is either "i" for install and update, or "r" for removal :param op: the operation performed :param package: the package information :param dst: where the package was installed """ if op=="i": installed_packages_list[package.name]=package; elif op=="r": # remove the package from the list del installed_packages_list[package.name] else: raise RuntimeError("[cf] fatal: invalid configuration op '%s'." % op) write_installed_packages_list() ######################################## def show_packages(): """ List all available packages, both installed or from remove sources """ logger.info("These packages are available:") for this_package in packages_sources.values(): if this_package.name in installed_packages_list: state="u" if installed_packages_list[this_package.name].timestamp<this_package.timestamp else 'i'; else: state="-" package_time = time.strftime("%a, %d %b %Y %H:%M:%S", time.gmtime(this_package.timestamp)) logger.info("{0} {1:<20} {2:<8} " "{3:<30} {4}".format(state, this_package.name, this_package.readable_size, package_time, this_package.source)) ######################################## def install_upgrade( package, upgrade=False,progress_hook=None ): """ This function installs or upgrades a package. :param package: package information :param upgrade: If True, performs and upgrade, installs underwise :param progress_hook: a download progress hook """ global hook_download_filename # hook-related if upgrade: operation = "[up] upgrading" else: operation = "[in] installing" logger.info("{0} '{1}' to {2}".format(operation, package.name, dataset_data_path)) remote_src=package.source # install location is determined by super-powers # (so a root package can be upgraded locally!) package.where=dataset_data_path; # TODO: to add caching, first lookup the # tarball in the package cache (but there's'nt # one for now) # cached=False; if not cached: hook_download_filename=remote_src # hook-related temp_filename=download_from_url(remote_src,filename=None,progress_hook=progress_hook) else: # assign filename to cached package pass logger.info("[in] running install scripts " "for package '{0}'".format(package.name)) # runs through the .../package_name/scripts/ # directory and executes the scripts in a # specific order (which shouldn't display # much unless they fail) # install_package(package,temp_filename,dataset_data_path) update_installed_list("i",package) ######################################## def upgrade_packages(packages_to_upgrade, hook=None ): """ Upgrades packages. If no packages are supplied, it will perform an "update-all" operation, finding all packages that are out of date. If packages names are supplied, only those are checked for upgrade (and upgraded if out of date) :param packages_to_upgrade: list of package names. :raises: IOErrors (from downloads/rights) """ # get names only if packages_to_upgrade==[]: packages_to_upgrade=installed_packages_list.keys() # all installed! all_packages=True else: all_packages=False # check what packages are in the list, # and really to be upgraded. # packages_really_to_upgrade=[] for this_package in packages_to_upgrade: if this_package in installed_packages_list: # check if there's a date installed_date=installed_packages_list[this_package].timestamp if this_package in packages_sources: repo_date=packages_sources[this_package].timestamp if installed_date < repo_date: # ok, there's a newer version logger.info(this_package) packages_really_to_upgrade.append(this_package) else: # no newer version, nothing to update pass else: logger.warning("[up] '{0}' is unknown " "(installed from file?).".format(this_package)) else: # not installed? if not all_packages: logger.warning("[up] '{0}' is not installed, " "cannot upgrade.".format(this_package)) pass # once we have determined which packages # are to be updated, we show them to the # user for him to confirm # if packages_really_to_upgrade!=[]: logger.info("[up] the following package(s) will be upgraded:") for this_package in packages_really_to_upgrade: readable_size = packages_sources[this_package].readable_size logger.info("{0} ({1})".format(this_package, readable_size)) r = input("Proceed? [yes/N] ") if r=='y' or r=='yes': for this_package in packages_really_to_upgrade: install_upgrade( packages_sources[this_package], upgrade=True, progress_hook=hook ) else: logger.info("[up] Taking '{0}' for no, so there.".format(r)) else: # ok, nothing to upgrade, # move along. pass ######################################## # # installs the packages, and forces if # they already exist # # packages must be supplied as argument. # # def install_packages( packages_to_install, force_install=False, hook=None ): """ Installs the packages, possibly forcing installs. :param packages_to_install: list of package names :param force_install: if True, re-installs even if installed. :param hook: download progress hook :raises: IOErrors """ if packages_to_install==[]: raise RuntimeError("[in] fatal: need packages names to install.") if force_install: logger.warning("[in] using the force") packages_really_to_install=[] for this_package in packages_to_install: if this_package in packages_sources: if force_install or not this_package in installed_packages_list: packages_really_to_install.append(this_package) else: logger.warning("[in] package '{0}' " "is already installed".format(this_package)) else: logger.warning("[in] unknown package '{0}'".format(this_package)) if packages_really_to_install!=[]: logger.info("[in] The following package(s) will be installed:") for this_package in packages_really_to_install: readable_size = packages_sources[this_package].readable_size logger.info("{0} ({1})".format(this_package, readable_size)) r = input("Proceed? [yes/N] ") if r=='y' or r=='yes': for this_package in packages_really_to_install: install_upgrade( packages_sources[this_package], upgrade=False, progress_hook=hook ) else: logger.info("[in] Taking '{0}' for no, so there.".format(r)) else: # ok, nothing to upgrade, # move along. pass ######################################## def install_packages_from_file( packages_to_install ): """ (Force)Installs packages from files, but does not update installed.lst files. caveat: not as tested as everything else. :param packages_to_install: list of files to install :raises: IOErrors """ if packages_to_install==[]: raise RuntimeError("[in] fatal: need packages names to install.") packages_really_to_install=[] for this_package in packages_to_install: if os.path.exists(this_package): packages_really_to_install.append(this_package) else: logger.warning("[in] package '{0}' not found".format(this_package)) if packages_really_to_install!=[]: logger.info("[in] The following package(s) will be installed:") packages = [] for this_package in packages_really_to_install: packages.append(corename(this_package)) logger.info(' '.join(packages)) r = input("Proceed? [yes/N] ") if r=='y' or r=='yes': for this_package in packages_really_to_install: #install_upgrade( this_package, upgrade=False, progress_hook=hook ) if os.path.exists(dataset_data_path+corename(this_package)): r = input("[in] '%s' already installed, overwrite? [yes/N] " % corename(this_package)) if r!='y' and r!='yes': logger.info("[in] skipping package " "'{0}'".format(corename(this_package))) continue install_package( corename(this_package), this_package, dataset_data_path) #update_installed_list("i",(make a package object here),dataset_data_path) else: logger.info("[in] Taking '{0}' for no, so there.".format(r)) ######################################## # # uninstall packages, whether or not they # are found in the sources.lst file (to # account for the packages installed from # file) # # like install, it expects a list, if there's # no list, nothing happens. It will test # whether or not the packages are installed, and # will ask the user for a confirmation. # def remove_packages( packages_to_remove ): """ Uninstall packages, whether or not they are found in the source.lst (so it can remove datasets installed from file). :param packages_to_remove: list of package names :raises: IOErrors """ if packages_to_remove==[]: raise RuntimeError("[rm] fatal: need packages names to remove.") packages_really_to_remove=[] for this_package in packages_to_remove: if this_package in packages_sources: #this_data_set_location=os.path.join(dataset_data_path,this_package) # check if in the installed.lst # then if directory actually exists # then if you have rights to remove it if this_package in installed_packages_list: this_data_set_location=os.path.join( installed_packages_list[this_package].where, installed_packages_list[this_package].name ) if os.path.exists(this_data_set_location): if (file_access_rights(this_data_set_location,os.W_OK)): # ok, you may have rights to delete it packages_really_to_remove.append(this_package) else: logger.warning("[rm] insufficient rights " "to remove '{0}'".format(this_package)) else: logger.warning("[rm] package '{0}' found in config file " "but not installed".format(this_package)) else: logger.warning("[rm] package '{0}' " "not installed".format(this_package)) else: logger.warning("[rm] unknown package '{0}'".format(this_package)) if packages_really_to_remove!=[]: logger.info("[rm] the following packages will be removed permanently:") packages = [] for this_package in packages_really_to_remove: packages.append(this_package) logger.info(' '.join(packages)) r = input("Proceed? [yes/N] ") if r=='y' or r=='yes': for this_package in packages_really_to_remove: remove_package( installed_packages_list[this_package], dataset_data_path ) else: logger.info("[up] Taking '{0}' for no, so there.".format(r)) else: # ok, nothing to remove, filenames where bad. pass ######################################## hook_download_filename="" def progress_bar( blocks, blocksize, totalsize ): """ Simple hook to show download progress. caveat: not that great-looking, fix later to a cooler progress bar or something. """ print("\r[dl] %6.2f%% %s" % (min(totalsize,blocks*blocksize)*100.0/totalsize, hook_download_filename), end='') sys.stdout.flush() ######################################## def process_arguments(): """ Processes the installation arguments (from the command line) The possible arguments are: list lists available datasets from sources.lst update updates sources.lst upgrade upgrades datasets that are out of date install <dataset1> <dataset2> ... <datasetn> uses sources.lst to locate the package and perform the installation force-install <dataset1> ... <datasetn> performs an install even if the data sets seem to be there. remove <dataset1> <dataset2> ... <datasetn> removes the dataset clean empties package cache (does nothing for now, because no cache.) """ if len(sys.argv)>1: # due to the relative simplicity of the # arguments, we won't use optparse (2.3-2.6) # nor argparse (2.7+), although in the future # it may pose problems if sys.argv[1]=="list": show_packages() elif sys.argv[1]=="update": atomic_update( os.path.join(dataset_web,dataset_sources), os.path.join(dataset_conf_path,dataset_sources), hook=progress_bar) elif sys.argv[1]=="upgrade": upgrade_packages(sys.argv[2:], hook=progress_bar) elif sys.argv[1]=="install": install_packages(sys.argv[2:], hook=progress_bar) elif sys.argv[1]=="install-from-file": install_packages_from_file(sys.argv[2:]) elif sys.argv[1]=="force-install": install_packages(sys.argv[2:], force_install=True, hook=progress_bar) elif sys.argv[1]=="remove": remove_packages(sys.argv[2:]) elif sys.argv[1]=="clean": # does nothing, no cache implemented # yet. pass elif sys.argv[1]=="version": logger.info(__version__) else: raise RuntimeError("[cl] unknown command '%s'" % sys.argv[1]) else: raise RuntimeError("[cl] missing command") ######################################## if __name__ == "__main__": # to remove RuntimeWarnings about how # tempfilename is unsafe. # warnings.simplefilter("ignore", RuntimeWarning) # OK, let's construct the environment # needed by dataset-get try: set_defaults() except Exception as e: logger.exception(e) exit(1) # fail! try: process_arguments() except Exception as e: logger.exception(e) exit(1)
bsd-3-clause
ba0c3a2ae7a7502fc39fbdb8cf6d19d3
31.528736
114
0.56336
4.251213
false
false
false
false
lisa-lab/pylearn2
pylearn2/datasets/binarizer.py
45
1455
""" .. todo:: WRITEME """ __authors__ = "Ian Goodfellow" __copyright__ = "Copyright 2010-2012, Universite de Montreal" __credits__ = ["Ian Goodfellow"] __license__ = "3-clause BSD" __maintainer__ = "LISA Lab" __email__ = "pylearn-dev@googlegroups" from pylearn2.datasets.transformer_dataset import TransformerDataset from pylearn2.expr.sampling import SampleBernoulli class Binarizer(TransformerDataset): """ A TransformerDataset that takes examples with features in the interval [0,1], and uses these as Bernoulli parameters to sample examples with features in {0,1}. Parameters ---------- raw : pylearn2 Dataset It must provide examples with features in the interval [0,1]. seed : integer or list of integers, optional The seed passed to MRG_RandomStreams to make the Bernoulli samples. If not specified, all class instances default to the same seed so two instances can be run synchronized side by side. """ def __init__(self, raw, seed=None): transformer = SampleBernoulli(seed=seed) super(Binarizer, self).__init__( raw, transformer, space_preserving=True) def get_design_matrix(self, topo=None): """ .. todo:: WRITEME """ if topo is not None: return self.raw.get_design_matrix(topo) X = self.raw.get_design_matrix() return self.transformer.perform(X)
bsd-3-clause
8351d572eacba029adee7f11f86d0fb7
28.1
74
0.644674
4.064246
false
false
false
false
lisa-lab/pylearn2
pylearn2/models/dbm/layer.py
31
127336
""" Common DBM Layer classes """ from __future__ import print_function __authors__ = ["Ian Goodfellow", "Vincent Dumoulin"] __copyright__ = "Copyright 2012-2013, Universite de Montreal" __credits__ = ["Ian Goodfellow"] __license__ = "3-clause BSD" __maintainer__ = "LISA Lab" import functools import logging import numpy as np import operator from theano.compat.six.moves import input, reduce, xrange import time import warnings from theano import tensor as T, function, config import theano from theano.gof.op import get_debug_values from theano.printing import Print from pylearn2.compat import OrderedDict from pylearn2.expr.nnet import sigmoid_numpy from pylearn2.expr.probabilistic_max_pooling import max_pool_channels, max_pool_b01c, max_pool, max_pool_c01b from pylearn2.linear.conv2d import make_random_conv2D, make_sparse_random_conv2D from pylearn2.linear.conv2d_c01b import setup_detector_layer_c01b from pylearn2.linear.matrixmul import MatrixMul from pylearn2.model_extensions.norm_constraint import MaxL2FilterNorm from pylearn2.models import Model from pylearn2.models.dbm import init_sigmoid_bias_from_marginals from pylearn2.space import VectorSpace, CompositeSpace, Conv2DSpace, Space from pylearn2.utils import is_block_gradient from pylearn2.utils import sharedX, safe_zip, py_integer_types, block_gradient from pylearn2.utils.exc import reraise_as from pylearn2.utils.rng import make_theano_rng from pylearn2.utils import safe_union logger = logging.getLogger(__name__) class Layer(Model): """ Abstract class. A layer of a DBM. May only belong to one DBM. Each layer has a state ("total state") that can be split into the piece that is visible to the layer above ("upward state") and the piece that is visible to the layer below ("downward state"). (Since visible layers don't have a downward state, the downward_state method only appears in the DBM_HiddenLayer subclass) For simple layers, all three of these are the same thing. """ def get_dbm(self): """ Returns the DBM that this layer belongs to, or None if it has not been assigned to a DBM yet. """ if hasattr(self, 'dbm'): return self.dbm return None def set_dbm(self, dbm): """ Assigns this layer to a DBM. Parameters ---------- dbm : WRITEME """ assert self.get_dbm() is None self.dbm = dbm def get_total_state_space(self): """ Returns the Space that the layer's total state lives in. """ raise NotImplementedError(str(type(self))+" does not implement " +\ "get_total_state_space()") def get_monitoring_channels(self): """ .. todo:: WRITEME """ return OrderedDict() def get_monitoring_channels_from_state(self, state): """ .. todo:: WRITEME """ return OrderedDict() def upward_state(self, total_state): """ Takes total_state and turns it into the state that layer_above should see when computing P( layer_above | this_layer). So far this has two uses: * If this layer consists of a detector sub-layer h that is pooled into a pooling layer p, then total_state = (p,h) but layer_above should only see p. * If the conditional P( layer_above | this_layer) depends on parameters of this_layer, sometimes you can play games with the state to avoid needing the layers to communicate. So far the only instance of this usage is when the visible layer is N( Wh, beta). This makes the hidden layer be sigmoid( v beta W + b). Rather than having the hidden layer explicitly know about beta, we can just pass v beta as the upward state. Parameters ---------- total_state : WRITEME Notes ----- This method should work both for computing sampling updates and for computing mean field updates. So far I haven't encountered a case where it needs to do different things for those two contexts. """ return total_state def make_state(self, num_examples, numpy_rng): """ Returns a shared variable containing an actual state (not a mean field state) for this variable. Parameters ---------- num_examples : WRITEME numpy_rng : WRITEME Returns ------- WRITEME """ raise NotImplementedError("%s doesn't implement make_state" % type(self)) def make_symbolic_state(self, num_examples, theano_rng): """ Returns a theano symbolic variable containing an actual state (not a mean field state) for this variable. Parameters ---------- num_examples : WRITEME numpy_rng : WRITEME Returns ------- WRITEME """ raise NotImplementedError("%s doesn't implement make_symbolic_state" % type(self)) def sample(self, state_below = None, state_above = None, layer_above = None, theano_rng = None): """ Returns an expression for samples of this layer's state, conditioned on the layers above and below Should be valid as an update to the shared variable returned by self.make_state Parameters ---------- state_below : WRITEME Corresponds to layer_below.upward_state(full_state_below), where full_state_below is the same kind of object as you get out of layer_below.make_state state_above : WRITEME Corresponds to layer_above.downward_state(full_state_above) theano_rng : WRITEME An MRG_RandomStreams instance Returns ------- WRITEME Notes ----- This can return multiple expressions if this layer's total state consists of more than one shared variable. """ if hasattr(self, 'get_sampling_updates'): raise AssertionError("Looks like "+str(type(self))+" needs to rename get_sampling_updates to sample.") raise NotImplementedError("%s doesn't implement sample" % type(self)) def expected_energy_term(self, state, average, state_below, average_below): """ Returns a term of the expected energy of the entire model. This term should correspond to the expected value of terms of the energy function that: - involve this layer only - if there is a layer below, include terms that involve both this layer and the layer below Do not include terms that involve the layer below only. Do not include any terms that involve the layer above, if it exists, in any way (the interface doesn't let you see the layer above anyway). Parameters ---------- state_below : WRITEME Upward state of the layer below. state : WRITEME Total state of this layer average_below : bool If True, the layer below is one of the variables to integrate over in the expectation, and state_below gives its variational parameters. If False, that layer is to be held constant and state_below gives a set of assignments to it average: like average_below, but for 'state' rather than 'state_below' Returns ------- rval : tensor_like A 1D theano tensor giving the expected energy term for each example """ raise NotImplementedError(str(type(self))+" does not implement expected_energy_term.") def finalize_initialization(self): """ Some layers' initialization depends on layer above being initialized, which is why this method is called after `set_input_space` has been called. """ pass class VisibleLayer(Layer): """ Abstract class. A layer of a DBM that may be used as a visible layer. Currently, all implemented layer classes may be either visible or hidden but not both. It may be worth making classes that can play both roles though. This would allow getting rid of the BinaryVector class. """ def get_total_state_space(self): """ Returns the total state of the layer. Returns ------- total_state : member of the input space The total state of the layer. """ return self.get_input_space() class HiddenLayer(Layer): """ Abstract class. A layer of a DBM that may be used as a hidden layer. """ def downward_state(self, total_state): """ .. todo:: WRITEME """ return total_state def get_stdev_rewards(self, state, coeffs): """ .. todo:: WRITEME """ raise NotImplementedError(str(type(self))+" does not implement get_stdev_rewards") def get_range_rewards(self, state, coeffs): """ .. todo:: WRITEME """ raise NotImplementedError(str(type(self))+" does not implement get_range_rewards") def get_l1_act_cost(self, state, target, coeff, eps): """ .. todo:: WRITEME """ raise NotImplementedError(str(type(self))+" does not implement get_l1_act_cost") def get_l2_act_cost(self, state, target, coeff): """ .. todo:: WRITEME """ raise NotImplementedError(str(type(self))+" does not implement get_l2_act_cost") class BinaryVector(VisibleLayer): """ A DBM visible layer consisting of binary random variables living in a VectorSpace. Parameters ---------- nvis : int Dimension of the space bias_from_marginals : pylearn2.datasets.dataset.Dataset Dataset, whose marginals are used to initialize the visible biases center : bool If True, use Gregoire Montavon's centering trick copies : int Use this number of virtual copies of the state. All the copies still share parameters. This can be useful for balancing the amount of influencing two neighboring layers have on each other if the layers have different numbers or different types of units. Without this replication, layers with more units or units with a greater dynamic range would dominate the interaction due to the symmetric nature of undirected interactions. """ def __init__(self, nvis, bias_from_marginals = None, center = False, copies = 1, learn_init_inpainting_state = False): super(BinaryVector, self).__init__() self.__dict__.update(locals()) del self.self # Don't serialize the dataset del self.bias_from_marginals self.space = VectorSpace(nvis) self.input_space = self.space origin = self.space.get_origin() if bias_from_marginals is None: init_bias = np.zeros((nvis,)) else: init_bias = init_sigmoid_bias_from_marginals(bias_from_marginals) self.bias = sharedX(init_bias, 'visible_bias') if center: self.offset = sharedX(sigmoid_numpy(init_bias)) def get_biases(self): """ Returns ------- biases : ndarray The numpy value of the biases """ return self.bias.get_value() def set_biases(self, biases, recenter=False): """ .. todo:: WRITEME """ self.bias.set_value(biases) if recenter: assert self.center self.offset.set_value(sigmoid_numpy(self.bias.get_value())) def upward_state(self, total_state): """ .. todo:: WRITEME """ if not hasattr(self, 'center'): self.center = False if self.center: rval = total_state - self.offset else: rval = total_state if not hasattr(self, 'copies'): self.copies = 1 return rval * self.copies def get_params(self): """ .. todo:: WRITEME """ return [self.bias] def sample(self, state_below = None, state_above = None, layer_above = None, theano_rng = None): """ .. todo:: WRITEME """ assert state_below is None if self.copies != 1: raise NotImplementedError() msg = layer_above.downward_message(state_above) bias = self.bias z = msg + bias phi = T.nnet.sigmoid(z) rval = theano_rng.binomial(size = phi.shape, p = phi, dtype = phi.dtype, n = 1 ) return rval def mf_update(self, state_above, layer_above): """ .. todo:: WRITEME """ msg = layer_above.downward_message(state_above) mu = self.bias z = msg + mu rval = T.nnet.sigmoid(z) return rval def make_state(self, num_examples, numpy_rng): """ .. todo:: WRITEME """ if not hasattr(self, 'copies'): self.copies = 1 if self.copies != 1: raise NotImplementedError() driver = numpy_rng.uniform(0.,1., (num_examples, self.nvis)) mean = sigmoid_numpy(self.bias.get_value()) sample = driver < mean rval = sharedX(sample, name = 'v_sample_shared') return rval def make_symbolic_state(self, num_examples, theano_rng): """ .. todo:: WRITEME """ if not hasattr(self, 'copies'): self.copies = 1 if self.copies != 1: raise NotImplementedError() mean = T.nnet.sigmoid(self.bias) rval = theano_rng.binomial(size=(num_examples, self.nvis), p=mean, dtype=theano.config.floatX) return rval def expected_energy_term(self, state, average, state_below = None, average_below = None): """ .. todo:: WRITEME """ if self.center: state = state - self.offset assert state_below is None assert average_below is None assert average in [True, False] self.space.validate(state) # Energy function is linear so it doesn't matter if we're averaging or not rval = -T.dot(state, self.bias) assert rval.ndim == 1 return rval * self.copies def init_inpainting_state(self, V, drop_mask, noise = False, return_unmasked = False): """ .. todo:: WRITEME """ assert drop_mask is None or drop_mask.ndim > 1 unmasked = T.nnet.sigmoid(self.bias.dimshuffle('x',0)) # this condition is needed later if unmasked is used as V_hat assert unmasked.ndim == 2 # this condition is also needed later if unmasked is used as V_hat assert hasattr(unmasked.owner.op, 'scalar_op') if drop_mask is not None: masked_mean = unmasked * drop_mask else: masked_mean = unmasked if not hasattr(self, 'learn_init_inpainting_state'): self.learn_init_inpainting_state = 0 if not self.learn_init_inpainting_state: masked_mean = block_gradient(masked_mean) masked_mean.name = 'masked_mean' if noise: theano_rng = theano.sandbox.rng_mrg.MRG_RandomStreams(42) # we want a set of random mean field parameters, not binary samples unmasked = T.nnet.sigmoid(theano_rng.normal(avg = 0., std = 1., size = masked_mean.shape, dtype = masked_mean.dtype)) masked_mean = unmasked * drop_mask masked_mean.name = 'masked_noise' if drop_mask is None: rval = masked_mean else: masked_V = V * (1-drop_mask) rval = masked_mean + masked_V rval.name = 'init_inpainting_state' if return_unmasked: assert unmasked.ndim > 1 return rval, unmasked return rval def inpaint_update(self, state_above, layer_above, drop_mask = None, V = None, return_unmasked = False): """ .. todo:: WRITEME """ msg = layer_above.downward_message(state_above) mu = self.bias z = msg + mu z.name = 'inpainting_z_[unknown_iter]' unmasked = T.nnet.sigmoid(z) if drop_mask is not None: rval = drop_mask * unmasked + (1-drop_mask) * V else: rval = unmasked rval.name = 'inpainted_V[unknown_iter]' if return_unmasked: owner = unmasked.owner assert owner is not None op = owner.op assert hasattr(op, 'scalar_op') assert isinstance(op.scalar_op, T.nnet.sigm.ScalarSigmoid) return rval, unmasked return rval def recons_cost(self, V, V_hat_unmasked, drop_mask = None, use_sum=False): """ .. todo:: WRITEME """ if use_sum: raise NotImplementedError() V_hat = V_hat_unmasked assert hasattr(V_hat, 'owner') owner = V_hat.owner assert owner is not None op = owner.op block_grad = False if is_block_gradient(op): assert isinstance(op.scalar_op, theano.scalar.Identity) block_grad = True real, = owner.inputs owner = real.owner op = owner.op if not hasattr(op, 'scalar_op'): raise ValueError("Expected V_hat_unmasked to be generated by an Elemwise op, got "+str(op)+" of type "+str(type(op))) assert isinstance(op.scalar_op, T.nnet.sigm.ScalarSigmoid) z ,= owner.inputs if block_grad: z = block_gradient(z) if V.ndim != V_hat.ndim: raise ValueError("V and V_hat_unmasked should have same ndim, but are %d and %d." % (V.ndim, V_hat.ndim)) unmasked_cost = V * T.nnet.softplus(-z) + (1 - V) * T.nnet.softplus(z) assert unmasked_cost.ndim == V_hat.ndim if drop_mask is None: masked_cost = unmasked_cost else: masked_cost = drop_mask * unmasked_cost return masked_cost.mean() class BinaryVectorMaxPool(HiddenLayer): """ A hidden layer that does max-pooling on binary vectors. It has two sublayers, the detector layer and the pooling layer. The detector layer is its downward state and the pooling layer is its upward state. Parameters ---------- detector_layer_dim : int Number of units in the detector layer pool_size : int Number of detector units per pooling unit (Pools are disjoint) layer_name : str Name of the layer irange : float If specified, initialize the weights in U(-irange, irange) include_prob : , optional Probability of including a weight element in the set of weights initialized to U(-irange, irange). If not included it is initialized to 0. sparse_init : int If specified, initialize this many weights in each column to be nonzero. sparse_stdev : float When using sparse_init, the non-zero weights are drawn from a Gaussian distribution with mean 0 and standard deviation `sparse_stdev` init_bias : float or ndarray Initialize the biases to this value W_lr_scale : float Multiply the learning rate on the weights by this number b_lr_scale : float Multiply the learning rate on the biases by this number center : bool If True, use Gregoire Montavon's centering trick mask_weights : WRITEME max_col_norm : float Constrain the columns of the weight matrix to have at most this norm copies : int See BinaryVector docstring for explanation """ # TODO: this layer uses (pooled, detector) as its total state, # which can be confusing when listing all the states in # the network left to right. Change this and # pylearn2.expr.probabilistic_max_pooling to use # (detector, pooled) def __init__(self, detector_layer_dim, pool_size, layer_name, irange = None, sparse_init = None, sparse_stdev = 1., include_prob = 1.0, init_bias = 0., W_lr_scale = None, b_lr_scale = None, center = False, mask_weights = None, max_col_norm = None, copies = 1): super(BinaryVectorMaxPool, self).__init__() self.__dict__.update(locals()) del self.self self.b = sharedX( np.zeros((self.detector_layer_dim,)) + init_bias, name = layer_name + '_b') if self.center: if self.pool_size != 1: raise NotImplementedError() self.offset = sharedX(sigmoid_numpy(self.b.get_value())) if max_col_norm is not None: self.extensions.append(MaxL2FilterNorm(max_col_norm, axis=0)) def get_lr_scalers(self): """ .. todo:: WRITEME """ if not hasattr(self, 'W_lr_scale'): self.W_lr_scale = None if not hasattr(self, 'b_lr_scale'): self.b_lr_scale = None rval = OrderedDict() if self.W_lr_scale is not None: W, = self.transformer.get_params() rval[W] = self.W_lr_scale if self.b_lr_scale is not None: rval[self.b] = self.b_lr_scale return rval def set_input_space(self, space): """ .. todo:: WRITEME Notes ----- This resets parameters! """ self.input_space = space if isinstance(space, VectorSpace): self.requires_reformat = False self.input_dim = space.dim else: self.requires_reformat = True self.input_dim = space.get_total_dimension() self.desired_space = VectorSpace(self.input_dim) if not (self.detector_layer_dim % self.pool_size == 0): raise ValueError("detector_layer_dim = %d, pool_size = %d. Should be divisible but remainder is %d" % (self.detector_layer_dim, self.pool_size, self.detector_layer_dim % self.pool_size)) self.h_space = VectorSpace(self.detector_layer_dim) self.pool_layer_dim = self.detector_layer_dim / self.pool_size self.output_space = VectorSpace(self.pool_layer_dim) rng = self.dbm.rng if self.irange is not None: assert self.sparse_init is None W = rng.uniform(-self.irange, self.irange, (self.input_dim, self.detector_layer_dim)) * \ (rng.uniform(0.,1., (self.input_dim, self.detector_layer_dim)) < self.include_prob) else: assert self.sparse_init is not None W = np.zeros((self.input_dim, self.detector_layer_dim)) def mask_rejects(idx, i): if self.mask_weights is None: return False return self.mask_weights[idx, i] == 0. for i in xrange(self.detector_layer_dim): assert self.sparse_init <= self.input_dim for j in xrange(self.sparse_init): idx = rng.randint(0, self.input_dim) while W[idx, i] != 0 or mask_rejects(idx, i): idx = rng.randint(0, self.input_dim) W[idx, i] = rng.randn() W *= self.sparse_stdev W = sharedX(W) W.name = self.layer_name + '_W' self.transformer = MatrixMul(W) W ,= self.transformer.get_params() assert W.name is not None if self.mask_weights is not None: expected_shape = (self.input_dim, self.detector_layer_dim) if expected_shape != self.mask_weights.shape: raise ValueError("Expected mask with shape "+str(expected_shape)+" but got "+str(self.mask_weights.shape)) self.mask = sharedX(self.mask_weights) @functools.wraps(Model._modify_updates) def _modify_updates(self, updates): # Patch old pickle files if not hasattr(self, 'mask_weights'): self.mask_weights = None if not hasattr(self, 'max_col_norm'): self.max_col_norm = None if self.mask_weights is not None: W ,= self.transformer.get_params() if W in updates: updates[W] = updates[W] * self.mask def get_total_state_space(self): """ .. todo:: WRITEME """ return CompositeSpace((self.output_space, self.h_space)) def get_params(self): """ .. todo:: WRITEME """ assert self.b.name is not None W ,= self.transformer.get_params() assert W.name is not None rval = self.transformer.get_params() assert not isinstance(rval, set) rval = list(rval) assert self.b not in rval rval.append(self.b) return rval def get_weight_decay(self, coeff): """ .. todo:: WRITEME """ if isinstance(coeff, str): coeff = float(coeff) assert isinstance(coeff, float) or hasattr(coeff, 'dtype') W ,= self.transformer.get_params() return coeff * T.sqr(W).sum() def get_weights(self): """ .. todo:: WRITEME """ if self.requires_reformat: # This is not really an unimplemented case. # We actually don't know how to format the weights # in design space. We got the data in topo space # and we don't have access to the dataset raise NotImplementedError() W ,= self.transformer.get_params() return W.get_value() def set_weights(self, weights): """ .. todo:: WRITEME """ W, = self.transformer.get_params() W.set_value(weights) def set_biases(self, biases, recenter = False): """ .. todo:: WRITEME """ self.b.set_value(biases) if recenter: assert self.center if self.pool_size != 1: raise NotImplementedError() self.offset.set_value(sigmoid_numpy(self.b.get_value())) def get_biases(self): """ .. todo:: WRITEME """ return self.b.get_value() def get_weights_format(self): """ .. todo:: WRITEME """ return ('v', 'h') def get_weights_view_shape(self): """ .. todo:: WRITEME """ total = self.detector_layer_dim cols = self.pool_size if cols == 1: # Let the PatchViewer decidew how to arrange the units # when they're not pooled raise NotImplementedError() # When they are pooled, make each pooling unit have one row rows = total / cols return rows, cols def get_weights_topo(self): """ .. todo:: WRITEME """ if not isinstance(self.input_space, Conv2DSpace): raise NotImplementedError() W ,= self.transformer.get_params() W = W.T W = W.reshape((self.detector_layer_dim, self.input_space.shape[0], self.input_space.shape[1], self.input_space.num_channels)) W = Conv2DSpace.convert(W, self.input_space.axes, ('b', 0, 1, 'c')) return function([], W)() def upward_state(self, total_state): """ .. todo:: WRITEME """ p,h = total_state self.h_space.validate(h) self.output_space.validate(p) if not hasattr(self, 'center'): self.center = False if self.center: return p - self.offset if not hasattr(self, 'copies'): self.copies = 1 return p * self.copies def downward_state(self, total_state): """ .. todo:: WRITEME """ p,h = total_state if not hasattr(self, 'center'): self.center = False if self.center: return h - self.offset return h * self.copies def get_monitoring_channels(self): """ .. todo:: WRITEME """ W ,= self.transformer.get_params() assert W.ndim == 2 sq_W = T.sqr(W) row_norms = T.sqrt(sq_W.sum(axis=1)) col_norms = T.sqrt(sq_W.sum(axis=0)) return OrderedDict([ ('row_norms_min' , row_norms.min()), ('row_norms_mean' , row_norms.mean()), ('row_norms_max' , row_norms.max()), ('col_norms_min' , col_norms.min()), ('col_norms_mean' , col_norms.mean()), ('col_norms_max' , col_norms.max()), ]) def get_monitoring_channels_from_state(self, state): """ .. todo:: WRITEME """ P, H = state rval = OrderedDict() if self.pool_size == 1: vars_and_prefixes = [ (P,'') ] else: vars_and_prefixes = [ (P, 'p_'), (H, 'h_') ] for var, prefix in vars_and_prefixes: v_max = var.max(axis=0) v_min = var.min(axis=0) v_mean = var.mean(axis=0) v_range = v_max - v_min # max_x.mean_u is "the mean over *u*nits of the max over e*x*amples" # The x and u are included in the name because otherwise its hard # to remember which axis is which when reading the monitor # I use inner.outer rather than outer_of_inner or something like that # because I want mean_x.* to appear next to each other in the alphabetical # list, as these are commonly plotted together for key, val in [ ('max_x.max_u', v_max.max()), ('max_x.mean_u', v_max.mean()), ('max_x.min_u', v_max.min()), ('min_x.max_u', v_min.max()), ('min_x.mean_u', v_min.mean()), ('min_x.min_u', v_min.min()), ('range_x.max_u', v_range.max()), ('range_x.mean_u', v_range.mean()), ('range_x.min_u', v_range.min()), ('mean_x.max_u', v_mean.max()), ('mean_x.mean_u', v_mean.mean()), ('mean_x.min_u', v_mean.min()) ]: rval[prefix+key] = val return rval def get_stdev_rewards(self, state, coeffs): """ .. todo:: WRITEME """ rval = 0. P, H = state self.output_space.validate(P) self.h_space.validate(H) if self.pool_size == 1: # If the pool size is 1 then pools = detectors # and we should not penalize pools and detectors separately assert len(state) == 2 if isinstance(coeffs, str): coeffs = float(coeffs) assert isinstance(coeffs, float) _, state = state state = [state] coeffs = [coeffs] else: assert all([len(elem) == 2 for elem in [state, coeffs]]) for s, c in safe_zip(state, coeffs): assert all([isinstance(elem, float) for elem in [c]]) if c == 0.: continue mn = s.mean(axis=0) dev = s - mn stdev = T.sqrt(T.sqr(dev).mean(axis=0)) rval += (0.5 - stdev).mean()*c return rval def get_range_rewards(self, state, coeffs): """ .. todo:: WRITEME """ rval = 0. P, H = state self.output_space.validate(P) self.h_space.validate(H) if self.pool_size == 1: # If the pool size is 1 then pools = detectors # and we should not penalize pools and detectors separately assert len(state) == 2 if isinstance(coeffs, str): coeffs = float(coeffs) assert isinstance(coeffs, float) _, state = state state = [state] coeffs = [coeffs] else: assert all([len(elem) == 2 for elem in [state, coeffs]]) for s, c in safe_zip(state, coeffs): assert all([isinstance(elem, float) for elem in [c]]) if c == 0.: continue mx = s.max(axis=0) assert hasattr(mx.owner.op, 'grad') assert mx.ndim == 1 mn = s.min(axis=0) assert hasattr(mn.owner.op, 'grad') assert mn.ndim == 1 r = mx - mn rval += (1 - r).mean()*c return rval def get_l1_act_cost(self, state, target, coeff, eps = None): """ .. todo:: WRITEME """ rval = 0. P, H = state self.output_space.validate(P) self.h_space.validate(H) if self.pool_size == 1: # If the pool size is 1 then pools = detectors # and we should not penalize pools and detectors separately assert len(state) == 2 if not isinstance(target, float): raise TypeError("BinaryVectorMaxPool.get_l1_act_cost expected target of type float " + \ " but an instance named "+self.layer_name + " got target "+str(target) + " of type "+str(type(target))) assert isinstance(coeff, float) or hasattr(coeff, 'dtype') _, state = state state = [state] target = [target] coeff = [coeff] if eps is None: eps = [0.] else: eps = [eps] else: assert all([len(elem) == 2 for elem in [state, target, coeff]]) if eps is None: eps = [0., 0.] if target[1] > target[0]: warnings.warn("Do you really want to regularize the detector units to be more active than the pooling units?") for s, t, c, e in safe_zip(state, target, coeff, eps): assert all([isinstance(elem, float) or hasattr(elem, 'dtype') for elem in [t, c, e]]) if c == 0.: continue m = s.mean(axis=0) assert m.ndim == 1 rval += T.maximum(abs(m-t)-e,0.).mean()*c return rval def get_l2_act_cost(self, state, target, coeff): """ .. todo:: WRITEME """ rval = 0. P, H = state self.output_space.validate(P) self.h_space.validate(H) if self.pool_size == 1: # If the pool size is 1 then pools = detectors # and we should not penalize pools and detectors separately assert len(state) == 2 if not isinstance(target, float): raise TypeError("BinaryVectorMaxPool.get_l1_act_cost expected target of type float " + \ " but an instance named "+self.layer_name + " got target "+str(target) + " of type "+str(type(target))) assert isinstance(coeff, float) or hasattr(coeff, 'dtype') _, state = state state = [state] target = [target] coeff = [coeff] else: assert all([len(elem) == 2 for elem in [state, target, coeff]]) if target[1] > target[0]: warnings.warn("Do you really want to regularize the detector units to be more active than the pooling units?") for s, t, c in safe_zip(state, target, coeff): assert all([isinstance(elem, float) or hasattr(elem, 'dtype') for elem in [t, c]]) if c == 0.: continue m = s.mean(axis=0) assert m.ndim == 1 rval += T.square(m-t).mean()*c return rval def sample(self, state_below = None, state_above = None, layer_above = None, theano_rng = None): """ .. todo:: WRITEME """ if self.copies != 1: raise NotImplementedError() if theano_rng is None: raise ValueError("theano_rng is required; it just defaults to None so that it may appear after layer_above / state_above in the list.") if state_above is not None: msg = layer_above.downward_message(state_above) else: msg = None if self.requires_reformat: state_below = self.input_space.format_as(state_below, self.desired_space) z = self.transformer.lmul(state_below) + self.b p, h, p_sample, h_sample = max_pool_channels(z, self.pool_size, msg, theano_rng) return p_sample, h_sample def downward_message(self, downward_state): """ .. todo:: WRITEME """ self.h_space.validate(downward_state) rval = self.transformer.lmul_T(downward_state) if self.requires_reformat: rval = self.desired_space.format_as(rval, self.input_space) return rval * self.copies def init_mf_state(self): """ .. todo:: WRITEME """ # work around theano bug with broadcasted vectors z = T.alloc(0., self.dbm.batch_size, self.detector_layer_dim).astype(self.b.dtype) + \ self.b.dimshuffle('x', 0) rval = max_pool_channels(z = z, pool_size = self.pool_size) return rval def make_state(self, num_examples, numpy_rng): """ .. todo:: WRITEME """ """ Returns a shared variable containing an actual state (not a mean field state) for this variable. """ if not hasattr(self, 'copies'): self.copies = 1 if self.copies != 1: raise NotImplementedError() empty_input = self.h_space.get_origin_batch(num_examples) empty_output = self.output_space.get_origin_batch(num_examples) h_state = sharedX(empty_input) p_state = sharedX(empty_output) theano_rng = make_theano_rng(None, numpy_rng.randint(2 ** 16), which_method="binomial") default_z = T.zeros_like(h_state) + self.b p_exp, h_exp, p_sample, h_sample = max_pool_channels( z = default_z, pool_size = self.pool_size, theano_rng = theano_rng) assert h_sample.dtype == default_z.dtype f = function([], updates = [ (p_state , p_sample), (h_state , h_sample) ]) f() p_state.name = 'p_sample_shared' h_state.name = 'h_sample_shared' return p_state, h_state def make_symbolic_state(self, num_examples, theano_rng): """ .. todo:: WRITEME """ """ Returns a theano symbolic variable containing an actual state (not a mean field state) for this variable. """ if not hasattr(self, 'copies'): self.copies = 1 if self.copies != 1: raise NotImplementedError() default_z = T.alloc(self.b, num_examples, self.detector_layer_dim) p_exp, h_exp, p_sample, h_sample = max_pool_channels(z=default_z, pool_size=self.pool_size, theano_rng=theano_rng) assert h_sample.dtype == default_z.dtype return p_sample, h_sample def expected_energy_term(self, state, average, state_below, average_below): """ .. todo:: WRITEME """ # Don't need to do anything special for centering, upward_state / downward state # make it all just work self.input_space.validate(state_below) if self.requires_reformat: if not isinstance(state_below, tuple): for sb in get_debug_values(state_below): if sb.shape[0] != self.dbm.batch_size: raise ValueError("self.dbm.batch_size is %d but got shape of %d" % (self.dbm.batch_size, sb.shape[0])) assert reduce(operator.mul, sb.shape[1:]) == self.input_dim state_below = self.input_space.format_as(state_below, self.desired_space) downward_state = self.downward_state(state) self.h_space.validate(downward_state) # Energy function is linear so it doesn't matter if we're averaging or not # Specifically, our terms are -u^T W d - b^T d where u is the upward state of layer below # and d is the downward state of this layer bias_term = T.dot(downward_state, self.b) weights_term = (self.transformer.lmul(state_below) * downward_state).sum(axis=1) rval = -bias_term - weights_term assert rval.ndim == 1 return rval * self.copies def linear_feed_forward_approximation(self, state_below): """ Used to implement TorontoSparsity. Unclear exactly what properties of it are important or how to implement it for other layers. Properties it must have: output is same kind of data structure (ie, tuple of theano 2-tensors) as mf_update. Properties it probably should have for other layer types: an infinitesimal change in state_below or the parameters should cause the same sign of change in the output of linear_feed_forward_approximation and in mf_update Should not have any non-linearities that cause the gradient to shrink Should disregard top-down feedback Parameters ---------- state_below : WRITEME """ z = self.transformer.lmul(state_below) + self.b if self.pool_size != 1: # Should probably implement sum pooling for the non-pooled version, # but in reality it's not totally clear what the right answer is raise NotImplementedError() return z, z def mf_update(self, state_below, state_above, layer_above = None, double_weights = False, iter_name = None): """ .. todo:: WRITEME """ self.input_space.validate(state_below) if self.requires_reformat: if not isinstance(state_below, tuple): for sb in get_debug_values(state_below): if sb.shape[0] != self.dbm.batch_size: raise ValueError("self.dbm.batch_size is %d but got shape of %d" % (self.dbm.batch_size, sb.shape[0])) assert reduce(operator.mul, sb.shape[1:]) == self.input_dim state_below = self.input_space.format_as(state_below, self.desired_space) if iter_name is None: iter_name = 'anon' if state_above is not None: assert layer_above is not None msg = layer_above.downward_message(state_above) msg.name = 'msg_from_'+layer_above.layer_name+'_to_'+self.layer_name+'['+iter_name+']' else: msg = None if double_weights: state_below = 2. * state_below state_below.name = self.layer_name + '_'+iter_name + '_2state' z = self.transformer.lmul(state_below) + self.b if self.layer_name is not None and iter_name is not None: z.name = self.layer_name + '_' + iter_name + '_z' p,h = max_pool_channels(z, self.pool_size, msg) p.name = self.layer_name + '_p_' + iter_name h.name = self.layer_name + '_h_' + iter_name return p, h class Softmax(HiddenLayer): """ A layer representing a single softmax distribution of a set of discrete categories. Parameters ---------- n_classes : int The number of discrete categories. layer_name : str The name of the layer. irange : float If not None, initialze the weights in U(-irange, irange) sparse_init : int If not None, initialize `sparse_init` weights per column to N(0, sparse_istdev^2) sparse_istdev : float see above W_lr_scale : float Scale the learning rate on the weights by this amount b_lr_scale : float Scale the learning rate on the biases by this amount max_col_norm : float If not None, constrain the columns of the weight matrix to have at most this L2 norm copies : int Make this many copies of the random variables, all sharing the same weights. This allows the undirected model to behave as if it has asymmetric connections. center : bool If True, use Gregoire Montavon's centering trick. learn_init_inpainting_state : bool If True, and using inpainting-based methods (MP-DBM), learn a parameter controlling the initial value of the mean field state for this layer. """ presynaptic_name = "presynaptic_Y_hat" def __init__(self, n_classes, layer_name, irange = None, sparse_init = None, sparse_istdev = 1., W_lr_scale = None, b_lr_scale = None, max_col_norm = None, copies = 1, center = False, learn_init_inpainting_state = True): super(Softmax, self).__init__() if isinstance(W_lr_scale, str): W_lr_scale = float(W_lr_scale) self.__dict__.update(locals()) del self.self assert isinstance(n_classes, py_integer_types) self.output_space = VectorSpace(n_classes) self.b = sharedX( np.zeros((n_classes,)), name = 'softmax_b') if self.center: b = self.b.get_value() self.offset = sharedX(np.exp(b) / np.exp(b).sum()) if max_col_norm is not None: self.extensions.append(MaxL2FilterNorm(max_col_norm, axis=0)) @functools.wraps(Model.get_lr_scalers) def get_lr_scalers(self): rval = OrderedDict() # Patch old pickle files if not hasattr(self, 'W_lr_scale'): self.W_lr_scale = None if self.W_lr_scale is not None: assert isinstance(self.W_lr_scale, float) rval[self.W] = self.W_lr_scale if not hasattr(self, 'b_lr_scale'): self.b_lr_scale = None if self.b_lr_scale is not None: assert isinstance(self.b_lr_scale, float) rval[self.b] = self.b_lr_scale return rval def get_total_state_space(self): """ .. todo:: WRITEME """ return self.output_space def get_monitoring_channels_from_state(self, state): """ .. todo:: WRITEME """ mx = state.max(axis=1) return OrderedDict([ ('mean_max_class' , mx.mean()), ('max_max_class' , mx.max()), ('min_max_class' , mx.min()) ]) def set_input_space(self, space): """ .. todo:: WRITEME """ self.input_space = space if not isinstance(space, Space): raise TypeError("Expected Space, got "+ str(space)+" of type "+str(type(space))) self.input_dim = space.get_total_dimension() self.needs_reformat = not isinstance(space, VectorSpace) self.desired_space = VectorSpace(self.input_dim) if not self.needs_reformat: assert self.desired_space == self.input_space rng = self.dbm.rng if self.irange is not None: assert self.sparse_init is None W = rng.uniform(-self.irange,self.irange, (self.input_dim,self.n_classes)) else: assert self.sparse_init is not None W = np.zeros((self.input_dim, self.n_classes)) for i in xrange(self.n_classes): for j in xrange(self.sparse_init): idx = rng.randint(0, self.input_dim) while W[idx, i] != 0.: idx = rng.randint(0, self.input_dim) W[idx, i] = rng.randn() * self.sparse_istdev self.W = sharedX(W, 'softmax_W' ) self._params = [ self.b, self.W ] def get_weights_topo(self): """ .. todo:: WRITEME """ if not isinstance(self.input_space, Conv2DSpace): raise NotImplementedError() desired = self.W.get_value().T ipt = self.desired_space.format_as(desired, self.input_space) rval = Conv2DSpace.convert_numpy(ipt, self.input_space.axes, ('b', 0, 1, 'c')) return rval def get_weights(self): """ .. todo:: WRITEME """ if not isinstance(self.input_space, VectorSpace): raise NotImplementedError() return self.W.get_value() def set_weights(self, weights): """ .. todo:: WRITEME """ self.W.set_value(weights) def set_biases(self, biases, recenter=False): """ .. todo:: WRITEME """ self.b.set_value(biases) if recenter: assert self.center self.offset.set_value( (np.exp(biases) / np.exp(biases).sum()).astype(self.offset.dtype)) def get_biases(self): """ .. todo:: WRITEME """ return self.b.get_value() def get_weights_format(self): """ .. todo:: WRITEME """ return ('v', 'h') def sample(self, state_below = None, state_above = None, layer_above = None, theano_rng = None): """ .. todo:: WRITEME """ if self.copies != 1: raise NotImplementedError("need to draw self.copies samples and average them together.") if state_above is not None: # If you implement this case, also add a unit test for it. # Or at least add a warning that it is not tested. raise NotImplementedError() if theano_rng is None: raise ValueError("theano_rng is required; it just defaults to None so that it may appear after layer_above / state_above in the list.") self.input_space.validate(state_below) # patch old pickle files if not hasattr(self, 'needs_reformat'): self.needs_reformat = self.needs_reshape del self.needs_reshape if self.needs_reformat: state_below = self.input_space.format_as(state_below, self.desired_space) self.desired_space.validate(state_below) z = T.dot(state_below, self.W) + self.b h_exp = T.nnet.softmax(z) h_sample = theano_rng.multinomial(pvals = h_exp, dtype = h_exp.dtype) return h_sample def mf_update(self, state_below, state_above = None, layer_above = None, double_weights = False, iter_name = None): """ .. todo:: WRITEME """ if state_above is not None: raise NotImplementedError() if double_weights: raise NotImplementedError() self.input_space.validate(state_below) # patch old pickle files if not hasattr(self, 'needs_reformat'): self.needs_reformat = self.needs_reshape del self.needs_reshape if self.needs_reformat: state_below = self.input_space.format_as(state_below, self.desired_space) for value in get_debug_values(state_below): if value.shape[0] != self.dbm.batch_size: raise ValueError("state_below should have batch size "+str(self.dbm.batch_size)+" but has "+str(value.shape[0])) self.desired_space.validate(state_below) assert self.W.ndim == 2 assert state_below.ndim == 2 b = self.b Z = T.dot(state_below, self.W) + b rval = T.nnet.softmax(Z) for value in get_debug_values(rval): assert value.shape[0] == self.dbm.batch_size return rval def downward_message(self, downward_state): """ .. todo:: WRITEME """ if not hasattr(self, 'copies'): self.copies = 1 rval = T.dot(downward_state, self.W.T) * self.copies rval = self.desired_space.format_as(rval, self.input_space) return rval def recons_cost(self, Y, Y_hat_unmasked, drop_mask_Y, scale): """ The cost of reconstructing `Y` as `Y_hat`. Specifically, the negative log probability. This cost is for use with multi-prediction training. Parameters ---------- Y : target space batch The data labels Y_hat_unmasked : target space batch The output of this layer's `mf_update`; the predicted values of `Y`. Even though the model is only predicting the dropped values, we take predictions for all the values here. drop_mask_Y : 1-D theano tensor A batch of 0s/1s, with 1s indicating that variables have been dropped, and should be included in the reconstruction cost. One indicator per example in the batch, since each example in this layer only has one random variable in it. scale : float Multiply the cost by this amount. We need to do this because the visible layer also goes into the cost. We use the mean over units and examples, so that the scale of the cost doesn't change too much with batch size or example size. We need to multiply this cost by scale to make sure that it is put on the same scale as the reconstruction cost for the visible units. ie, scale should be 1/nvis """ Y_hat = Y_hat_unmasked assert hasattr(Y_hat, 'owner') owner = Y_hat.owner assert owner is not None op = owner.op if isinstance(op, Print): assert len(owner.inputs) == 1 Y_hat, = owner.inputs owner = Y_hat.owner op = owner.op assert isinstance(op, T.nnet.Softmax) z ,= owner.inputs assert z.ndim == 2 z = z - z.max(axis=1).dimshuffle(0, 'x') log_prob = z - T.log(T.exp(z).sum(axis=1).dimshuffle(0, 'x')) # we use sum and not mean because this is really one variable per row log_prob_of = (Y * log_prob).sum(axis=1) masked = log_prob_of * drop_mask_Y assert masked.ndim == 1 rval = masked.mean() * scale * self.copies return - rval def init_mf_state(self): """ .. todo:: WRITEME """ rval = T.nnet.softmax(self.b.dimshuffle('x', 0)) + T.alloc(0., self.dbm.batch_size, self.n_classes).astype(config.floatX) return rval def make_state(self, num_examples, numpy_rng): """ .. todo:: WRITEME """ """ Returns a shared variable containing an actual state (not a mean field state) for this variable. """ if self.copies != 1: raise NotImplementedError("need to make self.copies samples and average them together.") t1 = time.time() empty_input = self.output_space.get_origin_batch(num_examples) h_state = sharedX(empty_input) default_z = T.zeros_like(h_state) + self.b theano_rng = make_theano_rng(None, numpy_rng.randint(2 ** 16), which_method="binomial") h_exp = T.nnet.softmax(default_z) h_sample = theano_rng.multinomial(pvals = h_exp, dtype = h_exp.dtype) h_state = sharedX( self.output_space.get_origin_batch( num_examples)) t2 = time.time() f = function([], updates = [( h_state , h_sample )]) t3 = time.time() f() t4 = time.time() logger.info('{0}.make_state took {1}'.format(self, t4-t1)) logger.info('\tcompose time: {0}'.format(t2-t1)) logger.info('\tcompile time: {0}'.format(t3-t2)) logger.info('\texecute time: {0}'.format(t4-t3)) h_state.name = 'softmax_sample_shared' return h_state def make_symbolic_state(self, num_examples, theano_rng): """ .. todo:: WRITEME """ """ Returns a symbolic variable containing an actual state (not a mean field state) for this variable. """ if self.copies != 1: raise NotImplementedError("need to make self.copies samples and average them together.") default_z = T.alloc(self.b, num_examples, self.n_classes) h_exp = T.nnet.softmax(default_z) h_sample = theano_rng.multinomial(pvals=h_exp, dtype=h_exp.dtype) return h_sample def get_weight_decay(self, coeff): """ .. todo:: WRITEME """ if isinstance(coeff, str): coeff = float(coeff) assert isinstance(coeff, float) or hasattr(coeff, 'dtype') return coeff * T.sqr(self.W).sum() def upward_state(self, state): """ .. todo:: WRITEME """ if self.center: return state - self.offset return state def downward_state(self, state): """ .. todo:: WRITEME """ if not hasattr(self, 'center'): self.center = False if self.center: """TODO: write a unit test verifying that inference or sampling below a centered Softmax layer works""" return state - self.offset return state def expected_energy_term(self, state, average, state_below, average_below): """ .. todo:: WRITEME """ if self.center: state = state - self.offset self.input_space.validate(state_below) if self.needs_reformat: state_below = self.input_space.format_as(state_below, self.desired_space) self.desired_space.validate(state_below) # Energy function is linear so it doesn't matter if we're averaging or not # Specifically, our terms are -u^T W d - b^T d where u is the upward state of layer below # and d is the downward state of this layer bias_term = T.dot(state, self.b) weights_term = (T.dot(state_below, self.W) * state).sum(axis=1) rval = -bias_term - weights_term rval *= self.copies assert rval.ndim == 1 return rval def init_inpainting_state(self, Y, noise): """ .. todo:: WRITEME """ if noise: theano_rng = make_theano_rng(None, 2012+10+30, which_method="binomial") return T.nnet.softmax(theano_rng.normal(avg=0., size=Y.shape, std=1., dtype='float32')) rval = T.nnet.softmax(self.b) if not hasattr(self, 'learn_init_inpainting_state'): self.learn_init_inpainting_state = 1 if not self.learn_init_inpainting_state: rval = block_gradient(rval) return rval def install_presynaptic_outputs(self, outputs_dict, batch_size): """ .. todo:: WRITEME """ assert self.presynaptic_name not in outputs_dict outputs_dict[self.presynaptic_name] = self.output_space.make_shared_batch(batch_size, self.presynaptic_name) class GaussianVisLayer(VisibleLayer): """ Implements a visible layer that is conditionally gaussian with diagonal variance. The layer lives in a Conv2DSpace. Parameters ---------- rows, cols, channels : WRITEME the shape of the space learn_init_inpainting : bool, optional WRITEME nvis : WRITEME init_beta : WRITEME the initial value of the precision parameter min_beta : WRITEME clip beta so it is at least this big (default 1) init_mu : WRITEME the initial value of the mean parameter tie_beta : WRITEME None or a string specifying how to tie beta 'locations' = tie beta across locations, ie beta should be a vector with one elem per channel tie_mu : WRITEME None or a string specifying how to tie mu 'locations' = tie mu across locations, ie mu should be a vector with one elem per channel bias_from_marginals : WRITEME beta_lr_scale : WRITEME axes : tuple WRITEME """ def __init__(self, rows = None, cols = None, learn_init_inpainting_state=True, channels = None, nvis = None, init_beta = 1., min_beta = 1., init_mu = None, tie_beta = None, tie_mu = None, bias_from_marginals = None, beta_lr_scale = 'by_sharing', axes = ('b', 0, 1, 'c')): super(type(self), self).__init__() warnings.warn("GaussianVisLayer math very faith based, need to finish working through gaussian.lyx") self.__dict__.update(locals()) del self.self if bias_from_marginals is not None: del self.bias_from_marginals if self.nvis is None: raise NotImplementedError() assert init_mu is None init_mu = bias_from_marginals.X.mean(axis=0) if init_mu is None: init_mu = 0. if nvis is None: assert rows is not None assert cols is not None assert channels is not None self.space = Conv2DSpace(shape=[rows,cols], num_channels=channels, axes=axes) # To make GaussianVisLayer compatible with any axis ordering self.batch_axis=list(axes).index('b') self.axes_to_sum = list(range(len(axes))) self.axes_to_sum.remove(self.batch_axis) else: assert rows is None assert cols is None assert channels is None self.space = VectorSpace(nvis) self.axes_to_sum = 1 self.batch_axis = None self.input_space = self.space origin = self.space.get_origin() beta_origin = origin.copy() assert tie_beta in [ None, 'locations'] if tie_beta == 'locations': assert nvis is None beta_origin = np.zeros((self.space.num_channels,)) self.beta = sharedX(beta_origin + init_beta,name = 'beta') assert self.beta.ndim == beta_origin.ndim mu_origin = origin.copy() assert tie_mu in [None, 'locations'] if tie_mu == 'locations': assert nvis is None mu_origin = np.zeros((self.space.num_channels,)) self.mu = sharedX( mu_origin + init_mu, name = 'mu') assert self.mu.ndim == mu_origin.ndim def get_monitoring_channels(self): """ .. todo:: WRITEME """ rval = OrderedDict() rval['beta_min'] = self.beta.min() rval['beta_mean'] = self.beta.mean() rval['beta_max'] = self.beta.max() return rval def get_params(self): """ .. todo:: WRITEME """ if self.mu is None: return [self.beta] return [self.beta, self.mu] def get_lr_scalers(self): """ .. todo:: WRITEME """ rval = OrderedDict() if self.nvis is None: rows, cols = self.space.shape num_loc = float(rows * cols) assert self.tie_beta in [None, 'locations'] if self.beta_lr_scale == 'by_sharing': if self.tie_beta == 'locations': assert self.nvis is None rval[self.beta] = 1. / num_loc elif self.beta_lr_scale == None: pass else: rval[self.beta] = self.beta_lr_scale assert self.tie_mu in [None, 'locations'] if self.tie_mu == 'locations': warn = True assert self.nvis is None rval[self.mu] = 1./num_loc logger.warning("mu lr_scaler hardcoded to 1/sharing") return rval @functools.wraps(Model._modify_updates) def _modify_updates(self, updates): if self.beta in updates: updated_beta = updates[self.beta] updates[self.beta] = T.clip(updated_beta, self.min_beta,1e6) def set_biases(self, bias): """ Set mean parameter Parameters ---------- bias: WRITEME Vector of size nvis """ self.mu = sharedX(bias, name = 'mu') def broadcasted_mu(self): """ Returns mu, broadcasted to have the same shape as a batch of data """ if self.tie_mu == 'locations': def f(x): if x == 'c': return 0 return 'x' axes = [f(ax) for ax in self.axes] rval = self.mu.dimshuffle(*axes) else: assert self.tie_mu is None if self.nvis is None: axes = [0, 1, 2] axes.insert(self.axes.index('b'), 'x') rval = self.mu.dimshuffle(*axes) else: rval = self.mu.dimshuffle('x', 0) self.input_space.validate(rval) return rval def broadcasted_beta(self): """ Returns beta, broadcasted to have the same shape as a batch of data """ return self.broadcast_beta(self.beta) def broadcast_beta(self, beta): """ .. todo:: WRITEME """ """ Returns beta, broadcasted to have the same shape as a batch of data """ if self.tie_beta == 'locations': def f(x): if x == 'c': return 0 return 'x' axes = [f(ax) for ax in self.axes] rval = beta.dimshuffle(*axes) else: assert self.tie_beta is None if self.nvis is None: axes = [0, 1, 2] axes.insert(self.axes.index('b'), 'x') rval = beta.dimshuffle(*axes) else: rval = beta.dimshuffle('x', 0) self.input_space.validate(rval) return rval def init_inpainting_state(self, V, drop_mask, noise = False, return_unmasked = False): """ .. todo:: WRITEME """ """for Vv, drop_mask_v in get_debug_values(V, drop_mask): assert Vv.ndim == 4 assert drop_mask_v.ndim in [3,4] for i in xrange(drop_mask.ndim): if Vv.shape[i] != drop_mask_v.shape[i]: print(Vv.shape) print(drop_mask_v.shape) assert False """ unmasked = self.broadcasted_mu() if drop_mask is None: assert not noise assert not return_unmasked return unmasked masked_mu = unmasked * drop_mask if not hasattr(self, 'learn_init_inpainting_state'): self.learn_init_inpainting_state = True if not self.learn_init_inpainting_state: masked_mu = block_gradient(masked_mu) masked_mu.name = 'masked_mu' if noise: theano_rng = make_theano_rng(None, 42, which_method="binomial") unmasked = theano_rng.normal(avg = 0., std = 1., size = masked_mu.shape, dtype = masked_mu.dtype) masked_mu = unmasked * drop_mask masked_mu.name = 'masked_noise' masked_V = V * (1-drop_mask) rval = masked_mu + masked_V rval.name = 'init_inpainting_state' if return_unmasked: return rval, unmasked return rval def expected_energy_term(self, state, average, state_below = None, average_below = None): """ .. todo:: WRITEME """ assert state_below is None assert average_below is None self.space.validate(state) if average: raise NotImplementedError(str(type(self))+" doesn't support integrating out variational parameters yet.") else: rval = 0.5 * (self.beta * T.sqr(state - self.mu)).sum(axis=self.axes_to_sum) assert rval.ndim == 1 return rval def inpaint_update(self, state_above, layer_above, drop_mask = None, V = None, return_unmasked = False): """ .. todo:: WRITEME """ msg = layer_above.downward_message(state_above) mu = self.broadcasted_mu() z = msg + mu z.name = 'inpainting_z_[unknown_iter]' if drop_mask is not None: rval = drop_mask * z + (1-drop_mask) * V else: rval = z rval.name = 'inpainted_V[unknown_iter]' if return_unmasked: return rval, z return rval def sample(self, state_below = None, state_above = None, layer_above = None, theano_rng = None): """ .. todo:: WRITEME """ assert state_below is None msg = layer_above.downward_message(state_above) mu = self.mu z = msg + mu rval = theano_rng.normal(size = z.shape, avg = z, dtype = z.dtype, std = 1. / T.sqrt(self.beta)) return rval def recons_cost(self, V, V_hat_unmasked, drop_mask = None, use_sum=False): """ .. todo:: WRITEME """ return self._recons_cost(V=V, V_hat_unmasked=V_hat_unmasked, drop_mask=drop_mask, use_sum=use_sum, beta=self.beta) def _recons_cost(self, V, V_hat_unmasked, beta, drop_mask=None, use_sum=False): """ .. todo:: WRITEME """ V_hat = V_hat_unmasked assert V.ndim == V_hat.ndim beta = self.broadcasted_beta() unmasked_cost = 0.5 * beta * T.sqr(V-V_hat) - 0.5*T.log(beta / (2*np.pi)) assert unmasked_cost.ndim == V_hat.ndim if drop_mask is None: masked_cost = unmasked_cost else: masked_cost = drop_mask * unmasked_cost if use_sum: return masked_cost.mean(axis=0).sum() return masked_cost.mean() return masked_cost.mean() def upward_state(self, total_state): """ .. todo:: WRITEME """ if self.nvis is None and total_state.ndim != 4: raise ValueError("total_state should have 4 dimensions, has "+str(total_state.ndim)) assert total_state is not None V = total_state self.input_space.validate(V) upward_state = (V - self.broadcasted_mu()) * self.broadcasted_beta() return upward_state def make_state(self, num_examples, numpy_rng): """ .. todo:: WRITEME """ shape = [num_examples] if self.nvis is None: rows, cols = self.space.shape channels = self.space.num_channels shape.append(rows) shape.append(cols) shape.append(channels) else: shape.append(self.nvis) sample = numpy_rng.randn(*shape) sample *= 1./np.sqrt(self.beta.get_value()) sample += self.mu.get_value() rval = sharedX(sample, name = 'v_sample_shared') return rval def install_presynaptic_outputs(self, outputs_dict, batch_size): """ .. todo:: WRITEME """ outputs_dict['output_V_weighted_pred_sum'] = self.space.make_shared_batch(batch_size) def ensemble_prediction(self, symbolic, outputs_dict, ensemble): """ .. todo:: WRITEME """ """ Output a symbolic expression for V_hat_unmasked based on taking the geometric mean over the ensemble and renormalizing. n - 1 members of the ensemble have modified outputs_dict and the nth gives its prediction in "symbolic". The parameters for the nth one are currently loaded in the model. """ weighted_pred_sum = outputs_dict['output_V_weighted_pred_sum'] \ + self.broadcasted_beta() * symbolic beta_sum = sum(ensemble.get_ensemble_variants(self.beta)) unmasked_V_hat = weighted_pred_sum / self.broadcast_beta(beta_sum) return unmasked_V_hat def ensemble_recons_cost(self, V, V_hat_unmasked, drop_mask=None, use_sum=False, ensemble=None): """ .. todo:: WRITEME """ beta = sum(ensemble.get_ensemble_variants(self.beta)) / ensemble.num_copies return self._recons_cost(V=V, V_hat_unmasked=V_hat_unmasked, beta=beta, drop_mask=drop_mask, use_sum=use_sum) class ConvMaxPool(HiddenLayer): """ Implements a hidden convolutional layer. The layer lives in a Conv2DSpace. Parameters ---------- output_channels : WRITEME kernel_rows : WRITEME kernel_cols : WRITEME pool_rows : WRITEME pool_cols : WRITEME layer_name : str Name of the layer center : bool If True, use Gregoire Montavon's centering trick irange : float If specified, initialize the weights in U(-irange, irange) sparse_init : WRITEME scale_by_sharing : WRITEME init_bias : WRITEME border_mode : WRITEME output_axes : WRITEME """ def __init__(self, output_channels, kernel_rows, kernel_cols, pool_rows, pool_cols, layer_name, center = False, irange = None, sparse_init = None, scale_by_sharing = True, init_bias = 0., border_mode = 'valid', output_axes = ('b', 'c', 0, 1)): super(ConvMaxPool, self).__init__() self.__dict__.update(locals()) del self.self assert (irange is None) != (sparse_init is None) self.b = sharedX( np.zeros((output_channels,)) + init_bias, name = layer_name + '_b') assert border_mode in ['full','valid'] def broadcasted_bias(self): """ .. todo:: WRITEME """ assert self.b.ndim == 1 shuffle = [ 'x' ] * 4 shuffle[self.output_axes.index('c')] = 0 return self.b.dimshuffle(*shuffle) def get_total_state_space(self): """ .. todo:: WRITEME """ return CompositeSpace((self.h_space, self.output_space)) def set_input_space(self, space): """ .. todo:: WRITEME """ """ Note: this resets parameters!""" if not isinstance(space, Conv2DSpace): raise TypeError("ConvMaxPool can only act on a Conv2DSpace, but received " + str(type(space))+" as input.") self.input_space = space self.input_rows, self.input_cols = space.shape self.input_channels = space.num_channels if self.border_mode == 'valid': self.h_rows = self.input_rows - self.kernel_rows + 1 self.h_cols = self.input_cols - self.kernel_cols + 1 else: assert self.border_mode == 'full' self.h_rows = self.input_rows + self.kernel_rows - 1 self.h_cols = self.input_cols + self.kernel_cols - 1 if not( self.h_rows % self.pool_rows == 0): raise ValueError("h_rows = %d, pool_rows = %d. Should be divisible but remainder is %d" % (self.h_rows, self.pool_rows, self.h_rows % self.pool_rows)) assert self.h_cols % self.pool_cols == 0 self.h_space = Conv2DSpace(shape = (self.h_rows, self.h_cols), num_channels = self.output_channels, axes = self.output_axes) self.output_space = Conv2DSpace(shape = (self.h_rows / self.pool_rows, self.h_cols / self.pool_cols), num_channels = self.output_channels, axes = self.output_axes) logger.info('{0}: detector shape: {1} ' 'pool shape: {2}'.format(self.layer_name, self.h_space.shape, self.output_space.shape)) if tuple(self.output_axes) == ('b', 0, 1, 'c'): self.max_pool = max_pool_b01c elif tuple(self.output_axes) == ('b', 'c', 0, 1): self.max_pool = max_pool else: raise NotImplementedError() if self.irange is not None: self.transformer = make_random_conv2D(self.irange, input_space = space, output_space = self.h_space, kernel_shape = (self.kernel_rows, self.kernel_cols), batch_size = self.dbm.batch_size, border_mode = self.border_mode, rng = self.dbm.rng) else: self.transformer = make_sparse_random_conv2D(self.sparse_init, input_space = space, output_space = self.h_space, kernel_shape = (self.kernel_rows, self.kernel_cols), batch_size = self.dbm.batch_size, border_mode = self.border_mode, rng = self.dbm.rng) self.transformer._filters.name = self.layer_name + '_W' W ,= self.transformer.get_params() assert W.name is not None if self.center: p_ofs, h_ofs = self.init_mf_state() self.p_offset = sharedX(self.output_space.get_origin(), 'p_offset') self.h_offset = sharedX(self.h_space.get_origin(), 'h_offset') f = function([], updates={self.p_offset: p_ofs[0,:,:,:], self.h_offset: h_ofs[0,:,:,:]}) f() def get_params(self): """ .. todo:: WRITEME """ assert self.b.name is not None W ,= self.transformer.get_params() assert W.name is not None return [ W, self.b] def state_to_b01c(self, state): """ .. todo:: WRITEME """ if tuple(self.output_axes) == ('b',0,1,'c'): return state return [ Conv2DSpace.convert(elem, self.output_axes, ('b', 0, 1, 'c')) for elem in state ] def get_range_rewards(self, state, coeffs): """ .. todo:: WRITEME """ rval = 0. if self.pool_rows == 1 and self.pool_cols == 1: # If the pool size is 1 then pools = detectors # and we should not penalize pools and detectors separately assert len(state) == 2 assert isinstance(coeffs, float) _, state = state state = [state] coeffs = [coeffs] else: assert all([len(elem) == 2 for elem in [state, coeffs]]) for s, c in safe_zip(state, coeffs): if c == 0.: continue # Range over everything but the channel index # theano can only take gradient through max if the max is over 1 axis or all axes # so I manually unroll the max for the case I use here assert self.h_space.axes == ('b', 'c', 0, 1) assert self.output_space.axes == ('b', 'c', 0, 1) mx = s.max(axis=3).max(axis=2).max(axis=0) assert hasattr(mx.owner.op, 'grad') mn = s.min(axis=3).max(axis=2).max(axis=0) assert hasattr(mn.owner.op, 'grad') assert mx.ndim == 1 assert mn.ndim == 1 r = mx - mn rval += (1. - r).mean() * c return rval def get_l1_act_cost(self, state, target, coeff, eps): """ .. todo:: WRITEME """ """ target: if pools contain more than one element, should be a list with two elements. the first element is for the pooling units and the second for the detector units. """ rval = 0. if self.pool_rows == 1 and self.pool_cols == 1: # If the pool size is 1 then pools = detectors # and we should not penalize pools and detectors separately assert len(state) == 2 assert isinstance(target, float) assert isinstance(coeff, float) _, state = state state = [state] target = [target] coeff = [coeff] if eps is None: eps = 0. eps = [eps] else: if eps is None: eps = [0., 0.] assert all([len(elem) == 2 for elem in [state, target, coeff]]) p_target, h_target = target if h_target > p_target and (coeff[0] != 0. and coeff[1] != 0.): # note that, within each group, E[p] is the sum of E[h] warnings.warn("Do you really want to regularize the detector units to be more active than the pooling units?") for s, t, c, e in safe_zip(state, target, coeff, eps): if c == 0.: continue # Average over everything but the channel index m = s.mean(axis= [ ax for ax in range(4) if self.output_axes[ax] != 'c' ]) assert m.ndim == 1 rval += T.maximum(abs(m-t)-e,0.).mean()*c return rval def get_lr_scalers(self): """ .. todo:: WRITEME """ if self.scale_by_sharing: # scale each learning rate by 1 / # times param is reused h_rows, h_cols = self.h_space.shape num_h = float(h_rows * h_cols) return OrderedDict([(self.transformer._filters, 1./num_h), (self.b, 1. / num_h)]) else: return OrderedDict() def upward_state(self, total_state): """ .. todo:: WRITEME """ p,h = total_state if not hasattr(self, 'center'): self.center = False if self.center: p -= self.p_offset h -= self.h_offset return p def downward_state(self, total_state): """ .. todo:: WRITEME """ p,h = total_state if not hasattr(self, 'center'): self.center = False if self.center: p -= self.p_offset h -= self.h_offset return h def get_monitoring_channels_from_state(self, state): """ .. todo:: WRITEME """ P, H = state if tuple(self.output_axes) == ('b',0,1,'c'): p_max = P.max(axis=(0,1,2)) p_min = P.min(axis=(0,1,2)) p_mean = P.mean(axis=(0,1,2)) else: assert tuple(self.output_axes) == ('b','c',0,1) p_max = P.max(axis=(0,2,3)) p_min = P.min(axis=(0,2,3)) p_mean = P.mean(axis=(0,2,3)) p_range = p_max - p_min rval = { 'p_max_max' : p_max.max(), 'p_max_mean' : p_max.mean(), 'p_max_min' : p_max.min(), 'p_min_max' : p_min.max(), 'p_min_mean' : p_min.mean(), 'p_min_max' : p_min.max(), 'p_range_max' : p_range.max(), 'p_range_mean' : p_range.mean(), 'p_range_min' : p_range.min(), 'p_mean_max' : p_mean.max(), 'p_mean_mean' : p_mean.mean(), 'p_mean_min' : p_mean.min() } return rval def get_weight_decay(self, coeffs): """ .. todo:: WRITEME """ W , = self.transformer.get_params() return coeffs * T.sqr(W).sum() def mf_update(self, state_below, state_above, layer_above = None, double_weights = False, iter_name = None): """ .. todo:: WRITEME """ self.input_space.validate(state_below) if iter_name is None: iter_name = 'anon' if state_above is not None: assert layer_above is not None msg = layer_above.downward_message(state_above) msg.name = 'msg_from_'+layer_above.layer_name+'_to_'+self.layer_name+'['+iter_name+']' else: msg = None if not hasattr(state_below, 'ndim'): raise TypeError("state_below should be a TensorType, got " + str(state_below) + " of type " + str(type(state_below))) if state_below.ndim != 4: raise ValueError("state_below should have ndim 4, has "+str(state_below.ndim)) if double_weights: state_below = 2. * state_below state_below.name = self.layer_name + '_'+iter_name + '_2state' z = self.transformer.lmul(state_below) + self.broadcasted_bias() if self.layer_name is not None and iter_name is not None: z.name = self.layer_name + '_' + iter_name + '_z' p,h = self.max_pool(z, (self.pool_rows, self.pool_cols), msg) p.name = self.layer_name + '_p_' + iter_name h.name = self.layer_name + '_h_' + iter_name return p, h def sample(self, state_below = None, state_above = None, layer_above = None, theano_rng = None): """ .. todo:: WRITEME """ if state_above is not None: msg = layer_above.downward_message(state_above) try: self.output_space.validate(msg) except TypeError as e: reraise_as(TypeError(str(type(layer_above))+".downward_message gave something that was not the right type: "+str(e))) else: msg = None z = self.transformer.lmul(state_below) + self.broadcasted_bias() p, h, p_sample, h_sample = self.max_pool(z, (self.pool_rows, self.pool_cols), msg, theano_rng) return p_sample, h_sample def downward_message(self, downward_state): """ .. todo:: WRITEME """ self.h_space.validate(downward_state) return self.transformer.lmul_T(downward_state) def set_batch_size(self, batch_size): """ .. todo:: WRITEME """ self.transformer.set_batch_size(batch_size) def get_weights_topo(self): """ .. todo:: WRITEME """ outp, inp, rows, cols = range(4) raw = self.transformer._filters.get_value() return np.transpose(raw,(outp,rows,cols,inp)) def init_mf_state(self): """ .. todo:: WRITEME """ default_z = self.broadcasted_bias() shape = { 'b': self.dbm.batch_size, 0: self.h_space.shape[0], 1: self.h_space.shape[1], 'c': self.h_space.num_channels } # work around theano bug with broadcasted stuff default_z += T.alloc(*([0.]+[shape[elem] for elem in self.h_space.axes])).astype(default_z.dtype) assert default_z.ndim == 4 p, h = self.max_pool( z = default_z, pool_shape = (self.pool_rows, self.pool_cols)) return p, h def make_state(self, num_examples, numpy_rng): """ .. todo:: WRITEME """ """ Returns a shared variable containing an actual state (not a mean field state) for this variable. """ t1 = time.time() empty_input = self.h_space.get_origin_batch(self.dbm.batch_size) h_state = sharedX(empty_input) default_z = T.zeros_like(h_state) + self.broadcasted_bias() theano_rng = make_theano_rng(None, numpy_rng.randint(2 ** 16), which_method="binomial") p_exp, h_exp, p_sample, h_sample = self.max_pool( z = default_z, pool_shape = (self.pool_rows, self.pool_cols), theano_rng = theano_rng) p_state = sharedX( self.output_space.get_origin_batch( self.dbm.batch_size)) t2 = time.time() f = function([], updates = [ (p_state, p_sample), (h_state, h_sample) ]) t3 = time.time() f() t4 = time.time() logger.info('{0}.make_state took'.format(self, t4-t1)) logger.info('\tcompose time: {0}'.format(t2-t1)) logger.info('\tcompile time: {0}'.format(t3-t2)) logger.info('\texecute time: {0}'.format(t4-t3)) p_state.name = 'p_sample_shared' h_state.name = 'h_sample_shared' return p_state, h_state def expected_energy_term(self, state, average, state_below, average_below): """ .. todo:: WRITEME """ self.input_space.validate(state_below) downward_state = self.downward_state(state) self.h_space.validate(downward_state) # Energy function is linear so it doesn't matter if we're averaging or not # Specifically, our terms are -u^T W d - b^T d where u is the upward state of layer below # and d is the downward state of this layer bias_term = (downward_state * self.broadcasted_bias()).sum(axis=(1,2,3)) weights_term = (self.transformer.lmul(state_below) * downward_state).sum(axis=(1,2,3)) rval = -bias_term - weights_term assert rval.ndim == 1 return rval class ConvC01B_MaxPool(HiddenLayer): """ .. todo:: WRITEME """ def __init__(self, output_channels, kernel_shape, pool_rows, pool_cols, layer_name, center = False, irange = None, sparse_init = None, scale_by_sharing = True, init_bias = 0., pad = 0, partial_sum = 1): self.__dict__.update(locals()) del self.self assert (irange is None) != (sparse_init is None) self.output_axes = ('c', 0, 1, 'b') self.detector_channels = output_channels self.tied_b = 1 def broadcasted_bias(self): """ .. todo:: WRITEME """ if self.b.ndim != 1: raise NotImplementedError() shuffle = [ 'x' ] * 4 shuffle[self.output_axes.index('c')] = 0 return self.b.dimshuffle(*shuffle) def get_total_state_space(self): """ .. todo:: WRITEME """ return CompositeSpace((self.h_space, self.output_space)) def set_input_space(self, space): """ .. todo:: WRITEME """ """ Note: this resets parameters!""" setup_detector_layer_c01b(layer=self, input_space=space, rng=self.dbm.rng,) if not tuple(space.axes) == ('c', 0, 1, 'b'): raise AssertionError("You're not using c01b inputs. Ian is enforcing c01b inputs while developing his pipeline to make sure it runs at maximal speed. If you really don't want to use c01b inputs, you can remove this check and things should work. If they don't work it's only because they're not tested.") if self.dummy_channels != 0: raise NotImplementedError(str(type(self))+" does not support adding dummy channels for cuda-convnet compatibility yet, you must implement that feature or use inputs with <=3 channels or a multiple of 4 channels") self.input_rows = self.input_space.shape[0] self.input_cols = self.input_space.shape[1] self.h_rows = self.detector_space.shape[0] self.h_cols = self.detector_space.shape[1] if not(self.h_rows % self.pool_rows == 0): raise ValueError(self.layer_name + ": h_rows = %d, pool_rows = %d. Should be divisible but remainder is %d" % (self.h_rows, self.pool_rows, self.h_rows % self.pool_rows)) assert self.h_cols % self.pool_cols == 0 self.h_space = Conv2DSpace(shape = (self.h_rows, self.h_cols), num_channels = self.output_channels, axes = self.output_axes) self.output_space = Conv2DSpace(shape = (self.h_rows / self.pool_rows, self.h_cols / self.pool_cols), num_channels = self.output_channels, axes = self.output_axes) logger.info('{0} : detector shape: {1} ' 'pool shape: {2}'.format(self.layer_name, self.h_space.shape, self.output_space.shape)) assert tuple(self.output_axes) == ('c', 0, 1, 'b') self.max_pool = max_pool_c01b if self.center: p_ofs, h_ofs = self.init_mf_state() self.p_offset = sharedX(self.output_space.get_origin(), 'p_offset') self.h_offset = sharedX(self.h_space.get_origin(), 'h_offset') f = function([], updates={self.p_offset: p_ofs[:,:,:,0], self.h_offset: h_ofs[:,:,:,0]}) f() def get_params(self): """ .. todo:: WRITEME """ assert self.b.name is not None W ,= self.transformer.get_params() assert W.name is not None return [ W, self.b] def state_to_b01c(self, state): """ .. todo:: WRITEME """ if tuple(self.output_axes) == ('b',0,1,'c'): return state return [ Conv2DSpace.convert(elem, self.output_axes, ('b', 0, 1, 'c')) for elem in state ] def get_range_rewards(self, state, coeffs): """ .. todo:: WRITEME """ rval = 0. if self.pool_rows == 1 and self.pool_cols == 1: # If the pool size is 1 then pools = detectors # and we should not penalize pools and detectors separately assert len(state) == 2 assert isinstance(coeffs, float) _, state = state state = [state] coeffs = [coeffs] else: assert all([len(elem) == 2 for elem in [state, coeffs]]) for s, c in safe_zip(state, coeffs): if c == 0.: continue # Range over everything but the channel index # theano can only take gradient through max if the max is over 1 axis or all axes # so I manually unroll the max for the case I use here assert self.h_space.axes == ('b', 'c', 0, 1) assert self.output_space.axes == ('b', 'c', 0, 1) mx = s.max(axis=3).max(axis=2).max(axis=0) assert hasattr(mx.owner.op, 'grad') mn = s.min(axis=3).max(axis=2).max(axis=0) assert hasattr(mn.owner.op, 'grad') assert mx.ndim == 1 assert mn.ndim == 1 r = mx - mn rval += (1. - r).mean() * c return rval def get_l1_act_cost(self, state, target, coeff, eps): """ .. todo:: WRITEME properly Parameters ---------- state : WRITEME target : WRITEME if pools contain more than one element, should be a list with two elements. the first element is for the pooling units and the second for the detector units. coeff : WRITEME eps : WRITEME """ rval = 0. if self.pool_rows == 1 and self.pool_cols == 1: # If the pool size is 1 then pools = detectors # and we should not penalize pools and detectors separately assert len(state) == 2 assert isinstance(target, float) assert isinstance(coeff, float) _, state = state state = [state] target = [target] coeff = [coeff] if eps is None: eps = 0. eps = [eps] else: if eps is None: eps = [0., 0.] assert all([len(elem) == 2 for elem in [state, target, coeff]]) p_target, h_target = target if h_target > p_target and (coeff[0] != 0. and coeff[1] != 0.): # note that, within each group, E[p] is the sum of E[h] warnings.warn("Do you really want to regularize the detector units to be more active than the pooling units?") for s, t, c, e in safe_zip(state, target, coeff, eps): if c == 0.: continue # Average over everything but the channel index m = s.mean(axis= [ ax for ax in range(4) if self.output_axes[ax] != 'c' ]) assert m.ndim == 1 rval += T.maximum(abs(m-t)-e,0.).mean()*c return rval def get_lr_scalers(self): """ .. todo:: WRITEME """ rval = OrderedDict() if self.scale_by_sharing: # scale each learning rate by 1 / # times param is reused h_rows, h_cols = self.h_space.shape num_h = float(h_rows * h_cols) rval[self.transformer._filters] = 1. /num_h rval[self.b] = 1. / num_h return rval def upward_state(self, total_state): """ .. todo:: WRITEME """ p,h = total_state if not hasattr(self, 'center'): self.center = False if self.center: p -= self.p_offset h -= self.h_offset return p def downward_state(self, total_state): """ .. todo:: WRITEME """ p,h = total_state if not hasattr(self, 'center'): self.center = False if self.center: p -= self.p_offset h -= self.h_offset return h def get_monitoring_channels_from_state(self, state): """ .. todo:: WRITEME """ P, H = state axes = tuple([i for i, ax in enumerate(self.output_axes) if ax != 'c']) p_max = P.max(axis=(0,1,2)) p_min = P.min(axis=(0,1,2)) p_mean = P.mean(axis=(0,1,2)) p_range = p_max - p_min rval = { 'p_max_max' : p_max.max(), 'p_max_mean' : p_max.mean(), 'p_max_min' : p_max.min(), 'p_min_max' : p_min.max(), 'p_min_mean' : p_min.mean(), 'p_min_max' : p_min.max(), 'p_range_max' : p_range.max(), 'p_range_mean' : p_range.mean(), 'p_range_min' : p_range.min(), 'p_mean_max' : p_mean.max(), 'p_mean_mean' : p_mean.mean(), 'p_mean_min' : p_mean.min() } return rval def get_weight_decay(self, coeffs): """ .. todo:: WRITEME """ W , = self.transformer.get_params() return coeffs * T.sqr(W).sum() def mf_update(self, state_below, state_above, layer_above = None, double_weights = False, iter_name = None): """ .. todo:: WRITEME """ self.input_space.validate(state_below) if iter_name is None: iter_name = 'anon' if state_above is not None: assert layer_above is not None msg = layer_above.downward_message(state_above) msg.name = 'msg_from_'+layer_above.layer_name+'_to_'+self.layer_name+'['+iter_name+']' else: msg = None if not hasattr(state_below, 'ndim'): raise TypeError("state_below should be a TensorType, got " + str(state_below) + " of type " + str(type(state_below))) if state_below.ndim != 4: raise ValueError("state_below should have ndim 4, has "+str(state_below.ndim)) if double_weights: state_below = 2. * state_below state_below.name = self.layer_name + '_'+iter_name + '_2state' z = self.transformer.lmul(state_below) + self.broadcasted_bias() if self.layer_name is not None and iter_name is not None: z.name = self.layer_name + '_' + iter_name + '_z' p,h = self.max_pool(z, (self.pool_rows, self.pool_cols), msg) p.name = self.layer_name + '_p_' + iter_name h.name = self.layer_name + '_h_' + iter_name return p, h def sample(self, state_below = None, state_above = None, layer_above = None, theano_rng = None): """ .. todo:: WRITEME """ raise NotImplementedError("Need to update for C01B") if state_above is not None: msg = layer_above.downward_message(state_above) try: self.output_space.validate(msg) except TypeError as e: reraise_as(TypeError(str(type(layer_above))+".downward_message gave something that was not the right type: "+str(e))) else: msg = None z = self.transformer.lmul(state_below) + self.broadcasted_bias() p, h, p_sample, h_sample = self.max_pool(z, (self.pool_rows, self.pool_cols), msg, theano_rng) return p_sample, h_sample def downward_message(self, downward_state): """ .. todo:: WRITEME """ self.h_space.validate(downward_state) return self.transformer.lmul_T(downward_state) def set_batch_size(self, batch_size): """ .. todo:: WRITEME """ self.transformer.set_batch_size(batch_size) def get_weights_topo(self): """ .. todo:: WRITEME """ return self.transformer.get_weights_topo() def init_mf_state(self): """ .. todo:: WRITEME """ default_z = self.broadcasted_bias() shape = { 'b': self.dbm.batch_size, 0: self.h_space.shape[0], 1: self.h_space.shape[1], 'c': self.h_space.num_channels } # work around theano bug with broadcasted stuff default_z += T.alloc(*([0.]+[shape[elem] for elem in self.h_space.axes])).astype(default_z.dtype) assert default_z.ndim == 4 p, h = self.max_pool( z = default_z, pool_shape = (self.pool_rows, self.pool_cols)) return p, h def make_state(self, num_examples, numpy_rng): """ .. todo:: WRITEME properly Returns a shared variable containing an actual state (not a mean field state) for this variable. """ raise NotImplementedError("Need to update for C01B") t1 = time.time() empty_input = self.h_space.get_origin_batch(self.dbm.batch_size) h_state = sharedX(empty_input) default_z = T.zeros_like(h_state) + self.broadcasted_bias() theano_rng = make_theano_rng(None, numpy_rng.randint(2 ** 16), which_method="binomial") p_exp, h_exp, p_sample, h_sample = self.max_pool( z = default_z, pool_shape = (self.pool_rows, self.pool_cols), theano_rng = theano_rng) p_state = sharedX( self.output_space.get_origin_batch( self.dbm.batch_size)) t2 = time.time() f = function([], updates = [ (p_state, p_sample), (h_state, h_sample) ]) t3 = time.time() f() t4 = time.time() logger.info('{0}.make_state took {1}'.format(self, t4-t1)) logger.info('\tcompose time: {0}'.format(t2-t1)) logger.info('\tcompile time: {0}'.format(t3-t2)) logger.info('\texecute time: {0}'.format(t4-t3)) p_state.name = 'p_sample_shared' h_state.name = 'h_sample_shared' return p_state, h_state def expected_energy_term(self, state, average, state_below, average_below): """ .. todo:: WRITEME """ raise NotImplementedError("Need to update for C01B") self.input_space.validate(state_below) downward_state = self.downward_state(state) self.h_space.validate(downward_state) # Energy function is linear so it doesn't matter if we're averaging or not # Specifically, our terms are -u^T W d - b^T d where u is the upward state of layer below # and d is the downward state of this layer bias_term = (downward_state * self.broadcasted_bias()).sum(axis=(1,2,3)) weights_term = (self.transformer.lmul(state_below) * downward_state).sum(axis=(1,2,3)) rval = -bias_term - weights_term assert rval.ndim == 1 return rval class BVMP_Gaussian(BinaryVectorMaxPool): """ Like BinaryVectorMaxPool, but must have GaussianVisLayer as its input. Uses its beta to bias the hidden units appropriately. See gaussian.lyx beta is *not* considered a parameter of this layer, it's just an external factor influencing how this layer behaves. Gradient can still flow to beta, but it will only be included in the parameters list if some class other than this layer includes it. .. todo:: WRITEME : parameter list """ def __init__(self, input_layer, detector_layer_dim, pool_size, layer_name, irange = None, sparse_init = None, sparse_stdev = 1., include_prob = 1.0, init_bias = 0., W_lr_scale = None, b_lr_scale = None, center = False, mask_weights = None, max_col_norm = None, copies = 1): warnings.warn("BVMP_Gaussian math is very faith-based, need to complete gaussian.lyx") args = locals() del args['input_layer'] del args['self'] super(BVMP_Gaussian, self).__init__(**args) self.input_layer = input_layer def get_weights(self): """ .. todo:: WRITEME """ if self.requires_reformat: # This is not really an unimplemented case. # We actually don't know how to format the weights # in design space. We got the data in topo space # and we don't have access to the dataset raise NotImplementedError() W ,= self.transformer.get_params() W = W.get_value() x = input("multiply by beta?") if x == 'y': beta = self.input_layer.beta.get_value() return (W.T * beta).T assert x == 'n' return W def set_weights(self, weights): """ .. todo:: WRITEME """ raise NotImplementedError("beta would make get_weights for visualization not correspond to set_weights") W, = self.transformer.get_params() W.set_value(weights) def set_biases(self, biases, recenter = False): """ .. todo:: WRITEME """ self.b.set_value(biases) if recenter: assert self.center if self.pool_size != 1: raise NotImplementedError() self.offset.set_value(sigmoid_numpy(self.b.get_value())) def get_biases(self): """ .. todo:: WRITEME """ return self.b.get_value() - self.beta_bias().eval() def sample(self, state_below = None, state_above = None, layer_above = None, theano_rng = None): """ .. todo:: WRITEME """ raise NotImplementedError("need to account for beta") if self.copies != 1: raise NotImplementedError() if theano_rng is None: raise ValueError("theano_rng is required; it just defaults to None so that it may appear after layer_above / state_above in the list.") if state_above is not None: msg = layer_above.downward_message(state_above) else: msg = None if self.requires_reformat: state_below = self.input_space.format_as(state_below, self.desired_space) z = self.transformer.lmul(state_below) + self.b p, h, p_sample, h_sample = max_pool_channels(z, self.pool_size, msg, theano_rng) return p_sample, h_sample def downward_message(self, downward_state): """ .. todo:: WRITEME """ rval = self.transformer.lmul_T(downward_state) if self.requires_reformat: rval = self.desired_space.format_as(rval, self.input_space) return rval * self.copies def init_mf_state(self): """ .. todo:: WRITEME """ # work around theano bug with broadcasted vectors z = T.alloc(0., self.dbm.batch_size, self.detector_layer_dim).astype(self.b.dtype) + \ self.b.dimshuffle('x', 0) + self.beta_bias() rval = max_pool_channels(z = z, pool_size = self.pool_size) return rval def make_state(self, num_examples, numpy_rng): """ .. todo:: WRITEME properly Returns a shared variable containing an actual state (not a mean field state) for this variable. """ raise NotImplementedError("need to account for beta") if not hasattr(self, 'copies'): self.copies = 1 if self.copies != 1: raise NotImplementedError() empty_input = self.h_space.get_origin_batch(num_examples) empty_output = self.output_space.get_origin_batch(num_examples) h_state = sharedX(empty_input) p_state = sharedX(empty_output) theano_rng = make_theano_rng(None, numpy_rng.randint(2 ** 16), which_method="binomial") default_z = T.zeros_like(h_state) + self.b p_exp, h_exp, p_sample, h_sample = max_pool_channels( z = default_z, pool_size = self.pool_size, theano_rng = theano_rng) assert h_sample.dtype == default_z.dtype f = function([], updates = [ (p_state , p_sample), (h_state , h_sample) ]) f() p_state.name = 'p_sample_shared' h_state.name = 'h_sample_shared' return p_state, h_state def expected_energy_term(self, state, average, state_below, average_below): """ .. todo:: WRITEME """ raise NotImplementedError("need to account for beta, and maybe some oether stuff") # Don't need to do anything special for centering, upward_state / downward state # make it all just work self.input_space.validate(state_below) if self.requires_reformat: if not isinstance(state_below, tuple): for sb in get_debug_values(state_below): if sb.shape[0] != self.dbm.batch_size: raise ValueError("self.dbm.batch_size is %d but got shape of %d" % (self.dbm.batch_size, sb.shape[0])) assert reduce(operator.mul, sb.shape[1:]) == self.input_dim state_below = self.input_space.format_as(state_below, self.desired_space) downward_state = self.downward_state(state) self.h_space.validate(downward_state) # Energy function is linear so it doesn't matter if we're averaging or not # Specifically, our terms are -u^T W d - b^T d where u is the upward state of layer below # and d is the downward state of this layer bias_term = T.dot(downward_state, self.b) weights_term = (self.transformer.lmul(state_below) * downward_state).sum(axis=1) rval = -bias_term - weights_term assert rval.ndim == 1 return rval * self.copies def linear_feed_forward_approximation(self, state_below): """ .. todo:: WRITEME properly Used to implement TorontoSparsity. Unclear exactly what properties of it are important or how to implement it for other layers. Properties it must have: output is same kind of data structure (ie, tuple of theano 2-tensors) as mf_update Properties it probably should have for other layer types: An infinitesimal change in state_below or the parameters should cause the same sign of change in the output of linear_feed_forward_approximation and in mf_update Should not have any non-linearities that cause the gradient to shrink Should disregard top-down feedback """ raise NotImplementedError("need to account for beta") z = self.transformer.lmul(state_below) + self.b if self.pool_size != 1: # Should probably implement sum pooling for the non-pooled version, # but in reality it's not totally clear what the right answer is raise NotImplementedError() return z, z def beta_bias(self): """ .. todo:: WRITEME """ W, = self.transformer.get_params() beta = self.input_layer.beta assert beta.ndim == 1 return - 0.5 * T.dot(beta, T.sqr(W)) def mf_update(self, state_below, state_above, layer_above = None, double_weights = False, iter_name = None): """ .. todo:: WRITEME """ self.input_space.validate(state_below) if self.requires_reformat: if not isinstance(state_below, tuple): for sb in get_debug_values(state_below): if sb.shape[0] != self.dbm.batch_size: raise ValueError("self.dbm.batch_size is %d but got shape of %d" % (self.dbm.batch_size, sb.shape[0])) assert reduce(operator.mul, sb.shape[1:]) == self.input_dim state_below = self.input_space.format_as(state_below, self.desired_space) if iter_name is None: iter_name = 'anon' if state_above is not None: assert layer_above is not None msg = layer_above.downward_message(state_above) msg.name = 'msg_from_'+layer_above.layer_name+'_to_'+self.layer_name+'['+iter_name+']' else: msg = None if double_weights: state_below = 2. * state_below state_below.name = self.layer_name + '_'+iter_name + '_2state' z = self.transformer.lmul(state_below) + self.b + self.beta_bias() if self.layer_name is not None and iter_name is not None: z.name = self.layer_name + '_' + iter_name + '_z' p,h = max_pool_channels(z, self.pool_size, msg) p.name = self.layer_name + '_p_' + iter_name h.name = self.layer_name + '_h_' + iter_name return p, h class CompositeLayer(HiddenLayer): """ A Layer constructing by aligning several other Layer objects side by side Parameters ---------- components : WRITEME A list of layers that are combined to form this layer inputs_to_components : None or dict mapping int to list of int Should be None unless the input space is a CompositeSpace If inputs_to_components[i] contains j, it means input i will be given as input to component j. If an input dodes not appear in the dictionary, it will be given to all components. This field allows one CompositeLayer to have another as input without forcing each component to connect to all members of the CompositeLayer below. For example, you might want to have both densely connected and convolutional units in all layers, but a convolutional unit is incapable of taking a non-topological input space. """ def __init__(self, layer_name, components, inputs_to_components = None): self.layer_name = layer_name self.components = list(components) assert isinstance(components, list) for component in components: assert isinstance(component, HiddenLayer) self.num_components = len(components) self.components = list(components) if inputs_to_components is None: self.inputs_to_components = None else: if not isinstance(inputs_to_components, dict): raise TypeError("CompositeLayer expected inputs_to_components to be a dict, got "+str(type(inputs_to_components))) self.inputs_to_components = OrderedDict() for key in inputs_to_components: assert isinstance(key, int) assert key >= 0 value = inputs_to_components[key] assert isinstance(value, list) assert all([isinstance(elem, int) for elem in value]) assert min(value) >= 0 assert max(value) < self.num_components self.inputs_to_components[key] = list(value) def set_input_space(self, space): """ .. todo:: WRITEME """ self.input_space = space if not isinstance(space, CompositeSpace): assert self.inputs_to_components is None self.routing_needed = False else: if self.inputs_to_components is None: self.routing_needed = False else: self.routing_needed = True assert max(self.inputs_to_components) < space.num_components # Invert the dictionary self.components_to_inputs = OrderedDict() for i in xrange(self.num_components): inputs = [] for j in xrange(space.num_components): if i in self.inputs_to_components[j]: inputs.append(i) if len(inputs) < space.num_components: self.components_to_inputs[i] = inputs for i, component in enumerate(self.components): if self.routing_needed and i in self.components_to_inputs: cur_space = space.restrict(self.components_to_inputs[i]) else: cur_space = space component.set_input_space(cur_space) self.output_space = CompositeSpace([ component.get_output_space() for component in self.components ]) def make_state(self, num_examples, numpy_rng): """ .. todo:: WRITEME """ return tuple(component.make_state(num_examples, numpy_rng) for component in self.components) def get_total_state_space(self): """ .. todo:: WRITEME """ return CompositeSpace([component.get_total_state_space() for component in self.components]) def set_batch_size(self, batch_size): """ .. todo:: WRITEME """ for component in self.components: component.set_batch_size(batch_size) def set_dbm(self, dbm): """ .. todo:: WRITEME """ for component in self.components: component.set_dbm(dbm) def mf_update(self, state_below, state_above, layer_above = None, double_weights = False, iter_name = None): """ .. todo:: WRITEME """ rval = [] for i, component in enumerate(self.components): if self.routing_needed and i in self.components_to_inputs: cur_state_below =self.input_space.restrict_batch(state_below, self.components_to_inputs[i]) else: cur_state_below = state_below class RoutingLayer(object): def __init__(self, idx, layer): self.__dict__.update(locals()) del self.self self.layer_name = 'route_'+str(idx)+'_'+layer.layer_name def downward_message(self, state): return self.layer.downward_message(state)[self.idx] if layer_above is not None: cur_layer_above = RoutingLayer(i, layer_above) else: cur_layer_above = None mf_update = component.mf_update(state_below = cur_state_below, state_above = state_above, layer_above = cur_layer_above, double_weights = double_weights, iter_name = iter_name) rval.append(mf_update) return tuple(rval) def init_mf_state(self): """ .. todo:: WRITEME """ return tuple([component.init_mf_state() for component in self.components]) def get_weight_decay(self, coeffs): """ .. todo:: WRITEME """ return sum([component.get_weight_decay(coeff) for component, coeff in safe_zip(self.components, coeffs)]) def upward_state(self, total_state): """ .. todo:: WRITEME """ return tuple([component.upward_state(elem) for component, elem in safe_zip(self.components, total_state)]) def downward_state(self, total_state): """ .. todo:: WRITEME """ return tuple([component.downward_state(elem) for component, elem in safe_zip(self.components, total_state)]) def downward_message(self, downward_state): """ .. todo:: WRITEME """ if isinstance(self.input_space, CompositeSpace): num_input_components = self.input_space.num_components else: num_input_components = 1 rval = [ None ] * num_input_components def add(x, y): if x is None: return y if y is None: return x return x + y for i, packed in enumerate(safe_zip(self.components, downward_state)): component, state = packed if self.routing_needed and i in self.components_to_inputs: input_idx = self.components_to_inputs[i] else: input_idx = range(num_input_components) partial_message = component.downward_message(state) if len(input_idx) == 1: partial_message = [ partial_message ] assert len(input_idx) == len(partial_message) for idx, msg in safe_zip(input_idx, partial_message): rval[idx] = add(rval[idx], msg) if len(rval) == 1: rval = rval[0] else: rval = tuple(rval) self.input_space.validate(rval) return rval def get_l1_act_cost(self, state, target, coeff, eps): """ .. todo:: WRITEME """ return sum([ comp.get_l1_act_cost(s, t, c, e) \ for comp, s, t, c, e in safe_zip(self.components, state, target, coeff, eps)]) def get_range_rewards(self, state, coeffs): """ .. todo:: WRITEME """ return sum([comp.get_range_rewards(s, c) for comp, s, c in safe_zip(self.components, state, coeffs)]) def get_params(self): """ .. todo:: WRITEME """ return reduce(lambda x, y: safe_union(x, y), [component.get_params() for component in self.components]) def get_weights_topo(self): """ .. todo:: WRITEME """ logger.info('Get topological weights for which layer?') for i, component in enumerate(self.components): logger.info('{0} {1}'.format(i, component.layer_name)) x = input() return self.components[int(x)].get_weights_topo() def get_monitoring_channels_from_state(self, state): """ .. todo:: WRITEME """ rval = OrderedDict() for layer, s in safe_zip(self.components, state): d = layer.get_monitoring_channels_from_state(s) for key in d: rval[layer.layer_name+'_'+key] = d[key] return rval def sample(self, state_below = None, state_above = None, layer_above = None, theano_rng = None): """ .. todo:: WRITEME """ rval = [] for i, component in enumerate(self.components): if self.routing_needed and i in self.components_to_inputs: cur_state_below =self.input_space.restrict_batch(state_below, self.components_to_inputs[i]) else: cur_state_below = state_below class RoutingLayer(object): def __init__(self, idx, layer): self.__dict__.update(locals()) del self.self self.layer_name = 'route_'+str(idx)+'_'+layer.layer_name def downward_message(self, state): return self.layer.downward_message(state)[self.idx] if layer_above is not None: cur_layer_above = RoutingLayer(i, layer_above) else: cur_layer_above = None sample = component.sample(state_below = cur_state_below, state_above = state_above, layer_above = cur_layer_above, theano_rng = theano_rng) rval.append(sample) return tuple(rval)
bsd-3-clause
b4a763d630728f22cf0d394d21c9bdf9
29.594906
315
0.53508
4.019698
false
false
false
false
lisa-lab/pylearn2
pylearn2/space/__init__.py
34
99992
""" Classes that define how vector spaces are formatted Most of our models can be viewed as linearly transforming one vector space to another. These classes define how the vector spaces should be represented as theano/numpy variables. For example, the VectorSpace class just represents a vector space with a vector, and the model can transform between spaces with a matrix multiply. The Conv2DSpace represents a vector space as an image, and the model can transform between spaces with a 2D convolution. To make models as general as possible, models should be written in terms of Spaces, rather than in terms of numbers of hidden units, etc. The model should also be written to transform between spaces using a generic linear transformer from the pylearn2.linear module. The Space class is needed so that the model can specify what kinds of inputs it needs and what kinds of outputs it will produce when communicating with other parts of the library. The model also uses Space objects internally to allocate parameters like hidden unit bias terms in the right space. """ __authors__ = "Ian Goodfellow" __copyright__ = "Copyright 2010-2012, Universite de Montreal" __credits__ = ["Ian Goodfellow"] __license__ = "3-clause BSD" __maintainer__ = "LISA Lab" __email__ = "pylearn-dev@googlegroups" import functools import warnings import numpy as np from theano.compat.six.moves import xrange import theano import theano.sparse from theano import tensor from theano.tensor import TensorType from theano.gof.op import get_debug_values from theano.sandbox.cuda.type import CudaNdarrayType from pylearn2.utils import py_integer_types, safe_zip, sharedX, wraps from pylearn2.format.target_format import OneHotFormatter if theano.sparse.enable_sparse: # We know scipy.sparse is available import scipy.sparse def _is_batch_all(batch, predicate): """ Implementation of is_symbolic_batch() and is_numeric_batch(). Returns True iff predicate() returns True for all components of (possibly composite) batch. Parameters ---------- batch : any numeric or symbolic batch. This includes numpy.ndarray, theano.gof.Variable, None, or a (nested) tuple thereof. predicate : function. A unary function of any non-composite batch that returns True or False. """ # Catches any CompositeSpace batches that were mistakenly hand-constructed # using nested lists rather than nested tuples. assert not isinstance(batch, list) # Data-less batches such as None or () are valid numeric and symbolic # batches. # # Justification: we'd like # is_symbolic_batch(space.make_theano_batch()) to always be True, even if # space is an empty CompositeSpace. if batch is None or (isinstance(batch, tuple) and len(batch) == 0): return True if isinstance(batch, tuple): subbatch_results = tuple(_is_batch_all(b, predicate) for b in batch) result = all(subbatch_results) # The subbatch_results must be all true, or all false, not a mix. assert result == any(subbatch_results), ("composite batch had a " "mixture of numeric and " "symbolic subbatches. This " "should never happen.") return result else: return predicate(batch) def is_symbolic_batch(batch): """ Returns True if batch is a symbolic variable. Note that a batch may be both a symbolic and numeric variable (e.g. () for empty CompositeSpaces, None for NullSpaces). """ return _is_batch_all(batch, lambda x: isinstance(x, theano.gof.Variable)) def is_numeric_batch(batch): """ Returns True if batch is a numeric variable. Note that a batch may be both a symbolic and numeric variable (e.g. () for empty CompositeSpaces, None for NullSpaces). """ def is_numeric(batch): # Uses the 'CudaNdarray' string to avoid importing # theano.sandbox.cuda when it is not available return (isinstance(batch, np.ndarray) or scipy.sparse.issparse(batch) or str(type(batch)) == "<type 'CudaNdarray'>") return _is_batch_all(batch, is_numeric) def _dense_to_sparse(batch): """ Casts dense batches to sparse batches (non-composite). Supports both symbolic and numeric variables. """ if isinstance(batch, tuple): raise TypeError("Composite batches not supported.") assert not isinstance(batch, list) if is_symbolic_batch(batch): assert isinstance(batch, theano.tensor.TensorVariable) return theano.sparse.csr_from_dense(batch) else: assert isinstance(batch, np.ndarray), "type of batch: %s" % type(batch) return scipy.sparse.csr_matrix(batch) def _reshape(arg, shape): """ Reshapes a tensor. Supports both symbolic and numeric variables. This is a hack that first converts from sparse to dense, reshapes the dense tensor, then re-converts from dense to sparse. It is therefore memory-inefficient and unsuitable for large tensors. It will be replaced by a proper sparse reshaping Op once Theano implements that. """ if isinstance(arg, tuple): raise TypeError("Composite batches not supported.") assert not isinstance(arg, list) if isinstance(arg, (np.ndarray, theano.tensor.TensorVariable)): return arg.reshape(shape) elif isinstance(arg, theano.sparse.SparseVariable): warnings.warn("Using pylearn2.space._reshape(), which is a " "memory-inefficient hack for reshaping sparse tensors. " "Do not use this on large tensors. This will eventually " "be replaced by a proper Theano Op for sparse " "reshaping, once that is written.") dense = theano.sparse.dense_from_sparse(arg) dense = dense.reshape(shape) if arg.format == 'csr': return theano.sparse.csr_from_dense(dense) elif arg.format == 'csc': return theano.sparse.csc_from_dense(dense) else: raise ValueError('Unexpected sparse format "%s".' % arg.format) else: raise TypeError('Unexpected batch type "%s"' % str(type(arg))) def _cast(arg, dtype): """ Does element-wise casting to dtype. Supports symbolic, numeric, simple, and composite batches. Returns <arg> untouched if <dtype> is None, or dtype is unchanged (i.e. casting a float32 batch to float32). (One exception: composite batches are never returned as-is. A new tuple will always be returned. However, any components with unchanged dtypes will be returned untouched.) """ if dtype is None: return arg assert dtype in tuple(t.dtype for t in theano.scalar.all_types) if isinstance(arg, tuple): return tuple(_cast(a, dtype) for a in arg) elif isinstance(arg, np.ndarray): # theano._asarray is a safer drop-in replacement to numpy.asarray. return theano._asarray(arg, dtype=dtype) elif str(type(arg)) == "<type 'CudaNdarray'>": # numeric CUDA array if str(dtype) != 'float32': raise TypeError("Can only cast a numeric CudaNdarray to " "float32, not %s" % dtype) else: return arg elif (isinstance(arg, theano.gof.Variable) and isinstance(arg.type, CudaNdarrayType)): # symbolic CUDA array if str(dtype) != 'float32': raise TypeError("Can only cast a theano CudaNdArrayType to " "float32, not %s" % dtype) else: return arg elif scipy.sparse.issparse(arg): return arg.astype(dtype) elif isinstance(arg, theano.tensor.TensorVariable): return theano.tensor.cast(arg, dtype) elif isinstance(arg, theano.sparse.SparseVariable): return theano.sparse.cast(arg, dtype) elif isinstance(arg, theano.sandbox.cuda.var.CudaNdarrayVariable): return arg else: raise TypeError("Unsupported arg type '%s'" % str(type(arg))) def _undo_op(arg, string, strict=False): """ Undo symbolic op if string is in str(op). Returns <arg> untouched if there was no symbolic op. Parameters ---------- arg : any symbolic variable. string : str String that specifies op. strict : bool Whether to force op undo or not (default False). """ if hasattr(arg.owner, 'op'): owner = arg.owner if string in str(owner.op): return owner.inputs[0] elif strict: raise ValueError(string + ' not found in op ' + str(owner.op) + '.') elif strict: raise ValueError(string + ' op not found in variable ' + str(arg) + '.') return arg class Space(object): """ A vector space that can be transformed by a linear operator. Space and its subclasses are used to transform a data batch's geometry (e.g. vectors <--> matrices) and optionally, its dtype (e.g. float <--> int). Batches may be one of the following types: - numpy.ndarray - scipy.sparse.csr_matrix - theano.gof.Variable - None (for NullSpace) - A (nested) tuple of the above, possibly empty (for CompositeSpace). Parameters ---------- validate_callbacks : list Callbacks that are run at the start of a call to validate. Each should be a callable with the same signature as validate. An example use case is installing an instance-specific error handler that provides extra instructions for how to correct an input that is in a bad space. np_validate_callacks : list similar to validate_callbacks, but run on calls to np_validate """ def __init__(self, validate_callbacks=None, np_validate_callbacks=None): if validate_callbacks is None: validate_callbacks = [] if np_validate_callbacks is None: np_validate_callbacks = [] self.validate_callbacks = validate_callbacks self.np_validate_callbacks = np_validate_callbacks # Forces subclasses to implement __eq__. # This is necessary for _format_as to work correctly. def __eq__(self, other): """ Returns true iff space.format_as(batch, self) and space.format_as(batch, other) return the same formatted batch. """ raise NotImplementedError("__eq__ not implemented in class %s." % type(self)) def get_batch_axis(self): """ Returns the batch axis of the output space. Returns ------- batch_axis : int the axis of the batch in the output space. """ return 0 def __ne__(self, other): """ .. todo:: WRITEME """ return not (self == other) def __repr__(self): """ .. todo:: WRITEME """ return str(self) @property def dtype(self): """ An object representing the data type used by this space. For simple spaces, this will be a dtype string, as used by numpy, scipy, and theano (e.g. 'float32'). For data-less spaces like NoneType, this will be some other string. For composite spaces, this will be a nested tuple of such strings. """ raise NotImplementedError() @dtype.setter def dtype(self, new_value): """ .. todo:: WRITEME """ raise NotImplementedError() @dtype.deleter def dtype(self): """ .. todo:: WRITEME """ raise RuntimeError("You may not delete the dtype of a space, " "though you can set it to None.") def get_origin(self): """ Returns the origin in this space. Returns ------- origin : ndarray An NumPy array, the shape of a single points in this space, representing the origin. """ raise NotImplementedError() def get_origin_batch(self, batch_size, dtype=None): """ Returns a batch containing `batch_size` copies of the origin. Parameters ---------- batch_size : int The number of examples in the batch to be returned. dtype : WRITEME The dtype of the batch to be returned. Default = None. If None, use self.dtype. Returns ------- batch : ndarray A NumPy array in the shape of a batch of `batch_size` points in this space (with points being indexed along the first axis), each `batch[i]` being a copy of the origin. """ raise NotImplementedError() def make_shared_batch(self, batch_size, name=None, dtype=None): """ .. todo:: WRITEME """ dtype = self._clean_dtype_arg(dtype) origin_batch = self.get_origin_batch(batch_size, dtype) return theano.shared(origin_batch, name=name) def make_theano_batch(self, name=None, dtype=None, batch_size=None): """ Returns a symbolic variable representing a batch of points in this space. Parameters ---------- name : str Variable name for the returned batch. dtype : str Data type for the returned batch. If omitted (None), self.dtype is used. batch_size : int Number of examples in the returned batch. Returns ------- batch : TensorVariable, SparseVariable, or tuple thereof A batch with the appropriate number of dimensions and appropriate broadcast flags to represent a batch of points in this space. """ raise NotImplementedError() def make_batch_theano(self, name=None, dtype=None, batch_size=None): """ An alias to make_theano_batch """ return self.make_theano_batch(name=name, dtype=dtype, batch_size=batch_size) @wraps(make_theano_batch) def get_theano_batch(self, *args, **kwargs): return self.make_theano_batch(*args, **kwargs) def get_total_dimension(self): """ Returns a Python int (not a theano iscalar) representing the dimensionality of a point in this space. If you format a batch of examples in this space as a design matrix (i.e., VectorSpace batch) then the number of columns will be equal to the total dimension. """ raise NotImplementedError(str(type(self)) + " does not implement get_total_dimension.") def np_format_as(self, batch, space): """ Returns a numeric batch (e.g. a numpy.ndarray or scipy.sparse sparse array), formatted to lie in this space. This is just a wrapper around self._format_as, with an extra check to throw an exception if <batch> is symbolic. Should be invertible, i.e. batch should equal `space.format_as(self.format_as(batch, space), self)` Parameters ---------- batch : numpy.ndarray, or one of the scipy.sparse matrices. Array which lies in this space. space : Space Target space to format batch to. Returns ------- WRITEME The formatted batch """ self._check_is_numeric(batch) return self._format_as(is_numeric=True, batch=batch, space=space) def _check_sizes(self, space): """ Called by self._format_as(space), to check whether self and space have compatible sizes. Throws a ValueError if they don't. """ my_dimension = self.get_total_dimension() other_dimension = space.get_total_dimension() if my_dimension != other_dimension: raise ValueError(str(self) + " with total dimension " + str(my_dimension) + " can't format a batch into " + str(space) + "because its total dimension is " + str(other_dimension)) def format_as(self, batch, space): """ .. todo:: WRITEME """ self._check_is_symbolic(batch) return self._format_as(is_numeric=False, batch=batch, space=space) def _format_as(self, is_numeric, batch, space): """ The shared implementation of format_as() and np_format_as(). Agnostic to whether batch is symbolic or numeric, which avoids duplicating a lot of code between format_as() and np_format_as(). Calls the appropriate callbacks, then calls self._format_as_impl(). Should be invertible, i.e. batch should equal `space._format_as(self._format_as(batch, space), self)` Parameters ---------- is_numeric : bool Set to True to call np_validate_callbacks(). Set to False to call validate_callbacks(). batch : WRITEME space : Space WRITEME Returns ------- WRITEME """ assert isinstance(is_numeric, bool) # Checks if batch belongs to this space self._validate(is_numeric, batch) # checks if self and space have compatible sizes for formatting. self._check_sizes(space) return self._format_as_impl(is_numeric, batch, space) def _format_as_impl(self, is_numeric, batch, target_space): """ Actual implementation of format_as/np_format_as. Formats batch to target_space. Should be invertible, i.e. batch should equal `space._format_as_impl(self._format_as_impl(batch, space), self)` Parameters ---------- is_numeric : bool Set to True to treat batch as a numeric batch, False to treat it as a symbolic batch. This is necessary because sometimes a batch's numeric/symbolicness can be ambiguous, i.e. when it's the empty tuple (). batch : a numpy.ndarray, scipy.sparse matrix, theano symbol, or a \ nested tuple thereof Implementations of this method may assume that batch lies in this space (i.e. that it passed self._validate(batch) without throwing an exception). target_space : A Space subclass The space to transform batch into. Returns ------- The batch, converted to the target_space. """ raise NotImplementedError("%s does not implement _format_as_impl()." % type(self)) def undo_np_format_as(self, batch, space): """ Returns a numeric batch (e.g. a numpy.ndarray or scipy.sparse sparse array), with formatting from space undone. This is just a wrapper around self._undo_format_as, with an extra check to throw an exception if <batch> is symbolic. Parameters ---------- batch : numpy.ndarray, or one of the scipy.sparse matrices. Array which lies in this space. space : Space Space to undo formatting from. Returns ------- numpy.ndarray or one of the scipy.sparse matrices The formatted batch. """ self._check_is_numeric(batch) return space.np_format_as(batch=batch, space=self) def undo_format_as(self, batch, space): """ Returns a symbolic batch (e.g. a theano.tensor or theano.sparse array), with formatting from space undone. This is just a wrapper around self._undo_format_as, with an extra check to throw an exception if <batch> is symbolic. Formatting to space Parameters ---------- batch : numpy.ndarray, or one of the scipy.sparse matrices. Array which lies in this space. space : Space Space to undo formatting from. Returns ------- A symbolic Theano variable The batch formatted as self. """ self._check_is_symbolic(batch) space.validate(batch) self._check_sizes(space) batch = self._undo_format_as_impl(batch=batch, space=space) # Checks if batch belongs to this space self.validate(batch) return batch def _undo_format_as_impl(self, batch, target_space): """ Actual implementation of undo_format_as. Undoes target_space_formatting. Note that undo_np_format_as calls np_format_as. Parameters ---------- batch : a theano symbol, or a nested tuple thereof Implementations of this method may assume that batch lies in space (i.e. that it passed self._validate(batch) without throwing an exception). target_space : A Space subclass The space to undo batch formatting from. Returns ------- A symbolic Theano variable The batch, converted from target_space, back to self. """ raise NotImplementedError("%s does not implement " "_undo_format_as_impl()." % type(self)) def validate(self, batch): """ Runs all validate_callbacks, then checks that batch lies in this space. Raises an exception if the batch isn't symbolic, or if any of these checks fails. Parameters ---------- batch : a symbolic (Theano) variable that lies in this space. """ self._check_is_symbolic(batch) self._validate(is_numeric=False, batch=batch) def np_validate(self, batch): """ Runs all np_validate_callbacks, then checks that batch lies in this space. Raises an exception if the batch isn't numeric, or if any of these checks fails. Parameters ---------- batch : a numeric (numpy/scipy.sparse) variable that lies in this \ space """ self._check_is_numeric(batch) self._validate(is_numeric=True, batch=batch) def _validate(self, is_numeric, batch): """ Shared implementation of validate() and np_validate(). Calls validate_callbacks or np_validate_callbacks as appropriate, then calls self._validate_impl(batch) to verify that batch belongs to this space. Parameters ---------- is_numeric : bool. Set to True to call np_validate_callbacks, False to call validate_callbacks. Necessary because it can be impossible to tell from the batch whether it should be treated as a numeric of symbolic batch, for example when the batch is the empty tuple (), or NullSpace batch None. batch : a theano variable, numpy ndarray, scipy.sparse matrix \ or a nested tuple thereof Represents a batch belonging to this space. """ if is_numeric: self._check_is_numeric(batch) callbacks_name = "np_validate_callbacks" else: self._check_is_symbolic(batch) callbacks_name = "validate_callbacks" if not hasattr(self, callbacks_name): raise TypeError("The " + str(type(self)) + " Space subclass " "is required to call the Space superclass " "constructor but does not.") else: callbacks = getattr(self, callbacks_name) for callback in callbacks: callback(batch) self._validate_impl(is_numeric, batch) def _validate_impl(self, is_numeric, batch): """ Subclasses must override this method so that it throws an exception if the batch is the wrong shape or dtype for this Space. Parameters ---------- is_numeric : bool Set to True to treat batch as a numeric type (numpy.ndarray or scipy.sparse matrix). Set to False to treat batch as a symbolic (Theano) variable. Necessary because batch could be (), which could be numeric or symbolic. batch : A numpy ndarray, scipy.sparse matrix, theano variable \ or a nested tuple thereof. Must be a valid batch belonging to this space. """ raise NotImplementedError('Class "%s" does not implement ' '_validate_impl()' % type(self)) def batch_size(self, batch): """ Returns the batch size of a symbolic batch. Parameters ---------- batch : WRITEME """ return self._batch_size(is_numeric=False, batch=batch) def np_batch_size(self, batch): """ Returns the batch size of a numeric (numpy/scipy.sparse) batch. Parameters ---------- batch : WRITEME """ return self._batch_size(is_numeric=True, batch=batch) def _batch_size(self, is_numeric, batch): """ .. todo:: WRITEME """ self._validate(is_numeric, batch) return self._batch_size_impl(is_numeric, batch) def _batch_size_impl(self, is_numeric, batch): """ Returns the batch size of a batch. Parameters ---------- batch : WRITEME """ raise NotImplementedError("%s does not implement batch_size" % type(self)) def get_batch(self, data, start, end): """ Returns a batch of data starting from index `start` to index `stop` Parameters ---------- data : WRITEME start : WRITEME end : WRITEME """ raise NotImplementedError(str(type(self)) + " does not implement " + "get_batch") @staticmethod def _check_is_numeric(batch): """ .. todo:: WRITEME """ if not is_numeric_batch(batch): raise TypeError('Expected batch to be a numeric variable, but ' 'instead it was of type "%s"' % type(batch)) @staticmethod def _check_is_symbolic(batch): """ .. todo:: WRITEME """ if not is_symbolic_batch(batch): raise TypeError('Expected batch to be a symbolic variable, but ' 'instead it was of type "%s"' % type(batch)) def _clean_dtype_arg(self, dtype): """ Checks dtype string for validity, and returns it if it is. If dtype is 'floatX', returns the theano.config.floatX dtype (this will either be 'float32' or 'float64'. """ if isinstance(dtype, np.dtype): dtype = str(dtype) if dtype == 'floatX': return theano.config.floatX if dtype is None or \ dtype in tuple(x.dtype for x in theano.scalar.all_types): return dtype raise TypeError('Unrecognized value "%s" (type %s) for dtype arg' % (dtype, type(dtype))) class SimplyTypedSpace(Space): """ An abstract base class for Spaces that use a numpy/theano dtype string for its .dtype property. """ def __init__(self, dtype='floatX', **kwargs): super(SimplyTypedSpace, self).__init__(**kwargs) self._dtype = super(SimplyTypedSpace, self)._clean_dtype_arg(dtype) def _clean_dtype_arg(self, dtype): """ if dtype is None, checks that self.dtype is not None. Otherwise, same as superclass' implementation. """ if dtype is None: if self.dtype is None: raise TypeError("self.dtype is None, so you must provide a " "non-None dtype argument to this method.") return self.dtype return super(SimplyTypedSpace, self)._clean_dtype_arg(dtype) def _validate_impl(self, is_numeric, batch): """ .. todo:: WRITEME """ if isinstance(batch, tuple): raise TypeError("This space only supports simple dtypes, but " "received a composite batch.") # Checks for information-destroying casts. # # To be maximally strict, we'd guard against all loss of precision by # checking if np.can_cast(batch.dtype, self.dtype). # # Because this prohibits float64->float32, it breaks too much of the # codebase (float64 is default float, float32 is default CUDA float for # many graphics cards). # # Therefore, we only prohibit the following: # * non-integral type to integral type # * complex to non-complex def is_complex(dtype): return np.issubdtype(dtype, np.complex) def is_integral(dtype): return np.issubdtype(dtype, np.integer) if self.dtype is not None: if (is_complex(batch.dtype) and not is_complex(self.dtype)) or \ (not is_integral(batch.dtype) and is_integral(self.dtype)): raise TypeError("Cannot safely cast batch dtype %s to " "space's dtype %s. " % (batch.dtype, self.dtype)) @property def dtype(self): """ .. todo:: WRITEME """ return self._dtype @dtype.setter def dtype(self, new_dtype): """ .. todo:: WRITEME """ self._dtype = super(SimplyTypedSpace, self)._clean_dtype_arg(new_dtype) def __setstate__(self, state_dict): """ .. todo:: WRITEME """ self.__dict__.update(state_dict) # When unpickling a Space that was pickled before Spaces had dtypes, # we need to set the _dtype to the default value. if '_dtype' not in state_dict: self._dtype = theano.config.floatX class IndexSpace(SimplyTypedSpace): """ A space representing indices, for example MNIST labels (0-10) or the indices of words in a dictionary for NLP tasks. A single space can contain multiple indices, for example the word indices of an n-gram. IndexSpaces can be converted to VectorSpaces in two ways: Either the labels are converted into one-hot vectors which are then concatenated, or they are converted into a single vector where 1s indicate labels present i.e. for 4 possible labels we have [0, 2] -> [1 0 1 0] or [0, 2] -> [1 0 0 0 0 0 1 0]. Parameters ---------- max_labels : int The number of possible classes/labels. This means that all labels should be < max_labels. Example: For MNIST there are 10 numbers and hence max_labels = 10. dim : int The number of indices in one space e.g. for MNIST there is one target label and hence dim = 1. If we have an n-gram of word indices as input to a neurel net language model, dim = n. dtype : str A numpy dtype string indicating this space's dtype. Must be an integer type e.g. int32 or int64. kwargs : dict Passes on to superclass constructor """ def __init__(self, max_labels, dim, dtype='int64', **kwargs): if 'int' not in dtype: raise ValueError("The dtype of IndexSpace must be an integer type") super(IndexSpace, self).__init__(dtype, **kwargs) self.max_labels = max_labels self.dim = dim self.formatter = OneHotFormatter(self.max_labels) def __str__(self): """Return a string representation""" return ('%(classname)s(dim=%(dim)s, max_labels=%(max_labels)s, ' 'dtype=%(dtype)s)') % dict(classname=self.__class__.__name__, dim=self.dim, max_labels=self.max_labels, dtype=self.dtype) def __hash__(self): return hash((type(self), self.dim, self.max_labels, self.dtype)) def __eq__(self, other): """ .. todo:: WRITEME """ return (type(self) == type(other) and self.max_labels == other.max_labels and self.dim == other.dim and self.dtype == other.dtype) def __ne__(self, other): """ .. todo:: WRITEME """ return (not self == other) @functools.wraps(Space.get_total_dimension) def get_total_dimension(self): return self.dim @functools.wraps(Space.get_origin) def get_origin(self): return np.zeros((1, self.dim,)) @functools.wraps(Space.get_origin_batch) def get_origin_batch(self, batch_size, dtype=None): dtype = self._clean_dtype_arg(dtype) return np.zeros((batch_size, self.dim), dtype=dtype) @functools.wraps(Space._check_sizes) def _check_sizes(self, space): if isinstance(space, VectorSpace): if space.dim not in (self.max_labels, # merged onehots self.dim * self.max_labels): # concatenated raise ValueError("Can't convert to VectorSpace of dim %d. " "Expected either dim=%d (merged one-hots) or " "%d (concatenated one-hots)" % (space.dim, self.max_labels, self.dim * self.max_labels)) elif isinstance(space, IndexSpace): if space.dim != self.dim or space.max_labels != self.max_labels: raise ValueError("Can't convert to IndexSpace of dim %d and " "max_labels %d." % (space.dim, self.max_labels)) else: raise ValueError("Can't convert to " + str(space.__class__)) @functools.wraps(Space._format_as_impl) def _format_as_impl(self, is_numeric, batch, space): if isinstance(space, VectorSpace): if self.max_labels == space.dim: mode = 'merge' elif self.dim * self.max_labels == space.dim: mode = 'concatenate' else: raise ValueError("There is a bug. Couldn't format to a " "VectorSpace because it had an incorrect " "size, but this should've been caught in " "IndexSpace._check_sizes().") format_func = (self.formatter.format if is_numeric else self.formatter.theano_expr) return _cast(format_func(batch, sparse=space.sparse, mode=mode), space.dtype) elif isinstance(space, IndexSpace): if space.dim != self.dim or space.max_labels != self.max_labels: raise ValueError("The two IndexSpaces' dim and max_labels " "values don't match. This should have been " "caught by IndexSpace._check_sizes().") return _cast(batch, space.dtype) else: raise ValueError("Can't convert %s to %s" % (self, space)) @functools.wraps(Space.make_theano_batch) def make_theano_batch(self, name=None, dtype=None, batch_size=None): if batch_size == 1: rval = tensor.lrow(name=name) else: rval = tensor.lmatrix(name=name) if theano.config.compute_test_value != 'off': if batch_size == 1: n = 1 else: # TODO: try to extract constant scalar value from batch_size n = 4 rval.tag.test_value = self.get_origin_batch(batch_size=n, dtype=dtype) return rval @functools.wraps(Space._batch_size_impl) def _batch_size_impl(self, is_numeric, batch): return batch.shape[0] @functools.wraps(Space._validate_impl) def _validate_impl(self, is_numeric, batch): """ .. todo:: WRITEME """ # checks that batch isn't a tuple, checks batch.type against self.dtype super(IndexSpace, self)._validate_impl(is_numeric, batch) if is_numeric: # Use the 'CudaNdarray' string to avoid importing # theano.sandbox.cuda when it is not available if not isinstance(batch, np.ndarray) \ and str(type(batch)) != "<type 'CudaNdarray'>": raise TypeError("The value of a IndexSpace batch should be a " "numpy.ndarray, or CudaNdarray, but is %s." % str(type(batch))) if batch.ndim != 2: raise ValueError("The value of a IndexSpace batch must be " "2D, got %d dimensions for %s." % (batch.ndim, batch)) if batch.shape[1] != self.dim: raise ValueError("The width of a IndexSpace batch must match " "with the space's dimension, but batch has " "shape %s and dim = %d." % (str(batch.shape), self.dim)) else: if not isinstance(batch, theano.gof.Variable): raise TypeError("IndexSpace batch should be a theano " "Variable, got " + str(type(batch))) if not isinstance(batch.type, (theano.tensor.TensorType, CudaNdarrayType)): raise TypeError("IndexSpace batch should be TensorType or " "CudaNdarrayType, got " + str(batch.type)) if batch.ndim != 2: raise ValueError('IndexSpace batches must be 2D, got %d ' 'dimensions' % batch.ndim) for val in get_debug_values(batch): self.np_validate(val) class VectorSpace(SimplyTypedSpace): """ A space whose points are defined as fixed-length vectors. Parameters ---------- dim : int Dimensionality of a vector in this space. sparse : bool, optional Sparse vector or not dtype : str, optional A numpy dtype string (e.g. 'float32') indicating this space's dtype, or None for a dtype-agnostic space. kwargs : dict Passed on to superclass constructor. """ def __init__(self, dim, sparse=False, dtype='floatX', **kwargs): super(VectorSpace, self).__init__(dtype, **kwargs) self.dim = dim self.sparse = sparse def __str__(self): """ .. todo:: WRITEME """ return ('%s(dim=%d%s, dtype=%s)' % (self.__class__.__name__, self.dim, ', sparse' if self.sparse else '', self.dtype)) @functools.wraps(Space.get_origin) def get_origin(self): return np.zeros((self.dim,)) @functools.wraps(Space.get_origin_batch) def get_origin_batch(self, batch_size, dtype=None): dtype = self._clean_dtype_arg(dtype) if self.sparse: return scipy.sparse.csr_matrix((batch_size, self.dim), dtype=dtype) else: return np.zeros((batch_size, self.dim), dtype=dtype) @functools.wraps(Space._batch_size_impl) def _batch_size_impl(self, is_numeric, batch): return batch.shape[0] @functools.wraps(Space.make_theano_batch) def make_theano_batch(self, name=None, dtype=None, batch_size=None): dtype = self._clean_dtype_arg(dtype) if self.sparse: if batch_size is not None: raise NotImplementedError("batch_size not implemented " "for sparse case") rval = theano.sparse.csr_matrix(name=name, dtype=dtype) else: if batch_size == 1: rval = tensor.row(name=name, dtype=dtype) else: rval = tensor.matrix(name=name, dtype=dtype) if theano.config.compute_test_value != 'off': if batch_size == 1: n = 1 else: # TODO: try to extract constant scalar value from batch_size n = 4 rval.tag.test_value = self.get_origin_batch(batch_size=n, dtype=dtype) return rval @functools.wraps(Space.get_total_dimension) def get_total_dimension(self): return self.dim @functools.wraps(Space._format_as_impl) def _format_as_impl(self, is_numeric, batch, space): to_type = None def is_sparse(batch): return (isinstance(batch, theano.sparse.SparseVariable) or scipy.sparse.issparse(batch)) if not isinstance(space, IndexSpace): my_dimension = self.get_total_dimension() other_dimension = space.get_total_dimension() if my_dimension != other_dimension: raise ValueError(str(self) + " with total dimension " + str(my_dimension) + " can't format a batch into " + str(space) + "because its total dimension is " + str(other_dimension)) if isinstance(space, CompositeSpace): if isinstance(batch, theano.sparse.SparseVariable): warnings.warn('Formatting from a sparse VectorSpace to a ' 'CompositeSpace is currently (2 Jan 2014) a ' 'non-differentiable action. This is because it ' 'calls slicing operations on a sparse batch ' '(e.g. "my_matrix[r:R, c:C]", which Theano does ' 'not yet have a gradient operator for. If ' 'autodifferentiation is reporting an error, ' 'this may be why. Formatting batch type %s ' 'from space %s to space %s' % (type(batch), self, space)) pos = 0 pieces = [] for component in space.components: width = component.get_total_dimension() subtensor = batch[:, pos:pos + width] pos += width vector_subspace = VectorSpace(dim=width, dtype=self.dtype, sparse=self.sparse) formatted = vector_subspace._format_as(is_numeric, subtensor, component) pieces.append(formatted) result = tuple(pieces) elif isinstance(space, Conv2DSpace): if is_sparse(batch): raise TypeError("Formatting a SparseVariable to a Conv2DSpace " "is not supported, since neither scipy nor " "Theano has sparse tensors with more than 2 " "dimensions. We need 4 dimensions to " "represent a Conv2DSpace batch") dims = {'b': batch.shape[0], 'c': space.num_channels, 0: space.shape[0], 1: space.shape[1]} if space.axes != space.default_axes: # Always use default_axes, so conversions like # Conv2DSpace(c01b) -> VectorSpace -> Conv2DSpace(b01c) work shape = [dims[ax] for ax in space.default_axes] batch = _reshape(batch, shape) batch = batch.transpose(*[space.default_axes.index(ax) for ax in space.axes]) result = batch else: shape = tuple([dims[elem] for elem in space.axes]) result = _reshape(batch, shape) to_type = space.dtype elif isinstance(space, VectorSpace): if self.dim != space.dim: raise ValueError("Can't convert between VectorSpaces of " "different sizes (%d to %d)." % (self.dim, space.dim)) if space.sparse != is_sparse(batch): if space.sparse: batch = _dense_to_sparse(batch) elif isinstance(batch, theano.sparse.SparseVariable): batch = theano.sparse.dense_from_sparse(batch) elif scipy.sparse.issparse(batch): batch = batch.todense() else: assert False, ("Unplanned-for branch in if-elif-elif " "chain. This is a bug in the code.") result = batch to_type = space.dtype else: raise NotImplementedError("%s doesn't know how to format as %s" % (self, space)) return _cast(result, dtype=to_type) @functools.wraps(Space._undo_format_as_impl) def _undo_format_as_impl(self, batch, space): def is_sparse(batch): return isinstance(batch, theano.sparse.SparseVariable) if not isinstance(space, IndexSpace): my_dimension = self.get_total_dimension() other_dimension = space.get_total_dimension() if my_dimension != other_dimension: raise ValueError(str(self) + " with total dimension " + str(my_dimension) + " can't undo format a batch from " + str(space) + "because its total dimension is " + str(other_dimension)) if isinstance(space, CompositeSpace): if isinstance(batch, theano.sparse.SparseVariable): warnings.warn('Undo formatting from a sparse VectorSpace to a ' 'CompositeSpace is currently (2 Jan 2014) a ' 'non-differentiable action. This is because it ' 'calls slicing operations on a sparse batch ' '(e.g. "my_matrix[r:R, c:C]", which Theano does ' 'not yet have a gradient operator for. If ' 'autodifferentiation is reporting an error, ' 'this may be why. Formatting batch type %s ' 'from space %s to space %s' % (type(batch), self, space)) # Recursively try and find a non-Composite, non-Null space # to extract underlying theano variable def extract_vector_variable(composite_space, batch_tuple): found = False for sp, el in safe_zip(composite_space.components, batch_tuple): dim = sp.get_total_dimension() if not isinstance(sp, NullSpace) and dim > 0: if isinstance(sp, CompositeSpace): var, found = extract_vector_variable(sp, el) var = var.owner.inputs[0] else: dummy_sp = VectorSpace(dim=dim, sparse=sp.sparse, dtype=sp.dtype ) var = dummy_sp.undo_format_as(el, sp) found = True if found: break return var, found var, found = extract_vector_variable(space, batch) batch = var if not found: raise TypeError("Could not find a valid space " "to undo format from in the " "CompositeSpace.") else: # Undo subtensor slice owner = batch.owner assert 'Subtensor' in str(owner.op) batch = owner.inputs[0] elif isinstance(space, Conv2DSpace): if is_sparse(batch): raise TypeError("Undo formatting a SparseVariable to a " "Conv2DSpace is not supported, since " "neither scipy nor Theano has sparse " "tensors with more than 2 dimensions. " "We need 4 dimensions to represent a " "Conv2DSpace batch") # Check for cast batch = _undo_op(batch, 'Cast') # Undo axes shuffle if space.axes != space.default_axes: batch = _undo_op(batch, 'DimShuffle', strict=True) # Undo reshape batch = _undo_op(batch, 'Reshape{4}', strict=True) elif isinstance(space, VectorSpace): if self.dim != space.dim: raise ValueError("Can't convert between VectorSpaces of " "different sizes (%d to %d)." % (self.dim, space.dim)) # Check for cast batch = _undo_op(batch, 'Cast') # Undo any sparse-dense switches if self.sparse != is_sparse(batch): if space.sparse: batch = _undo_op(batch, 'SparseFromDense', strict=True) elif isinstance(batch, theano.sparse.SparseVariable): batch = _undo_op(batch, 'DenseFromSparse', strict=True) else: assert False, ("Unplanned-for branch in if-elif " "chain. This is a bug in the code.") else: raise NotImplementedError("%s doesn't know how to format as %s" % (self, space)) return batch def __eq__(self, other): """ .. todo:: WRITEME """ return (type(self) == type(other) and self.dim == other.dim and self.sparse == other.sparse and self.dtype == other.dtype) def __hash__(self): """ .. todo:: WRITEME """ return hash((type(self), self.dim, self.sparse, self.dtype)) @functools.wraps(Space._validate_impl) def _validate_impl(self, is_numeric, batch): """ .. todo:: WRITEME """ # checks that batch isn't a tuple, checks batch.type against self.dtype super(VectorSpace, self)._validate_impl(is_numeric, batch) if isinstance(batch, theano.gof.Variable): if self.sparse: if not isinstance(batch.type, theano.sparse.SparseType): raise TypeError('This VectorSpace is%s sparse, but the ' 'provided batch is not. (batch type: "%s")' % ('' if self.sparse else ' not', type(batch))) elif not isinstance(batch.type, (theano.tensor.TensorType, CudaNdarrayType)): raise TypeError("VectorSpace batch should be TensorType or " "CudaNdarrayType, got " + str(batch.type)) if batch.ndim != 2: raise ValueError('VectorSpace batches must be 2D, got %d ' 'dimensions' % batch.ndim) for val in get_debug_values(batch): self.np_validate(val) # sic; val is numeric, not symbolic else: # Use the 'CudaNdarray' string to avoid importing # theano.sandbox.cuda when it is not available if (not self.sparse and not isinstance(batch, np.ndarray) and type(batch) != 'CudaNdarray'): raise TypeError("The value of a VectorSpace batch should be a " "numpy.ndarray, or CudaNdarray, but is %s." % str(type(batch))) if self.sparse: if not theano.sparse.enable_sparse: raise TypeError("theano.sparse is not enabled, cannot " "have a value for a sparse VectorSpace.") if not scipy.sparse.issparse(batch): raise TypeError("The value of a sparse VectorSpace batch " "should be a sparse scipy matrix, got %s " "of type %s." % (batch, type(batch))) if batch.ndim != 2: raise ValueError("The value of a VectorSpace batch must be " "2D, got %d dimensions for %s." % (batch.ndim, batch)) if batch.shape[1] != self.dim: raise ValueError("The width of a VectorSpace batch must match " "with the space's dimension, but batch has " "shape %s and dim = %d." % (str(batch.shape), self.dim)) class VectorSequenceSpace(SimplyTypedSpace): """ A space representing a single, variable-length sequence of fixed-sized vectors. Parameters ---------- dim : int Vector size dtype : str, optional A numpy dtype string indicating this space's dtype. kwargs : dict Passes on to superclass constructor """ def __init__(self, dim, dtype='floatX', **kwargs): super(VectorSequenceSpace, self).__init__(dtype, **kwargs) self.dim = dim def __str__(self): """Return a string representation""" return ('%(classname)s(dim=%(dim)s, dtype=%(dtype)s)' % dict(classname=self.__class__.__name__, dim=self.dim, dtype=self.dtype)) @wraps(Space.__eq__) def __eq__(self, other): return (type(self) == type(other) and self.dim == other.dim and self.dtype == other.dtype) @wraps(Space._check_sizes) def _check_sizes(self, space): if not isinstance(space, VectorSequenceSpace): raise ValueError("Can't convert to " + str(space.__class__)) else: if space.dim != self.dim: raise ValueError("Can't convert to VectorSequenceSpace of " "dim %d" % (space.dim,)) @wraps(Space._format_as_impl) def _format_as_impl(self, is_numeric, batch, space): if isinstance(space, VectorSequenceSpace): if space.dim != self.dim: raise ValueError("The two VectorSequenceSpaces' dim " "values don't match. This should have been " "caught by " "VectorSequenceSpace._check_sizes().") return _cast(batch, space.dtype) else: raise ValueError("Can't convert %s to %s" % (self, space)) @wraps(Space.make_theano_batch) def make_theano_batch(self, name=None, dtype=None, batch_size=None): if batch_size == 1: return tensor.matrix(name=name) else: return ValueError("VectorSequenceSpace does not support batches " "of sequences.") @wraps(Space._batch_size_impl) def _batch_size_impl(self, is_numeric, batch): # Only batch size of 1 is supported return 1 @wraps(Space._validate_impl) def _validate_impl(self, is_numeric, batch): # checks that batch isn't a tuple, checks batch.type against self.dtype super(VectorSequenceSpace, self)._validate_impl(is_numeric, batch) if is_numeric: # Use the 'CudaNdarray' string to avoid importing # theano.sandbox.cuda when it is not available if not isinstance(batch, np.ndarray) \ and str(type(batch)) != "<type 'CudaNdarray'>": raise TypeError("The value of a VectorSequenceSpace batch " "should be a numpy.ndarray, or CudaNdarray, " "but is %s." % str(type(batch))) if batch.ndim != 2: raise ValueError("The value of a VectorSequenceSpace batch " "must be 2D, got %d dimensions for %s." % (batch.ndim, batch)) if batch.shape[1] != self.dim: raise ValueError("The width of a VectorSequenceSpace 'batch' " "must match with the space's window" "dimension, but batch has dim %d and " "this space's dim is %d." % (batch.shape[1], self.dim)) else: if not isinstance(batch, theano.gof.Variable): raise TypeError("VectorSequenceSpace batch should be a theano " "Variable, got " + str(type(batch))) if not isinstance(batch.type, (theano.tensor.TensorType, CudaNdarrayType)): raise TypeError("VectorSequenceSpace batch should be " "TensorType or CudaNdarrayType, got " + str(batch.type)) if batch.ndim != 2: raise ValueError("VectorSequenceSpace 'batches' must be 2D, " "got %d dimensions" % batch.ndim) for val in get_debug_values(batch): self.np_validate(val) class IndexSequenceSpace(SimplyTypedSpace): """ A space representing a single, variable-length sequence of indexes. Parameters ---------- max_labels : int The number of possible classes/labels. This means that all labels should be < max_labels. dim : int The number of indices in one element of the sequence dtype : str A numpy dtype string indicating this space's dtype. Must be an integer type e.g. int32 or int64. kwargs : dict Passes on to superclass constructor """ def __init__(self, max_labels, dim, dtype='int64', **kwargs): if 'int' not in dtype: raise ValueError("The dtype of IndexSequenceSpace must be an " "integer type") super(IndexSequenceSpace, self).__init__(dtype, **kwargs) self.max_labels = max_labels self.dim = dim self.formatter = OneHotFormatter(self.max_labels) def __str__(self): """Return a string representation""" return ('%(classname)s(dim=%(dim)s, max_labels=%(max_labels)s, ' 'dtype=%(dtype)s)') % dict(classname=self.__class__.__name__, dim=self.dim, max_labels=self.max_labels, dtype=self.dtype) def __eq__(self, other): """ .. todo:: WRITEME """ return (type(self) == type(other) and self.max_labels == other.max_labels and self.dim == other.dim and self.dtype == other.dtype) @wraps(Space._check_sizes) def _check_sizes(self, space): if isinstance(space, VectorSequenceSpace): # self.max_labels -> merged onehots # self.dim * self.max_labels -> concatenated if space.dim not in (self.max_labels, self.dim * self.max_labels): raise ValueError("Can't convert to VectorSequenceSpace of " "dim %d. Expected either " "dim=%d (merged one-hots) or %d " "(concatenated one-hots)" % (space.dim, self.max_labels, self.dim * self.max_labels)) elif isinstance(space, IndexSequenceSpace): if space.dim != self.dim or space.max_labels != self.max_labels: raise ValueError("Can't convert to IndexSequenceSpace of " "dim %d and max_labels %d." % (space.dim, self.max_labels)) else: raise ValueError("Can't convert to " + str(space.__class__)) @wraps(Space._format_as_impl) def _format_as_impl(self, is_numeric, batch, space): if isinstance(space, VectorSequenceSpace): if self.max_labels == space.dim: mode = 'merge' elif self.dim * self.max_labels == space.dim: mode = 'concatenate' else: raise ValueError("There is a bug. Couldn't format to a " "VectorSequenceSpace because it had an " "incorrect size, but this should've been " "caught in " "IndexSequenceSpace._check_sizes().") format_func = (self.formatter.format if is_numeric else self.formatter.theano_expr) return _cast(format_func(batch, mode=mode), space.dtype) elif isinstance(space, IndexSequenceSpace): if space.dim != self.dim or space.max_labels != self.max_labels: raise ValueError("The two IndexSequenceSpaces' dim and " "max_labels values don't match. This should " "have been caught by " "IndexSequenceSpace._check_sizes().") return _cast(batch, space.dtype) else: raise ValueError("Can't convert %s to %s" % (self, space)) @wraps(Space.make_theano_batch) def make_theano_batch(self, name=None, dtype=None, batch_size=None): if batch_size == 1: return tensor.matrix(name=name) else: return ValueError("IndexSequenceSpace does not support batches " "of sequences.") @wraps(Space._batch_size_impl) def _batch_size_impl(self, is_numeric, batch): # Only batch size of 1 is supported return 1 @wraps(Space._validate_impl) def _validate_impl(self, is_numeric, batch): # checks that batch isn't a tuple, checks batch.type against self.dtype super(IndexSequenceSpace, self)._validate_impl(is_numeric, batch) if is_numeric: # Use the 'CudaNdarray' string to avoid importing # theano.sandbox.cuda when it is not available if not isinstance(batch, np.ndarray) \ and str(type(batch)) != "<type 'CudaNdarray'>": raise TypeError("The value of a IndexSequenceSpace batch " "should be a numpy.ndarray, or CudaNdarray, " "but is %s." % str(type(batch))) if batch.ndim != 2: raise ValueError("The value of a IndexSequenceSpace batch " "must be 2D, got %d dimensions for %s." % (batch.ndim, batch)) if batch.shape[1] != self.dim: raise ValueError("The width of a IndexSequenceSpace batch " "must match with the space's dimension, but " "batch has shape %s and dim = %d." % (str(batch.shape), self.dim)) else: if not isinstance(batch, theano.gof.Variable): raise TypeError("IndexSequenceSpace batch should be a theano " "Variable, got " + str(type(batch))) if not isinstance(batch.type, (theano.tensor.TensorType, CudaNdarrayType)): raise TypeError("IndexSequenceSpace batch should be " "TensorType or CudaNdarrayType, got " + str(batch.type)) if batch.ndim != 2: raise ValueError('IndexSequenceSpace batches must be 2D, got ' '%d dimensions' % batch.ndim) for val in get_debug_values(batch): self.np_validate(val) class Conv2DSpace(SimplyTypedSpace): """ A space whose points are 3-D tensors representing (potentially multi-channel) images. Parameters ---------- shape : sequence, length 2 The shape of a single image, i.e. (rows, cols). num_channels : int (synonym: channels) Number of channels in the image, i.e. 3 if RGB. axes : tuple A tuple indicating the semantics of each axis, containing the following elements in some order: - 'b' : this axis is the batch index of a minibatch. - 'c' : this axis the channel index of a minibatch. - 0 : topological axis 0 (rows) - 1 : topological axis 1 (columns) For example, a PIL image has axes (0, 1, 'c') or (0, 1). The pylearn2 image displaying functionality uses ('b', 0, 1, 'c') for batches and (0, 1, 'c') for images. theano's conv2d operator uses ('b', 'c', 0, 1) images. dtype : str A numpy dtype string (e.g. 'float32') indicating this space's dtype, or None for a dtype-agnostic space. kwargs : dict Passed on to superclass constructor """ # Assume pylearn2's get_topological_view format, since this is how # data is currently served up. If we make better iterators change # default to ('b', 'c', 0, 1) for theano conv2d default_axes = ('b', 0, 1, 'c') def __init__(self, shape, channels=None, num_channels=None, axes=None, dtype='floatX', **kwargs): super(Conv2DSpace, self).__init__(dtype, **kwargs) assert (channels is None) + (num_channels is None) == 1 if num_channels is None: num_channels = channels assert isinstance(num_channels, py_integer_types) if not hasattr(shape, '__len__'): raise ValueError("shape argument for Conv2DSpace must have a " "length. Got %s." % str(shape)) if len(shape) != 2: raise ValueError("shape argument to Conv2DSpace must be length 2, " "not %d" % len(shape)) assert all(isinstance(elem, py_integer_types) for elem in shape) assert all(elem > 0 for elem in shape) assert isinstance(num_channels, py_integer_types) assert num_channels > 0 # Converts shape to a tuple, so it can be hashable, and self can be too self.shape = tuple(shape) self.num_channels = num_channels if axes is None: axes = self.default_axes assert len(axes) == 4 self.axes = tuple(axes) def __str__(self): """ .. todo:: WRITEME """ return ("%s(shape=%s, num_channels=%d, axes=%s, dtype=%s)" % (self.__class__.__name__, str(self.shape), self.num_channels, str(self.axes), self.dtype)) def __eq__(self, other): """ .. todo:: WRITEME """ assert isinstance(self.axes, tuple) if isinstance(other, Conv2DSpace): assert isinstance(other.axes, tuple) return (type(self) == type(other) and self.shape == other.shape and self.num_channels == other.num_channels and self.axes == other.axes and self.dtype == other.dtype) def __hash__(self): """ .. todo:: WRITEME """ return hash((type(self), self.shape, self.num_channels, self.axes, self.dtype)) @functools.wraps(Space.get_batch_axis) def get_batch_axis(self): return self.axes.index('b') @functools.wraps(Space.get_origin) def get_origin(self): dims = {0: self.shape[0], 1: self.shape[1], 'c': self.num_channels} shape = [dims[elem] for elem in self.axes if elem != 'b'] return np.zeros(shape, dtype=self.dtype) @functools.wraps(Space.get_origin_batch) def get_origin_batch(self, batch_size, dtype=None): dtype = self._clean_dtype_arg(dtype) if not isinstance(batch_size, py_integer_types): raise TypeError("Conv2DSpace.get_origin_batch expects an int, " "got %s of type %s" % (str(batch_size), type(batch_size))) assert batch_size > 0 dims = {'b': batch_size, 0: self.shape[0], 1: self.shape[1], 'c': self.num_channels} shape = [dims[elem] for elem in self.axes] return np.zeros(shape, dtype=dtype) @functools.wraps(Space.make_theano_batch) def make_theano_batch(self, name=None, dtype=None, batch_size=None): dtype = self._clean_dtype_arg(dtype) broadcastable = [False] * 4 broadcastable[self.axes.index('c')] = (self.num_channels == 1) broadcastable[self.axes.index('b')] = (batch_size == 1) broadcastable = tuple(broadcastable) rval = TensorType(dtype=dtype, broadcastable=broadcastable )(name=name) if theano.config.compute_test_value != 'off': if batch_size == 1: n = 1 else: # TODO: try to extract constant scalar value from batch_size n = 4 rval.tag.test_value = self.get_origin_batch(batch_size=n, dtype=dtype) return rval @functools.wraps(Space._batch_size_impl) def _batch_size_impl(self, is_numeric, batch): return batch.shape[self.axes.index('b')] @staticmethod def convert(tensor, src_axes, dst_axes): """ Returns a view of tensor using the axis semantics defined by dst_axes. (If src_axes matches dst_axes, returns tensor itself) Useful for transferring tensors between different Conv2DSpaces. Parameters ---------- tensor : tensor_like A 4-tensor representing a batch of images src_axes : WRITEME Axis semantics of tensor dst_axes : WRITEME WRITEME """ src_axes = tuple(src_axes) dst_axes = tuple(dst_axes) assert len(src_axes) == 4 assert len(dst_axes) == 4 if src_axes == dst_axes: return tensor shuffle = [src_axes.index(elem) for elem in dst_axes] if is_symbolic_batch(tensor): return tensor.dimshuffle(*shuffle) else: return tensor.transpose(*shuffle) @staticmethod def convert_numpy(tensor, src_axes, dst_axes): """ .. todo:: WRITEME """ return Conv2DSpace.convert(tensor, src_axes, dst_axes) @functools.wraps(Space.get_total_dimension) def get_total_dimension(self): # Patch old pickle files if not hasattr(self, 'num_channels'): self.num_channels = self.nchannels return self.shape[0] * self.shape[1] * self.num_channels @functools.wraps(Space._validate_impl) def _validate_impl(self, is_numeric, batch): # checks batch.type against self.dtype super(Conv2DSpace, self)._validate_impl(is_numeric, batch) if not is_numeric: if isinstance(batch, theano.sparse.SparseVariable): raise TypeError("Conv2DSpace cannot use SparseVariables, " "since as of this writing (28 Dec 2013), " "there is not yet a SparseVariable type with " "4 dimensions") if not isinstance(batch, theano.gof.Variable): raise TypeError("Conv2DSpace batches must be theano " "Variables, got " + str(type(batch))) if not isinstance(batch.type, (theano.tensor.TensorType, CudaNdarrayType)): raise TypeError('Expected TensorType or CudaNdArrayType, got ' '"%s"' % type(batch.type)) if batch.ndim != 4: raise ValueError("The value of a Conv2DSpace batch must be " "4D, got %d dimensions for %s." % (batch.ndim, batch)) for val in get_debug_values(batch): self.np_validate(val) else: if scipy.sparse.issparse(batch): raise TypeError("Conv2DSpace cannot use sparse batches, since " "scipy.sparse does not support 4 dimensional " "tensors currently (28 Dec 2013).") if (not isinstance(batch, np.ndarray)) \ and type(batch) != 'CudaNdarray': raise TypeError("The value of a Conv2DSpace batch should be a " "numpy.ndarray, or CudaNdarray, but is %s." % str(type(batch))) if batch.ndim != 4: raise ValueError("The value of a Conv2DSpace batch must be " "4D, got %d dimensions for %s." % (batch.ndim, batch)) d = self.axes.index('c') actual_channels = batch.shape[d] if actual_channels != self.num_channels: raise ValueError("Expected axis %d to be number of channels " "(%d) but it is %d" % (d, self.num_channels, actual_channels)) assert batch.shape[self.axes.index('c')] == self.num_channels for coord in [0, 1]: d = self.axes.index(coord) actual_shape = batch.shape[d] expected_shape = self.shape[coord] if actual_shape != expected_shape: raise ValueError("Conv2DSpace with shape %s and axes %s " "expected dimension %s of a batch (%s) " "to have length %s but it has %s" % (str(self.shape), str(self.axes), str(d), str(batch), str(expected_shape), str(actual_shape))) @functools.wraps(Space._format_as_impl) def _format_as_impl(self, is_numeric, batch, space): if isinstance(space, VectorSpace): # We need to ensure that the resulting batch will always be # the same in `space`, no matter what the axes of `self` are. if self.axes != self.default_axes: # The batch index goes on the first axis assert self.default_axes[0] == 'b' batch = batch.transpose(*[self.axes.index(axis) for axis in self.default_axes]) result = batch.reshape((batch.shape[0], self.get_total_dimension())) if space.sparse: result = _dense_to_sparse(result) elif isinstance(space, Conv2DSpace): result = Conv2DSpace.convert(batch, self.axes, space.axes) else: raise NotImplementedError("%s doesn't know how to format as %s" % (str(self), str(space))) return _cast(result, space.dtype) @functools.wraps(Space._undo_format_as_impl) def _undo_format_as_impl(self, batch, space): # Check for cast batch = _undo_op(batch, 'Cast') if isinstance(space, VectorSpace): # Check for SparseFromDense batch = _undo_op(batch, 'SparseFromDense') # Undo reshape op batch = _undo_op(batch, 'Reshape', strict=True) # Check to see if axis ordering was changed if self.axes != self.default_axes: batch = _undo_op(batch, 'DimShuffle', strict=True) elif isinstance(space, Conv2DSpace): # Check to see if axis ordering was changed if space.axes != self.axes: batch = _undo_op(batch, 'DimShuffle', strict=True) else: raise NotImplementedError("%s doesn't know how to format as %s" % (str(self), str(space))) return batch class CompositeSpace(Space): """ A Space whose points are tuples of points in other spaces. May be nested, in which case the points are nested tuples. Parameters ---------- components : WRITEME kwargs : dict WRITEME """ def __init__(self, components, **kwargs): super(CompositeSpace, self).__init__(**kwargs) assert isinstance(components, (list, tuple)) for i, component in enumerate(components): if not isinstance(component, Space): raise TypeError("component %d is %s of type %s, expected " "Space instance. " % (i, str(component), str(type(component)))) self.components = list(components) def __eq__(self, other): """ .. todo:: WRITEME """ return (type(self) == type(other) and len(self.components) == len(other.components) and all(my_component == other_component for my_component, other_component in zip(self.components, other.components))) def __hash__(self): """ .. todo:: WRITEME """ return hash((type(self), tuple(self.components))) def __str__(self): """ .. todo:: WRITEME """ return '%(classname)s(%(components)s)' % \ dict(classname=self.__class__.__name__, components=', '.join([str(c) for c in self.components])) @property def dtype(self): """ Returns a nested tuple of dtype strings. NullSpaces will yield a bogus dtype string (see NullSpace.dtype). """ def get_dtype_of_space(space): if isinstance(space, CompositeSpace): return tuple(get_dtype_of_space(c) for c in space.components) elif isinstance(space, NullSpace): return NullSpace().dtype else: return space.dtype return get_dtype_of_space(self) @dtype.setter def dtype(self, new_dtype): """ If new_dtype is None or a string, it will be applied to all components (except any NullSpaces). If new_dtype is a (nested) tuple, its elements will be applied to corresponding components. """ if isinstance(new_dtype, tuple): for component, new_dt in safe_zip(self.components, new_dtype): component.dtype = new_dt elif new_dtype is None or isinstance(new_dtype, str): for component in self.components: if not isinstance(component, NullSpace): component.dtype = new_dtype def restrict(self, subset): """ Returns a new Space containing only the components whose indices are given in subset. The new space will contain the components in the order given in the subset list. Parameters ---------- subset : WRITEME Notes ----- The returned Space may not be a CompositeSpace if `subset` contains only one index. """ assert isinstance(subset, (list, tuple)) if len(subset) == 1: idx, = subset return self.components[idx] return CompositeSpace([self.components[i] for i in subset]) def restrict_batch(self, batch, subset): """ Returns a batch containing only the components whose indices are present in subset. May not be a tuple anymore if there is only one index. Outputs will be ordered in the order that they appear in subset. Only supports symbolic batches. Parameters ---------- batch : WRITEME subset : WRITEME """ self._validate(is_numeric=False, batch=batch) assert isinstance(subset, (list, tuple)) if len(subset) == 1: idx, = subset return batch[idx] return tuple([batch[i] for i in subset]) @functools.wraps(Space.get_total_dimension) def get_total_dimension(self): return sum([component.get_total_dimension() for component in self.components]) @functools.wraps(Space.make_shared_batch) def make_shared_batch(self, batch_size, name=None, dtype=None): dtype = self._clean_dtype_arg(dtype) batch = self.get_origin_batch(batch_size, dtype) def recursive_shared(batch): if isinstance(batch, tuple): return tuple(recursive_shared(b) for b in batch) else: return theano.shared(batch, name=name) return recursive_shared(batch) @functools.wraps(Space._format_as_impl) def _format_as_impl(self, is_numeric, batch, space): """ Supports formatting to a single VectorSpace, or to a CompositeSpace. CompositeSpace->VectorSpace: Traverses the nested components in depth-first order, serializing the leaf nodes (i.e. the non-composite subspaces) into the VectorSpace. CompositeSpace->CompositeSpace: Only works for two CompositeSpaces that have the same nested structure. Traverses both CompositeSpaces' nested components in parallel, converting between corresponding non-composite components in <self> and <space> as: `self_component._format_as(is_numeric, batch_component, space_component)` Parameters ---------- batch : WRITEME space : WRITEME Returns ------- WRITEME """ if isinstance(space, VectorSpace): pieces = [] for component, input_piece in zip(self.components, batch): subspace = VectorSpace(dim=component.get_total_dimension(), dtype=space.dtype, sparse=space.sparse) pieces.append(component._format_as(is_numeric, input_piece, subspace)) # Pieces should all have the same dtype, before we concatenate them if len(pieces) > 0: for piece in pieces[1:]: if pieces[0].dtype != piece.dtype: assert space.dtype is None raise TypeError("Tried to format components with " "differing dtypes into a VectorSpace " "with no dtype of its own. " "dtypes: %s" % str(tuple(str(p.dtype) for p in pieces))) if is_symbolic_batch(batch): if space.sparse: return theano.sparse.hstack(pieces) else: return tensor.concatenate(pieces, axis=1) else: if space.sparse: return scipy.sparse.hstack(pieces) else: return np.concatenate(pieces, axis=1) if isinstance(space, CompositeSpace): def recursive_format_as(orig_space, batch, dest_space): if not (isinstance(orig_space, CompositeSpace) == isinstance(dest_space, CompositeSpace)): raise TypeError("Can't convert between CompositeSpaces " "with different tree structures") # No need to check batch's tree structure. Space._format_as() # already did that by calling _validate(), before calling this # method. if isinstance(orig_space, CompositeSpace): return tuple(recursive_format_as(os, bt, ds) for os, bt, ds in safe_zip(orig_space.components, batch, dest_space.components)) else: return orig_space._format_as(is_numeric, batch, dest_space) return recursive_format_as(self, batch, space) raise NotImplementedError(str(self) + " does not know how to format as " + str(space)) @functools.wraps(Space._undo_format_as_impl) def _undo_format_as_impl(self, batch, space): """ Undoes the formatting to a single VectorSpace, or to a CompositeSpace. CompositeSpace->VectorSpace: Traverses the nested components in depth-first order, serializing the leaf nodes (i.e. the non-composite subspaces) into the VectorSpace. CompositeSpace->CompositeSpace: Only works for two CompositeSpaces that have the same nested structure. Traverses both CompositeSpaces' nested components in parallel, converting between corresponding non-composite components in <self> and <space> as: `self_component._format_as(is_numeric, batch_component, space_component)` Parameters ---------- batch : WRITEME space : WRITEME Returns ------- WRITEME """ if isinstance(space, VectorSpace): # Undo join if space.sparse: owner = batch.owner assert owner is not None assert 'HStack' in str(owner.op) batch = owner.inputs else: owner = batch.owner assert owner is not None assert str(owner.op) == 'Join' # First component is join axis batch = owner.inputs[1:] def extract_dtype(dtype): if isinstance(dtype, tuple): return extract_dtype(dtype[0]) else: return dtype def compose_batch(composite_space, batch_list): rval = () for sp, bt in safe_zip(composite_space.components, batch_list): if False and isinstance(sp, CompositeSpace): composed, batch_list = compose_batch(sp, batch_list) rval += (composed,) else: sparse = getattr(sp, 'sparse', False) dtype = extract_dtype(sp.dtype) new_sp = VectorSpace(dim=sp.get_total_dimension(), dtype=dtype, sparse=sparse ) new_batch = sp.undo_format_as(bt, new_sp) rval += (new_batch,) return rval composed = compose_batch(self, batch) return composed if isinstance(space, CompositeSpace): def recursive_undo_format_as(orig_space, batch, dest_space): if not (isinstance(orig_space, CompositeSpace) == isinstance(dest_space, CompositeSpace)): raise TypeError("Can't convert between CompositeSpaces " "with different tree structures") # No need to check batch's tree structure. # Space.undo_format_as() already did that # by calling _validate(), before calling this # method. if isinstance(orig_space, CompositeSpace): return tuple(recursive_undo_format_as(os, bt, ds) for os, bt, ds in safe_zip(orig_space.components, batch, dest_space.components)) else: return orig_space.undo_format_as(batch, dest_space) return recursive_undo_format_as(self, batch, space) raise NotImplementedError(str(self) + " does not know how to format as " + str(space)) @functools.wraps(Space._validate_impl) def _validate_impl(self, is_numeric, batch): if not isinstance(batch, tuple): raise TypeError("The value of a CompositeSpace batch should be a " "tuple, but is %s of type %s." % (batch, type(batch))) if len(batch) != len(self.components): raise ValueError("Expected %d elements in batch, got %d" % (len(self.components), len(batch))) for batch_elem, component in zip(batch, self.components): component._validate(is_numeric, batch_elem) def get_origin_batch(self, batch_size, dtype=None): """ Calls get_origin_batch on all subspaces, and returns a (nested) tuple containing their return values. Parameters ---------- batch_size : int Batch size. dtype : str the dtype to use for all the get_origin_batch() calls on subspaces. If dtype is None, or a single dtype string, that will be used for all calls. If dtype is a (nested) tuple, it must mirror the tree structure of this CompositeSpace. """ dtype = self._clean_dtype_arg(dtype) return tuple(component.get_origin_batch(batch_size, dt) for component, dt in safe_zip(self.components, dtype)) @functools.wraps(Space.make_theano_batch) def make_theano_batch(self, name=None, dtype=None, batch_size=None): """ Calls make_theano_batch on all subspaces, and returns a (nested) tuple containing their return values. Parameters ---------- name : str Name of the symbolic variable dtype : str The dtype of the returned batch. If dtype is a string, it will be applied to all components. If dtype is None, C.dtype will be used for each component C. If dtype is a nested tuple, its elements will be applied to corresponding elements in the components. batch_size : int Batch size. """ if name is None: name = [None] * len(self.components) elif not isinstance(name, (list, tuple)): name = ['%s[%i]' % (name, i) for i in xrange(len(self.components))] dtype = self._clean_dtype_arg(dtype) assert isinstance(name, (list, tuple)) assert isinstance(dtype, (list, tuple)) rval = tuple([x.make_theano_batch(name=n, dtype=d, batch_size=batch_size) for x, n, d in safe_zip(self.components, name, dtype)]) return rval @functools.wraps(Space._batch_size_impl) def _batch_size_impl(self, is_numeric, batch): def has_no_data(space): """ Returns True if space can contain no data. """ return (isinstance(subspace, NullSpace) or (isinstance(subspace, CompositeSpace) and len(subspace.components) == 0)) if is_symbolic_batch(batch): for subspace, subbatch in safe_zip(self.components, batch): if not has_no_data(subspace): return subspace._batch_size(is_numeric, subbatch) return 0 # TODO: shouldn't this line return a Theano object? else: result = None for subspace, subbatch in safe_zip(self.components, batch): batch_size = subspace._batch_size(is_numeric, subbatch) if has_no_data(subspace): assert batch_size == 0 else: if result is None: result = batch_size elif batch_size != result: raise ValueError("All non-empty components of a " "CompositeSpace should have the same " "batch size, but we encountered " "components with size %s, then %s." % (result, batch_size)) return 0 if result is None else result def _clean_dtype_arg(self, dtype): """ If dtype is None or a string, this returns a nested tuple that mirrors the tree structure of this CompositeSpace, with dtype at the leaves. If dtype is a nested tuple, this checks that it has the same tree structure as this CompositeSpace. """ super_self = super(CompositeSpace, self) def make_dtype_tree(dtype, space): """ Creates a nested tuple tree that mirrors the tree structure of <space>, populating the leaves with <dtype>. """ if isinstance(space, CompositeSpace): return tuple(make_dtype_tree(dtype, component) for component in space.components) else: return super_self._clean_dtype_arg(dtype) def check_dtype_tree(dtype, space): """ Verifies that a dtype tree mirrors the tree structure of <space>, calling Space._clean_dtype_arg on the leaves. """ if isinstance(space, CompositeSpace): if not isinstance(dtype, tuple): raise TypeError("Tree structure mismatch.") return tuple(check_dtype_tree(dt, c) for dt, c in safe_zip(dtype, space.components)) else: if not (dtype is None or isinstance(dtype, str)): raise TypeError("Tree structure mismatch.") return super_self._clean_dtype_arg(dtype) if dtype is None or isinstance(dtype, str): dtype = super_self._clean_dtype_arg(dtype) return make_dtype_tree(dtype, self) else: return check_dtype_tree(dtype, self) class NullSpace(Space): """ A space that contains no data. As such, it has the following quirks: * Its validate()/np_validate() methods only accept None. * Its dtype string is "Nullspace's dtype". * The source name associated to this Space is the empty string (''). """ # NullSpaces don't support validation callbacks, since they only take None # as data batches. def __init__(self): super(NullSpace, self).__init__() def __str__(self): """ .. todo:: WRITEME """ return "NullSpace" def __eq__(self, other): """ .. todo:: WRITEME """ return type(self) == type(other) def __hash__(self): """ .. todo:: WRITEME """ return hash(type(self)) @property def dtype(self): """ .. todo:: WRITEME """ return "%s's dtype" % self.__class__.__name__ @dtype.setter def dtype(self, new_dtype): """ .. todo:: WRITEME """ if new_dtype != self.dtype: raise TypeError('%s can only take the bogus dtype "%s"' % (self.__class__.__name__, self.dtype)) # otherwise, do nothing @functools.wraps(Space.make_theano_batch) def make_theano_batch(self, name=None, dtype=None): return None @functools.wraps(Space._validate_impl) def _validate_impl(self, is_numeric, batch): if batch is not None: raise TypeError('NullSpace only accepts None, as a dummy data ' 'batch. Instead, got %s of type %s' % (batch, type(batch))) @functools.wraps(Space._format_as_impl) def _format_as_impl(self, is_numeric, batch, space): assert isinstance(space, NullSpace) return None @functools.wraps(Space._batch_size_impl) def _batch_size_impl(self, is_numeric, batch): # There is no way to know how many examples would actually # have been in the batch, since it is empty. We return 0. self._validate(is_numeric, batch) return 0
bsd-3-clause
4f4914929d8a60fdf57a0d2f988e2839
36.590977
79
0.528532
4.694681
false
false
false
false
lisa-lab/pylearn2
pylearn2/packaged_dependencies/theano_linear/unshared_conv/gpu_unshared_conv.py
44
23194
""" WRITEME """ from __future__ import print_function import inspect import os import StringIO import theano from theano.sandbox.cuda import CudaNdarrayType from theano.gof import local_optimizer from theano.sandbox.cuda.opt import register_opt from theano.sandbox.cuda import gpu_from_host, host_from_gpu from theano.sandbox.cuda.basic_ops import gpu_contiguous from .unshared_conv import FilterActs from .unshared_conv import WeightActs from .unshared_conv import ImgActs _this_dir = os.path.dirname(inspect.getfile(inspect.currentframe())) # XXX: move to cuda.opt and refactor there def any_from_gpu(*vv): """ .. todo:: WRITEME """ for v in vv: if v.owner and v.owner.op == host_from_gpu: return True return False # XXX: move to cuda.opt and refactor there def any_gpu_client(*vv): """ .. todo:: WRITEME """ for v in vv: for (cl, pos) in v.clients: if cl.op == gpu_from_host: return True return False class Base(theano.Op): """ .. todo:: WRITEME Parameters ---------- module_stride : WRITEME partial_sum : WRITEME """ def __init__(self, module_stride, partial_sum): self.module_stride = module_stride self.partial_sum = partial_sum def _attributes(self): """ .. todo:: WRITEME """ return ( self.module_stride, self.partial_sum, ) def __eq__(self, other): """ .. todo:: WRITEME """ return (type(self) == type(other) and self._attributes() == other._attributes()) def __hash__(self): """ .. todo:: WRITEME """ return hash((type(self), self._attributes())) def __str__(self): """ .. todo:: WRITEME """ return '%s{module_stride=%i,partial_sum=%i}' % ( self.__class__.__name__, self.module_stride, self.partial_sum, ) class GpuFilterActs(Base): """ .. todo:: WRITEME """ def make_node(self, images, filters): """ .. todo:: WRITEME """ ibcast = images.broadcastable fbcast = filters.broadcastable igroups, icolors_per_group, irows, icols, icount = ibcast fmodulesR, fmodulesC, fcolors, frows, fcols = fbcast[:-2] fgroups, filters_per_group = fbcast[-2:] hbcast = (fgroups, filters_per_group, fmodulesR, fmodulesC, icount) if not isinstance(images.type, CudaNdarrayType): raise TypeError('gpu_filter_acts requires CudaNdarray images', images) if not isinstance(filters.type, CudaNdarrayType): raise TypeError('gpu_filter_acts requires CudaNdarray filters', filters) htype = CudaNdarrayType(broadcastable=hbcast) return theano.gof.Apply(self, [images, filters], [htype()]) def c_support_code(self): """ .. todo:: WRITEME """ cufile = open(os.path.join(_this_dir, 'filter_acts.cu')) return cufile.read() def c_code_cache_version(self): """ .. todo:: WRITEME """ return () def c_code(self, node, nodename, inputs, outputs, sub): """ .. todo:: WRITEME """ #z_out = alpha * dot(x,y) + beta * z_in #inplace version, set set z_out = z_in #not inplace version, we copy z_in to z_out. images, filters, = inputs responses, = outputs fail = sub['fail'] moduleStride = str(self.module_stride) sio = StringIO.StringIO() print(""" //XXX: actually the rightmost images dimension can be strided if (!CudaNdarray_is_c_contiguous(%(images)s)) { PyErr_Format(PyExc_NotImplementedError, "images not c contiguous"); %(fail)s; } if (!CudaNdarray_is_c_contiguous(%(filters)s)) { PyErr_Format(PyExc_NotImplementedError, "filters not c contiguous"); %(fail)s; } if (%(images)s->nd != 5) { PyErr_Format(PyExc_TypeError, "images ndim (%%i) must be 5", %(images)s->nd); %(fail)s; } if (%(filters)s->nd != 7) { PyErr_Format(PyExc_TypeError, "filters ndim (%%i) must be 7", %(filters)s->nd); %(fail)s; } //fprintf(stderr, "really running on GPU\\n"); { // new scope, new vars int igroups = CudaNdarray_HOST_DIMS(%(images)s)[0]; int icolors_per_group = CudaNdarray_HOST_DIMS(%(images)s)[1]; int irows = CudaNdarray_HOST_DIMS(%(images)s)[2]; int icols = CudaNdarray_HOST_DIMS(%(images)s)[3]; int icount = CudaNdarray_HOST_DIMS(%(images)s)[4]; int fmodulesR = CudaNdarray_HOST_DIMS(%(filters)s)[0]; int fmodulesC = CudaNdarray_HOST_DIMS(%(filters)s)[1]; int fcolors = CudaNdarray_HOST_DIMS(%(filters)s)[2]; int frows = CudaNdarray_HOST_DIMS(%(filters)s)[3]; int fcols = CudaNdarray_HOST_DIMS(%(filters)s)[4]; int fgroups = CudaNdarray_HOST_DIMS(%(filters)s)[5]; int filters_per_group = CudaNdarray_HOST_DIMS(%(filters)s)[6]; // XXX: use this parameter properly int paddingStart = 0; int imgStride = icount; float scaleTargets = 0.0; float scaleOutput = 1.0; bool conv = false; if (igroups != fgroups) { PyErr_Format(PyExc_ValueError, "igroups != fgroups (%%i != %%i)", igroups, fgroups); %(fail)s; } if (icolors_per_group != fcolors) { PyErr_Format(PyExc_ValueError, "icolors_per_group != fcolors (%%i != %%i)", icolors_per_group, fcolors); %(fail)s; } if (!%(responses)s) { Py_XDECREF(%(responses)s); int dims[5]; dims[0] = fgroups; dims[1] = filters_per_group; dims[2] = fmodulesR; dims[3] = fmodulesC; dims[4] = icount; %(responses)s = (CudaNdarray*)CudaNdarray_NewDims(5, dims); if (!%(responses)s) { %(fail)s; } } assert(CudaNdarray_is_c_contiguous(%(responses)s)); if (_filterActs( igroups, icolors_per_group, irows, icols, icount, fmodulesR, fmodulesC, frows, fcols, filters_per_group, CudaNdarray_DEV_DATA(%(images)s), CudaNdarray_DEV_DATA(%(filters)s), CudaNdarray_DEV_DATA(%(responses)s), paddingStart, %(moduleStride)s, imgStride, scaleTargets, scaleOutput, conv)) { %(fail)s; } } // end bogus scope used for vars """, file=sio) return sio.getvalue() % locals() @register_opt() @local_optimizer([FilterActs]) def insert_gpu_filter_acts(node): """ .. todo:: WRITEME """ if isinstance(node.op, FilterActs): images, filters = node.inputs if any_from_gpu(images, filters) or any_gpu_client(*node.outputs): gpu_filter_acts = GpuFilterActs( module_stride=node.op.module_stride, partial_sum=1) return [host_from_gpu(gpu_filter_acts( gpu_from_host(images), gpu_from_host(filters)))] class GpuWeightActs(Base): """ .. todo:: WRITEME """ def make_node(self, images, hidacts, frows, fcols): """ .. todo:: WRITEME """ if self.partial_sum != 1: # this corresponds to grad when doing convolution raise NotImplementedError('partial sum') frows = theano.tensor.as_tensor_variable(frows) fcols = theano.tensor.as_tensor_variable(fcols) if frows.dtype[:3] not in ('int', 'uin'): raise TypeError(frows) if fcols.dtype[:3] not in ('int', 'uin'): raise TypeError(frows) if frows.ndim: raise TypeError('frows should be scalar', frows) if fcols.ndim: raise TypeError('fcols should be scalar', fcols) igroups, icolors, irows, icols, icount = images.type.broadcastable hgroups, hcolors, hrows, hcols, hcount = hidacts.type.broadcastable otype = theano.sandbox.cuda.CudaNdarrayType( broadcastable=(hrows, hcols, icolors, False, False, hgroups, hcolors)) return theano.Apply(self, [images, hidacts, frows, fcols], [otype()]) def c_support_code(self): """ .. todo:: WRITEME """ cufile = open(os.path.join(_this_dir, 'weight_acts.cu')) return cufile.read() def c_code_cache_version(self): """ .. todo:: WRITEME """ return () def c_code(self, node, nodename, inames, onames, sub): """ .. todo:: WRITEME """ images, hidacts, frows, fcols = inames dweights, = onames fail = sub['fail'] moduleStride = str(self.module_stride) sio = StringIO.StringIO() print(""" if (!CudaNdarray_is_c_contiguous(%(images)s)) { //XXX: Alex's code actually supports the rightmost images // dimension strided PyErr_Format(PyExc_NotImplementedError, "images not c contiguous"); %(fail)s; } if (!CudaNdarray_is_c_contiguous(%(hidacts)s)) { PyErr_Format(PyExc_NotImplementedError, "hidacts not c contiguous"); %(fail)s; } if (%(images)s->nd != 5) { PyErr_Format(PyExc_TypeError, "images ndim (%%i) must be 5", %(images)s->nd); %(fail)s; } if (%(hidacts)s->nd != 5) { PyErr_Format(PyExc_TypeError, "hidacts ndim (%%i) must be 5", %(images)s->nd); %(fail)s; } if (PyArray_NDIM(%(frows)s) != 0) { PyErr_Format(PyExc_TypeError, "frows ndim (%%i) must be 0", PyArray_NDIM(%(frows)s)); %(fail)s; } if (PyArray_NDIM(%(fcols)s) != 0) { PyErr_Format(PyExc_TypeError, "fcols ndim (%%i) must be 0", PyArray_NDIM(%(fcols)s)); %(fail)s; } { // new scope, new vars int igroups = CudaNdarray_HOST_DIMS(%(images)s)[0]; int icolors_per_group = CudaNdarray_HOST_DIMS(%(images)s)[1]; int irows = CudaNdarray_HOST_DIMS(%(images)s)[2]; int icols = CudaNdarray_HOST_DIMS(%(images)s)[3]; int icount = CudaNdarray_HOST_DIMS(%(images)s)[4]; int hgroups = CudaNdarray_HOST_DIMS(%(hidacts)s)[0]; int hcolors_per_group = CudaNdarray_HOST_DIMS(%(hidacts)s)[1]; int hrows = CudaNdarray_HOST_DIMS(%(hidacts)s)[2]; int hcols = CudaNdarray_HOST_DIMS(%(hidacts)s)[3]; int hcount = CudaNdarray_HOST_DIMS(%(hidacts)s)[4]; int fmodulesR = hrows; int fmodulesC = hcols; int fcolors = icolors_per_group; int frows = ((dtype_%(frows)s *) PyArray_DATA(%(frows)s))[0]; int fcols = ((dtype_%(fcols)s *) PyArray_DATA(%(fcols)s))[0]; int fgroups = hgroups; int filters_per_group = hcolors_per_group; // XXX: use this parameter properly int paddingStart = 0; int imgStride = icount; float scaleTargets = 0.0; float scaleOutput = 1.0; int moduleStride = %(moduleStride)s; int partialSum = 1; // set to 0 for convolution. if (igroups != hgroups) { PyErr_Format(PyExc_ValueError, "igroups != hgroups (%%i != %%i)", igroups, hgroups); %(fail)s; } if (icolors_per_group != fcolors) { PyErr_Format(PyExc_ValueError, "icolors_per_group != fcolors (%%i != %%i)", icolors_per_group, fcolors); %(fail)s; } if (icount != hcount) { PyErr_Format(PyExc_ValueError, "icount != hcount (%%i != %%i)", icount, hcount); %(fail)s; } // XXX: CHECK SHAPE IS CORRECT if (!%(dweights)s) { Py_XDECREF(%(dweights)s); int dims[7]; dims[0] = fmodulesR; dims[1] = fmodulesC; dims[2] = fcolors; dims[3] = frows; dims[4] = fcols; dims[5] = fgroups; dims[6] = filters_per_group; %(dweights)s = (CudaNdarray*)CudaNdarray_NewDims(7, dims); if (!%(dweights)s) { %(fail)s; } } assert(CudaNdarray_is_c_contiguous(%(dweights)s)); if (_weightActs( igroups, icolors_per_group, irows, icols, icount, fmodulesR, fmodulesC, frows, fcols, filters_per_group, CudaNdarray_DEV_DATA(%(images)s), CudaNdarray_DEV_DATA(%(hidacts)s), CudaNdarray_DEV_DATA(%(dweights)s), paddingStart, moduleStride, imgStride, scaleTargets, scaleOutput, partialSum)) { %(fail)s; } } // end bogus scope used for vars """, file=sio) return sio.getvalue() % locals() @register_opt() @local_optimizer([WeightActs]) def insert_gpu_weight_acts(node): """ .. todo:: WRITEME """ if isinstance(node.op, WeightActs): """ .. todo:: WRITEME """ images, hidacts, frows, fcols = node.inputs if any_from_gpu(images, hidacts) or any_gpu_client(*node.outputs): gpu_weight_acts = GpuWeightActs( module_stride=node.op.module_stride, partial_sum=1) return [host_from_gpu(gpu_weight_acts( gpu_from_host(images), gpu_contiguous(hidacts), frows, fcols, ))] class GpuImgActs(Base): """ .. todo:: WRITEME """ def make_node(self, filters, hidacts, irows, icols): """ .. todo:: WRITEME """ irows = theano.tensor.as_tensor_variable(irows) icols = theano.tensor.as_tensor_variable(icols) if irows.dtype[:3] not in ('int', 'uin'): raise TypeError(irows) if icols.dtype[:3] not in ('int', 'uin'): raise TypeError(irows) if irows.ndim: raise TypeError('irows should be scalar', irows) if icols.ndim: raise TypeError('icols should be scalar', icols) return theano.gof.Apply(self, [filters, hidacts, irows, icols], [hidacts.type()]) def c_support_code(self): """ .. todo:: WRITEME """ cufile = open(os.path.join(_this_dir, 'raw_img_acts.cu')) return cufile.read() def c_code_cache_version(self): """ .. todo:: WRITEME """ return () def c_code(self, node, nodename, inames, onames, sub): """ .. todo:: WRITEME """ filters, hidacts, irows, icols = inames dimages, = onames fail = sub['fail'] moduleStride = str(self.module_stride) sio = StringIO.StringIO() print(""" if (!CudaNdarray_is_c_contiguous(%(filters)s)) { //XXX: Alex's code actually supports the rightmost images // dimension strided PyErr_Format(PyExc_NotImplementedError, "images not c contiguous"); %(fail)s; } if (!CudaNdarray_is_c_contiguous(%(hidacts)s)) { PyErr_Format(PyExc_NotImplementedError, "hidacts not c contiguous"); %(fail)s; } if (%(filters)s->nd != 7) { PyErr_Format(PyExc_TypeError, "images ndim (%%i) must be 7", %(filters)s->nd); %(fail)s; } if (%(hidacts)s->nd != 5) { PyErr_Format(PyExc_TypeError, "hidacts ndim (%%i) must be 5", %(hidacts)s->nd); %(fail)s; } if (PyArray_NDIM(%(irows)s) != 0) { PyErr_Format(PyExc_TypeError, "frows ndim (%%i) must be 0", PyArray_NDIM(%(irows)s)); %(fail)s; } if (PyArray_NDIM(%(icols)s) != 0) { PyErr_Format(PyExc_TypeError, "fcols ndim (%%i) must be 0", PyArray_NDIM(%(icols)s)); %(fail)s; } { // new scope, new vars int fmodulesR = CudaNdarray_HOST_DIMS(%(filters)s)[0]; int fmodulesC = CudaNdarray_HOST_DIMS(%(filters)s)[1]; int fcolors = CudaNdarray_HOST_DIMS(%(filters)s)[2]; int frows = CudaNdarray_HOST_DIMS(%(filters)s)[3]; int fcols = CudaNdarray_HOST_DIMS(%(filters)s)[4]; int fgroups = CudaNdarray_HOST_DIMS(%(filters)s)[5]; int filters_per_group = CudaNdarray_HOST_DIMS(%(filters)s)[6]; int hgroups = CudaNdarray_HOST_DIMS(%(hidacts)s)[0]; int hcolors_per_group = CudaNdarray_HOST_DIMS(%(hidacts)s)[1]; int hrows = CudaNdarray_HOST_DIMS(%(hidacts)s)[2]; int hcols = CudaNdarray_HOST_DIMS(%(hidacts)s)[3]; int hcount = CudaNdarray_HOST_DIMS(%(hidacts)s)[4]; int igroups = fgroups; int icolors_per_group = fcolors; int irows = ((dtype_%(irows)s *) PyArray_DATA(%(irows)s))[0]; int icols = ((dtype_%(icols)s *) PyArray_DATA(%(icols)s))[0]; int icount = hcount; // TODO: use this parameter properly int paddingStart = 0; float scaleTargets = 0.0; float scaleOutput = 1.0; int moduleStride = %(moduleStride)s; bool conv = 0; if (hgroups != fgroups) { PyErr_Format(PyExc_ValueError, "hgroups != fgroups (%%i != %%i)", hgroups, fgroups); %(fail)s; } if (hcolors_per_group != filters_per_group) { PyErr_Format(PyExc_ValueError, "hcolors_per_group != filters_per_group (%%i != %%i)", hcolors_per_group, filters_per_group); %(fail)s; } // XXX: CHECK SHAPE IS CORRECT if (!%(dimages)s) { Py_XDECREF(%(dimages)s); int dims[5]; dims[0] = igroups; dims[1] = icolors_per_group; dims[2] = irows; dims[3] = icols; dims[4] = icount; %(dimages)s = (CudaNdarray*)CudaNdarray_NewDims(5, dims); if (!%(dimages)s) { %(fail)s; } } assert(CudaNdarray_is_c_contiguous(%(dimages)s)); if (paddingStart + (fmodulesR - 1) * moduleStride + frows < irows) { PyErr_Format(PyExc_ValueError, "uhoh123: %%i %%i %%i %%i %%i", paddingStart, fmodulesR, moduleStride, frows, irows); %(fail)s; } if (_imgActs( fgroups, filters_per_group, fcolors, hcount, fmodulesR, fmodulesC, frows, fcols, irows, icols, CudaNdarray_DEV_DATA(%(filters)s), CudaNdarray_DEV_DATA(%(hidacts)s), CudaNdarray_DEV_DATA(%(dimages)s), paddingStart, moduleStride, scaleTargets, scaleOutput, conv)) { %(fail)s; } } // end bogus scope used for vars """, file=sio) return sio.getvalue() % locals() @register_opt() @local_optimizer([ImgActs]) def insert_gpu_img_acts(node): """ .. todo:: WRITEME """ if isinstance(node.op, ImgActs): filters, hidacts, irows, icols = node.inputs if any_from_gpu(filters, hidacts) or any_gpu_client(*node.outputs): gpu_img_acts = GpuImgActs( module_stride=node.op.module_stride, partial_sum=1) return [host_from_gpu(gpu_img_acts( gpu_from_host(filters), gpu_contiguous(hidacts), irows, icols, ))]
bsd-3-clause
9eb51f198fab1c73c61cd311e77f85e8
28.248424
85
0.454902
4.087769
false
false
false
false
lisa-lab/pylearn2
pylearn2/datasets/cos_dataset.py
45
3759
""" .. todo:: WRITEME """ import numpy as N import copy from theano import config import theano.tensor as T from pylearn2.utils.rng import make_np_rng class CosDataset(object): """ Makes a dataset that streams randomly generated 2D examples. The first coordinate is sampled from a uniform distribution. The second coordinate is the cosine of the first coordinate, plus some gaussian noise. """ def __init__(self, min_x=-6.28, max_x=6.28, std=.05, rng=None): """ .. todo:: WRITEME """ self.min_x, self.max_x, self.std = min_x, max_x, std rng = make_np_rng(rng, [17, 2, 946], which_method=['uniform', 'randn']) self.default_rng = copy.copy(rng) self.rng = rng def energy(self, mat): """ .. todo:: WRITEME """ x = mat[:, 0] y = mat[:, 1] rval = (y - N.cos(x)) ** 2. / (2. * (self.std ** 2.)) return rval def pdf_func(self, mat): """ .. todo:: WRITEME properly This dataset can generate an infinite amount of examples. This function gives the pdf from which the examples are drawn. """ x = mat[:, 0] y = mat[:, 1] rval = N.exp(-(y - N.cos(x)) ** 2. / (2. * (self.std ** 2.))) rval /= N.sqrt(2.0 * N.pi * (self.std ** 2.)) rval /= (self.max_x - self.min_x) rval *= x < self.max_x rval *= x > self.min_x return rval def free_energy(self, X): """ .. todo:: WRITEME properly This dataset can generate an infinite amount of examples. This function gives the energy function for the distribution from which the examples are drawn. """ x = X[:, 0] y = X[:, 1] rval = T.sqr(y - T.cos(x)) / (2. * (self.std ** 2.)) mask = x < self.max_x mask = mask * (x > self.min_x) rval = mask * rval + (1 - mask) * 1e30 return rval def pdf(self, X): """ .. todo:: WRITEME properly This dataset can generate an infinite amount of examples. This function gives the pdf from which the examples are drawn. """ x = X[:, 0] y = X[:, 1] rval = T.exp(-T.sqr(y - T.cos(x)) / (2. * (self.std ** 2.))) rval /= N.sqrt(2.0 * N.pi * (self.std ** 2.)) rval /= (self.max_x - self.min_x) rval *= x < self.max_x rval *= x > self.min_x return rval def get_stream_position(self): """ .. todo:: WRITEME """ return copy.copy(self.rng) def set_stream_position(self, s): """ .. todo:: WRITEME """ self.rng = copy.copy(s) def restart_stream(self): """ .. todo:: WRITEME """ self.reset_RNG() def reset_RNG(self): """ .. todo:: WRITEME """ if 'default_rng' not in dir(self): self.default_rng = N.random.RandomState([17, 2, 946]) self.rng = copy.copy(self.default_rng) def apply_preprocessor(self, preprocessor, can_fit=False): """ .. todo:: WRITEME """ raise NotImplementedError() def get_batch_design(self, batch_size): """ .. todo:: WRITEME """ x = N.cast[config.floatX](self.rng.uniform(self.min_x, self.max_x, (batch_size, 1))) y = N.cos(x) + (N.cast[config.floatX](self.rng.randn(*x.shape)) * self.std) rval = N.hstack((x, y)) return rval
bsd-3-clause
b476ad697dd5c8e64f01fed8fd09de53
23.89404
79
0.478053
3.725471
false
false
false
false
lisa-lab/pylearn2
pylearn2/distributions/uniform_hypersphere.py
49
2116
""" .. todo:: WRITEME """ uthors__ = "Ian Goodfellow" __copyright__ = "Copyright 2010-2012, Universite de Montreal" __credits__ = ["Ian Goodfellow"] __license__ = "3-clause BSD" __maintainer__ = "LISA Lab" __email__ = "pylearn-dev@googlegroups" import numpy as N import theano.tensor as T from theano import config from scipy.special import gammaln from pylearn2.utils.rng import make_theano_rng class UniformHypersphere(object): """ .. todo:: WRITEME """ def __init__(self, dim, radius): self.dim = dim self.radius = radius self.s_rng = make_theano_rng(None, 42, which_method='normal') log_C = ((float(self.dim) / 2.) * N.log(N.pi) - gammaln(1. + float(self.dim) / 2.)) self.logZ = N.log(self.dim) + log_C + (self.dim - 1) * N.log(radius) assert not N.isnan(self.logZ) assert not N.isinf(self.logZ) def free_energy(self, X): """ .. todo:: WRITEME properly Parameters ---------- X : WRITEME Must contain only examples that lie on the hypersphere """ #design matrix format return T.zeros_like(X[:, 0]) def log_prob(self, X): """ .. todo:: WRITEME """ return - self.free_energy(X) - self.logZ def random_design_matrix(self, m): """ .. todo:: WRITEME """ Z = self.s_rng.normal(size=(m, self.dim), avg=0., std=1., dtype=config.floatX) Z.name = 'UH.rdm.Z' sq_norm_Z = T.sum(T.sqr(Z), axis=1) sq_norm_Z.name = 'UH.rdm.sq_norm_Z' eps = 1e-6 mask = sq_norm_Z < eps mask.name = 'UH.rdm.mask' Z = (Z.T * (1. - mask) + mask).T Z.name = 'UH.rdm.Z2' sq_norm_Z = sq_norm_Z * (1. - mask) + self.dim * mask sq_norm_Z.name = 'UH.rdm.sq_norm_Z2' norm_Z = T.sqrt(sq_norm_Z) norm_Z.name = 'UH.rdm.sq_norm_Z2' rval = self.radius * (Z.T / norm_Z).T rval.name = 'UH.rdm.rval' return rval
bsd-3-clause
e28bcfe1033155a492539b831bbcdc8d
25.45
76
0.512287
3.084548
false
false
false
false
lisa-lab/pylearn2
pylearn2/optimization/linear_cg.py
49
2914
""" .. todo:: WRITEME """ import theano from theano import tensor from theano.ifelse import ifelse def linear_cg(fn, params, tol=1e-3, max_iters=1000, floatX=None): """ Minimizes a POSITIVE DEFINITE quadratic function via linear conjugate gradient using the R operator to avoid explicitly representing the Hessian. If you have several variables, this is cheaper than Newton's method, which would need to invert the Hessian. It is also cheaper than standard linear conjugate gradient, which works with an explicit representation of the Hessian. It is also cheaper than nonlinear conjugate gradient which does a line search by repeatedly evaluating f. For more information about linear conjugate gradient, you may look at http://en.wikipedia.org/wiki/Conjugate_gradient_method . (This reference describes linear CG but not converting it to use the R operator instead of an explicit representation of the Hessian) Parameters ---------- params : WRITEME f : theano_like A theano expression which is quadratic with POSITIVE DEFINITE hessian in x x : list List of theano shared variables that influence f tol : float Minimization halts when the norm of the gradient is smaller than tol Returns ------- rval : theano_like The solution in form of a symbolic expression (or list of symbolic expressions) """ provided_as_list = True if not isinstance(params, (list,tuple)): params = [params] provided_as_list = False n_params = len(params) def loop(rsold, *args): ps = args[:n_params] rs = args[n_params:2*n_params] xs = args[2*n_params:] Aps = [] for param in params: rval = tensor.Rop(tensor.grad(fn,param), params, ps) if isinstance(rval, (list, tuple)): Aps.append(rval[0]) else: Aps.append(rval) alpha = rsold/ sum( (x*y).sum() for x,y in zip(Aps, ps) ) xs = [ x - alpha*p for x,p in zip(xs,ps)] rs = [ r - alpha*Ap for r,Ap in zip(rs,Aps)] rsnew = sum( (r*r).sum() for r in rs) ps = [ r + rsnew/rsold*p for r,p in zip(rs,ps)] return [rsnew]+ps+rs+xs, theano.scan_module.until(rsnew < tol) r0s = tensor.grad(fn, params) if not isinstance(r0s, (list,tuple)): r0s = [r0s] p0s = [x for x in r0s] x0s = params rsold = sum( (r*r).sum() for r in r0s) outs, updates = theano.scan( loop, outputs_info = [rsold] + p0s+r0s+x0s, n_steps = max_iters, name = 'linear_conjugate_gradient') fxs = outs[1+2*n_params:] fxs = [ifelse(rsold <tol, x0, x[-1]) for x0,x in zip(x0s, fxs)] if not provided_as_list: return fxs[0] else: return fxs
bsd-3-clause
83ce8550fcf4b2e7dda993e9d8b92c58
33.690476
79
0.604324
3.610905
false
false
false
false
lisa-lab/pylearn2
pylearn2/expr/sampling.py
49
1815
""" .. todo:: WRITEME """ __authors__ = "Ian Goodfellow" __copyright__ = "Copyright 2010-2012, Universite de Montreal" __credits__ = ["Ian Goodfellow"] __license__ = "3-clause BSD" __maintainer__ = "LISA Lab" __email__ = "pylearn-dev@googlegroups" from theano.sandbox.rng_mrg import MRG_RandomStreams from pylearn2.blocks import Block from pylearn2.utils.rng import make_theano_rng class SampleBernoulli(Block): """ .. todo:: WRITEME Parameters ---------- theano_rng : WRITEME seed : WRITEME input_space : WRITEME """ def __init__(self, theano_rng = None, seed=None, input_space=None): super(SampleBernoulli, self).__init__() assert theano_rng is None or seed is None theano_rng = make_theano_rng(theano_rng if theano_rng is not None else seed, 2012+11+22, which_method='binomial') self.__dict__.update(locals()) del self.self def __call__(self, inputs): """ .. todo:: WRITEME """ if self.input_space: self.input_space.validate(inputs) return self.theano_rng.binomial(p=inputs, size=inputs.shape, dtype=inputs.dtype) def set_input_space(self, space): """ .. todo:: WRITEME """ self.input_space = space def get_input_space(self): """ .. todo:: WRITEME """ if self.input_space is not None: return self.input_space raise ValueError("No input space was specified for this Block (%s). " "You can call set_input_space to correct that." % str(self)) def get_output_space(self): """ .. todo:: WRITEME """ return self.get_input_space()
bsd-3-clause
8a72070624293b71489521890df95e36
23.527027
88
0.552617
3.937093
false
false
false
false
lisa-lab/pylearn2
pylearn2/sandbox/cuda_convnet/stochastic_pool.py
2
19249
""" GPU op for Stochastic max pooling as defined in: Stochastic Pooling for Regularization of Deep Convolutional Neural Networks Matthew D. Zeiler, Rob Fergus, ICLR 2013 The code is written around Alex Krizhevsky's cuda-convnet """ __authors__ = "Mehdi Mirza" __copyright__ = "Copyright 2010-2013, Universite de Montreal" __credits__ = ["Mehdi Mirza", "David Warde-Farley"] __license__ = "3-clause BSD" __maintainer__ = "Mehdi Mirza" __email__ = "mirzamom@iro" import warnings import numpy from theano import shared from theano.gof import Apply from theano.sandbox.cuda import CudaNdarrayType from theano.sandbox.cuda.basic_ops import as_cuda_ndarray_variable from theano.sandbox.cuda.basic_ops import gpu_contiguous from theano.sandbox.cuda import GpuOp from theano.tensor import get_scalar_constant_value, NotScalarConstantError, zeros_like from pylearn2.sandbox.cuda_convnet.base_acts import UnimplementedError from pylearn2.sandbox.cuda_convnet.convnet_compile import convnet_available from pylearn2.sandbox.cuda_convnet.convnet_compile import cuda_convnet_loc from pylearn2.sandbox.cuda_convnet.shared_code import this_dir from pylearn2.sandbox.cuda_convnet.pool import MaxPoolGrad def stochastic_max_pool_c01b(c01b, pool_shape, pool_stride, start=0, seed = 1234): """ .. todo:: WRITEME """ assert pool_shape[0] == pool_shape[1] assert pool_stride[0] == pool_stride[1] op = StochasticMaxPool(pool_shape[0], pool_stride[0], start, seed) c01b = gpu_contiguous(c01b) return op(c01b) def weighted_max_pool_c01b(c01b, pool_shape, pool_stride, start=0): """ .. todo:: WRITEME """ assert pool_shape[0] == pool_shape[1] assert pool_stride[0] == pool_stride[1] op = WeightedMaxPool(pool_shape[0], pool_stride[0], start) c01b = gpu_contiguous(c01b) return op(c01b) class StochasticMaxPool(GpuOp): """ Stochastic MaxPool op code on the GPU. The input are in the order (channel, image rows, image cols, batch) Works only on square images and the grad works only when channel % 16 == 0. Parameters ---------- ds : int defines the size of the pooling region in the x (equivalently, y) dimension. Squares of size (ds)2 get reduced to one value by this layer. There are no restrictions on the value of this parameter. It's fine for a pooling square to fall off the boundary of the image. Named SizeX in Alex's code. stride : int defines the stride size between successive pooling squares. Setting this parameter smaller than sizeX produces overlapping pools. Setting it equal to sizeX gives the usual, non-overlapping pools. Values greater than sizeX are not allowed. start : int, optional tells the net where in the input image to start the pooling (in x,y coordinates). In principle, you can start anywhere you want. Setting this to a positive number will cause the net to discard some pixels at the top and at the left of the image. Setting this to a negative number will cause it to include pixels that don't exist (which is fine). start=0 is the usual setting. outputs : int, optional allows you to control how many output values in the x (equivalently, y) dimension this operation will produce. This parameter is analogous to the start parameter, in that it allows you to discard some portion of the image by setting it to a value small enough to leave part of the image uncovered. Setting it to zero instructs the net to produce as many outputs as is necessary to ensure that the whole image is covered. default 0 seed : WRITEME """ def __init__(self, ds, stride, start=0, outputs=0, seed = 1234): self.ds = ds self.stride = stride self.start = start self.copy_non_contiguous = 0 self.seed_state = shared(numpy.asarray(seed).astype('float32')) self.seed_state.default_update = self.seed_state + 1 assert stride > 0 and stride <= ds, (stride, ds) assert ds > 0, ds # We check in the code if ds <= imgSizeX def __eq__(self, other): """ .. todo:: WRITEME """ #Dont put copy_non_contigous as this doesn't change the output return (type(self) == type(other) and self.ds == other.ds and self.stride == other.stride and self.start == other.start) def __hash__(self): """ .. todo:: WRITEME """ #Dont put copy_non_contigous as this doesn't change the output return (hash(type(self)) ^ hash(self.ds) ^ hash(self.stride) ^ hash(self.start)) def c_header_dirs(self): """ .. todo:: WRITEME """ return [this_dir] def c_headers(self): """ .. todo:: WRITEME """ return ['nvmatrix.cuh', 'conv_util.cuh'] def c_lib_dirs(self): """ .. todo:: WRITEME """ return [cuda_convnet_loc] def c_libraries(self): """ .. todo:: WRITEME """ return ['cuda_convnet'] def c_code_cache_version(self): """ .. todo:: WRITEME """ return (1,) def _argument_contiguity_check(self, arg_name): """ .. todo:: WRITEME """ return """ if (!CudaNdarray_is_c_contiguous(%%(%(arg_name)s)s)) { if (!(%(class_name_caps)s_COPY_NON_CONTIGUOUS)) { PyErr_SetString(PyExc_ValueError, "%(class)s: %(arg_name)s must be C contiguous"); %%(fail)s; } } """ % { 'class': self.__class__.__name__, 'arg_name': arg_name, 'class_name_caps': self.__class__.__name__.upper(), } def make_node(self, images): """ .. todo:: WRITEME """ images = as_cuda_ndarray_variable(images) assert images.ndim == 4 channels_broadcastable = images.type.broadcastable[0] batch_broadcastable = images.type.broadcastable[3] rows_broadcastable = False cols_broadcastable = False targets_broadcastable = (channels_broadcastable, rows_broadcastable, cols_broadcastable, batch_broadcastable) targets_type = CudaNdarrayType(broadcastable=targets_broadcastable) targets = targets_type() seed = self.seed_state seed = as_cuda_ndarray_variable(seed) return Apply(self, [images, seed], [targets]) def c_code(self, node, name, inputs, outputs, sub): """ .. todo:: WRITEME """ images, seed = inputs targets, = outputs fail = sub['fail'] # The amount of braces that must be closed at the end num_braces = 0 if self.copy_non_contiguous: raise UnimplementedError() else: basic_setup = "#define STOCHASTICMAXPOOL_COPY_NON_CONTIGUOUS 0\n" # Convert images in nv_images, an NVMatrix, for compatibility # with the cuda-convnet functions setup_nv_images = self._argument_contiguity_check("images") + """ if (%(images)s->nd != 4) { PyErr_Format(PyExc_ValueError, "images must have nd=4, got nd=%%i", %(images)s->nd); %(fail)s; } { //setup_nv_images brace 1 const int * images_dims = CudaNdarray_HOST_DIMS(%(images)s); const int img_channels = images_dims[0]; const int imgSizeY = images_dims[1]; const int imgSizeX = images_dims[2]; const int batch_size = images_dims[3]; if(imgSizeY != imgSizeX){ PyErr_Format(PyExc_ValueError, "images must be square(dims[1] == dims[2]). Shape (%%i,%%i,%%i,%%i)", img_channels, imgSizeY, imgSizeX, batch_size); %(fail)s; } if(%(ds)s > imgSizeY){ PyErr_Format(PyExc_ValueError, "ds(%%d) must be <= imgSizeX(%%d) and imgSizeY(%%d).", %(ds)s, imgSizeX, imgSizeY); %(fail)s; } if(%(start)s >= imgSizeX){ PyErr_Format(PyExc_ValueError, "start is %%d but must be smaller then the images size of %%d x %%d.", %(start)s, imgSizeX, imgSizeY); %(fail)s; } NVMatrix nv_images(%(images)s, img_channels * imgSizeY * imgSizeX, batch_size, "MaxPool:nv_images"); //int * seed = CudaNdarray_HOST_DIMS%(seed)s; float * seed = CudaNdarray_DEV_DATA(%(seed)s); //int * seed = %(seed)s; """ num_braces += 1 setup_nv_targets = """ //int _outputsX = int(ceil((dic['imgSize'] - dic['start'] - dic['sizeX']) / float(dic['stride']))) + 1; int _outputsX = ((int)(ceil((imgSizeY - %(start)s - %(ds)s) / ((float)%(stride)s)))) + 1; int target_dims [] = { img_channels, _outputsX, _outputsX, batch_size }; if (CudaNdarray_prep_output(& %(targets)s, 4, target_dims)) { %(fail)s; } { // setup_nv_target brace # 1 NVMatrix nv_targets(%(targets)s, target_dims[0] * target_dims[1] * target_dims[2], target_dims[3], "MaxPool:nv_targets"); """ num_braces += 1 do_pool = """ convLocalStochasticMaxPool(nv_images, nv_targets, img_channels, %(ds)s, %(start)s, %(stride)s, _outputsX, MaxPooler(), seed); """ braces = '}' * num_braces rval = (basic_setup + setup_nv_images + setup_nv_targets + do_pool + braces) start = self.start stride = self.stride ds = self.ds rval = rval % locals() return rval def grad(self, inp, grads): """ .. todo:: WRITEME """ x, seed = inp gz, = grads gz = gpu_contiguous(gz) maxout = self(x) return [MaxPoolGrad(self.ds, self.stride, self.start)(x, maxout, gz), zeros_like(seed)] # Make sure the cuda_convnet library is compiled and up-to-date def make_thunk(self, *args, **kwargs): """ .. todo:: WRITEME """ if not convnet_available(): raise RuntimeError('Could not compile cuda_convnet') return super(StochasticMaxPool, self).make_thunk(*args, **kwargs) class WeightedMaxPool(GpuOp): """ This op wrap Alex's MaxPool code on the GPU. The input are in the order (channel, image rows, image cols, batch) Works only on square images and the grad works only when channel % 16 == 0. Parameters ---------- ds : int defines the size of the pooling region in the x (equivalently, y) dimension. Squares of size (ds)2 get reduced to one value by this layer. There are no restrictions on the value of this parameter. It's fine for a pooling square to fall off the boundary of the image. Named SizeX in Alex's code. stride : int defines the stride size between successive pooling squares. Setting this parameter smaller than sizeX produces overlapping pools. Setting it equal to sizeX gives the usual, non-overlapping pools. Values greater than sizeX are not allowed. start : int, optional tells the net where in the input image to start the pooling (in x,y coordinates). In principle, you can start anywhere you want. Setting this to a positive number will cause the net to discard some pixels at the top and at the left of the image. Setting this to a negative number will cause it to include pixels that don't exist (which is fine). start=0 is the usual setting. outputs : int, optional allows you to control how many output values in the x (equivalently, y) dimension this operation will produce. This parameter is analogous to the start parameter, in that it allows you to discard some portion of the image by setting it to a value small enough to leave part of the image uncovered. Setting it to zero instructs the net to produce as many outputs as is necessary to ensure that the whole image is covered. default 0 """ def __init__(self, ds, stride, start=0, outputs=0): self.ds = ds self.stride = stride self.start = start self.copy_non_contiguous = 0 assert stride > 0 and stride <= ds, (stride, ds) assert ds > 0, ds # We check in the code if ds <= imgSizeX def __eq__(self, other): """ .. todo:: WRITEME """ #Dont put copy_non_contigous as this doesn't change the output return (type(self) == type(other) and self.ds == other.ds and self.stride == other.stride and self.start == other.start) def __hash__(self): """ .. todo:: WRITEME """ #Dont put copy_non_contigous as this doesn't change the output return (hash(type(self)) ^ hash(self.ds) ^ hash(self.stride) ^ hash(self.start)) def c_header_dirs(self): """ .. todo:: WRITEME """ return [this_dir] def c_headers(self): """ .. todo:: WRITEME """ return ['nvmatrix.cuh', 'conv_util.cuh'] def c_lib_dirs(self): """ .. todo:: WRITEME """ return [cuda_convnet_loc] def c_libraries(self): """ .. todo:: WRITEME """ return ['cuda_convnet'] def c_code_cache_version(self): """ .. todo:: WRITEME """ return (1,) def _argument_contiguity_check(self, arg_name): """ .. todo:: WRITEME """ return """ if (!CudaNdarray_is_c_contiguous(%%(%(arg_name)s)s)) { if (!(%(class_name_caps)s_COPY_NON_CONTIGUOUS)) { PyErr_SetString(PyExc_ValueError, "%(class)s: %(arg_name)s must be C contiguous"); %%(fail)s; } } """ % { 'class': self.__class__.__name__, 'arg_name': arg_name, 'class_name_caps': self.__class__.__name__.upper(), } def make_node(self, images): """ .. todo:: WRITEME """ images = as_cuda_ndarray_variable(images) assert images.ndim == 4 channels_broadcastable = images.type.broadcastable[0] batch_broadcastable = images.type.broadcastable[3] rows_broadcastable = False cols_broadcastable = False targets_broadcastable = (channels_broadcastable, rows_broadcastable, cols_broadcastable, batch_broadcastable) targets_type = CudaNdarrayType(broadcastable=targets_broadcastable) targets = targets_type() return Apply(self, [images], [targets]) def c_code(self, node, name, inputs, outputs, sub): """ .. todo:: WRITEME """ images, = inputs targets, = outputs fail = sub['fail'] # The amount of braces that must be closed at the end num_braces = 0 if self.copy_non_contiguous: raise UnimplementedError() else: basic_setup = "#define WEIGHTEDMAXPOOL_COPY_NON_CONTIGUOUS 0\n" # Convert images in nv_images, an NVMatrix, for compatibility # with the cuda-convnet functions setup_nv_images = self._argument_contiguity_check("images") + """ if (%(images)s->nd != 4) { PyErr_Format(PyExc_ValueError, "images must have nd=4, got nd=%%i", %(images)s->nd); %(fail)s; } { //setup_nv_images brace 1 const int * images_dims = CudaNdarray_HOST_DIMS(%(images)s); const int img_channels = images_dims[0]; const int imgSizeY = images_dims[1]; const int imgSizeX = images_dims[2]; const int batch_size = images_dims[3]; if(imgSizeY != imgSizeX){ PyErr_Format(PyExc_ValueError, "images must be square(dims[1] == dims[2]). Shape (%%i,%%i,%%i,%%i)", img_channels, imgSizeY, imgSizeX, batch_size); %(fail)s; } if(%(ds)s > imgSizeY){ PyErr_Format(PyExc_ValueError, "ds(%%d) must be <= imgSizeX(%%d) and imgSizeY(%%d).", %(ds)s, imgSizeX, imgSizeY); %(fail)s; } if(%(start)s >= imgSizeX){ PyErr_Format(PyExc_ValueError, "start is %%d but must be smaller then the images size of %%d x %%d.", %(start)s, imgSizeX, imgSizeY); %(fail)s; } NVMatrix nv_images(%(images)s, img_channels * imgSizeY * imgSizeX, batch_size, "MaxPool:nv_images"); """ num_braces += 1 setup_nv_targets = """ //int _outputsX = int(ceil((dic['imgSize'] - dic['start'] - dic['sizeX']) / float(dic['stride']))) + 1; int _outputsX = ((int)(ceil((imgSizeY - %(start)s - %(ds)s) / ((float)%(stride)s)))) + 1; int target_dims [] = { img_channels, _outputsX, _outputsX, batch_size }; if (CudaNdarray_prep_output(& %(targets)s, 4, target_dims)) { %(fail)s; } { // setup_nv_target brace # 1 NVMatrix nv_targets(%(targets)s, target_dims[0] * target_dims[1] * target_dims[2], target_dims[3], "MaxPool:nv_targets"); """ num_braces += 1 do_pool = """ convLocalWeightedPool(nv_images, nv_targets, img_channels, %(ds)s, %(start)s, %(stride)s, _outputsX, MaxPooler()); """ braces = '}' * num_braces rval = (basic_setup + setup_nv_images + setup_nv_targets + do_pool + braces) start = self.start stride = self.stride ds = self.ds rval = rval % locals() return rval def grad(self, inp, grads): """ .. todo:: WRITEME """ raise NotImplementedError() # Make sure the cuda_convnet library is compiled and up-to-date def make_thunk(self, node, storage_map, compute_map, no_recycling): """ .. todo:: WRITEME """ if not convnet_available(): raise RuntimeError('Could not compile cuda_convnet') return super(WeightedMaxPool, self).make_thunk( node, storage_map, compute_map, no_recycling)
bsd-3-clause
62810aac50316353f43dcc8b54442bee
30.197731
111
0.55312
3.926765
false
false
false
false
lisa-lab/pylearn2
pylearn2/scripts/jobman/tester.py
44
3126
""" This an example script inserting a pylearn2 yaml code into a jobman database. The code below defines a yaml template string in state.yaml_template, and the values of its hyper-parameters in state.hyper_parameters, and run the code that is located in state.extract_results on this model using jobman. Actually, we add the job here and it can be launched later as usual (please check how to start jobs using jobman from the jobman tutorial website) """ from nose.plugins.skip import SkipTest try: from jobman.tools import DD, flatten from jobman import api0, sql except ImportError: raise SkipTest() from pylearn2.scripts.jobman import experiment def result_extractor(train_obj): """ This is a user specific function, that is used by jobman to extract results The returned dictionary will be saved in state.results """ import numpy channels = train_obj.model.monitor.channels train_cost = channels['sgd_cost(ExhaustiveSGD[X])'] best_epoch = numpy.argmin(train_cost.val_record) best_rec_error = train_cost.val_record[best_epoch] batch_num = train_cost.batch_record[best_epoch] return dict( best_epoch=best_epoch, train_rec_error=best_rec_error, batch_num=batch_num) if __name__ == '__main__': db = api0.open_db('sqlite:///test.db?table=test_jobman_pylearn2') state = DD() state.yaml_template = ''' !obj:pylearn2.train.Train { "dataset": !obj:pylearn2.datasets.npy_npz.NpyDataset &dataset { "file" : "%(file)s" }, "model": !obj:pylearn2.autoencoder.ContractiveAutoencoder { "nvis" : %(nvis)d, "nhid" : %(nhid)d, "irange" : 0.05, "act_enc": "sigmoid", #for some reason only sigmoid function works "act_dec": "sigmoid", }, "algorithm": !obj:pylearn2.training_algorithms.sgd.SGD { "learning_rate" : %(learning_rate)f, "batch_size" : %(batch_size)d, "monitoring_batches" : 5, "monitoring_dataset" : *dataset, "cost" : !obj:pylearn2.costs.cost.SumOfCosts { "costs": [ [1.0, !obj:pylearn2.costs.autoencoder.MeanBinaryCrossEntropy {} ], [%(coefficient)f, !obj:pylearn2.costs.cost.MethodCost { method: 'contraction_penalty' } ] ] }, "termination_criterion" : %(term_crit)s, } } ''' state.hyper_parameters = { "file": "${PYLEARN2_DATA_PATH}/UTLC/pca/sylvester_train_x_pca32.npy", "nvis": 32, "nhid": 6, "learning_rate": 0.1, "batch_size": 10, "coefficient": 0.5, "term_crit": { "__builder__": "pylearn2.training_algorithms.sgd.EpochCounter", "max_epochs": 2 } } state.extract_results = "pylearn2.scripts.jobman.tester.result_extractor" sql.insert_job( experiment.train_experiment, flatten(state), db, force_dup=True)
bsd-3-clause
07eb2e770d37237007527101bed2d2da
31.226804
109
0.590211
3.647608
false
true
false
false
lisa-lab/pylearn2
pylearn2/models/tests/test_dbm.py
44
37752
from __future__ import print_function from pylearn2.models.dbm.dbm import DBM from pylearn2.models.dbm.layer import BinaryVector, BinaryVectorMaxPool, Softmax, GaussianVisLayer __authors__ = "Ian Goodfellow" __copyright__ = "Copyright 2012, Universite de Montreal" __credits__ = ["Ian Goodfellow", "Devon Hjelm"] __license__ = "3-clause BSD" __maintainer__ = "LISA Lab" import numpy as np import random assert hasattr(np, 'exp') from theano.compat.six.moves import xrange from theano import config from theano import function from theano.sandbox.rng_mrg import MRG_RandomStreams from theano import tensor as T from pylearn2.expr.basic import is_binary from pylearn2.expr.nnet import inverse_sigmoid_numpy from pylearn2.costs.dbm import VariationalCD import pylearn2.testing.datasets as datasets from pylearn2.space import VectorSpace from pylearn2.utils import sharedX from pylearn2.utils import safe_zip from pylearn2.utils.data_specs import DataSpecsMapping class DummyLayer(object): """ A layer that we build for the test that just uses a state as its downward message. """ def downward_state(self, state): return state def downward_message(self, state): return state class DummyDBM(object): """ A dummy DBM for some of the tests below. """ def __init__(self, rng): self.rng = rng class TestBinaryVector: """ Testing class for DBM BinaryVector. """ def setUp(self): pass @staticmethod def check_samples(value, expected_shape, expected_mean, tol): """ Tests that a matrix of binary samples (observations in rows, variables in columns) 1) Has the right shape 2) Is binary 3) Converges to the right mean """ assert value.shape == expected_shape assert is_binary(value) mean = value.mean(axis=0) max_error = np.abs(mean-expected_mean).max() print('Actual mean:') print(mean) print('Expected mean:') print(expected_mean) print('Maximal error:', max_error) if max_error > tol: raise ValueError("Samples don't seem to have the right mean.") def test_make_state(self): # Verifies that BinaryVector.make_state creates # a shared variable whose value passes check_samples n = 5 num_samples = 1000 tol = .04 layer = BinaryVector(nvis = n) rng = np.random.RandomState([2012,11,1]) mean = rng.uniform(1e-6, 1. - 1e-6, (n,)) z = inverse_sigmoid_numpy(mean) layer.set_biases(z.astype(config.floatX)) init_state = layer.make_state(num_examples=num_samples, numpy_rng=rng) value = init_state.get_value() TestBinaryVector.check_samples(value, (num_samples, n), mean, tol) def test_sample(self): # Verifies that BinaryVector.sample returns an expression # whose value passes check_samples assert hasattr(np, 'exp') n = 5 num_samples = 1000 tol = .04 vis = BinaryVector(nvis=n) hid = DummyLayer() rng = np.random.RandomState([2012,11,1,259]) mean = rng.uniform(1e-6, 1. - 1e-6, (n,)) ofs = rng.randn(n) vis.set_biases(ofs.astype(config.floatX)) z = inverse_sigmoid_numpy(mean) - ofs z_var = sharedX(np.zeros((num_samples, n)) + z) theano_rng = MRG_RandomStreams(2012+11+1) sample = vis.sample(state_above=z_var, layer_above=hid, theano_rng=theano_rng) sample = sample.eval() TestBinaryVector.check_samples(sample, (num_samples, n), mean, tol) def check_gaussian_samples(value, nsamples, nvis, rows, cols, channels, expected_mean, tol): """ Tests that a matrix of Gaussian samples (observations in rows, variables in columns) 1) Has the right shape 2) Is not binary 3) Converges to the right mean """ if nvis: expected_shape = (nsamples, nvis) else: expected_shape = (nsamples,rows,cols,channels) assert value.shape == expected_shape assert not is_binary(value) mean = value.mean(axis=0) max_error = np.abs(mean-expected_mean).max() print('Actual mean:') print(mean) print('Expected mean:') print(expected_mean) print('Maximal error:', max_error) print('Tolerable variance:', tol) if max_error > tol: raise ValueError("Samples don't seem to have the right mean.") else: print('Mean is within expected range') def test_gaussian_vis_layer_make_state(): """ Verifies that GaussianVisLayer.make_state creates a shared variable whose value passes check_gaussian_samples In this case the layer lives in a VectorSpace """ n = 5 rows = None cols = None channels = None num_samples = 1000 tol = .042 # tolerated variance beta = 1/tol # precision parameter layer = GaussianVisLayer(nvis = n, init_beta=beta) rng = np.random.RandomState([2012,11,1]) mean = rng.uniform(1e-6, 1. - 1e-6, (n,)) z= mean layer.set_biases(z.astype(config.floatX)) init_state = layer.make_state(num_examples=num_samples, numpy_rng=rng) value = init_state.get_value() check_gaussian_samples(value, num_samples, n, rows, cols, channels, mean, tol) def test_gaussian_vis_layer_make_state_conv(): """ Verifies that GaussianVisLayer.make_state creates a shared variable whose value passes check_gaussian_samples In this case the layer lives in a Conv2DSpace """ n = None rows = 3 cols = 3 channels = 3 num_samples = 1000 tol = .042 # tolerated variance beta = 1/tol # precision parameter # axes for batch, rows, cols, channels, can be given in any order axes = ['b', 0, 1, 'c'] random.shuffle(axes) axes = tuple(axes) print('axes:', axes) layer = GaussianVisLayer(rows=rows, cols=cols, channels=channels, init_beta=beta, axes=axes) # rng = np.random.RandomState([2012,11,1]) rng = np.random.RandomState() mean = rng.uniform(1e-6, 1. - 1e-6, (rows, cols, channels)) #z = inverse_sigmoid_numpy(mean) z= mean layer.set_biases(z.astype(config.floatX)) init_state = layer.make_state(num_examples=num_samples, numpy_rng=rng) value = init_state.get_value() check_gaussian_samples(value, num_samples, n, rows, cols, channels, mean, tol) def test_gaussian_vis_layer_sample(): """ Verifies that GaussianVisLayer.sample returns an expression whose value passes check_gaussian_samples In this case the layer lives in a VectorSpace """ assert hasattr(np, 'exp') n = 5 num_samples = 1000 tol = .042 # tolerated variance beta = 1/tol # precision parameter rows = None cols = None channels = None class DummyLayer(object): """ A layer that we build for the test that just uses a state as its downward message. """ def downward_state(self, state): return state def downward_message(self, state): return state vis = GaussianVisLayer(nvis=n, init_beta=beta) hid = DummyLayer() rng = np.random.RandomState([2012,11,1,259]) mean = rng.uniform(1e-6, 1. - 1e-6, (n,)) ofs = rng.randn(n) vis.set_biases(ofs.astype(config.floatX)) #z = inverse_sigmoid_numpy(mean) - ofs z=mean -ofs # linear activation function z_var = sharedX(np.zeros((num_samples, n)) + z) # mean will be z_var + mu theano_rng = MRG_RandomStreams(2012+11+1) sample = vis.sample(state_above=z_var, layer_above=hid, theano_rng=theano_rng) sample = sample.eval() check_gaussian_samples(sample, num_samples, n, rows, cols, channels, mean, tol) def test_gaussian_vis_layer_sample_conv(): """ Verifies that GaussianVisLayer.sample returns an expression whose value passes check_gaussian_samples. In this case the layer lives in a Conv2DSpace """ assert hasattr(np, 'exp') n = None num_samples = 1000 tol = .042 # tolerated variance beta = 1/tol # precision parameter rows = 3 cols = 3 channels = 3 # axes for batch, rows, cols, channels, can be given in any order axes = ['b', 0, 1, 'c'] random.shuffle(axes) axes = tuple(axes) print('axes:', axes) class DummyLayer(object): """ A layer that we build for the test that just uses a state as its downward message. """ def downward_state(self, state): return state def downward_message(self, state): return state vis = GaussianVisLayer(nvis=None,rows=rows, cols=cols, channels=channels, init_beta=beta, axes=axes) hid = DummyLayer() rng = np.random.RandomState([2012,11,1,259]) mean = rng.uniform(1e-6, 1. - 1e-6, (rows, cols, channels)) ofs = rng.randn(rows,cols,channels) vis.set_biases(ofs.astype(config.floatX)) #z = inverse_sigmoid_numpy(mean) - ofs z = mean -ofs z_var = sharedX(np.zeros((num_samples, rows, cols, channels)) + z) theano_rng = MRG_RandomStreams(2012+11+1) sample = vis.sample(state_above=z_var, layer_above=hid, theano_rng=theano_rng) sample = sample.eval() check_gaussian_samples(sample, num_samples, n, rows, cols, channels, mean, tol) def check_bvmp_samples(value, num_samples, n, pool_size, mean, tol): """ bvmp=BinaryVectorMaxPool value: a tuple giving (pooled batch, detector batch) (all made with same params) num_samples: number of samples there should be in the batch n: detector layer dimension pool_size: size of each pool region mean: (expected value of pool unit, expected value of detector units) tol: amount the emprical mean is allowed to deviate from the analytical expectation checks that: 1) all values are binary 2) detector layer units are mutually exclusive 3) pooled unit is max of the detector units 4) correct number of samples is present 5) variables are of the right shapes 6) samples converge to the right expected value """ pv, hv = value assert n % pool_size == 0 num_pools = n // pool_size assert pv.ndim == 2 assert pv.shape[0] == num_samples assert pv.shape[1] == num_pools assert hv.ndim == 2 assert hv.shape[0] == num_samples assert hv.shape[1] == n assert is_binary(pv) assert is_binary(hv) for i in xrange(num_pools): sub_p = pv[:,i] assert sub_p.shape == (num_samples,) sub_h = hv[:,i*pool_size:(i+1)*pool_size] assert sub_h.shape == (num_samples, pool_size) if not np.all(sub_p == sub_h.max(axis=1)): for j in xrange(num_samples): print(sub_p[j], sub_h[j,:]) assert sub_p[j] == sub_h[j,:] assert False assert np.max(sub_h.sum(axis=1)) == 1 p, h = mean assert p.ndim == 1 assert h.ndim == 1 emp_p = pv.mean(axis=0) emp_h = hv.mean(axis=0) max_diff = np.abs(p - emp_p).max() if max_diff > tol: print('expected value of pooling units: ',p) print('empirical expectation: ',emp_p) print('maximum difference: ',max_diff) raise ValueError("Pooling unit samples have an unlikely mean.") max_diff = np.abs(h - emp_h).max() if max_diff > tol: assert False def test_bvmp_make_state(): # Verifies that BinaryVector.make_state creates # a shared variable whose value passes check_binary_samples num_pools = 3 num_samples = 1000 tol = .04 rng = np.random.RandomState([2012,11,1,9]) # pool_size=1 is an important corner case for pool_size in [1, 2, 5]: n = num_pools * pool_size layer = BinaryVectorMaxPool( detector_layer_dim=n, layer_name='h', irange=1., pool_size=pool_size) # This is just to placate mf_update below input_space = VectorSpace(1) class DummyDBM(object): def __init__(self): self.rng = rng layer.set_dbm(DummyDBM()) layer.set_input_space(input_space) layer.set_biases(rng.uniform(-pool_size, 1., (n,)).astype(config.floatX)) # To find the mean of the samples, we use mean field with an input of 0 mean = layer.mf_update( state_below=T.alloc(0., 1, 1), state_above=None, layer_above=None) mean = function([], mean)() mean = [ mn[0,:] for mn in mean ] state = layer.make_state(num_examples=num_samples, numpy_rng=rng) value = [elem.get_value() for elem in state] check_bvmp_samples(value, num_samples, n, pool_size, mean, tol) def make_random_basic_binary_dbm( rng, pool_size_1, num_vis = None, num_pool_1 = None, num_pool_2 = None, pool_size_2 = None, center = False ): """ Makes a DBM with BinaryVector for the visible layer, and two hidden layers of type BinaryVectorMaxPool. The weights and biases are initialized randomly with somewhat large values (i.e., not what you'd want to use for learning) rng: A numpy RandomState. pool_size_1: The size of the pools to use in the first layer. """ if num_vis is None: num_vis = rng.randint(1,11) if num_pool_1 is None: num_pool_1 = rng.randint(1,11) if num_pool_2 is None: num_pool_2 = rng.randint(1,11) if pool_size_2 is None: pool_size_2 = rng.randint(1,6) num_h1 = num_pool_1 * pool_size_1 num_h2 = num_pool_2 * pool_size_2 v = BinaryVector(num_vis, center=center) v.set_biases(rng.uniform(-1., 1., (num_vis,)).astype(config.floatX), recenter=center) h1 = BinaryVectorMaxPool( detector_layer_dim = num_h1, pool_size = pool_size_1, layer_name = 'h1', center = center, irange = 1.) h1.set_biases(rng.uniform(-1., 1., (num_h1,)).astype(config.floatX), recenter=center) h2 = BinaryVectorMaxPool( center = center, detector_layer_dim = num_h2, pool_size = pool_size_2, layer_name = 'h2', irange = 1.) h2.set_biases(rng.uniform(-1., 1., (num_h2,)).astype(config.floatX), recenter=center) dbm = DBM(visible_layer = v, hidden_layers = [h1, h2], batch_size = 1, niter = 50) return dbm def test_bvmp_mf_energy_consistent(): # A test of the BinaryVectorMaxPool class # Verifies that the mean field update is consistent with # the energy function # Specifically, in a DBM consisting of (v, h1, h2), the # lack of intra-layer connections means that # P(h1|v, h2) is factorial so mf_update tells us the true # conditional. # We also know P(h1[i] | h1[-i], v) # = P(h, v) / P(h[-i], v) # = P(h, v) / sum_h[i] P(h, v) # = exp(-E(h, v)) / sum_h[i] exp(-E(h, v)) # So we can check that computing P(h[i] | v) with both # methods works the same way rng = np.random.RandomState([2012,11,1,613]) def do_test(pool_size_1): # Make DBM and read out its pieces dbm = make_random_basic_binary_dbm( rng = rng, pool_size_1 = pool_size_1, ) v = dbm.visible_layer h1, h2 = dbm.hidden_layers num_p = h1.get_output_space().dim # Choose which unit we will test p_idx = rng.randint(num_p) # Randomly pick a v, h1[-p_idx], and h2 to condition on # (Random numbers are generated via dbm.rng) layer_to_state = dbm.make_layer_to_state(1) v_state = layer_to_state[v] h1_state = layer_to_state[h1] h2_state = layer_to_state[h2] # Debugging checks num_h = h1.detector_layer_dim assert num_p * pool_size_1 == num_h pv, hv = h1_state assert pv.get_value().shape == (1, num_p) assert hv.get_value().shape == (1, num_h) # Infer P(h1[i] | h2, v) using mean field expected_p, expected_h = h1.mf_update( state_below = v.upward_state(v_state), state_above = h2.downward_state(h2_state), layer_above = h2) expected_p = expected_p[0, p_idx] expected_h = expected_h[0, p_idx * pool_size : (p_idx + 1) * pool_size] expected_p, expected_h = function([], [expected_p, expected_h])() # Infer P(h1[i] | h2, v) using the energy function energy = dbm.energy(V = v_state, hidden = [h1_state, h2_state]) unnormalized_prob = T.exp(-energy) assert unnormalized_prob.ndim == 1 unnormalized_prob = unnormalized_prob[0] unnormalized_prob = function([], unnormalized_prob) p_state, h_state = h1_state def compute_unnormalized_prob(which_detector): write_h = np.zeros((pool_size_1,)) if which_detector is None: write_p = 0. else: write_p = 1. write_h[which_detector] = 1. h_value = h_state.get_value() p_value = p_state.get_value() h_value[0, p_idx * pool_size : (p_idx + 1) * pool_size] = write_h p_value[0, p_idx] = write_p h_state.set_value(h_value) p_state.set_value(p_value) return unnormalized_prob() off_prob = compute_unnormalized_prob(None) on_probs = [compute_unnormalized_prob(idx) for idx in xrange(pool_size)] denom = off_prob + sum(on_probs) off_prob /= denom on_probs = [on_prob / denom for on_prob in on_probs] assert np.allclose(1., off_prob + sum(on_probs)) # np.asarray(on_probs) doesn't make a numpy vector, so I do it manually wtf_numpy = np.zeros((pool_size_1,)) for i in xrange(pool_size_1): wtf_numpy[i] = on_probs[i] on_probs = wtf_numpy # Check that they match if not np.allclose(expected_p, 1. - off_prob): print('mean field expectation of p:',expected_p) print('expectation of p based on enumerating energy function values:',1. - off_prob) print('pool_size_1:',pool_size_1) assert False if not np.allclose(expected_h, on_probs): print('mean field expectation of h:',expected_h) print('expectation of h based on enumerating energy function values:',on_probs) assert False # 1 is an important corner case # We must also run with a larger number to test the general case for pool_size in [1, 2, 5]: do_test(pool_size) def test_bvmp_mf_energy_consistent_center(): """ A test of the BinaryVectorMaxPool class Verifies that the mean field update is consistent with the energy function when using Gregoire Montavon's centering trick. Specifically, in a DBM consisting of (v, h1, h2), the lack of intra-layer connections means that P(h1|v, h2) is factorial so mf_update tells us the true conditional. We also know P(h1[i] | h1[-i], v) = P(h, v) / P(h[-i], v) = P(h, v) / sum_h[i] P(h, v) = exp(-E(h, v)) / sum_h[i] exp(-E(h, v)) So we can check that computing P(h[i] | v) with both methods works the same way :return: """ rng = np.random.RandomState([2012,11,1,613]) def do_test(pool_size_1): # Make DBM and read out its pieces dbm = make_random_basic_binary_dbm( rng = rng, pool_size_1 = pool_size_1, pool_size_2 = 1, # centering is only updated for pool size 1 center = True ) v = dbm.visible_layer h1, h2 = dbm.hidden_layers num_p = h1.get_output_space().dim # Choose which unit we will test p_idx = rng.randint(num_p) # Randomly pick a v, h1[-p_idx], and h2 to condition on # (Random numbers are generated via dbm.rng) layer_to_state = dbm.make_layer_to_state(1) v_state = layer_to_state[v] h1_state = layer_to_state[h1] h2_state = layer_to_state[h2] # Debugging checks num_h = h1.detector_layer_dim assert num_p * pool_size_1 == num_h pv, hv = h1_state assert pv.get_value().shape == (1, num_p) assert hv.get_value().shape == (1, num_h) # Infer P(h1[i] | h2, v) using mean field expected_p, expected_h = h1.mf_update( state_below = v.upward_state(v_state), state_above = h2.downward_state(h2_state), layer_above = h2) expected_p = expected_p[0, p_idx] expected_h = expected_h[0, p_idx * pool_size_1 : (p_idx + 1) * pool_size_1] expected_p, expected_h = function([], [expected_p, expected_h])() # Infer P(h1[i] | h2, v) using the energy function energy = dbm.energy(V = v_state, hidden = [h1_state, h2_state]) unnormalized_prob = T.exp(-energy) assert unnormalized_prob.ndim == 1 unnormalized_prob = unnormalized_prob[0] unnormalized_prob = function([], unnormalized_prob) p_state, h_state = h1_state def compute_unnormalized_prob(which_detector): write_h = np.zeros((pool_size_1,)) if which_detector is None: write_p = 0. else: write_p = 1. write_h[which_detector] = 1. h_value = h_state.get_value() p_value = p_state.get_value() h_value[0, p_idx * pool_size_1 : (p_idx + 1) * pool_size_1] = write_h p_value[0, p_idx] = write_p h_state.set_value(h_value) p_state.set_value(p_value) return unnormalized_prob() off_prob = compute_unnormalized_prob(None) on_probs = [compute_unnormalized_prob(idx) for idx in xrange(pool_size_1)] denom = off_prob + sum(on_probs) off_prob /= denom on_probs = [on_prob / denom for on_prob in on_probs] assert np.allclose(1., off_prob + sum(on_probs)) # np.asarray(on_probs) doesn't make a numpy vector, so I do it manually wtf_numpy = np.zeros((pool_size_1,)) for i in xrange(pool_size_1): wtf_numpy[i] = on_probs[i] on_probs = wtf_numpy # Check that they match if not np.allclose(expected_p, 1. - off_prob): print('mean field expectation of p:',expected_p) print('expectation of p based on enumerating energy function values:',1. - off_prob) print('pool_size_1:',pool_size_1) assert False if not np.allclose(expected_h, on_probs): print('mean field expectation of h:',expected_h) print('expectation of h based on enumerating energy function values:',on_probs) assert False # 1 is the only pool size for which centering is implemented do_test(1) def test_bvmp_mf_sample_consistent(): # A test of the BinaryVectorMaxPool class # Verifies that the mean field update is consistent with # the sampling function # Specifically, in a DBM consisting of (v, h1, h2), the # lack of intra-layer connections means that # P(h1|v, h2) is factorial so mf_update tells us the true # conditional. # We can thus use mf_update to compute the expected value # of a sample of h1 from v and h2, and check that samples # drawn using the layer's sample method convert to that # value. rng = np.random.RandomState([2012,11,1,1016]) theano_rng = MRG_RandomStreams(2012+11+1+1036) num_samples = 1000 tol = .042 def do_test(pool_size_1): # Make DBM and read out its pieces dbm = make_random_basic_binary_dbm( rng = rng, pool_size_1 = pool_size_1, ) v = dbm.visible_layer h1, h2 = dbm.hidden_layers num_p = h1.get_output_space().dim # Choose which unit we will test p_idx = rng.randint(num_p) # Randomly pick a v, h1[-p_idx], and h2 to condition on # (Random numbers are generated via dbm.rng) layer_to_state = dbm.make_layer_to_state(1) v_state = layer_to_state[v] h1_state = layer_to_state[h1] h2_state = layer_to_state[h2] # Debugging checks num_h = h1.detector_layer_dim assert num_p * pool_size_1 == num_h pv, hv = h1_state assert pv.get_value().shape == (1, num_p) assert hv.get_value().shape == (1, num_h) # Infer P(h1[i] | h2, v) using mean field expected_p, expected_h = h1.mf_update( state_below = v.upward_state(v_state), state_above = h2.downward_state(h2_state), layer_above = h2) expected_p = expected_p[0, :] expected_h = expected_h[0, :] expected_p, expected_h = function([], [expected_p, expected_h])() # copy all the states out into a batch size of num_samples cause_copy = sharedX(np.zeros((num_samples,))).dimshuffle(0,'x') v_state = v_state[0,:] + cause_copy p, h = h1_state h1_state = (p[0,:] + cause_copy, h[0,:] + cause_copy) p, h = h2_state h2_state = (p[0,:] + cause_copy, h[0,:] + cause_copy) h1_samples = h1.sample(state_below = v.upward_state(v_state), state_above = h2.downward_state(h2_state), layer_above = h2, theano_rng = theano_rng) h1_samples = function([], h1_samples)() check_bvmp_samples(h1_samples, num_samples, num_h, pool_size, (expected_p, expected_h), tol) # 1 is an important corner case # We must also run with a larger number to test the general case for pool_size in [1, 2, 5]: do_test(pool_size) def check_multinomial_samples(value, expected_shape, expected_mean, tol): """ Tests that a matrix of multinomial samples (observations in rows, variables in columns) 1) Has the right shape 2) Is binary 3) Has one 1 per row 4) Converges to the right mean """ assert value.shape == expected_shape assert is_binary(value) assert np.all(value.sum(axis=1) == 1) mean = value.mean(axis=0) max_error = np.abs(mean-expected_mean).max() if max_error > tol: print('Actual mean:') print(mean) print('Expected mean:') print(expected_mean) print('Maximal error:', max_error) raise ValueError("Samples don't seem to have the right mean.") def test_softmax_make_state(): # Verifies that BinaryVector.make_state creates # a shared variable whose value passes check_multinomial_samples n = 5 num_samples = 1000 tol = .04 layer = Softmax(n_classes = n, layer_name = 'y') rng = np.random.RandomState([2012, 11, 1, 11]) z = 3 * rng.randn(n) mean = np.exp(z) mean /= mean.sum() layer.set_biases(z.astype(config.floatX)) state = layer.make_state(num_examples=num_samples, numpy_rng=rng) value = state.get_value() check_multinomial_samples(value, (num_samples, n), mean, tol) def test_softmax_mf_energy_consistent(): # A test of the Softmax class # Verifies that the mean field update is consistent with # the energy function # Since a Softmax layer contains only one random variable # (with n_classes possible values) the mean field assumption # does not impose any restriction so mf_update simply gives # the true expected value of h given v. # We also know P(h | v) # = P(h, v) / P( v) # = P(h, v) / sum_h P(h, v) # = exp(-E(h, v)) / sum_h exp(-E(h, v)) # So we can check that computing P(h | v) with both # methods works the same way rng = np.random.RandomState([2012,11,1,1131]) # Make DBM num_vis = rng.randint(1,11) n_classes = rng.randint(1, 11) v = BinaryVector(num_vis) v.set_biases(rng.uniform(-1., 1., (num_vis,)).astype(config.floatX)) y = Softmax( n_classes = n_classes, layer_name = 'y', irange = 1.) y.set_biases(rng.uniform(-1., 1., (n_classes,)).astype(config.floatX)) dbm = DBM(visible_layer = v, hidden_layers = [y], batch_size = 1, niter = 50) # Randomly pick a v to condition on # (Random numbers are generated via dbm.rng) layer_to_state = dbm.make_layer_to_state(1) v_state = layer_to_state[v] y_state = layer_to_state[y] # Infer P(y | v) using mean field expected_y = y.mf_update( state_below = v.upward_state(v_state)) expected_y = expected_y[0, :] expected_y = expected_y.eval() # Infer P(y | v) using the energy function energy = dbm.energy(V = v_state, hidden = [y_state]) unnormalized_prob = T.exp(-energy) assert unnormalized_prob.ndim == 1 unnormalized_prob = unnormalized_prob[0] unnormalized_prob = function([], unnormalized_prob) def compute_unnormalized_prob(which): write_y = np.zeros((n_classes,)) write_y[which] = 1. y_value = y_state.get_value() y_value[0, :] = write_y y_state.set_value(y_value) return unnormalized_prob() probs = [compute_unnormalized_prob(idx) for idx in xrange(n_classes)] denom = sum(probs) probs = [on_prob / denom for on_prob in probs] # np.asarray(probs) doesn't make a numpy vector, so I do it manually wtf_numpy = np.zeros((n_classes,)) for i in xrange(n_classes): wtf_numpy[i] = probs[i] probs = wtf_numpy if not np.allclose(expected_y, probs): print('mean field expectation of h:',expected_y) print('expectation of h based on enumerating energy function values:',probs) assert False def test_softmax_mf_energy_consistent_centering(): # A test of the Softmax class # Verifies that the mean field update is consistent with # the energy function when using the centering trick # Since a Softmax layer contains only one random variable # (with n_classes possible values) the mean field assumption # does not impose any restriction so mf_update simply gives # the true expected value of h given v. # We also know P(h | v) # = P(h, v) / P( v) # = P(h, v) / sum_h P(h, v) # = exp(-E(h, v)) / sum_h exp(-E(h, v)) # So we can check that computing P(h | v) with both # methods works the same way rng = np.random.RandomState([2012,11,1,1131]) # Make DBM num_vis = rng.randint(1,11) n_classes = rng.randint(1, 11) v = BinaryVector(num_vis, center=True) v.set_biases(rng.uniform(-1., 1., (num_vis,)).astype(config.floatX), recenter=True) y = Softmax( n_classes = n_classes, layer_name = 'y', irange = 1., center=True) y.set_biases(rng.uniform(-1., 1., (n_classes,)).astype(config.floatX), recenter=True) dbm = DBM(visible_layer = v, hidden_layers = [y], batch_size = 1, niter = 50) # Randomly pick a v to condition on # (Random numbers are generated via dbm.rng) layer_to_state = dbm.make_layer_to_state(1) v_state = layer_to_state[v] y_state = layer_to_state[y] # Infer P(y | v) using mean field expected_y = y.mf_update( state_below = v.upward_state(v_state)) expected_y = expected_y[0, :] expected_y = expected_y.eval() # Infer P(y | v) using the energy function energy = dbm.energy(V = v_state, hidden = [y_state]) unnormalized_prob = T.exp(-energy) assert unnormalized_prob.ndim == 1 unnormalized_prob = unnormalized_prob[0] unnormalized_prob = function([], unnormalized_prob) def compute_unnormalized_prob(which): write_y = np.zeros((n_classes,)) write_y[which] = 1. y_value = y_state.get_value() y_value[0, :] = write_y y_state.set_value(y_value) return unnormalized_prob() probs = [compute_unnormalized_prob(idx) for idx in xrange(n_classes)] denom = sum(probs) probs = [on_prob / denom for on_prob in probs] # np.asarray(probs) doesn't make a numpy vector, so I do it manually wtf_numpy = np.zeros((n_classes,)) for i in xrange(n_classes): wtf_numpy[i] = probs[i] probs = wtf_numpy if not np.allclose(expected_y, probs): print('mean field expectation of h:',expected_y) print('expectation of h based on enumerating energy function values:',probs) assert False def test_softmax_mf_sample_consistent(): # A test of the Softmax class # Verifies that the mean field update is consistent with # the sampling function # Since a Softmax layer contains only one random variable # (with n_classes possible values) the mean field assumption # does not impose any restriction so mf_update simply gives # the true expected value of h given v. # We can thus use mf_update to compute the expected value # of a sample of y conditioned on v, and check that samples # drawn using the layer's sample method convert to that # value. rng = np.random.RandomState([2012,11,1,1154]) theano_rng = MRG_RandomStreams(2012+11+1+1154) num_samples = 1000 tol = .042 # Make DBM num_vis = rng.randint(1,11) n_classes = rng.randint(1, 11) v = BinaryVector(num_vis) v.set_biases(rng.uniform(-1., 1., (num_vis,)).astype(config.floatX)) y = Softmax( n_classes = n_classes, layer_name = 'y', irange = 1.) y.set_biases(rng.uniform(-1., 1., (n_classes,)).astype(config.floatX)) dbm = DBM(visible_layer = v, hidden_layers = [y], batch_size = 1, niter = 50) # Randomly pick a v to condition on # (Random numbers are generated via dbm.rng) layer_to_state = dbm.make_layer_to_state(1) v_state = layer_to_state[v] y_state = layer_to_state[y] # Infer P(y | v) using mean field expected_y = y.mf_update( state_below = v.upward_state(v_state)) expected_y = expected_y[0, :] expected_y = expected_y.eval() # copy all the states out into a batch size of num_samples cause_copy = sharedX(np.zeros((num_samples,))).dimshuffle(0,'x') v_state = v_state[0,:] + cause_copy y_state = y_state[0,:] + cause_copy y_samples = y.sample(state_below = v.upward_state(v_state), theano_rng=theano_rng) y_samples = function([], y_samples)() check_multinomial_samples(y_samples, (num_samples, n_classes), expected_y, tol) def test_make_symbolic_state(): # Tests whether the returned p_sample and h_sample have the right # dimensions num_examples = 40 theano_rng = MRG_RandomStreams(2012+11+1) visible_layer = BinaryVector(nvis=100) rval = visible_layer.make_symbolic_state(num_examples=num_examples, theano_rng=theano_rng) hidden_layer = BinaryVectorMaxPool(detector_layer_dim=500, pool_size=1, layer_name='h', irange=0.05, init_bias=-2.0) p_sample, h_sample = hidden_layer.make_symbolic_state(num_examples=num_examples, theano_rng=theano_rng) softmax_layer = Softmax(n_classes=10, layer_name='s', irange=0.05) h_sample_s = softmax_layer.make_symbolic_state(num_examples=num_examples, theano_rng=theano_rng) required_shapes = [(40, 100), (40, 500), (40, 500), (40, 10)] f = function(inputs=[], outputs=[rval, p_sample, h_sample, h_sample_s]) for s, r in zip(f(), required_shapes): assert s.shape == r def test_variational_cd(): # Verifies that VariationalCD works well with make_layer_to_symbolic_state visible_layer = BinaryVector(nvis=100) hidden_layer = BinaryVectorMaxPool(detector_layer_dim=500, pool_size=1, layer_name='h', irange=0.05, init_bias=-2.0) model = DBM(visible_layer=visible_layer, hidden_layers=[hidden_layer], batch_size=100, niter=1) cost = VariationalCD(num_chains=100, num_gibbs_steps=2) data_specs = cost.get_data_specs(model) mapping = DataSpecsMapping(data_specs) space_tuple = mapping.flatten(data_specs[0], return_tuple=True) source_tuple = mapping.flatten(data_specs[1], return_tuple=True) theano_args = [] for space, source in safe_zip(space_tuple, source_tuple): name = '%s' % (source) arg = space.make_theano_batch(name=name) theano_args.append(arg) theano_args = tuple(theano_args) nested_args = mapping.nest(theano_args) grads, updates = cost.get_gradients(model, nested_args) def test_extra(): """ Test functionality that remains private, if available. """ try: import galatea except ImportError: return from galatea.dbm.pylearn2_bridge import run_unit_tests run_unit_tests()
bsd-3-clause
2790db943f51af73e98a42f002ad8531
30.2
104
0.592896
3.465712
false
false
false
false
jupyterhub/oauthenticator
oauthenticator/tests/test_github.py
1
7568
import functools import json import logging import re from io import BytesIO from urllib.parse import parse_qs, urlparse from pytest import fixture, mark from tornado.httpclient import HTTPResponse from tornado.httputil import HTTPHeaders from traitlets.config import Config from ..github import GitHubOAuthenticator from .mocks import setup_oauth_mock def user_model(username): """Return a user model""" return { 'email': 'dinosaurs@space', 'id': 5, 'login': username, 'name': 'Hoban Washburn', } @fixture def github_client(client): setup_oauth_mock( client, host=['github.com', 'api.github.com'], access_token_path='/login/oauth/access_token', user_path='/user', token_type='token', ) return client async def test_github(github_client): authenticator = GitHubOAuthenticator() handler = github_client.handler_for_user(user_model('wash')) user_info = await authenticator.authenticate(handler) name = user_info['name'] assert name == 'wash' auth_state = user_info['auth_state'] assert 'access_token' in auth_state assert 'github_user' in auth_state assert auth_state["github_user"] == { 'email': 'dinosaurs@space', 'id': 5, 'login': name, 'name': 'Hoban Washburn', } def make_link_header(urlinfo, page): return { 'Link': '<{}://{}{}?page={}>;rel="next"'.format( urlinfo.scheme, urlinfo.netloc, urlinfo.path, page ) } async def test_allowed_org_membership(github_client): client = github_client authenticator = GitHubOAuthenticator() ## Mock Github API orgs = { 'red': ['grif', 'simmons', 'donut', 'sarge', 'lopez'], 'blue': ['tucker', 'caboose', 'burns', 'sheila', 'texas'], } org_teams = {'blue': {'alpha': ['tucker', 'caboose', 'burns']}} member_regex = re.compile(r'/orgs/(.*)/members') def org_members(paginate, request): urlinfo = urlparse(request.url) org = member_regex.match(urlinfo.path).group(1) if org not in orgs: return HTTPResponse(request, 404) if not paginate: return [user_model(m) for m in orgs[org]] else: page = parse_qs(urlinfo.query).get('page', ['1']) page = int(page[0]) return org_members_paginated( org, page, urlinfo, functools.partial(HTTPResponse, request) ) def org_members_paginated(org, page, urlinfo, response): if page < len(orgs[org]): headers = make_link_header(urlinfo, page + 1) elif page == len(orgs[org]): headers = {} else: return response(400) headers.update({'Content-Type': 'application/json'}) ret = [user_model(orgs[org][page - 1])] return response( 200, headers=HTTPHeaders(headers), buffer=BytesIO(json.dumps(ret).encode('utf-8')), ) org_membership_regex = re.compile(r'/orgs/(.*)/members/(.*)') def org_membership(request): urlinfo = urlparse(request.url) urlmatch = org_membership_regex.match(urlinfo.path) org = urlmatch.group(1) username = urlmatch.group(2) print('Request org = %s, username = %s' % (org, username)) if org not in orgs: print('Org not found: org = %s' % (org)) return HTTPResponse(request, 404) if username not in orgs[org]: print('Member not found: org = %s, username = %s' % (org, username)) return HTTPResponse(request, 404) return HTTPResponse(request, 204) team_membership_regex = re.compile(r'/orgs/(.*)/teams/(.*)/members/(.*)') def team_membership(request): urlinfo = urlparse(request.url) urlmatch = team_membership_regex.match(urlinfo.path) org = urlmatch.group(1) team = urlmatch.group(2) username = urlmatch.group(3) print('Request org = %s, team = %s username = %s' % (org, team, username)) if org not in orgs: print('Org not found: org = %s' % (org)) return HTTPResponse(request, 404) if team not in org_teams[org]: print('Team not found in org: team = %s, org = %s' % (team, org)) return HTTPResponse(request, 404) if username not in org_teams[org][team]: print( 'Member not found: org = %s, team = %s, username = %s' % (org, team, username) ) return HTTPResponse(request, 404) return HTTPResponse(request, 204) ## Perform tests for paginate in (False, True): client_hosts = client.hosts['api.github.com'] client_hosts.append((team_membership_regex, team_membership)) client_hosts.append((org_membership_regex, org_membership)) client_hosts.append((member_regex, functools.partial(org_members, paginate))) authenticator.allowed_organizations = ['blue'] handler = client.handler_for_user(user_model('caboose')) user = await authenticator.authenticate(handler) assert user['name'] == 'caboose' handler = client.handler_for_user(user_model('donut')) user = await authenticator.authenticate(handler) assert user is None # reverse it, just to be safe authenticator.allowed_organizations = ['red'] handler = client.handler_for_user(user_model('caboose')) user = await authenticator.authenticate(handler) assert user is None handler = client.handler_for_user(user_model('donut')) user = await authenticator.authenticate(handler) assert user['name'] == 'donut' # test team membership authenticator.allowed_organizations = ['blue:alpha', 'red'] handler = client.handler_for_user(user_model('tucker')) user = await authenticator.authenticate(handler) assert user['name'] == 'tucker' handler = client.handler_for_user(user_model('grif')) user = await authenticator.authenticate(handler) assert user['name'] == 'grif' handler = client.handler_for_user(user_model('texas')) user = await authenticator.authenticate(handler) assert user is None client_hosts.pop() client_hosts.pop() @mark.parametrize( "org, username, expected", [ ("blue", "texas", "https://api.github.com/orgs/blue/members/texas"), ( "blue:alpha", "tucker", "https://api.github.com/orgs/blue/teams/alpha/members/tucker", ), ("red", "grif", "https://api.github.com/orgs/red/members/grif"), ], ) def test_build_check_membership_url(org, username, expected): output = GitHubOAuthenticator()._build_check_membership_url(org, username) assert output == expected def test_deprecated_config(caplog): cfg = Config() cfg.GitHubOAuthenticator.github_organization_whitelist = ["jupy"] cfg.Authenticator.whitelist = {"user1"} log = logging.getLogger("testlog") authenticator = GitHubOAuthenticator(config=cfg, log=log) assert ( log.name, logging.WARNING, 'GitHubOAuthenticator.github_organization_whitelist is deprecated in GitHubOAuthenticator 0.12.0, use ' 'GitHubOAuthenticator.allowed_organizations instead', ) in caplog.record_tuples assert authenticator.allowed_organizations == {"jupy"} assert authenticator.allowed_users == {"user1"}
bsd-3-clause
6c6986946a81f239131e8589e3de9262
31.62069
111
0.607294
3.782109
false
false
false
false
jupyterhub/oauthenticator
oauthenticator/globus.py
1
14263
""" Custom Authenticator to use Globus OAuth2 with JupyterHub """ import base64 import os import pickle import urllib from jupyterhub.auth import LocalAuthenticator from tornado.httpclient import HTTPRequest from tornado.web import HTTPError from traitlets import Bool, List, Set, Unicode, default from .oauth2 import OAuthenticator, OAuthLogoutHandler class GlobusLogoutHandler(OAuthLogoutHandler): """ Handle custom logout URLs and token revocation. If a custom logout url is specified, the 'logout' button will log the user out of that identity provider in addition to clearing the session with Jupyterhub, otherwise only the Jupyterhub session is cleared. """ async def get(self): # Ensure self.handle_logout() is called before self.default_handle_logout() # If default_handle_logout() is called first, the user session is popped and # it's not longer possible to call get_auth_state() to revoke tokens. # See https://github.com/jupyterhub/jupyterhub/blob/HEAD/jupyterhub/handlers/login.py # noqa await self.handle_logout() await self.default_handle_logout() if self.authenticator.logout_redirect_url: # super().get() will attempt to render a logout page. Make sure we # return after the redirect to avoid exceptions. self.redirect(self.authenticator.logout_redirect_url) return await super().get() async def handle_logout(self): """Overridden method for custom logout functionality. Should be called by Jupyterhub on logout just before destroying the users session to log them out.""" await super().handle_logout() if self.current_user and self.authenticator.revoke_tokens_on_logout: await self.clear_tokens(self.current_user) async def clear_tokens(self, user): """Revoke and clear user tokens from the database""" state = await user.get_auth_state() if state: await self.authenticator.revoke_service_tokens(state.get('tokens')) self.log.info( 'Logout: Revoked tokens for user "{}" services: {}'.format( user.name, ','.join(state['tokens'].keys()) ) ) state['tokens'] = {} await user.save_auth_state(state) class GlobusOAuthenticator(OAuthenticator): """The Globus OAuthenticator handles both authorization and passing transfer tokens to the spawner.""" login_service = 'Globus' logout_handler = GlobusLogoutHandler @default("user_auth_state_key") def _user_auth_state_key_default(self): return "globus_user" @default("userdata_url") def _userdata_url_default(self): return "https://auth.globus.org/v2/oauth2/userinfo" @default("authorize_url") def _authorize_url_default(self): return "https://auth.globus.org/v2/oauth2/authorize" @default("revocation_url") def _revocation_url_default(self): return "https://auth.globus.org/v2/oauth2/token/revoke" revocation_url = Unicode(help="Globus URL to revoke live tokens.").tag(config=True) @default("token_url") def _token_url_default(self): return "https://auth.globus.org/v2/oauth2/token" globus_groups_url = Unicode(help="Globus URL to get list of user's Groups.").tag( config=True ) @default("globus_groups_url") def _globus_groups_url_default(self): return "https://groups.api.globus.org/v2/groups/my_groups" identity_provider = Unicode( help="""Restrict which institution a user can use to login (GlobusID, University of Hogwarts, etc.). This should be set in the app at developers.globus.org, but this acts as an additional check to prevent unnecessary account creation.""" ).tag(config=True) def _identity_provider_default(self): return os.getenv('IDENTITY_PROVIDER', '') username_from_email = Bool( help="""Create username from email address, not preferred username. If an identity provider is specified, email address must be from the same domain. Email scope will be set automatically.""" ).tag(config=True) @default("username_from_email") def _username_from_email_default(self): return False @default("username_claim") def _username_claim_default(self): if self.username_from_email: return "email" return "preferred_username" exclude_tokens = List( help="""Exclude tokens from being passed into user environments when they start notebooks, Terminals, etc.""" ).tag(config=True) def _exclude_tokens_default(self): return ['auth.globus.org', 'groups.api.globus.org'] def _scope_default(self): scopes = [ 'openid', 'profile', 'urn:globus:auth:scope:transfer.api.globus.org:all', ] if self.allowed_globus_groups or self.admin_globus_groups: scopes.append( 'urn:globus:auth:scope:groups.api.globus.org:view_my_groups_and_memberships' ) if self.username_from_email: scopes.append('email') return scopes globus_local_endpoint = Unicode( help="""If Jupyterhub is also a Globus endpoint, its endpoint id can be specified here.""" ).tag(config=True) def _globus_local_endpoint_default(self): return os.getenv('GLOBUS_LOCAL_ENDPOINT', '') revoke_tokens_on_logout = Bool( help="""Revoke tokens so they cannot be used again. Single-user servers MUST be restarted after logout in order to get a fresh working set of tokens.""" ).tag(config=True) def _revoke_tokens_on_logout_default(self): return False allowed_globus_groups = Set( help="""Allow members of defined Globus Groups to access JupyterHub. Users in an admin Globus Group are also automatically allowed. Groups are specified with their UUIDs. Setting this will add the Globus Groups scope.""" ).tag(config=True) admin_globus_groups = Set( help="""Set members of defined Globus Groups as JupyterHub admin users. These users are automatically allowed to login to JupyterHub. Groups are specified with their UUIDs. Setting this will add the Globus Groups scope.""" ).tag(config=True) @staticmethod def check_user_in_groups(member_groups, allowed_groups): return bool(set(member_groups) & set(allowed_groups)) async def pre_spawn_start(self, user, spawner): """Add tokens to the spawner whenever the spawner starts a notebook. This will allow users to create a transfer client: globus-sdk-python.readthedocs.io/en/stable/tutorial/#tutorial-step4 """ spawner.environment['GLOBUS_LOCAL_ENDPOINT'] = self.globus_local_endpoint state = await user.get_auth_state() if state: globus_data = base64.b64encode(pickle.dumps(state)) spawner.environment['GLOBUS_DATA'] = globus_data.decode('utf-8') def get_globus_tokens(self, token_info): # Each token should have these attributes. Resource server is optional, # and likely won't be present. token_attrs = [ 'expires_in', 'resource_server', 'scope', 'token_type', 'refresh_token', 'access_token', ] # The Auth Token is a bit special, it comes back at the top level with the # id token. The id token has some useful information in it, but nothing that # can't be retrieved with an Auth token. # Repackage the Auth token into a dict that looks like the other tokens auth_token_dict = { attr_name: token_info.get(attr_name) for attr_name in token_attrs } # Make sure only the essentials make it into tokens. Other items, such as 'state' are # not needed after authentication and can be discarded. other_tokens = [ {attr_name: token_dict.get(attr_name) for attr_name in token_attrs} for token_dict in token_info['other_tokens'] ] return other_tokens + [auth_token_dict] def build_auth_state_dict(self, token_info, user_info): """ Usernames (and therefore Jupyterhub accounts) will correspond to a Globus User ID, so foouser@globusid.org will have the 'foouser' account in Jupyterhub. """ tokens = self.get_globus_tokens(token_info) # historically, tokens have been organized by resource server for convenience. # If multiple scopes are requested from the same resource server, they will be # combined into a single token from Globus Auth. by_resource_server = { token_dict['resource_server']: token_dict for token_dict in tokens if token_dict['resource_server'] not in self.exclude_tokens } return { 'client_id': self.client_id, 'tokens': by_resource_server, 'token_response': token_info, self.user_auth_state_key: user_info, } async def get_users_groups_ids(self, tokens): user_group_ids = set() # Get Groups access token, may not be in dict headed to auth state for token_dict in tokens: if token_dict['resource_server'] == 'groups.api.globus.org': groups_token = token_dict['access_token'] # Get list of user's Groups groups_headers = self.get_default_headers() groups_headers['Authorization'] = 'Bearer {}'.format(groups_token) req = HTTPRequest(self.globus_groups_url, method='GET', headers=groups_headers) groups_resp = await self.fetch(req) # Build set of Group IDs for group in groups_resp: user_group_ids.add(group['id']) return user_group_ids async def user_is_authorized(self, auth_model): tokens = self.get_globus_tokens(auth_model["auth_state"]["token_response"]) if self.allowed_globus_groups or self.admin_globus_groups: # If any of these configurations are set, user must be in the allowed or admin Globus Group user_group_ids = await self.get_users_groups_ids(tokens) if not self.check_user_in_groups( user_group_ids, self.allowed_globus_groups ): if not self.check_user_in_groups( user_group_ids, self.admin_globus_groups ): self.log.warning( '{} not in an allowed Globus Group'.format( self.user_info_to_username( auth_model["auth_state"][self.user_auth_state_key] ) ) ) return False return True async def update_auth_model(self, auth_model): username = self.user_info_to_username( auth_model["auth_state"][self.user_auth_state_key] ) tokens = self.get_globus_tokens(auth_model["auth_state"]["token_response"]) if self.admin_globus_groups: # If any of these configurations are set, user must be in the allowed or admin Globus Group user_group_ids = await self.get_users_groups_ids(tokens) # Admin users are being managed via Globus Groups # Default to False auth_model['admin'] = False if self.check_user_in_groups(user_group_ids, self.admin_globus_groups): auth_model['admin'] = True return auth_model def user_info_to_username(self, user_info): """ Usernames (and therefore Jupyterhub accounts) will correspond to a Globus User ID, so foouser@globusid.org will have the 'foouser' account in Jupyterhub. """ # It's possible for identity provider domains to be namespaced # https://docs.globus.org/api/auth/specification/#identity_provider_namespaces # noqa username, domain = user_info.get(self.username_claim).split('@', 1) if self.identity_provider and domain != self.identity_provider: raise HTTPError( 403, 'This site is restricted to {} accounts. Please link your {}' ' account at {}.'.format( self.identity_provider, self.identity_provider, 'globus.org/app/account', ), ) return username def get_default_headers(self): return {"Accept": "application/json", "User-Agent": "JupyterHub"} def get_client_credential_headers(self): headers = self.get_default_headers() b64key = base64.b64encode( bytes("{}:{}".format(self.client_id, self.client_secret), "utf8") ) headers["Authorization"] = "Basic {}".format(b64key.decode("utf8")) return headers async def revoke_service_tokens(self, services): """Revoke live Globus access and refresh tokens. Revoking inert or non-existent tokens does nothing. Services are defined by dicts returned by tokens.by_resource_server, for example: services = { 'transfer.api.globus.org': {'access_token': 'token'}, ... <Additional services>... } """ access_tokens = [ token_dict.get('access_token') for token_dict in services.values() ] refresh_tokens = [ token_dict.get('refresh_token') for token_dict in services.values() ] all_tokens = [tok for tok in access_tokens + refresh_tokens if tok is not None] for token in all_tokens: req = HTTPRequest( self.revocation_url, method="POST", headers=self.get_client_credential_headers(), body=urllib.parse.urlencode({'token': token}), ) await self.fetch(req) class LocalGlobusOAuthenticator(LocalAuthenticator, GlobusOAuthenticator): """A version that mixes in local system user creation""" pass
bsd-3-clause
ba777f7c10e3b84ae32c3c1ee9c05ce4
38.729805
115
0.62294
4.070491
false
false
false
false