code
stringlengths
101
5.91M
class FlaxDistilBertPreTrainedModel(metaclass=DummyObject): _backends = ['flax'] def __init__(self, *args, **kwargs): requires_backends(self, ['flax'])
def pretty_eta(seconds_left): minutes_left = (seconds_left // 60) seconds_left %= 60 hours_left = (minutes_left // 60) minutes_left %= 60 days_left = (hours_left // 24) hours_left %= 24 def helper(cnt, name): return '{} {}{}'.format(str(cnt), name, ('s' if (cnt > 1) else '')) if (days_left > 0): msg = helper(days_left, 'day') if (hours_left > 0): msg += (' and ' + helper(hours_left, 'hour')) return msg if (hours_left > 0): msg = helper(hours_left, 'hour') if (minutes_left > 0): msg += (' and ' + helper(minutes_left, 'minute')) return msg if (minutes_left > 0): return helper(minutes_left, 'minute') return 'less than a minute'
def save_figure(destination, obj=None): plt.tight_layout() plt.savefig(destination) plt.close()
class FeatureExtractor(ch.nn.Module): def __init__(self, submod, layers): super(FeatureExtractor, self).__init__() self.submod = submod self.layers = layers self.n = 0 for layer_func in layers: layer = layer_func(self.submod) def hook(module, _, output): module.register_buffer('activations', output) layer.register_forward_hook(hook) def forward(self, *args, **kwargs): out = self.submod(*args, **kwargs) activs = [layer_fn(self.submod).activations for layer_fn in self.layers] return ([out] + activs)
class install_lib(orig.install_lib): def run(self): self.build() outfiles = self.install() if (outfiles is not None): self.byte_compile(outfiles) def get_exclusions(self): all_packages = (pkg for ns_pkg in self._get_SVEM_NSPs() for pkg in self._all_packages(ns_pkg)) excl_specs = product(all_packages, self._gen_exclusion_paths()) return set(starmap(self._exclude_pkg_path, excl_specs)) def _exclude_pkg_path(self, pkg, exclusion_path): parts = (pkg.split('.') + [exclusion_path]) return os.path.join(self.install_dir, *parts) def _all_packages(pkg_name): while pkg_name: (yield pkg_name) (pkg_name, sep, child) = pkg_name.rpartition('.') def _get_SVEM_NSPs(self): if (not self.distribution.namespace_packages): return [] install_cmd = self.get_finalized_command('install') svem = install_cmd.single_version_externally_managed return (self.distribution.namespace_packages if svem else []) def _gen_exclusion_paths(): (yield '__init__.py') (yield '__init__.pyc') (yield '__init__.pyo') if (not hasattr(imp, 'get_tag')): return base = os.path.join('__pycache__', ('__init__.' + imp.get_tag())) (yield (base + '.pyc')) (yield (base + '.pyo')) (yield (base + '.opt-1.pyc')) (yield (base + '.opt-2.pyc')) def copy_tree(self, infile, outfile, preserve_mode=1, preserve_times=1, preserve_symlinks=0, level=1): assert (preserve_mode and preserve_times and (not preserve_symlinks)) exclude = self.get_exclusions() if (not exclude): return orig.install_lib.copy_tree(self, infile, outfile) from setuptools.archive_util import unpack_directory from distutils import log outfiles = [] def pf(src, dst): if (dst in exclude): log.warn('Skipping installation of %s (namespace package)', dst) return False log.info('copying %s -> %s', src, os.path.dirname(dst)) outfiles.append(dst) return dst unpack_directory(infile, outfile, pf) return outfiles def get_outputs(self): outputs = orig.install_lib.get_outputs(self) exclude = self.get_exclusions() if exclude: return [f for f in outputs if (f not in exclude)] return outputs
class MarioDataset(Dataset): def __init__(self, tokenizer: Optional[PreTrainedTokenizer]=None, level_string: Optional[str]=None, context_len: int=700, height: int=14, remove_start_end_tokens: bool=False, sample_all_indices: bool=False): if (level_string is None): print('No level string specified, using default string FULL_LEVEL_STR_WITH_PATHS...') level_string = FULL_LEVEL_STR_WITH_PATHS elif ('.txt' in level_string): with open(level_string, 'r') as file: level_string = file.read() self.character_set = set(level_string) if ('\n' in self.character_set): self.character_set.remove('\n') self.vocab_size = len(self.character_set) self.sample_all_indices = sample_all_indices def get_training_corpus(): (yield list(level_string)) if (tokenizer is None): tokenizer = AutoTokenizer.from_pretrained(DEFAULT_MODEL) self.tokenizer = tokenizer if (getattr(tokenizer, 'train_new_from_iterator', None) is not None): self.tokenizer = self.tokenizer.train_new_from_iterator(get_training_corpus(), 52000) elif (getattr(tokenizer, 'train_from_iterator', None) is not None): self.tokenizer = PreTrainedTokenizerFast(tokenizer_object=self.tokenizer) self.tokenizer = self.tokenizer.train_new_from_iterator(get_training_corpus(), self.vocab_size) self.context_len = context_len self.height = height (x, self.str_arr) = self.convert_level_to_tensor(level_string.split('\n')) self.input_ids = x['input_ids'].squeeze() self.attention_masks = x['attention_mask'].squeeze() if remove_start_end_tokens: self.input_ids = self.input_ids[1:(- 1)] self.attention_masks = self.attention_masks[1:(- 1)] self.indices = self.generate_indices() (self.unique_tokens, self.unique_counts) = self.input_ids.unique(return_counts=True) self.weighted_unique_counts = ((1.0 / self.unique_counts) / torch.sum(self.unique_counts)) self.token_dict = {} string_tokens = list(self.tokenizer.decode(self.unique_tokens)) for (int_token, string_token) in zip(self.unique_tokens, string_tokens): self.token_dict[string_token] = int_token def convert_level_to_tensor(self, level: List[str]): str_arr = flip_and_transpose(np.array(characterize(level))) str_arr = ''.join(join_list_of_list(str_arr)) x = self.tokenizer(str_arr, return_tensors='pt') return (x, str_arr) def __len__(self): return self.indices.shape[0] def __getitem__(self, idx): if isinstance(idx, int): indices = self.indices[idx] else: indices = torch.stack([self.indices[i] for i in idx]) return (self.input_ids[indices], self.attention_masks[indices]) def generate_indices(self): out = [] for idx in range((self.input_ids.shape[0] - self.context_len)): if (((idx % self.height) == 0) or self.sample_all_indices): arange = torch.arange(idx, (idx + self.context_len)) out.append(arange) return torch.stack(out) def sample_indices(self, batch_size): out = [] for _ in range(batch_size): start_idx = np.random.randint(0, (self.__len__() - self.context_len)) indices = torch.arange(start_idx, (start_idx + self.context_len)) out.append(indices) return torch.stack(out) def __str__(self): str_list = characterize(self.tokenizer.batch_decode(self.x['input_ids'])) string = '\n'.join(join_list_of_list(flip_and_transpose(np.array(str_list), True))) return string
def get_cgroup_path(private=True): if private: return '/' p = None with open('/proc/1/cpuset', 'r') as f: p = f.read().strip() assert p return p
class LabeledPatients(MutableMapping[(int, List[Label])]): def __init__(self, patients_to_labels: Dict[(int, List[Label])], labeler_type: LabelType): self.patients_to_labels: Dict[(int, List[Label])] = patients_to_labels self.labeler_type: LabelType = labeler_type def save(self, target_filename) -> None: with open(target_filename, 'w') as f: writer = csv.writer(f) header = ['patient_id', 'prediction_time', 'label_type', 'value'] if (self.labeler_type == 'survival'): header.append('is_censored') writer.writerow(header) for (patient, labels) in self.patients_to_labels.items(): for label in labels: if (self.labeler_type == 'survival'): assert isinstance(label.value, SurvivalValue) writer.writerow([patient, label.time.isoformat(), self.labeler_type, (label.value.time_to_event / datetime.timedelta(minutes=1)), label.value.is_censored]) else: writer.writerow([patient, label.time.isoformat(), self.labeler_type, label.value]) def get_labels_from_patient_idx(self, idx: int) -> List[Label]: return self.patients_to_labels[idx] def get_all_patient_ids(self) -> List[int]: return sorted(list(self.patients_to_labels.keys())) def get_patients_to_labels(self) -> Dict[(int, List[Label])]: return self.patients_to_labels def get_labeler_type(self) -> LabelType: return self.labeler_type def as_numpy_arrays(self) -> Tuple[(NDArray[(Literal['n_patients, 1'], np.int64)], NDArray[(Literal['n_patients, 1 or 2'], Any)], NDArray[(Literal['n_patients, 1'], np.datetime64)])]: patient_ids: List[int] = [] label_values: List[Any] = [] label_times: List[datetime.datetime] = [] if (self.labeler_type in ['boolean', 'numerical', 'categorical']): for (patient_id, labels) in self.patients_to_labels.items(): for label in labels: patient_ids.append(patient_id) label_values.append(label.value) label_times.append(label.time) elif (self.labeler_type in ['survival']): for (patient_id, labels) in self.patients_to_labels.items(): for label in labels: survival_value: SurvivalValue = cast(SurvivalValue, label.value) patient_ids.append(patient_id) label_values.append([survival_value.time_to_event, survival_value.is_censored]) label_times.append(label.time) else: raise ValueError('Other label types are not implemented yet for this method') return (np.array(patient_ids), np.array(label_values), np.array(label_times)) def get_num_patients(self) -> int: return len(self) def get_num_labels(self) -> int: total: int = 0 for labels in self.patients_to_labels.values(): total += len(labels) return total def as_list_of_label_tuples(self) -> List[Tuple[(int, Label)]]: result: List[Tuple[(int, Label)]] = [] for (patient_id, labels) in self.patients_to_labels.items(): for label in labels: result.append((int(patient_id), label)) return result def load_from_numpy(cls, patient_ids: NDArray[(Literal['n_patients, 1'], np.int64)], label_values: NDArray[(Literal['n_patients, 1 or 2'], Any)], label_times: NDArray[(Literal['n_patients, 1'], datetime.datetime)], labeler_type: LabelType) -> LabeledPatients: patients_to_labels: DefaultDict[(int, List[Label])] = collections.defaultdict(list) for (patient_id, l_value, l_time) in zip(patient_ids, label_values, label_times): if (labeler_type in ['boolean', 'numerical', 'categorical']): patients_to_labels[patient_id].append(Label(time=l_time, value=l_value)) elif (labeler_type in ['survival']): patients_to_labels[patient_id].append(Label(time=l_time, value=SurvivalValue(time_to_event=l_value[0], is_censored=l_value[1]))) else: raise ValueError('Other label types are not implemented yet for this method') return LabeledPatients(dict(patients_to_labels), labeler_type) def __str__(self): return ('LabeledPatients:\n' + pprint.pformat(self.patients_to_labels)) def __getitem__(self, key): return self.patients_to_labels[key] def __setitem__(self, key, item): self.patients_to_labels[key] = item def __delitem__(self, key): del self.patients_to_labels[key] def __iter__(self): return iter(self.patients_to_labels) def __len__(self): return len(self.patients_to_labels)
class TensorflowImporter(): def __init__(self, *args, **kwargs): self._tf_file = args[0] self._tf_format = kwargs.get('tf_format') self._outputs = kwargs.get('outputs') self._inputs = kwargs.get('inputs') def convert_to_onnx(self, graph_def, inputs, outputs): with tf.Graph().as_default() as tf_graph: tf.import_graph_def(graph_def, name='') with tf_loader.tf_session(graph=tf_graph): g = process_tf_graph(tf_graph, continue_on_error=False, target=','.join(constants.DEFAULT_TARGET), opset=11, input_names=inputs, output_names=outputs, inputs_as_nchw=None) onnx_graph = optimizer.optimize_graph(g) model_proto = onnx_graph.make_model('converted from {}'.format(self._tf_file)) return model_proto def load_checkpoint_v1(self): ckpt_path = os.path.dirname(self._tf_file) latest_ckpt = tf.train.latest_checkpoint(ckpt_path) saver = tf.train.import_meta_graph((latest_ckpt + '.meta')) with tf.Session() as session: session.run([tf.global_variables_initializer(), tf.local_variables_initializer()]) saver.restore(session, latest_ckpt) graph_def = session.graph.as_graph_def(add_shapes=True) frozen_graph = freeze_graph.freeze_graph_with_def_protos(input_graph_def=graph_def, input_saver_def=None, input_checkpoint=latest_ckpt, output_node_names=self._outputs, restore_op_name='', filename_tensor_name='', output_graph=None, clear_devices=True, initializer_nodes='') return frozen_graph def execute(self): if (self._tf_format == 'TF_PB'): graph_def = graph_pb2.GraphDef() with tf.io.gfile.GFile(self._tf_file, 'rb') as f: graph_def.ParseFromString(f.read()) (inputs, outputs) = find_out_terminal_node(graph_def, postfix=True) (graph_def, inputs, outputs) = tf_loader.from_graphdef(self._tf_file, inputs, outputs) elif (self._tf_format == 'SAVED_MODEL'): (graph_def, inputs, outputs) = tf_loader.from_saved_model(self._tf_file, None, None) else: if (self._outputs is None): raise ImportError("Missing '--outputs' parameter.") if (self._inputs is None): raise ImportError("Missing '--inputs' parameter.") inputs = [(i + ':0') for i in self._inputs.split(',')] outputs = [(i + ':0') for i in self._outputs.split(',')] if (self._tf_format == 'TF_CKPT_V1'): graph_def = self.load_checkpoint_v1() elif (self._tf_format == 'TF_CKPT_V2'): (graph_def, inputs, outputs) = tf_loader.from_checkpoint(self._tf_file, inputs, outputs) onnx_model = self.convert_to_onnx(graph_def, inputs, outputs) onnx_importer = OnnxImporter() onnx_importer.import_from_onnx_model(onnx_model) return onnx_importer.execute()
.skip def test_lambda_nested_call(): def lamb2(A, B, C, f): A[:] = f(B, C) def lamb1(A: dace.float64[20], B: dace.float64[20], C: dace.float64[20]): f = (lambda a, b: (a + b)) lamb2(A, B, C, f) A = np.random.rand(20) B = np.random.rand(20) C = np.random.rand(20) lamb1(A, B, C) assert np.allclose(A, (B + C))
class UnboundedMemory(BaseMemory): def __init__(self, **kwargs): super(UnboundedMemory, self).__init__(**kwargs) def initialize_memory(self): mem = torch.zeros(1, self.mem_size).to(self.device) ent_counter = torch.tensor([0.0]).to(self.device) last_mention_idx = torch.zeros(1).long().to(self.device) return (mem, ent_counter, last_mention_idx) def predict_action(self, query_vector, ment_score, mem_vectors, ent_counter, feature_embs): coref_new_scores = self.get_coref_new_scores(query_vector, ment_score, mem_vectors, ent_counter, feature_embs) not_a_ment_score = (- ment_score) over_ign_score = torch.cat([torch.tensor([0.0]).to(self.device), not_a_ment_score], dim=0).to(self.device) return (coref_new_scores, over_ign_score) def interpret_scores(self, coref_new_scores, overwrite_ign_scores, first_overwrite): if first_overwrite: num_ents = 0 num_cells = 1 else: num_ents = (coref_new_scores.shape[0] - 1) num_cells = num_ents pred_max_idx = torch.argmax(coref_new_scores).item() if (pred_max_idx < num_cells): return (pred_max_idx, 'c') elif (pred_max_idx == num_cells): over_max_idx = torch.argmax(overwrite_ign_scores).item() if (over_max_idx == 0): return (num_ents, 'o') else: return ((- 1), 'i') else: raise NotImplementedError def forward(self, mention_emb_list, mention_scores, gt_actions, metadata, rand_fl_list, teacher_forcing=False): (mem_vectors, ent_counter, last_mention_idx) = self.initialize_memory() action_logit_list = [] action_list = [] first_overwrite = True last_action_str = '<s>' follow_gt = (self.training or teacher_forcing) for (ment_idx, (ment_emb, ment_score, (gt_cell_idx, gt_action_str))) in enumerate(zip(mention_emb_list, mention_scores, gt_actions)): query_vector = ment_emb metadata['last_action'] = self.action_str_to_idx[last_action_str] feature_embs = self.get_feature_embs(ment_idx, last_mention_idx, ent_counter, metadata) if (not (follow_gt and (gt_action_str == 'i') and (rand_fl_list[ment_idx] > self.sample_invalid))): (coref_new_scores, overwrite_ign_scores) = self.predict_action(query_vector, ment_score, mem_vectors, ent_counter, feature_embs) (pred_cell_idx, pred_action_str) = self.interpret_scores(coref_new_scores, overwrite_ign_scores, first_overwrite) action_logit_list.append((coref_new_scores, overwrite_ign_scores)) action_list.append((pred_cell_idx, pred_action_str)) else: continue if follow_gt: action_str = gt_action_str cell_idx = gt_cell_idx else: action_str = pred_action_str cell_idx = pred_cell_idx last_action_str = action_str if (first_overwrite and (action_str == 'o')): first_overwrite = False mem_vectors = torch.unsqueeze(query_vector, dim=0) ent_counter = torch.tensor([1.0]).to(self.device) last_mention_idx[0] = ment_idx else: num_ents = mem_vectors.shape[0] cell_mask = (torch.arange(0, num_ents) == cell_idx).float().to(self.device) mask = torch.unsqueeze(cell_mask, dim=1) mask = mask.repeat(1, self.mem_size) if (action_str == 'c'): mem_vectors = self.coref_update(mem_vectors, query_vector, cell_idx, mask, ent_counter) ent_counter = (ent_counter + cell_mask) last_mention_idx[cell_idx] = ment_idx elif (action_str == 'o'): mem_vectors = torch.cat([mem_vectors, torch.unsqueeze(query_vector, dim=0)], dim=0) ent_counter = torch.cat([ent_counter, torch.tensor([1.0]).to(self.device)], dim=0) last_mention_idx = torch.cat([last_mention_idx, torch.tensor([ment_idx]).to(self.device)], dim=0) return (action_logit_list, action_list)
def test_clean_up_not_old(nparray, tensor_key): db = TensorDB() db.cache_tensor({tensor_key: nparray}) db.clean_up() cached_nparray = db.get_tensor_from_cache(tensor_key) assert np.array_equal(nparray, cached_nparray)
def is_integer(s): try: s = int(s) return True except Exception: return False
class TestLocalProjectCheckout(): def setup(self): self.shell = Shell() self.temp_dir = mkdtemp(prefix='mubench-checkout-local_') self.local_url = join(self.temp_dir, 'origin') os.makedirs(self.local_url) open(join(self.local_url, 'some.file'), 'w').close() self.checkouts_dir = join(self.temp_dir, 'checkouts') def teardown(self): rmtree(self.temp_dir) def test_create(self): uut = LocalProjectCheckout(self.local_url, self.checkouts_dir, '-project-') uut.create(0) expected_checkout_path = join(self.checkouts_dir, '-project-', 'checkout') assert_equals(expected_checkout_path, uut.checkout_dir) assert exists(join(expected_checkout_path, 'some.file')) def test_not_exists(self): uut = LocalProjectCheckout(self.local_url, self.checkouts_dir, '-project-') assert (not uut.exists()) def test_exists_after_create(self): uut = LocalProjectCheckout(self.local_url, self.checkouts_dir, '-project-') uut.create(0) assert uut.exists() def test_not_exists_empty(self): uut = LocalProjectCheckout(self.local_url, self.checkouts_dir, '-project-') os.makedirs(uut.checkout_dir) assert (not uut.exists()) def test_delete(self): uut = LocalProjectCheckout(self.local_url, self.checkouts_dir, '-project-') uut.create(0) uut.delete() assert (not exists(uut.checkout_dir)) def test_to_string(self): uut = LocalProjectCheckout(self.local_url, self.checkouts_dir, '-project-') assert_equals('local:{}'.format(self.local_url), str(uut))
(frozen=True) class CritiqueTaskTemplate(): name: str instructions: str num_respondents: int questions: List[CritiqueQuestionTemplate]
class SerializationMixin(object): def save_instance(self, filepath): save_dict = dict(auto_fix_time_shifts=self._state_data.auto_fix_time_shifts, power_signals_d=self._state_data.power_signals_d.tolist(), rank_k=self._state_data.rank_k, matrix_l0=self._state_data.matrix_l0.tolist(), matrix_r0=self._state_data.matrix_r0.tolist(), l_value=self._state_data.l_value.tolist(), r_value=self._state_data.r_value.tolist(), beta_value=float(self._state_data.beta_value), component_r0=self._state_data.component_r0.tolist(), mu_l=self._state_data.mu_l, mu_r=self._state_data.mu_r, tau=self._state_data._tau, is_solver_error=self._state_data.is_solver_error, is_problem_status_error=self._state_data.is_problem_status_error, f1_increase=self._state_data.f1_increase, obj_increase=self._state_data.obj_increase, residuals_median=self._state_data.residuals_median, residuals_variance=self._state_data.residuals_variance, residual_l0_norm=self._state_data.residual_l0_norm, weights=self._state_data.weights.tolist()) with open(filepath, 'w') as file: json.dump(save_dict, file) def load_instance(cls, filepath): with open(filepath, 'r') as file: load_dict = json.load(file) power_signals_d = np.array(load_dict['power_signals_d']) rank_k = load_dict['rank_k'] instance = cls(np.array(power_signals_d), rank_k=rank_k) instance.state_data.power_signals_d = power_signals_d instance.state_data.rank_k = rank_k instance.state_data.matrix_l0 = np.array(load_dict['matrix_l0']) instance.state_data.matrix_r0 = np.array(load_dict['matrix_r0']) instance.state_data.l_cs_value = np.array(load_dict['l_value']) instance.state_data.r_cs_value = np.array(load_dict['r_value']) instance.state_data.beta_value = load_dict['beta_value'] instance.state_data.component_r0 = np.array(load_dict['component_r0']) instance.state_data.mu_l = load_dict['mu_l'] instance.state_data.mu_r = load_dict['mu_r'] instance.state_data.tau = load_dict['tau'] instance.state_data.is_solver_error = load_dict['is_solver_error'] instance.state_data.is_problem_status_error = load_dict['is_problem_status_error'] instance.state_data.f1_increase = load_dict['f1_increase'] instance.state_data.obj_increase = load_dict['obj_increase'] instance.state_data.residuals_median = load_dict['residuals_median'] instance.state_data.residuals_variance = load_dict['residuals_variance'] instance.state_data.residual_l0_norm = load_dict['residual_l0_norm'] instance.state_data.weights = np.array(load_dict['weights']) instance._keep_result_variables_as_properties(instance.state_data.l_cs_value, instance.state_data.r_cs_value, instance.state_data.beta_value) instance._keep_supporting_parameters_as_properties(instance.state_data.weights) return instance
('prefix') class PrefixFactory(SingleFeatureFactory): def feature_name(self): return ('prefix_%s' % self.prefix_size) def prefix_size(self): return self.args['prefix_size'] def compute_feature(self, tokens, token_index): return get_word_chunk(normalize_token(tokens[token_index]), self.prefix_size, 0)
class _VisibleDeprecationTestCase(_DeprecationTestCase): warning_cls = np.VisibleDeprecationWarning
def getProductions(code): stream = antlr4.InputStream(code) lexer = JavaLexer(stream) toks = antlr4.CommonTokenStream(lexer) parser = JavaParserModified(toks) tree = parser.memberDeclaration() st = [] st.append(tree) rule_seq = [] while (len(st) > 0): top = st.pop() (name, typ) = nname(top) if (name == 'ErrorN'): return None if (typ == 'T'): pass else: rule = getRuleAtNode(top) rule_seq.append(rule) for i in range((top.getChildCount() - 1), (- 1), (- 1)): st.append(top.getChild(i)) return rule_seq[6:]
def convert_to_color(arr_2d, palette): arr_3d = np.zeros((arr_2d.shape[0], arr_2d.shape[1], 3), dtype=np.uint8) for (c, i) in palette.items(): m = (arr_2d == c) arr_3d[m] = i return arr_3d
def get_noise(data, dist='G', noise_std=(float(25) / 255.0), mode='S', min_noise=(float(5) / 255.0), max_noise=(float(55) / 255.0)): if (dist == 'G'): noise_std /= 255.0 min_noise /= 255.0 max_noise /= 255.0 noise = torch.randn_like(data) if (mode == 'B'): n = noise.shape[0] noise_tensor_array = (((max_noise - min_noise) * torch.rand(n)) + min_noise) for i in range(n): noise.data[i] = (noise.data[i] * noise_tensor_array[i]) else: noise.data = (noise.data * noise_std) elif (dist == 'P'): noise = torch.randn_like(data) if (mode == 'S'): noise_std /= 255.0 noise = ((torch.poisson((data * noise_std)) / noise_std) - data) return noise
class GaloisGroup_v2(GaloisGroup_perm): def __init__(self, number_field, algorithm='pari', names=None, gc_numbering=None, _type=None): if (not number_field.is_absolute()): deprecation(28782, 'Use .absolute_field().galois_group() if you want the Galois group of the absolute field') if (gc_numbering is None): gc_numbering = (algorithm != 'magma') self._type = _type super().__init__(number_field, algorithm, names, gc_numbering) _method(key=GaloisGroup_perm._get_algorithm) def _pol_galgp(self, algorithm=None): algorithm = self._get_algorithm(algorithm) f = self._field.absolute_polynomial() pari_group = (self._type != 'gap') return f.galois_group(pari_group=pari_group, algorithm=algorithm) def group(self): deprecation(28782, 'the group method is deprecated; you can use _pol_galgp if you really need it') return self._pol_galgp() _method(key=_alg_key) def order(self, algorithm=None, recompute=False): algorithm = self._get_algorithm(algorithm) K = self._field if ((K.absolute_degree() < 12) or (algorithm != 'pari')): return self._pol_galgp(algorithm=algorithm).order() else: return self._galois_closure.absolute_degree() def easy_order(self, algorithm=None): algorithm = self._get_algorithm(algorithm) if self.order.cache: return next(iter(self.order.cache.values())) K = self._field if ((K.absolute_degree() < 12) or (algorithm != 'pari')): size = self._pol_galgp(algorithm=algorithm).order() self.order.cache[None] = size return size _method(key=_alg_key) def transitive_number(self, algorithm=None, recompute=False): algorithm = self._get_algorithm(algorithm) K = self._field if ((K.absolute_degree() < 12) or (algorithm != 'pari')): return self._pol_galgp(algorithm=algorithm).transitive_number() else: if self._gc_numbering: G = self._field.galois_group(algorithm=self._default_algorithm, names=self._gc_names, gc_numbering=False) else: G = self return ZZ(G.gap().TransitiveIdentification()) def pari_label(self): return self._pol_galgp().label() _method def signature(self): if (self._field.absolute_degree() < 12): return self._pol_galgp().signature() elif self._field.absolute_polynomial().discriminant().is_square(): return ZZ(1) else: return ZZ((- 1)) _attribute def _gcdata(self): K = self._field if self.is_galois(): return (K, K.hom(K.gen(), K)) else: if K.is_relative(): K = K.absolute_field((K.variable_name() + 'a')) (from_abs, to_abs) = K.structure() else: to_abs = None (L, emb) = K.galois_closure(names=self._gc_names, map=True) if (to_abs is not None): emb = (emb * to_abs) return (L, emb) _attribute def _pari_data(self): return self._galois_closure.__pari__().galoisinit() _attribute def _elts(self): if self._gc_numbering: return sorted([self(x, check=False) for x in self._pari_data[5]]) else: return sorted(list(self.iteration())) _attribute def _gens(self): if self._gc_numbering: gens = [standardize_generator(x, as_cycles=True) for x in self._pari_data[6]] if (not gens): gens = [()] gens = [self.element_class(x, self, check=False) for x in gens] return sorted(set(gens)) else: G = self._field.galois_group(algorithm=self._default_algorithm, names=self._gc_names, gc_numbering=True) self._galois_closure = L = G._galois_closure gens = [g.as_hom() for g in G._gens] if gens: roots = ([None] + self._field.absolute_polynomial().roots(L, multiplicities=False)) new_gens = [] for g in gens: seen = set() cycles = [] for start in range(1, len(roots)): if (start in seen): continue cycle = [start] r = roots[start] while True: r = g(r) i = roots.index(r) seen.add(i) if (i == start): break cycle.append(i) cycles.append(tuple(cycle)) new_gens.append(cycles) else: new_gens = [()] return [self.element_class(x, self, check=False) for x in new_gens] def _element_constructor_(self, x, check=True): if (x == 1): return self.identity() if (isinstance(x, NumberFieldHomomorphism_im_gens) and (x.parent() == self.number_field().Hom(self.number_field()))): l = [g for g in self if (g.as_hom() == x)] if (len(l) != 1): raise ArithmeticError return l[0] return self.element_class(x, self, check=check) def is_galois(self): K = self._field d = K.absolute_degree() if (d < 12): return (self._pol_galgp().order() == d) else: return (len(K.automorphisms()) == d) def _repr_(self): K = self.number_field() f = K.defining_polynomial() d = K.absolute_degree() if (d < 12): plabel = self.pari_label().split('=')[(- 1)].strip() tlabel = ('%sT%s (%s) with order %s ' % (d, self.transitive_number(), plabel, self.order())) else: tlabel = '' if ((d < 12) or self.is_galois()): return ('Galois group %sof %s' % (tlabel, f)) else: return ('Galois group %sof (non-Galois) %s' % (tlabel, f)) def number_field(self): return self._field def list(self): return self._elts def unrank(self, i): return self._elts[i] def __iter__(self): return iter(self._elts) _method def _ramgroups(self, P): K = self.number_field() P = K.ideal_monoid()(P).pari_prime() return pari(K).idealramgroups(self._pari_data, P) def decomposition_group(self, P): if (not self.is_galois()): raise TypeError('Decomposition groups only defined for Galois extensions') if isinstance(P, NumberFieldHomomorphism_im_gens): if self.number_field().is_totally_real(): return self.subgroup([]) else: return self.subgroup([self.complex_conjugation(P)]) else: return self.ramification_group(P, (- 1)) def complex_conjugation(self, P=None): if (P is None): Q = self.number_field().specified_complex_embedding() if (Q is None): raise ValueError('No default complex embedding specified') P = Q P = refine_embedding(P, infinity) if (not self.number_field().is_galois()): raise TypeError('Extension is not Galois') if self.number_field().is_totally_real(): raise TypeError('No complex conjugation (field is real)') g = self.number_field().gen() gconj = P(g).conjugate() elts = [s for s in self if (P(s(g)) == gconj)] if (len(elts) != 1): raise ArithmeticError('Something has gone very wrong here') return elts[0] def ramification_group(self, P, v): if (not self.is_galois()): raise TypeError('Ramification groups only defined for Galois extensions') ramdata = self._ramgroups(P) if (v < (- 1)): raise ValueError('v must be at least -1') elif ((v + 1) >= len(ramdata)): return self.subgroup([]) else: return self.subgroup(ramdata[(v + 1)][0]) def inertia_group(self, P): if (not self.is_galois()): raise TypeError('Inertia groups only defined for Galois extensions') return self.ramification_group(P, 0) def ramification_breaks(self, P): if (not self.is_galois()): raise TypeError('Ramification breaks only defined for Galois extensions') ramdata = self._ramgroups(P) n = len(ramdata) from sage.sets.set import Set return Set(([(i - 1) for i in range((n - 1)) if (ramdata[i][1] != ramdata[(i + 1)][1])] + [(n - 2)])) def artin_symbol(self, P): if (not self.is_galois()): raise TypeError('Artin symbols only defined for Galois extensions') P = self.number_field().ideal_monoid()(P) if (not P.is_prime()): raise ValueError(('%s is not prime' % P)) p = P.smallest_integer() t = [] gens = self.number_field().ring_of_integers().ring_generators() for s in self.decomposition_group(P): w = [(s(g) - (g ** p)).valuation(P) for g in gens] if (min(w) >= 1): t.append(s) if (len(t) > 1): raise ValueError(('%s is ramified' % P)) return t[0]
def add_comm_rewrites(ctx: LeanGenContext, expr: Expression) -> List[str]: return [((('add_comm ' + a_expr) + ' ') + b_expr) for (a_expr, b_expr) in get_reversed_add_exprs(expr=expr, simplifier=ctx.simplifier)]
def create_wham_whamr_csv(datapath, savepath, fs, version='min', savename='whamr_', set_types=['tr', 'cv', 'tt'], add_reverb=True, task='separation', dereverberate=True): if (fs == 8000): sample_rate = '8k' elif (fs == 16000): sample_rate = '16k' else: raise ValueError('Unsupported sampling rate') for set_type in set_types: if add_reverb: mix_both = ('mix_both_reverb/' if (task == 'separation') else 'mix_single_reverb/') if (dereverberate and (set_type != 'tr')): s1 = 's1_reverb/' s2 = 's2_reverb/' else: s1 = 's1_anechoic/' s2 = 's2_anechoic/' else: mix_both = ('mix_both/' if (task == 'separation') else 'mix_single/') s1 = 's1/' s2 = 's2/' mix_path = os.path.join(datapath, 'wav{}'.format(sample_rate), version, set_type, mix_both) s1_path = os.path.join(datapath, 'wav{}'.format(sample_rate), version, set_type, s1) s2_path = os.path.join(datapath, 'wav{}'.format(sample_rate), version, set_type, s2) noise_path = os.path.join(datapath, 'wav{}'.format(sample_rate), version, set_type, 'noise/') files = os.listdir(mix_path) mix_fl_paths = [(mix_path + fl) for fl in files] s1_fl_paths = [(s1_path + fl) for fl in files] s2_fl_paths = [(s2_path + fl) for fl in files] noise_fl_paths = [(noise_path + fl) for fl in files] csv_columns = ['ID', 'duration', 'mix_wav', 'mix_wav_format', 'mix_wav_opts', 's1_wav', 's1_wav_format', 's1_wav_opts', 's2_wav', 's2_wav_format', 's2_wav_opts', 'noise_wav', 'noise_wav_format', 'noise_wav_opts'] with open(os.path.join(savepath, ((savename + set_type) + '.csv')), 'w') as csvfile: writer = csv.DictWriter(csvfile, fieldnames=csv_columns) writer.writeheader() for (i, (mix_path, s1_path, s2_path, noise_path)) in enumerate(zip(mix_fl_paths, s1_fl_paths, s2_fl_paths, noise_fl_paths)): row = {'ID': i, 'duration': 1.0, 'mix_wav': mix_path, 'mix_wav_format': 'wav', 'mix_wav_opts': None, 's1_wav': s1_path, 's1_wav_format': 'wav', 's1_wav_opts': None, 's2_wav': s2_path, 's2_wav_format': 'wav', 's2_wav_opts': None, 'noise_wav': noise_path, 'noise_wav_format': 'wav', 'noise_wav_opts': None} writer.writerow(row)
class AutoTokenCounter(TokenCounter): def __init__(self, huggingface_tokenizer: HuggingFaceTokenizer): self.token_counters: Dict[(str, TokenCounter)] = {} self.huggingface_tokenizer: HuggingFaceTokenizer = huggingface_tokenizer def get_token_counter(self, organization: str) -> TokenCounter: token_counter = self.token_counters.get(organization) if (token_counter is None): if (organization == 'openai'): token_counter = OpenAITokenCounter(self.huggingface_tokenizer) elif (organization == 'ai21'): token_counter = AI21TokenCounter() elif (organization == 'gooseai'): token_counter = GooseAITokenCounter() elif (organization == 'cohere'): token_counter = CohereTokenCounter() else: token_counter = FreeTokenCounter() self.token_counters[organization] = token_counter return token_counter def count_tokens(self, request: Request, completions: List[Sequence]) -> int: token_counter: TokenCounter = self.get_token_counter(request.model_host) return token_counter.count_tokens(request, completions)
class RandomDataset(data.Dataset): def __init__(self, num_random=10000, shape=(3, 224, 224)): self.size = num_random self.shape = shape def __len__(self): return self.size def __repr__(self): return self.__class__.__name__ def __getitem__(self, index): img = torch.rand(*self.shape) target = 0 return (F.normalize(img, normalizing_mean, normalizing_std), target)
def new_query(table: Table, ncols) -> Query: return Query(predicates=OrderedDict.fromkeys(table.data.columns, None), ncols=ncols)
def read_from_hdf5(fd, group, cache=None): if (cache is None): cache = {} from sfepy.discrete.iga.domain import IGDomain from sfepy.discrete.fem.meshio import HDF5MeshIO types = {b'True': True, b'False': False, b'None': None} def _read_from_hdf5(group): while isinstance(group, pt.link.SoftLink): group = group() path = path_of_hdf5_group(group) if (path in cache): return cache[path] cache_ids = tuple() if ('cache' in group): cache_ids = group.cache.read().reshape((- 1)) cache_ids = [dec(i) for i in cache_ids] for cache_id in cache_ids: if (cache_id in cache): out = cache[cache_id] break else: out = _read_value_from_hdf5(group) else: out = _read_value_from_hdf5(group) if (not isinstance(out, (int, float, str, bool, None.__class__))): cache[path] = out for cache_id in cache_ids: cache[cache_id] = out return out def _read_value_from_hdf5(group): type = group.type.read().item() if (type in types): return types[type] data = group.data while isinstance(data, pt.link.SoftLink): data = data() def load_list(): out = ([None] * group.len.read()) for i in data: out[int(i._v_name)] = _read_from_hdf5(i) return out def load_dict(): out = {} for i in fd.iter_nodes(data): out[i._v_name] = _read_from_hdf5(i) return out def load_object(fn=None): if fn: out = fn(fd, data) else: out = data.read() return out if (type == b'raw'): return data.read() if (type == b'object'): return load_object() if (type == b'str'): return dec(data.read().item()) if (type == b'pickle'): return pickle.loads(data.read().item()) if (type == b'dict'): return load_dict() if (type == b'Struct'): return Struct(**load_dict()) if (type == b'list'): return load_list() if (type == b'tuple'): return tuple(load_list()) if (type == b'sparse_matrix'): return load_object(read_sparse_matrix_from_hdf5) if (type == b'IGDomain'): return load_object(IGDomain.read_domain_from_hdf5) if (type == b'Mesh'): return load_object(HDF5MeshIO.read_mesh_from_hdf5) raise Exception('Unknown h5 group type {}'.format(type.decode('utf8'))) return _read_from_hdf5(group)
def batch_transformer(U, thetas, out_size, name='BatchSpatialTransformer'): with tf.variable_scope(name): (num_batch, num_transforms) = map(int, thetas.get_shape().as_list()[:2]) indices = [([i] * num_transforms) for i in xrange(num_batch)] input_repeated = tf.gather(U, tf.reshape(indices, [(- 1)])) return transformer(input_repeated, thetas, out_size)
def Jacobian(C): try: return C.jacobian() except AttributeError: return Jacobian_generic(C)
def sum_task(mixture_or_task_name, dataset_split='train', add_percentiles=True): sequence_length = {'inputs': 512, 'targets': 512} df_packing = analyze_packing(mixture_or_task_name=mixture_or_task_name, sequence_length=sequence_length, dataset_split=dataset_split) df_padding = analyze_padding(mixture_or_task_name=mixture_or_task_name, sequence_length=sequence_length, dataset_split=dataset_split) print((40 * '=')) print('-I- mixture_or_task_name', mixture_or_task_name) print('-I- packing:') print(df_packing.describe(percentiles=[0.5, 0.75, 0.9, 0.99])) print('-I- padding:') described_padding = df_padding.describe(percentiles=[0.5, 0.75, 0.9, 0.99]) print(described_padding) splits = ['train'] npacked = df_packing['npacked'].mean() ntrain = len(df_padding) sequence_length_req = infer_no_truncation_padding_seq_length(df_padding) record = {'mixture_or_task_name': mixture_or_task_name, 'max_input': sequence_length_req['inputs'], 'max_targets': sequence_length_req['targets'], 'npacked': npacked, 'examples': ntrain} if add_percentiles: percs_input = {f'input_seq_length_{i}%': described_padding['input_seq_length'][f'{i}%'] for i in [50, 75, 90, 99]} percs_target = {f'target_seq_length_{i}%': described_padding['target_seq_length'][f'{i}%'] for i in [50, 75, 90, 99]} record.update(percs_input) record.update(percs_target) print('-I summary:') pprint(record) print((40 * '=')) return record
def test_phi_plus_phi_plus(): for i in range(400): (k1, k2, k3, k4, a3) = create_scenario(phi_plus, phi_plus, i) state = correct_order(k1.state, k1.keys) assert numpy.array_equal(state, phi_plus)
def read_relation_from_id(filename='./data/WN18RR/relation2id.txt'): relation2id = {} with open(filename, 'r') as f: for line in f: if (len(line.strip().split()) > 1): (relation, relation_id) = (line.strip().split()[0].strip(), line.strip().split()[1].strip()) relation2id[relation] = int(relation_id) return relation2id
def __lagrange_bounds_phc(n, m, a, tmpfile=None): S = coefficients_to_power_sums(n, m, a) (fi, fo) = os.popen2('which phc') find_phc = fo.readlines() fi.close() fo.close() if (find_phc == []): raise RuntimeError('PHCpack not installed.') if (tmpfile is None): tmpfile = sage.misc.misc.tmp_filename() f = open((tmpfile + '.phc'), 'w') f.close() output_data = [] for P in sage.combinat.partition.Partitions((n - 1), length=(m - 1)): f = open(tmpfile, 'w') f.write((('%d' % m) + '\n')) for j in range(1, (m + 1)): for i in range((m - 1)): f.write((((('%d' % P[i]) + ('*x%d' % i)) + ('**%d' % j)) + ' + ')) f.write(((('xn**%d' % j) + (' - (%d' % S[j])) + ');\n')) f.close() os.remove((tmpfile + '.phc')) os.popen((((('phc -b ' + tmpfile) + ' ') + tmpfile) + '.phc')) f = open((tmpfile + '.phc'), 'r') f_str = f.read() pos = f_str.find('= real ') crits = [] while (pos != (- 1)): posl = f_str.rfind('xn', 0, pos) f_str_split = f_str[posl:pos].split() crits += [float(f_str_split[2])] pos = f_str.find('= real ', (pos + 1)) if (len(crits) > 0): output_data += [[P, min(crits), max(crits)]] if (len(output_data) > 0): return [min([v[1] for v in output_data]), max([v[2] for v in output_data])] else: return []
def _nls_subproblem(X, W, H, tol, max_iter, alpha=0.0, l1_ratio=0.0, sigma=0.01, beta=0.1): WtX = safe_sparse_dot(W.T, X) WtW = np.dot(W.T, W) gamma = 1 for n_iter in range(1, (max_iter + 1)): grad = (np.dot(WtW, H) - WtX) if ((alpha > 0) and (l1_ratio == 1.0)): grad += alpha elif (alpha > 0): grad += (alpha * (l1_ratio + ((1 - l1_ratio) * H))) if (_norm((grad * np.logical_or((grad < 0), (H > 0)))) < tol): break Hp = H for inner_iter in range(20): Hn = (H - (gamma * grad)) Hn *= (Hn > 0) d = (Hn - H) gradd = np.dot(grad.ravel(), d.ravel()) dQd = np.dot(np.dot(WtW, d).ravel(), d.ravel()) suff_decr = ((((1 - sigma) * gradd) + (0.5 * dQd)) < 0) if (inner_iter == 0): decr_gamma = (not suff_decr) if decr_gamma: if suff_decr: H = Hn break else: gamma *= beta elif ((not suff_decr) or (Hp == Hn).all()): H = Hp break else: gamma /= beta Hp = Hn if (n_iter == max_iter): warnings.warn('Iteration limit reached in nls subproblem.', ConvergenceWarning) return (H, grad, n_iter)
class PathScaleCCompiler(UnixCCompiler): compiler_type = 'pathcc' cc_exe = 'pathcc' cxx_exe = 'pathCC' def __init__(self, verbose=0, dry_run=0, force=0): UnixCCompiler.__init__(self, verbose, dry_run, force) cc_compiler = self.cc_exe cxx_compiler = self.cxx_exe self.set_executables(compiler=cc_compiler, compiler_so=cc_compiler, compiler_cxx=cxx_compiler, linker_exe=cc_compiler, linker_so=(cc_compiler + ' -shared'))
def load_checkpoint(step, model, optimizer, scheduler): global global_step global global_epoch checkpoint_path = os.path.join(args.save, args.model_name, 'checkpoint_step{:09d}.pth'.format(step)) print('Load checkpoint from: {}'.format(checkpoint_path)) checkpoint = torch.load(checkpoint_path) try: model.load_state_dict(checkpoint['state_dict']) except RuntimeError: print('INFO: this model is trained with DataParallel. Creating new state_dict without module...') state_dict = checkpoint['state_dict'] from collections import OrderedDict new_state_dict = OrderedDict() for (k, v) in state_dict.items(): name = k[7:] new_state_dict[name] = v model.load_state_dict(new_state_dict) optimizer.load_state_dict(checkpoint['optimizer']) scheduler.load_state_dict(checkpoint['scheduler']) global_step = checkpoint['global_step'] global_epoch = checkpoint['global_epoch'] return (model, optimizer, scheduler)
def save_args(original_args): reversed_trainer_log_levels = {v: k for (k, v) in trainer_log_levels.items()} original_args['log_level'] = reversed_trainer_log_levels[original_args['log_level']] original_args['log_level_replica'] = reversed_trainer_log_levels[original_args['log_level_replica']] for arg in ['_n_gpu', 'local_rank']: if (arg in original_args): del original_args[arg] to_delete = [] for (k, v) in original_args.items(): if (v is None): to_delete.append(k) if (original_args[k] == ''): to_delete.append(k) for k in to_delete: del original_args[k] with open(os.path.join(original_args['output_dir'], 'args.json'), mode='w') as f: json.dump(original_args, f, indent=4)
def get_from_to_our_keys(model_name: str) -> Dict[(str, str)]: our_config = RegNetConfig(depths=[2, 7, 17, 1], hidden_sizes=[8, 8, 8, 8], groups_width=8) if ('in1k' in model_name): our_model = RegNetForImageClassification(our_config) else: our_model = RegNetModel(our_config) from_model = FakeRegNetVisslWrapper(RegNet(FakeRegNetParams(depth=27, group_width=1010, w_0=1744, w_a=620.83, w_m=2.52))) with torch.no_grad(): from_model = from_model.eval() our_model = our_model.eval() x = torch.randn((1, 3, 32, 32)) dest_tracker = Tracker(our_model) dest_traced = dest_tracker(x).parametrized pprint(dest_tracker.name2module) src_tracker = Tracker(from_model) src_traced = src_tracker(x).parametrized def to_params_dict(dict_with_modules): params_dict = OrderedDict() for (name, module) in dict_with_modules.items(): for (param_name, param) in module.state_dict().items(): params_dict[f'{name}.{param_name}'] = param return params_dict from_to_ours_keys = {} src_state_dict = to_params_dict(src_traced) dst_state_dict = to_params_dict(dest_traced) for ((src_key, src_param), (dest_key, dest_param)) in zip(src_state_dict.items(), dst_state_dict.items()): from_to_ours_keys[src_key] = dest_key logger.info(f'{src_key} -> {dest_key}') if ('in1k' in model_name): from_to_ours_keys['0.clf.0.weight'] = 'classifier.1.weight' from_to_ours_keys['0.clf.0.bias'] = 'classifier.1.bias' return from_to_ours_keys
class TransductiveFinetuning(Finetune): def __init__(self, *args, fine_tuning_steps: int=25, fine_tuning_lr: float=5e-05, temperature: float=1.0, **kwargs): super().__init__(*args, fine_tuning_steps=fine_tuning_steps, fine_tuning_lr=fine_tuning_lr, temperature=temperature, **kwargs) def forward(self, query_images: Tensor) -> Tensor: query_features = self.compute_features(query_images) with torch.enable_grad(): self.prototypes.requires_grad_() optimizer = torch.optim.Adam([self.prototypes], lr=self.fine_tuning_lr) for _ in range(self.fine_tuning_steps): support_cross_entropy = nn.functional.cross_entropy((self.temperature * self.l2_distance_to_prototypes(self.support_features)), self.support_labels) query_conditional_entropy = entropy((self.temperature * self.l2_distance_to_prototypes(query_features))) loss = (support_cross_entropy + query_conditional_entropy) optimizer.zero_grad() loss.backward() optimizer.step() return self.softmax_if_specified(self.l2_distance_to_prototypes(query_features), temperature=self.temperature).detach() def is_transductive() -> bool: return True
def test_argcombinations(): array = ak.Array([[0.0, 1.1, 2.2, 3.3], [], [4.4, 5.5, 6.6], [7.7], [8.8, 9.9, 10.0, 11.1, 12.2]]) assert (to_list(ak.operations.argcombinations(array, 2, replacement=False)) == [[(0, 1), (0, 2), (0, 3), (1, 2), (1, 3), (2, 3)], [], [(0, 1), (0, 2), (1, 2)], [], [(0, 1), (0, 2), (0, 3), (0, 4), (1, 2), (1, 3), (1, 4), (2, 3), (2, 4), (3, 4)]])
class AI21TokenCostEstimator(TokenCostEstimator): def estimate_tokens(self, request: Request, metric_service: MetricService) -> int: return (request.num_completions * request.max_tokens)
class CosineClassifier(_Classifier): def __init__(self, feat_dim=None, num_classes=None, dtype=None, scale=30, **kwargs): super().__init__(feat_dim, num_classes, dtype) self.scale = scale def forward(self, x): x = F.normalize(x, dim=(- 1)) weight = F.normalize(self.weight, dim=(- 1)) return (F.linear(x, weight) * self.scale)
class BQCorpusBertPipe(MatchingBertPipe): def __init__(self, tokenizer='cn-char'): super().__init__(tokenizer=tokenizer) def process_from_file(self, paths=None): data_bundle = BQCorpusLoader().load(paths) data_bundle = RenamePipe(task='cn-nli-bert').process(data_bundle) data_bundle = self.process(data_bundle) data_bundle = TruncateBertPipe(task='cn').process(data_bundle) data_bundle = RenamePipe(task='cn-nli-bert').process(data_bundle) return data_bundle
class Action(Enum): opened = 'opened' reopened = 'reopened' closed = 'closed' labeled = 'labeled' unlabeled = 'unlabeled' ready_for_review = 'ready_for_review' synchronize = 'synchronize' review_requested = 'review_requested' converted_to_draft = 'converted_to_draft' submitted = 'submitted'
def train(args, trainer, task, epoch_itr): update_freq = (args.update_freq[(epoch_itr.epoch - 1)] if (epoch_itr.epoch <= len(args.update_freq)) else args.update_freq[(- 1)]) itr = epoch_itr.next_epoch_itr(fix_batches_to_gpus=args.fix_batches_to_gpus, shuffle=(epoch_itr.epoch >= args.curriculum)) itr = iterators.GroupedIterator(itr, update_freq) progress = progress_bar.build_progress_bar(args, itr, epoch_itr.epoch, no_progress_bar='simple') extra_meters = collections.defaultdict((lambda : AverageMeter())) valid_subsets = args.valid_subset.split(',') max_update = (args.max_update or math.inf) for (i, samples) in enumerate(progress, start=epoch_itr.iterations_in_epoch): log_output = trainer.train_step(samples) if (log_output is None): continue stats = get_training_stats(trainer) for (k, v) in log_output.items(): if (k in ['loss', 'nll_loss', 'ntokens', 'nsentences', 'sample_size']): continue if (('loss' in k) or (k == 'accuracy')): extra_meters[k].update(v, log_output['sample_size']) else: extra_meters[k].update(v) stats[k] = extra_meters[k].avg progress.log(stats, tag='train', step=stats['num_updates']) if (i == 0): trainer.get_meter('wps').reset() trainer.get_meter('ups').reset() num_updates = trainer.get_num_updates() if ((not args.disable_validation) and (args.save_interval_updates > 0) and ((num_updates % args.save_interval_updates) == 0) and (num_updates > 0)): valid_losses = validate(args, trainer, task, epoch_itr, valid_subsets) checkpoint_utils.save_checkpoint(args, trainer, epoch_itr, valid_losses[0]) if (num_updates >= max_update): break stats = get_training_stats(trainer) for (k, meter) in extra_meters.items(): stats[k] = meter.avg progress.print(stats, tag='train', step=stats['num_updates']) for k in ['train_loss', 'train_nll_loss', 'wps', 'ups', 'wpb', 'bsz', 'gnorm', 'clip']: meter = trainer.get_meter(k) if (meter is not None): meter.reset()
def register_Ns3DsrDsrMaintainBuffEntry_methods(root_module, cls): cls.add_constructor([param('ns3::dsr::DsrMaintainBuffEntry const &', 'arg0')]) cls.add_constructor([param('ns3::Ptr< ns3::Packet const >', 'pa', default_value='0'), param('ns3::Ipv4Address', 'us', default_value='ns3::Ipv4Address()'), param('ns3::Ipv4Address', 'n', default_value='ns3::Ipv4Address()'), param('ns3::Ipv4Address', 's', default_value='ns3::Ipv4Address()'), param('ns3::Ipv4Address', 'dst', default_value='ns3::Ipv4Address()'), param('uint16_t', 'ackId', default_value='0'), param('uint8_t', 'segs', default_value='0'), param('ns3::Time', 'exp', default_value='ns3::Simulator::Now()')]) cls.add_method('GetAckId', 'uint16_t', [], is_const=True) cls.add_method('GetDst', 'ns3::Ipv4Address', [], is_const=True) cls.add_method('GetExpireTime', 'ns3::Time', [], is_const=True) cls.add_method('GetNextHop', 'ns3::Ipv4Address', [], is_const=True) cls.add_method('GetOurAdd', 'ns3::Ipv4Address', [], is_const=True) cls.add_method('GetPacket', 'ns3::Ptr< ns3::Packet const >', [], is_const=True) cls.add_method('GetSegsLeft', 'uint8_t', [], is_const=True) cls.add_method('GetSrc', 'ns3::Ipv4Address', [], is_const=True) cls.add_method('SetAckId', 'void', [param('uint16_t', 'ackId')]) cls.add_method('SetDst', 'void', [param('ns3::Ipv4Address', 'n')]) cls.add_method('SetExpireTime', 'void', [param('ns3::Time', 'exp')]) cls.add_method('SetNextHop', 'void', [param('ns3::Ipv4Address', 'n')]) cls.add_method('SetOurAdd', 'void', [param('ns3::Ipv4Address', 'us')]) cls.add_method('SetPacket', 'void', [param('ns3::Ptr< ns3::Packet const >', 'p')]) cls.add_method('SetSegsLeft', 'void', [param('uint8_t', 'segs')]) cls.add_method('SetSrc', 'void', [param('ns3::Ipv4Address', 's')]) return
def keypoint_rcnn(model): logger.warn('Deprecated: use `MODEL.TYPE: generalized_rcnn` with `MODEL.KEYPOINTS_ON: True`') return generalized_rcnn(model)
class Attention(nn.Module): def __init__(self, input_dim, hidden_dim, attn_channel, kernel_size): super(Attention, self).__init__() self.kernel_size = kernel_size self.padding = (kernel_size // 2) self.H = nn.Conv2d(in_channels=hidden_dim, out_channels=attn_channel, kernel_size=kernel_size, padding=self.padding, bias=False) self.W = nn.Conv2d(in_channels=input_dim, out_channels=attn_channel, kernel_size=kernel_size, padding=self.padding, bias=False) self.V = nn.Conv2d(in_channels=attn_channel, out_channels=1, kernel_size=kernel_size, padding=self.padding, bias=False) def forward(self, input_tensor, hidden): hid_conv_out = self.H(hidden[0]) in_conv_out = self.W(input_tensor) energy = self.V((hid_conv_out + in_conv_out).tanh()) return energy
def main(): data = load_pickle(args.data_path) query_cam = data['query_cam'] query_label = data['query_label'] gallery_cam = data['gallery_cam'] gallery_label = data['gallery_label'] gallery_feature = torch.FloatTensor(data['gallery_f']) query_feature = torch.FloatTensor(data['query_f']) query_feature = query_feature.cuda() gallery_feature = gallery_feature.cuda() indices = gnn_reranking(query_feature, gallery_feature, args.k1, args.k2) evaluate_ranking_list(indices, query_label, query_cam, gallery_label, gallery_cam)
((not workspace.C.has_mkldnn), 'Skipping as we do not have mkldnn.') class MKLReluTest(hu.HypothesisTestCase): (size=st.integers(8, 20), input_channels=st.integers(1, 3), batch_size=st.integers(1, 3), inplace=st.booleans(), **mu.gcs) def test_mkl_relu(self, size, input_channels, batch_size, inplace, gc, dc): op = core.CreateOperator('Relu', ['X'], (['Y'] if (not inplace) else ['X'])) X = (np.random.rand(batch_size, input_channels, size, size).astype(np.float32) - 0.5) self.assertDeviceChecks(dc, op, [X], [0])
def dag2pag(dag, islatent): udg = nx.Graph() nodes = dag.get_nodes() nodes_ids = {node: i for (i, node) in enumerate(nodes)} n = len(nodes) for (x, y) in combinations(range(n), 2): if dag.get_edge(nodes[x], nodes[y]): udg.add_edge(x, y) observed_nodes = list((set(nodes) - set(islatent))) PAG = GeneralGraph(observed_nodes) for (nodex, nodey) in combinations(observed_nodes, 2): edge = Edge(nodex, nodey, Endpoint.CIRCLE, Endpoint.CIRCLE) edge.set_endpoint1(Endpoint.CIRCLE) edge.set_endpoint2(Endpoint.CIRCLE) PAG.add_edge(edge) sepset = {(nodex, nodey): set() for (nodex, nodey) in permutations(observed_nodes, 2)} for (nodex, nodey) in combinations(observed_nodes, 2): if (nodex in islatent): continue if (nodey in islatent): continue all_paths = nx.all_simple_paths(udg, nodes_ids[nodex], nodes_ids[nodey]) noncolider_path = [] is_connected = False for path in all_paths: path_sep = True has_nonlatent = False for i in range(1, (len(path) - 1)): if (nodes[path[i]] in observed_nodes): has_nonlatent = True has_collider = (is_endpoint(dag.get_edge(nodes[path[(i - 1)]], nodes[path[i]]), nodes[path[i]], Endpoint.ARROW) and is_endpoint(dag.get_edge(nodes[path[(i + 1)]], nodes[path[i]]), nodes[path[i]], Endpoint.ARROW)) if has_collider: path_sep = False if (not path_sep): continue if has_nonlatent: noncolider_path.append(path) else: is_connected = True break if (not is_connected): edge = PAG.get_edge(nodex, nodey) if edge: PAG.remove_edge(edge) for path in noncolider_path: for i in range(1, (len(path) - 1)): if (nodes[path[i]] in islatent): continue sepset[(nodex, nodey)] |= {nodes[path[i]]} sepset[(nodey, nodex)] |= {nodes[path[i]]} for (nodex, nodey) in combinations(observed_nodes, 2): if PAG.get_edge(nodex, nodey): continue for nodez in observed_nodes: if (nodez == nodex): continue if (nodez == nodey): continue if (nodez not in sepset[(nodex, nodey)]): edge_xz = PAG.get_edge(nodex, nodez) edge_yz = PAG.get_edge(nodey, nodez) if (edge_xz and edge_yz): PAG.remove_edge(edge_xz) mod_endpoint(edge_xz, nodez, Endpoint.ARROW) PAG.add_edge(edge_xz) PAG.remove_edge(edge_yz) mod_endpoint(edge_yz, nodez, Endpoint.ARROW) PAG.add_edge(edge_yz) changeFlag = True while changeFlag: changeFlag = False changeFlag = rulesR1R2cycle(PAG, None, changeFlag, False) changeFlag = ruleR3(PAG, sepset, None, changeFlag, False) return PAG
def list_s3_objects(bucket, name): s3_client = boto3.client('s3', aws_access_key_id=access_key, aws_secret_access_key=secret_key) res = s3_client.list_objects(Bucket=bucket, Prefix=name, MaxKeys=1) return res
def make_beta_schedule(schedule, n_timestep, linear_start=0.0001, linear_end=0.02, cosine_s=0.008): if (schedule == 'linear'): betas = (torch.linspace((linear_start ** 0.5), (linear_end ** 0.5), n_timestep, dtype=torch.float64) ** 2) elif (schedule == 'cosine'): timesteps = ((torch.arange((n_timestep + 1), dtype=torch.float64) / n_timestep) + cosine_s) alphas = (((timesteps / (1 + cosine_s)) * np.pi) / 2) alphas = torch.cos(alphas).pow(2) alphas = (alphas / alphas[0]) betas = (1 - (alphas[1:] / alphas[:(- 1)])) betas = np.clip(betas, a_min=0, a_max=0.999) elif (schedule == 'sqrt_linear'): betas = torch.linspace(linear_start, linear_end, n_timestep, dtype=torch.float64) elif (schedule == 'sqrt'): betas = (torch.linspace(linear_start, linear_end, n_timestep, dtype=torch.float64) ** 0.5) else: raise ValueError(f"schedule '{schedule}' unknown.") return betas.numpy()
def __getattr__(name): if (name == 'load_boston'): msg = textwrap.dedent('\n `load_boston` has been removed from scikit-learn since version 1.2.\n\n The Boston housing prices dataset has an ethical problem: as\n investigated in [1], the authors of this dataset engineered a\n non-invertible variable "B" assuming that racial self-segregation had a\n positive impact on house prices [2]. Furthermore the goal of the\n research that led to the creation of this dataset was to study the\n impact of air quality but it did not give adequate demonstration of the\n validity of this assumption.\n\n The scikit-learn maintainers therefore strongly discourage the use of\n this dataset unless the purpose of the code is to study and educate\n about ethical issues in data science and machine learning.\n\n In this special case, you can fetch the dataset from the original\n source::\n\n import pandas as pd\n import numpy as np\n\n data_url = " raw_df = pd.read_csv(data_url, sep="\\s+", skiprows=22, header=None)\n data = np.hstack([raw_df.values[::2, :], raw_df.values[1::2, :2]])\n target = raw_df.values[1::2, 2]\n\n Alternative datasets include the California housing dataset and the\n Ames housing dataset. You can load the datasets as follows::\n\n from sklearn.datasets import fetch_california_housing\n housing = fetch_california_housing()\n\n for the California housing dataset and::\n\n from sklearn.datasets import fetch_openml\n housing = fetch_openml(name="house_prices", as_frame=True)\n\n for the Ames housing dataset.\n\n [1] M Carlisle.\n "Racist data destruction?"\n < [2] Harrison Jr, David, and Daniel L. Rubinfeld.\n "Hedonic housing prices and the demand for clean air."\n Journal of environmental economics and management 5.1 (1978): 81-102.\n < ') raise ImportError(msg) try: return globals()[name] except KeyError: raise AttributeError
class model(): def __init__(self, curr_param): self.sess = tf.Session() self.training_step = 0 self.par = curr_param def load_data(self, mode): if (mode == 'train'): file_name = self.par.train_file self.training_data = [] else: file_name = self.par.test_file self.test_data = [] with open(file_name) as f: scans = f.readlines() scans = [s.rstrip() for s in scans] if (mode == 'train'): self.training_scans = scans else: self.test_scans = scans scans = [os.path.join(self.par.pre_output_dir, s.rstrip()) for s in scans] cnt = 0 for s_path in scans: if (mode == 'train'): rot = random.randint(0, (self.par.num_rotations - 1)) else: rot = 0 s = ScanData() s.load(os.path.join(s_path, str(rot)), self.par.num_scales) s.remap_depth(vmin=(- self.par.conv_rad[0]), vmax=self.par.conv_rad[0]) s.remap_normals() if (mode == 'train'): self.training_data.append(s) else: self.test_data.append(s) cnt += 1 def precompute_validation_batches(self): self.validation_batches = [] for test_scan in self.test_data: if (self.par.data_sampling_type == 'part'): batch_array = get_batch_array(test_scan, self.par) for b in batch_array: if (np.shape(b.colors[0])[0] <= self.par.batch_size): self.validation_batches.append(b) else: b = get_batch_from_full_scan(test_scan, self.par.num_scales, self.par.d_par.class_weights) if (np.shape(b.colors[0])[0] <= self.par.batch_size): self.validation_batches.append(b) def get_training_batch(self, iter_num): if (self.par.data_sampling_type == 'full'): num_train_scans = len(self.training_data) scan_num = (iter_num % num_train_scans) return get_batch_from_full_scan(self.training_data[scan_num], self.par.num_scales, self.par.d_par.class_weights) else: scan_num = (iter_num % self.par.batch_array_size) if (scan_num == 0): random_scan = random.randint(0, (len(self.training_data) - 1)) self.tr_batch_array = get_batch_array(self.training_data[random_scan], self.par) return self.tr_batch_array[scan_num] def get_feed_dict(self, b): bs = self.par.batch_size mask1 = get_pooling_mask(b.pool_ind[1]) mask2 = get_pooling_mask(b.pool_ind[2]) ret_dict = {self.c1_ind: expand_dim_to_batch2(b.conv_ind[0], bs), self.c2_ind: expand_dim_to_batch2(b.conv_ind[1], (bs // 2)), self.c3_ind: expand_dim_to_batch2(b.conv_ind[2], (bs // 4)), self.p12_ind: expand_dim_to_batch2(b.pool_ind[1], (bs // 2)), self.p12_mask: expand_dim_to_batch2(mask1, (bs // 2), dummy_val=0), self.p23_ind: expand_dim_to_batch2(b.pool_ind[2], (bs // 4)), self.p23_mask: expand_dim_to_batch2(mask2, (bs // 4), dummy_val=0), self.label: expand_dim_to_batch1(b.labels[0], bs), self.loss_weight: expand_dim_to_batch1(b.loss_weights[0], bs)} if ('d' in self.par.input_type): ret_dict.update({self.input_depth1: expand_dim_to_batch2(b.depth[0].T, bs)}) ret_dict.update({self.input_depth2: expand_dim_to_batch2(b.depth[1].T, (bs // 2))}) ret_dict.update({self.input_depth3: expand_dim_to_batch2(b.depth[2].T, (bs // 4))}) if ('n' in self.par.input_type): ret_dict.update({self.input_normals: expand_dim_to_batch2(b.normals[0], bs)}) if ('h' in self.par.input_type): ret_dict.update({self.input_h: expand_dim_to_batch2(b.height[0], bs)}) if ('c' in self.par.input_type): ret_dict.update({self.input_colors: expand_dim_to_batch2(b.colors[0], bs)}) return ret_dict def build_model(self, batch_size): self.best_accuracy = 0.0 fs = self.par.filter_size bs = batch_size num_input_ch = 0 input_list = [] if ('d' in self.par.input_type): num_input_ch += 1 self.input_depth1 = tf.placeholder(tf.float32, [bs, (fs * fs)]) self.input_depth2 = tf.placeholder(tf.float32, [(bs // 2), (fs * fs)]) self.input_depth3 = tf.placeholder(tf.float32, [(bs // 4), (fs * fs)]) if ('n' in self.par.input_type): num_input_ch += 3 self.input_normals = tf.placeholder(tf.float32, [bs, 3]) input_list.append(self.input_normals) if ('h' in self.par.input_type): num_input_ch += 1 self.input_h = tf.placeholder(tf.float32, [bs, 1]) input_list.append(self.input_h) if ('c' in self.par.input_type): num_input_ch += 3 self.input_colors = tf.placeholder(tf.float32, [bs, 3]) input_list.append(self.input_colors) self.c1_ind = tf.placeholder(tf.int32, [bs, (fs * fs)]) self.p12_ind = tf.placeholder(tf.int32, [(bs // 2), 8]) self.p12_mask = tf.placeholder(tf.float32, [(bs // 2), 8]) self.c2_ind = tf.placeholder(tf.int32, [(bs // 2), (fs * fs)]) self.p23_ind = tf.placeholder(tf.int32, [(bs // 4), 8]) self.p23_mask = tf.placeholder(tf.float32, [(bs // 4), 8]) self.c3_ind = tf.placeholder(tf.int32, [(bs // 4), (fs * fs)]) self.label = tf.placeholder(tf.int32, [bs]) self.loss_weight = tf.placeholder(tf.float32, [bs]) label_mask = tf.cast(self.label, tf.bool) shape_unpool2 = tf.constant([(bs // 2), 64]) shape_unpool1 = tf.constant([bs, 32]) if ('d' in self.par.input_type): if (num_input_ch > 1): signal_input = tf.concat(input_list, axis=1) h_conv1 = lrelu(point_conv('conv1', signal_input, self.c1_ind, (fs * fs), num_input_ch, 32, extra_chan=self.input_depth1)) else: signal_input = tf.expand_dims(tf.expand_dims(self.input_depth1, axis=2), axis=0) h_conv1 = lrelu(conv_2d_layer('conv1', signal_input, 1, 32, 1, (fs * fs), 1, 1, padding='VALID')) else: signal_input = tf.concat(input_list, axis=1) h_conv1 = lrelu(point_conv('conv1', signal_input, self.c1_ind, (fs * fs), num_input_ch, 32)) h_conv1 = tf.squeeze(h_conv1) h_conv11 = lrelu(point_conv('conv11', h_conv1, self.c1_ind, (fs * fs), 32, 32)) h_pool1 = point_pool(h_conv11, self.p12_ind, self.p12_mask) if ('d' in self.par.input_type): h_conv2 = lrelu(point_conv('conv2', h_pool1, self.c2_ind, (fs * fs), 33, 64, extra_chan=self.input_depth2)) else: h_conv2 = lrelu(point_conv('conv2', h_pool1, self.c2_ind, (fs * fs), 32, 64)) h_conv22 = lrelu(point_conv('conv22', h_conv2, self.c2_ind, (fs * fs), 64, 64)) h_pool2 = point_pool(h_conv22, self.p23_ind, self.p23_mask) if ('d' in self.par.input_type): h_conv3 = lrelu(point_conv('conv3', h_pool2, self.c3_ind, (fs * fs), 65, 128, extra_chan=self.input_depth3)) else: h_conv3 = lrelu(point_conv('conv3', h_pool2, self.c3_ind, (fs * fs), 64, 128)) h_conv33 = lrelu(point_conv('conv33', h_conv3, self.c3_ind, (fs * fs), 128, 64)) h_unpool2 = point_unpool(h_conv33, self.p23_ind, shape_unpool2) uconv2_in = tf.concat([h_conv22, h_unpool2], axis=1) h_uconv2 = lrelu(point_conv('uconv2', uconv2_in, self.c2_ind, (fs * fs), 128, 64)) h_uconv22 = lrelu(point_conv('uconv22', h_uconv2, self.c2_ind, (fs * fs), 64, 32)) h_unpool1 = point_unpool(h_uconv22, self.p12_ind, shape_unpool1) uconv1_in = tf.concat([h_conv11, h_unpool1], axis=1) h_uconv1 = lrelu(point_conv('uconv1', uconv1_in, self.c1_ind, (fs * fs), 64, 32)) h_uconv11 = tf.squeeze(point_conv('uconv11', h_uconv1, self.c1_ind, (fs * fs), 32, 32)) pred_input = tf.expand_dims(tf.expand_dims(h_uconv11, axis=1), axis=0) h_pred = tf.squeeze(conv_2d_layer('pred1', pred_input, 32, self.par.d_par.num_classes, 1, 1, 1, 1)) self.output = tf.argmax(h_pred, axis=1, output_type=tf.int32) masked_output = tf.boolean_mask(h_pred, label_mask) masked_label = tf.boolean_mask(self.label, label_mask) masked_weights = tf.boolean_mask(self.loss_weight, label_mask) tr_var = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES, '') self.loss = tf.reduce_mean(tf.multiply(masked_weights, tf.nn.sparse_softmax_cross_entropy_with_logits(labels=masked_label, logits=masked_output))) self.train_step = tf.train.AdamOptimizer(0.0001).minimize(self.loss) correct_prediction = tf.equal(tf.argmax(masked_output, axis=1, output_type=tf.int32), masked_label) self.accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32)) self.test_loss_placeholder = tf.placeholder(dtype=tf.float32, shape=[]) self.test_loss_summary = tf.summary.scalar('accuracy', self.test_loss_placeholder) self.train_loss_placeholder = tf.placeholder(dtype=tf.float32, shape=[]) self.train_loss_summary = tf.summary.scalar('train_loss', self.train_loss_placeholder) curr_time = strftime('%Y-%m-%d %H:%M:%S', gmtime()) self.writer = tf.summary.FileWriter(os.path.join(self.par.log_dir, curr_time)) self.saver = tf.train.Saver(tr_var, max_to_keep=self.par.max_snapshots) def initialize_model(self): self.sess.run(tf.global_variables_initializer()) def save_snapshot(self): self.saver.save(self.sess, os.path.join(self.par.snapshot_dir, 'model'), global_step=self.training_step) def load_snapshot(self): snapshot_name = tf.train.latest_checkpoint(self.par.snapshot_dir) if (snapshot_name is not None): model_file_name = os.path.basename(snapshot_name) print(('Loading snapshot ' + model_file_name)) itn = int(model_file_name.split('-')[1]) self.training_step = itn self.saver.restore(self.sess, snapshot_name) def train(self): bs = self.par.batch_size for iter_i in range(self.training_step, self.par.max_iter_count): if ((iter_i > 0) and ((iter_i % self.par.reload_iter) == 0)): self.load_data('train') if ((iter_i % self.par.test_iter) == 0): self.validate(iter_i) b = self.get_training_batch(iter_i) if (b.num_points() > bs): continue out = self.sess.run([self.train_step, self.loss, self.output], feed_dict=self.get_feed_dict(b)) print(((str(iter_i) + ' : ') + str(out[1]))) summary = self.sess.run(self.train_loss_summary, feed_dict={self.train_loss_placeholder: out[1]}) self.writer.add_summary(summary, iter_i) def validate(self, step): pixel_count = 0 acc = [] pix = [] bs = self.par.batch_size for b in self.validation_batches: out = self.sess.run([self.accuracy, self.output], feed_dict=self.get_feed_dict(b)) valid_out = np.multiply(out[1], np.asarray(expand_dim_to_batch1(b.labels[0], bs), dtype=bool)) acc.append(out[0]) pix.append(np.count_nonzero(b.labels[0])) pixel_count += np.count_nonzero(b.labels[0]) avg_acc = 0.0 for i in range(0, len(acc)): avg_acc += ((acc[i] * pix[i]) / pixel_count) print(('Accuracy: ' + str(avg_acc))) if (avg_acc > self.best_accuracy): self.best_accuracy = avg_acc self.save_snapshot() summary = self.sess.run(self.test_loss_summary, feed_dict={self.test_loss_placeholder: avg_acc}) self.writer.add_summary(summary, step) def test(self): scan_id = 0 cs = self.par.cube_size print('Testing...') for val_scan in self.test_data: if (self.par.data_sampling_type == 'full'): scan_batches = [get_batch_from_full_scan(val_scan, self.par.num_scales, self.par.d_par.class_weights)] else: global scan min_bound = (val_scan.clouds[0].get_min_bound() - (cs[0] * 0.5)) max_bound = (val_scan.clouds[0].get_max_bound() + (cs[0] * 0.5)) scan = val_scan rad = self.par.valid_rad points = [] x_s = (min_bound[0] + (rad / 2.0)) while (x_s < (max_bound[0] - rad)): y_s = (min_bound[1] + (rad / 2.0)) while (y_s < (max_bound[1] - rad)): z_s = (min_bound[2] + (rad / 2.0)) while (z_s < (max_bound[2] - rad)): if val_scan.has_points([x_s, y_s, z_s], rad): points.append([x_s, y_s, z_s]) z_s += rad y_s += rad x_s += rad arr_size = len(points) print(('Number of test batches: ' + str(arr_size))) print('Loading batches...') scan_batches = get_batch_array(val_scan, self.par, points) print('Done.') for b in scan_batches: out = self.sess.run(self.output, feed_dict=self.get_feed_dict(b)) valid_out = np.multiply(out, np.asarray(expand_dim_to_batch1(b.labels[0], self.par.batch_size), dtype=bool)) if (self.par.data_sampling_type == 'full'): val_scan.assign_labels(valid_out) else: val_scan.assign_labels_part(valid_out, b.index_maps[0]) make_dir(os.path.join(self.par.output_dir, self.test_scans[scan_id])) val_scan.save(os.path.join(self.par.output_dir, self.test_scans[scan_id])) print(self.test_scans[scan_id]) scan_id += 1
def main(): parser = argparse.ArgumentParser() parser.add_argument('--train_corpus', default=None, type=str, required=True, help='The input train corpus, each line contains numbers that are the roberta tokenized indices.') parser.add_argument('--train_eval_corpus', default=None, type=str, required=False, help='The input train eval corpus, each line contains numbers that are the roberta tokenized indices.') parser.add_argument('--eval_corpus', default=None, type=str, required=False, help='The input eval corpus, each line contains numbers that are the roberta tokenized indices.') parser.add_argument('--bert_model', default=None, type=str, required=True, help='Bert pre-trained model selected in the list: roberta-base, roberta-base, roberta-large') parser.add_argument('--output_dir', default=None, type=str, required=True, help='The output directory where the model checkpoints will be written.') parser.add_argument('--max_seq_length', default=208, type=int, help='The maximum total input sequence length after WordPiece tokenization. \nSequences longer than this will be truncated, and sequences shorter \nthan this will be padded.') parser.add_argument('--do_train', action='store_true', help='Whether to run training.') parser.add_argument('--do_eval', action='store_true', help='Whether to run evaluation.') parser.add_argument('--mlm_loss', action='store_true', help='Whether to add mlm loss.') parser.add_argument('--concat_tabcol', action='store_true', help='Whether to concatenate table and column representations.') parser.add_argument('--train_batch_size', default=48, type=int, help='Total batch size for training.') parser.add_argument('--learning_rate', default=1e-05, type=float, help='The initial learning rate for Adam.') parser.add_argument('--adam_epsilon', default=1e-08, type=float, help='Epsilon for Adam optimizer.') parser.add_argument('--num_train_epochs', default=1000.0, type=float, help='Total number of training epochs to perform.') parser.add_argument('--warmup_steps', default=0, type=int, help='Linear warmup over warmup_steps.') parser.add_argument('--no_cuda', action='store_true', help='Whether not to use CUDA when available') parser.add_argument('--on_memory', action='store_true', help='Whether to load train samples into memory or use disk') parser.add_argument('--do_lower_case', action='store_true', help='Whether to lower case the input text. True for uncased models, False for cased models.') parser.add_argument('--local_rank', type=int, default=(- 1), help='local_rank for distributed training on gpus') parser.add_argument('--seed', type=int, default=42, help='random seed for initialization') parser.add_argument('--gradient_accumulation_steps', type=int, default=1, help='Number of updates steps to accumualte before performing a backward/update pass.') parser.add_argument('--fp16', action='store_true', help='Whether to use 16-bit float precision instead of 32-bit') parser.add_argument('--loss_scale', type=float, default=0, help='Loss scaling to improve fp16 numeric stability. Only used when fp16 set to True.\n0 (default value): dynamic loss scaling.\nPositive power of 2: static loss scaling value.\n') args = parser.parse_args() wandb.init(project='column_roberta', name=args.output_dir) if ((args.local_rank == (- 1)) or args.no_cuda): device = torch.device(('cuda' if (torch.cuda.is_available() and (not args.no_cuda)) else 'cpu')) n_gpu = torch.cuda.device_count() else: torch.cuda.set_device(args.local_rank) device = torch.device('cuda', args.local_rank) n_gpu = 1 torch.distributed.init_process_group(init_method='env://', backend='nccl') print('\n=====Using device: {} n_gpu: {}, distributed training: {}, 16-bits training: {}'.format(device, n_gpu, bool((args.local_rank != (- 1))), args.fp16)) if (args.gradient_accumulation_steps < 1): raise ValueError('Invalid gradient_accumulation_steps parameter: {}, should be >= 1'.format(args.gradient_accumulation_steps)) args.train_batch_size = (args.train_batch_size // args.gradient_accumulation_steps) random.seed(args.seed) np.random.seed(args.seed) torch.manual_seed(args.seed) if (n_gpu > 0): torch.cuda.manual_seed_all(args.seed) if (not args.do_train): raise ValueError('Training is currently the only implemented execution option. Please set `do_train`.') tokenizer = RobertaTokenizer.from_pretrained(args.bert_model, do_lower_case=args.do_lower_case) num_train_optimization_steps = None if args.do_train: print('Loading Train Dataset', args.train_corpus) train_dataset = BERTDataset(args.train_corpus, tokenizer, seq_len=args.max_seq_length, corpus_lines=None, on_memory=args.on_memory, mlm_loss=args.mlm_loss) num_train_optimization_steps = (int(((len(train_dataset) / args.train_batch_size) / args.gradient_accumulation_steps)) * args.num_train_epochs) if (args.local_rank != (- 1)): num_train_optimization_steps = (num_train_optimization_steps // torch.distributed.get_world_size()) if args.do_eval: print('Loading Eval Dataset', args.eval_corpus) eval_dataset = BERTDataset(args.eval_corpus, tokenizer, seq_len=args.max_seq_length, corpus_lines=None, on_memory=args.on_memory) train_eval_dataset = BERTDataset(args.train_eval_corpus, tokenizer, seq_len=args.max_seq_length, corpus_lines=None, on_memory=args.on_memory) model = RobertaForMaskedLM.from_pretrained(args.bert_model) if args.fp16: model.half() model.to(device) if (args.local_rank != (- 1)): model = nn.parallel.DistributedDataParallel(model, device_ids=[args.local_rank], output_device=args.local_rank, broadcast_buffers=True, find_unused_parameters=True) elif (n_gpu > 1): model = torch.nn.DataParallel(model) wandb.watch(model) if args.do_train: param_optimizer = list(model.named_parameters()) no_decay = ['bias', 'LayerNorm.bias', 'LayerNorm.weight'] optimizer_grouped_parameters = [{'params': [p for (n, p) in param_optimizer if (not any(((nd in n) for nd in no_decay)))], 'weight_decay': 0.01}, {'params': [p for (n, p) in param_optimizer if any(((nd in n) for nd in no_decay))], 'weight_decay': 0.0}] if args.fp16: try: from apex.optimizers import FP16_Optimizer from apex.optimizers import FusedAdam except ImportError: raise ImportError('Please install apex from to use distributed and fp16 training.') optimizer = FusedAdam(optimizer_grouped_parameters, lr=args.learning_rate, bias_correction=False, max_grad_norm=1.0) if (args.loss_scale == 0): optimizer = FP16_Optimizer(optimizer, dynamic_loss_scale=True) else: optimizer = FP16_Optimizer(optimizer, static_loss_scale=args.loss_scale) else: optimizer = AdamW(optimizer_grouped_parameters, lr=args.learning_rate, eps=args.adam_epsilon) scheduler = WarmupLinearSchedule(optimizer, warmup_steps=args.warmup_steps, t_total=num_train_optimization_steps) global_step = 0 if args.do_train: print('***** Running training *****') print(' Num examples:', len(train_dataset)) print(' Batch size:', args.train_batch_size) print(' Num steps:', num_train_optimization_steps) if (args.local_rank == (- 1)): train_sampler = RandomSampler(train_dataset) else: train_sampler = DistributedSampler(train_dataset) train_dataloader = DataLoader(train_dataset, sampler=train_sampler, batch_size=args.train_batch_size) if args.do_eval: eval_sampler = SequentialSampler(eval_dataset) eval_dataloader = DataLoader(eval_dataset, sampler=eval_sampler, batch_size=args.train_batch_size) train_eval_sampler = SequentialSampler(train_eval_dataset) train_eval_dataloader = DataLoader(train_eval_dataset, sampler=train_eval_sampler, batch_size=args.train_batch_size) best_acc = 0 for epoch in range(int(args.num_train_epochs)): tr_loss = 0 (nb_tr_examples, nb_tr_steps) = (0, 0) total_steps = len(train_dataloader) step_check = (int((total_steps * 0.1)) if (int((total_steps * 0.1)) != 0) else 1) save_check = (int((total_steps * 0.5)) if (int((total_steps * 0.5)) != 0) else 1) print('\nEpoch: ', epoch) for (step, batch) in enumerate(train_dataloader): model.train() (input_ids, input_mask, col_label_ids, lm_label_ids, q_tab_inds) = batch (input_ids, input_mask, col_label_ids, lm_label_ids, q_tab_inds) = to_device(device, input_ids, input_mask, col_label_ids, lm_label_ids, q_tab_inds) if (not args.concat_tabcol): q_tab_inds = None if args.mlm_loss: loss = model(input_ids=input_ids, attention_mask=input_mask, masked_lm_labels=lm_label_ids, masked_col_labels=col_label_ids, q_tab_inds=q_tab_inds) else: loss = model(input_ids=input_ids, attention_mask=input_mask, masked_col_labels=col_label_ids, q_tab_inds=q_tab_inds) if (n_gpu > 1): loss = loss.mean() if (args.gradient_accumulation_steps > 1): loss = (loss / args.gradient_accumulation_steps) if args.fp16: optimizer.backward(loss) else: loss.backward() tr_loss += loss.item() wandb.log({'batch_training_loss': loss.item()}) nb_tr_examples += input_ids.size(0) nb_tr_steps += 1 if ((step % step_check) == 0): print(('Finishing training for current epoch:\t' + str(round((step / total_steps), 3)))) if (((step + 1) % args.gradient_accumulation_steps) == 0): optimizer.step() scheduler.step() optimizer.zero_grad() global_step += 1 epoch_loss = (tr_loss / total_steps) print(('Train epoch loss:\t' + str(epoch_loss))) wandb.log({'epoch_training_loss': epoch_loss}) if (args.do_eval and (args.local_rank in [(- 1), 0])): print('***** Running evaluation on train *****') train_eval_loss = 0.0 train_preds = None train_nb_eval_steps = 0 train_out_label_ids = None for (step, batch) in enumerate(train_eval_dataloader): model.eval() (input_ids, input_mask, col_label_ids, _, q_tab_inds) = batch (input_ids, input_mask, col_label_ids, q_tab_inds) = to_device(device, input_ids, input_mask, col_label_ids, q_tab_inds) if (not args.concat_tabcol): q_tab_inds = None with torch.no_grad(): (tmp_eval_loss, logits) = model(input_ids=input_ids, attention_mask=input_mask, masked_col_labels=col_label_ids, q_tab_inds=q_tab_inds, is_train=False) col_logits = logits[1] train_eval_loss += tmp_eval_loss.mean().item() train_nb_eval_steps += 1 if (train_preds is None): train_preds = col_logits.detach().cpu().numpy() train_out_label_ids = col_label_ids.detach().cpu().numpy() else: train_preds = np.append(train_preds, col_logits.detach().cpu().numpy(), axis=0) train_out_label_ids = np.append(train_out_label_ids, col_label_ids.detach().cpu().numpy(), axis=0) train_eval_loss = (train_eval_loss / train_nb_eval_steps) train_preds = np.argmax(train_preds, axis=2) train_cur_acc = simple_accuracy(train_preds, train_out_label_ids) print(('Train eval epoch loss:\t' + str(train_eval_loss))) print(('Train eval accuracy:\t' + str(train_cur_acc))) wandb.log({'epoch_train_eval_loss': train_eval_loss}) wandb.log({'train_eval_accuracy': train_cur_acc}) print('***** Running evaluation on dev *****') eval_loss = 0.0 preds = None nb_eval_steps = 0 out_label_ids = None for (step, batch) in enumerate(eval_dataloader): model.eval() (input_ids, input_mask, col_label_ids, _, q_tab_inds) = batch (input_ids, input_mask, col_label_ids, q_tab_inds) = to_device(device, input_ids, input_mask, col_label_ids, q_tab_inds) if (not args.concat_tabcol): q_tab_inds = None with torch.no_grad(): (tmp_eval_loss, logits) = model(input_ids=input_ids, attention_mask=input_mask, masked_col_labels=col_label_ids, q_tab_inds=q_tab_inds, is_train=False) col_logits = logits[1] eval_loss += tmp_eval_loss.mean().item() nb_eval_steps += 1 if (preds is None): preds = col_logits.detach().cpu().numpy() out_label_ids = col_label_ids.detach().cpu().numpy() else: preds = np.append(preds, col_logits.detach().cpu().numpy(), axis=0) out_label_ids = np.append(out_label_ids, col_label_ids.detach().cpu().numpy(), axis=0) eval_loss = (eval_loss / nb_eval_steps) preds = np.argmax(preds, axis=2) cur_acc = simple_accuracy(preds, out_label_ids) print(('Eval epoch loss:\t' + str(eval_loss))) print(('Eval accuracy:\t' + str(cur_acc))) wandb.log({'dev_eval_accuracy': cur_acc}) wandb.log({'epoch_dev_eval_loss': eval_loss}) if (cur_acc > best_acc): best_acc = cur_acc print(('** ** * Saving fine-tuned model for epoch ' + str(epoch))) model_to_save = (model.module if hasattr(model, 'module') else model) output_model_file = os.path.join(args.output_dir, WEIGHTS_NAME) output_config_file = os.path.join(args.output_dir, CONFIG_NAME) if args.do_train: torch.save(model_to_save.state_dict(), output_model_file) model_to_save.config.to_json_file(output_config_file) tokenizer.save_vocabulary(args.output_dir) if ((epoch < 10) or ((epoch % 10) == 0)): print(('** ** * Saving fine-tuned model for epoch ' + str(epoch))) model_to_save = (model.module if hasattr(model, 'module') else model) output_model_file = os.path.join(args.output_dir, WEIGHTS_NAME.replace('.bin', (('_' + str(epoch)) + '.bin'))) output_config_file = os.path.join(args.output_dir, CONFIG_NAME.replace('.json', (('_' + str(epoch)) + '.json'))) if args.do_train: torch.save(model_to_save.state_dict(), output_model_file) model_to_save.config.to_json_file(output_config_file) tokenizer.save_vocabulary(args.output_dir)
def weights_init_kaiming(m): classname = m.__class__.__name__ if (classname.find('Linear') != (- 1)): nn.init.normal_(m.weight, 0, 0.01) if (m.bias is not None): nn.init.constant_(m.bias, 0.0) elif (classname.find('Conv') != (- 1)): nn.init.kaiming_normal_(m.weight, mode='fan_out') if (m.bias is not None): nn.init.constant_(m.bias, 0.0) elif (classname.find('BatchNorm') != (- 1)): if m.affine: nn.init.constant_(m.weight, 1.0) nn.init.constant_(m.bias, 0.0)
class Mlp3Layer256UnitLongerTrainingDecreaseBatchSize(NeuralNetworkTrainingDecreaseBatchSize, Mlp3Layer256Unit): pass
class TestGradients(TestCase): exact_dtype = True def _get_safe_inplace(self, inplace_variant): (inplace_variant) def _fn(t, *args, **kwargs): return inplace_variant(t.clone(), *args, **kwargs) return _fn def _check_helper(self, device, dtype, op, variant, check): if (variant is None): self.skipTest('Skipped! Variant not implemented.') if (not op.supports_dtype(dtype, torch.device(device).type)): self.skipTest(f'Skipped! {op.name} does not support dtype {str(dtype)}') samples = op.sample_inputs(device, dtype, requires_grad=True) for sample in samples: partial_fn = partial(variant, **sample.kwargs) if (check == 'gradcheck'): self.assertTrue(gradcheck(partial_fn, ((sample.input,) + sample.args), check_grad_dtypes=True)) elif (check == 'gradgradcheck'): self.assertTrue(gradgradcheck(partial_fn, ((sample.input,) + sample.args), gen_non_contig_grad_outputs=False, check_grad_dtypes=True)) self.assertTrue(gradgradcheck(partial_fn, ((sample.input,) + sample.args), gen_non_contig_grad_outputs=True, check_grad_dtypes=True)) else: self.assertTrue(False, msg='Unknown check requested!') def _grad_test_helper(self, device, dtype, op, variant): return self._check_helper(device, dtype, op, variant, 'gradcheck') def _gradgrad_test_helper(self, device, dtype, op, variant): return self._check_helper(device, dtype, op, variant, 'gradgradcheck') (torch.double) (op_db) def test_fn_grad(self, device, dtype, op): self._grad_test_helper(device, dtype, op, op.get_op()) (torch.double) (op_db) def test_method_grad(self, device, dtype, op): self._grad_test_helper(device, dtype, op, op.get_method()) (torch.double) (op_db) def test_inplace_grad(self, device, dtype, op): if (not op.test_inplace_grad): self.skipTest('Skipped! Inplace gradcheck marked to skip.') self._grad_test_helper(device, dtype, op, self._get_safe_inplace(op.get_inplace())) (torch.double) (op_db) def test_fn_gradgrad(self, device, dtype, op): self._gradgrad_test_helper(device, dtype, op, op.get_op()) (torch.double) (op_db) def test_method_gradgrad(self, device, dtype, op): self._gradgrad_test_helper(device, dtype, op, op.get_method()) (torch.double) (op_db) def test_inplace_gradgrad(self, device, dtype, op): if (not op.test_inplace_grad): self.skipTest('Skipped! Inplace gradgradcheck marked to skip.') self._gradgrad_test_helper(device, dtype, op, self._get_safe_inplace(op.get_inplace()))
def quantize_linear_modules(module, dtype=torch.int8): warnings.warn('quantize_linear_modules function has been deprecated. Please use torch.quantization.quantize_dynamic API instead.') reassign = {} for (name, mod) in module.named_modules(): if (mod is module): continue new_mod = quantize_linear_modules(mod, dtype) if (new_mod is not mod): reassign[name] = new_mod for (name, mod) in reassign.items(): setattr(module, name, mod) if isinstance(module, torch.nn.Linear): if (dtype == torch.int8): return QuantizedLinear(module) elif (dtype == torch.float16): return QuantizedLinearFP16(module) else: raise RuntimeError('Unsupported dtype: {}'.format(dtype)) return module
def env_desc_gen(**config): env = MDPEnvironment(**config) env_desc = {'creator': MDPEnvironment, 'possible_agents': env.possible_agents, 'action_spaces': env.action_spaces, 'observation_spaces': env.observation_spaces, 'config': config} env.close() return env_desc
class SingleTaskSVGP(BaseGPSurrogate, SingleTaskVariationalGP): def __init__(self, feature_dim, out_dim, num_inducing_points, encoder, noise_constraint=None, lengthscale_prior=None, outcome_transform=None, input_transform=None, learn_inducing_points=True, mll_beta=1.0, *args, **kwargs): BaseGPSurrogate.__init__(self, *args, encoder=encoder, **kwargs) self.num_inducing_points = num_inducing_points if (out_dim == 1): covar_module = kernels.MaternKernel(ard_num_dims=feature_dim, lengthscale_prior=lengthscale_prior) covar_module.initialize(lengthscale=self.lengthscale_init) likelihood = likelihoods.GaussianLikelihood(noise_constraint=noise_constraint) likelihood.initialize(noise=self.task_noise_init) else: covar_module = kernels.MaternKernel(batch_shape=(out_dim,), ard_num_dims=feature_dim, lengthscale_prior=lengthscale_prior) covar_module.initialize(lengthscale=self.lengthscale_init) likelihood = likelihoods.MultitaskGaussianLikelihood(num_tasks=out_dim, has_global_noise=False, noise_constraint=noise_constraint) likelihood.initialize(task_noises=self.task_noise_init) dummy_X = (2 * (torch.rand(num_inducing_points, feature_dim).to(self.device, self.dtype) - 0.5)) dummy_Y = torch.randn(num_inducing_points, out_dim).to(self.device, self.dtype) covar_module = (covar_module if (covar_module is None) else covar_module.to(self.device, self.dtype)) self.base_cls = SingleTaskVariationalGP self.base_cls.__init__(self, dummy_X, dummy_Y, likelihood, out_dim, learn_inducing_points, covar_module=covar_module, inducing_points=dummy_X, outcome_transform=outcome_transform, input_transform=input_transform) self.encoder = encoder.to(self.device, self.dtype) self.mll_beta = mll_beta def clear_cache(self): clear_cache_hook(self) clear_cache_hook(self.model) clear_cache_hook(self.model.variational_strategy) if hasattr(self.model.variational_strategy, 'base_variational_strategy'): clear_cache_hook(self.model.variational_strategy.base_variational_strategy) def forward(self, inputs): features = (self.get_features(inputs, self.bs) if isinstance(inputs, np.ndarray) else inputs) res = self.base_cls.forward(self, features) return res def posterior(self, inputs, output_indices=None, observation_noise=False, **kwargs): self.clear_cache() features = (self.get_features(inputs, self.bs) if isinstance(inputs, np.ndarray) else inputs) return self.base_cls.posterior(self, features, output_indices, observation_noise, **kwargs) def set_train_data(self, inputs=None, targets=None, strict=True): self.clear_cache() def fit(self, X_train, Y_train, X_val, Y_val, X_test, Y_test, reset=False, log_prefix='single_task_svgp', **kwargs): if reset: raise NotImplementedError fit_kwargs = dict(surrogate=self, mll=VariationalELBO(self.likelihood, self.model, num_data=X_train.shape[0]), X_train=X_train, Y_train=Y_train, X_val=X_val, Y_val=Y_val, X_test=X_test, Y_test=Y_test, train_bs=self.bs, eval_bs=self.bs, shuffle_train=True, log_prefix=log_prefix) fit_kwargs.update(kwargs) return fit_gp_surrogate(**fit_kwargs) def reshape_targets(self, targets): if (targets.shape[(- 1)] > 1): return targets else: return targets.squeeze((- 1))
def clip_grad_by_value_backward(grad_inputs, inputs, input_shapes, outputs, output_shapes): dy = grad_inputs[0] x0 = inputs[0] assert False, 'This function is not called since the function is the composite of other functions.'
class StepOff(BaseVRMWaveform): def __init__(self, t0=0.0): self.t0 = t0 def t0(self): return self._t0 .setter def t0(self, value): self._t0 = validate_float('t0', value) def getCharDecay(self, fieldType, times): fieldType = validate_string('fieldType', fieldType, ['dhdt', 'dbdt']) if (self.t0 >= np.min(times)): raise ValueError(('Earliest time channel must be after beginning of off-time (t0 = %.2e s)' % self.t0)) t0 = self.t0 if (fieldType == 'dbdt'): mu0 = ((4 * np.pi) * 1e-07) eta = ((- mu0) / (times - t0)) elif (fieldType == 'dhdt'): eta = ((- 1) / (times - t0)) return eta def getLogUniformDecay(self, fieldType, times, chi0, dchi, tau1, tau2): fieldType = validate_string('fieldType', fieldType, ['dhdt', 'dbdt']) nT = len(times) nC = len(dchi) t0 = self.t0 times = np.kron(np.ones((nC, 1)), times) chi0 = np.kron(np.reshape(chi0, newshape=(nC, 1)), np.ones((1, nT))) dchi = np.kron(np.reshape(dchi, newshape=(nC, 1)), np.ones((1, nT))) tau1 = np.kron(np.reshape(tau1, newshape=(nC, 1)), np.ones((1, nT))) tau2 = np.kron(np.reshape(tau2, newshape=(nC, 1)), np.ones((1, nT))) if (fieldType == 'h'): eta = (((0.5 * (1 - np.sign((times - t0)))) * chi0) + (((0.5 * (1 + np.sign((times - t0)))) * (dchi / np.log((tau2 / tau1)))) * (spec.expi(((- (times - t0)) / tau2)) - spec.expi(((- (times - t0)) / tau1))))) elif (fieldType == 'b'): mu0 = ((4 * np.pi) * 1e-07) eta = (((0.5 * (1 - np.sign((times - t0)))) * chi0) + (((0.5 * (1 + np.sign((times - t0)))) * (dchi / np.log((tau2 / tau1)))) * (spec.expi(((- (times - t0)) / tau2)) - spec.expi(((- (times - t0)) / tau1))))) eta = (mu0 * eta) elif (fieldType == 'dhdt'): eta = (0.0 + ((((0.5 * (1 + np.sign((times - t0)))) * (dchi / np.log((tau2 / tau1)))) * (np.exp(((- (times - t0)) / tau1)) - np.exp(((- (times - t0)) / tau2)))) / (times - t0))) elif (fieldType == 'dbdt'): mu0 = ((4 * np.pi) * 1e-07) eta = (0.0 + ((((0.5 * (1 + np.sign((times - t0)))) * (dchi / np.log((tau2 / tau1)))) * (np.exp(((- (times - t0)) / tau1)) - np.exp(((- (times - t0)) / tau2)))) / (times - t0))) eta = (mu0 * eta) return eta
_grad() def concat_all_gather(tensor): tensors_gather = [torch.ones_like(tensor) for _ in range(torch.distributed.get_world_size())] torch.distributed.all_gather(tensors_gather, tensor, async_op=False) output = torch.cat(tensors_gather, dim=0) return output
def register(target: Target) -> Target: from . import cli global ALL_TARGETS ALL_TARGETS += (target,) cli.TARGETS_TYPE.choices += (target.__name__,) return target
def _frobenius_shift(K, generators, check_only=False): if (len(generators) == 1): return generators p = K.characteristic() n = K.degree() compatible = {} from .integer_mod import mod for m in n.divisors(): compatible[m] = {} for (q, x) in generators.items(): for m in (n // q).divisors(): compatible[m][q] = (x ** (((p ** (n // q)) - 1) // ((p ** m) - 1))) if check_only: for m in n.divisors(): try: (q, x) = compatible[m].popitem() except KeyError: break for (qq, xx) in compatible[m].items(): assert (x == xx) return crt = {} qlist = sorted(generators.keys()) for j in range(1, len(qlist)): for i in range(j): crt[(i, j)] = [] for m in n.divisors(): mqlist = sorted(compatible[m].keys()) for k in range(1, len(mqlist)): j = qlist.index(mqlist[k]) i = qlist.index(mqlist[(k - 1)]) crt[(i, j)].append(_find_pow_of_frobenius(p, m, compatible[m][qlist[j]], compatible[m][qlist[i]])) for (i, j) in list(crt): L = crt[(i, j)] running = mod(0, 1) for a in L: running = _crt_non_coprime(running, a) crt[(i, j)] = [(mod(running, (qq ** running.modulus().valuation(qq))), running.modulus().valuation(qq)) for qq in qlist] crt[(j, i)] = [((- a), level) for (a, level) in crt[(i, j)]] import bisect frob_powers = [mod(0, 1) for _ in qlist] def find_leveller(qindex, level, x, xleveled, searched, i): searched[i] = True crt_possibles = [] for j in range(1, len(qlist)): if (i == j): continue if (crt[(i, j)][qindex][1] >= level): if xleveled[j]: return [j] elif (j not in searched): crt_possibles.append(j) for j in crt_possibles: path = find_leveller(qindex, level, x, xleveled, searched, j) if (path is not None): path.append(j) return path return None def propagate_levelling(qindex, level, x, xleveled, i): for j in range(1, len(qlist)): if (i == j): continue if ((not xleveled[j]) and (crt[(i, j)][qindex][1] >= level)): newxj = (x[i][0] + crt[(i, j)][qindex][0]) x[j] = (newxj, min(x[i][1], crt[(i, j)][qindex][1])) xleveled[j] = True propagate_levelling(qindex, level, x, xleveled, j) for qindex in range(len(qlist)): q = qlist[qindex] x = ([0] + [crt[(0, j)][qindex] for j in range(1, len(qlist))]) levels = [] for j in range(2, len(qlist)): for i in range(j): if (i != 0): assert (x[j][0] == (x[i][0] + crt[(i, j)][qindex][0])) level = crt[(i, j)][qindex][1] if (level > 0): ins = bisect.bisect_left(levels, level) if (ins == len(levels)): levels.append(level) elif (levels[ins] != level): levels.insert(ins, level) for level in levels: xleveled = ([0] + [(x[i][1] >= level) for i in range(1, len(qlist))]) while True: try: i = xleveled.index(False, 1) searched = {} levelling_path = find_leveller(qindex, level, x, xleveled, searched, i) if (levelling_path is None): x[i] = (mod(x[i][0].lift(), (q ** level)), level) xleveled[i] = True propagate_levelling(qindex, level, x, xleveled, i) else: levelling_path.append(i) for m in range(1, len(path)): if (not xleveled[path[m]]): newx = (x[path[(m - 1)]][0] + crt[(path[(m - 1)], path[m])][qindex][0]) x[path[m]] = (newx, min(x[path[(m - 1)]][1], crt[(path[(m - 1)], path[m])][qindex][1])) xleveled[path[m]] = True propagate_levelling(qindex, level, x, xleveled, path[m]) except ValueError: break for j in range(1, len(qlist)): frob_powers[j] = frob_powers[j].crt(x[j][0]) for j in range(1, len(qlist)): generators[qlist[j]] = (generators[qlist[j]] ** (p ** (- frob_powers[j]).lift())) _frobenius_shift(K, generators, check_only=True)
def changeContagion(G, A, i): delta = 0 for u in G.neighbourIterator(i): if (A[u] == 1): delta += 1 return delta
def download_a_url(dl_folder, url): (url, filename) = get_downloaded_file(dl_folder, url) if os.path.exists(filename): print(f'{filename} has already been downloaded so skip') return filename print(f'downloading {url} to {filename}') if (isinstance(url, list) or isinstance(url, tuple)): download_parts_and_combine(dl_folder, url, filename) else: wget.download(url, filename, bar=bar_custom) print(f'dowloaded: {filename}') return filename
def test_queryrequest4(): url = (brokerIp + '/ngsi10/queryContext') headers = {'Content-Type': 'appliction/json'} r = requests.post(url, data=json.dumps(data_ngsi10.subdata45), headers=headers) assert (r.status_code == 200)
class AggregatorGRPCClient(): def __init__(self, agg_addr, agg_port, tls, disable_client_auth, root_certificate, certificate, private_key, aggregator_uuid=None, federation_uuid=None, single_col_cert_common_name=None, **kwargs): self.uri = f'{agg_addr}:{agg_port}' self.tls = tls self.disable_client_auth = disable_client_auth self.root_certificate = root_certificate self.certificate = certificate self.private_key = private_key self.logger = getLogger(__name__) if (not self.tls): self.logger.warn('gRPC is running on insecure channel with TLS disabled.') self.channel = self.create_insecure_channel(self.uri) else: self.channel = self.create_tls_channel(self.uri, self.root_certificate, self.disable_client_auth, self.certificate, self.private_key) self.header = None self.aggregator_uuid = aggregator_uuid self.federation_uuid = federation_uuid self.single_col_cert_common_name = single_col_cert_common_name self.interceptors = (RetryOnRpcErrorClientInterceptor(sleeping_policy=ConstantBackoff(logger=self.logger, reconnect_interval=int(kwargs.get('client_reconnect_interval', 1)), uri=self.uri), status_for_retry=(grpc.StatusCode.UNAVAILABLE,)),) self.stub = aggregator_pb2_grpc.AggregatorStub(grpc.intercept_channel(self.channel, *self.interceptors)) def create_insecure_channel(self, uri): return grpc.insecure_channel(uri, options=channel_options) def create_tls_channel(self, uri, root_certificate, disable_client_auth, certificate, private_key): with open(root_certificate, 'rb') as f: root_certificate_b = f.read() if disable_client_auth: self.logger.warn('Client-side authentication is disabled.') private_key_b = None certificate_b = None else: with open(private_key, 'rb') as f: private_key_b = f.read() with open(certificate, 'rb') as f: certificate_b = f.read() credentials = grpc.ssl_channel_credentials(root_certificates=root_certificate_b, private_key=private_key_b, certificate_chain=certificate_b) return grpc.secure_channel(uri, credentials, options=channel_options) def _set_header(self, collaborator_name): self.header = aggregator_pb2.MessageHeader(sender=collaborator_name, receiver=self.aggregator_uuid, federation_uuid=self.federation_uuid, single_col_cert_common_name=(self.single_col_cert_common_name or '')) def validate_response(self, reply, collaborator_name): check_equal(reply.header.receiver, collaborator_name, self.logger) check_equal(reply.header.sender, self.aggregator_uuid, self.logger) check_equal(reply.header.federation_uuid, self.federation_uuid, self.logger) check_equal(reply.header.single_col_cert_common_name, (self.single_col_cert_common_name or ''), self.logger) def disconnect(self): self.logger.debug(f'Disconnecting from gRPC server at {self.uri}') self.channel.close() def reconnect(self): self.disconnect() if (not self.tls): self.channel = self.create_insecure_channel(self.uri) else: self.channel = self.create_tls_channel(self.uri, self.root_certificate, self.disable_client_auth, self.certificate, self.private_key) self.logger.debug(f'Connecting to gRPC at {self.uri}') self.stub = aggregator_pb2_grpc.AggregatorStub(grpc.intercept_channel(self.channel, *self.interceptors)) _atomic_connection _resend_data_on_reconnection def get_tasks(self, collaborator_name): self._set_header(collaborator_name) request = aggregator_pb2.GetTasksRequest(header=self.header) response = self.stub.GetTasks(request) self.validate_response(response, collaborator_name) return (response.tasks, response.round_number, response.sleep_time, response.quit) _atomic_connection _resend_data_on_reconnection def get_aggregated_tensor(self, collaborator_name, tensor_name, round_number, report, tags, require_lossless): self._set_header(collaborator_name) request = aggregator_pb2.GetAggregatedTensorRequest(header=self.header, tensor_name=tensor_name, round_number=round_number, report=report, tags=tags, require_lossless=require_lossless) response = self.stub.GetAggregatedTensor(request) self.validate_response(response, collaborator_name) return response.tensor _atomic_connection _resend_data_on_reconnection def send_local_task_results(self, collaborator_name, round_number, task_name, data_size, named_tensors): self._set_header(collaborator_name) request = aggregator_pb2.TaskResults(header=self.header, round_number=round_number, task_name=task_name, data_size=data_size, tensors=named_tensors) stream = [] stream += utils.proto_to_datastream(request, self.logger) response = self.stub.SendLocalTaskResults(iter(stream)) self.validate_response(response, collaborator_name) def _get_trained_model(self, experiment_name, model_type): get_model_request = self.stub.GetTrainedModelRequest(experiment_name=experiment_name, model_type=model_type) model_proto_response = self.stub.GetTrainedModel(get_model_request) (tensor_dict, _) = utils.deconstruct_model_proto(model_proto_response.model_proto, NoCompressionPipeline()) return tensor_dict
_warnings(category=sklearn.exceptions.ConvergenceWarning) .filterwarnings('ignore:The SAMME.R algorithm') .parametrize('name, Estimator', all_estimators()) def test_fit_docstring_attributes(name, Estimator): pytest.importorskip('numpydoc') from numpydoc import docscrape doc = docscrape.ClassDoc(Estimator) attributes = doc['Attributes'] if (Estimator.__name__ in ('HalvingRandomSearchCV', 'RandomizedSearchCV', 'HalvingGridSearchCV', 'GridSearchCV')): est = _construct_searchcv_instance(Estimator) elif (Estimator.__name__ in ('ColumnTransformer', 'Pipeline', 'FeatureUnion')): est = _construct_compose_pipeline_instance(Estimator) elif (Estimator.__name__ == 'SparseCoder'): est = _construct_sparse_coder(Estimator) else: est = _construct_instance(Estimator) if (Estimator.__name__ == 'SelectKBest'): est.set_params(k=2) elif (Estimator.__name__ == 'DummyClassifier'): est.set_params(strategy='stratified') elif ((Estimator.__name__ == 'CCA') or Estimator.__name__.startswith('PLS')): est.set_params(n_components=1) elif (Estimator.__name__ in ('GaussianRandomProjection', 'SparseRandomProjection')): est.set_params(n_components=2) elif (Estimator.__name__ == 'TSNE'): est.set_params(perplexity=2) if (Estimator.__name__ in ('LinearSVC', 'LinearSVR')): est.set_params(dual='auto') if (Estimator.__name__ in ('NMF', 'MiniBatchNMF')): est.set_params(n_components='auto') if (Estimator.__name__ == 'QuantileRegressor'): solver = ('highs' if (sp_version >= parse_version('1.6.0')) else 'interior-point') est.set_params(solver=solver) if ('max_iter' in est.get_params()): est.set_params(max_iter=2) if ('random_state' in est.get_params()): est.set_params(random_state=0) skipped_attributes = {} if Estimator.__name__.endswith('Vectorizer'): if (Estimator.__name__ in ('CountVectorizer', 'HashingVectorizer', 'TfidfVectorizer')): X = ['This is the first document.', 'This document is the second document.', 'And this is the third one.', 'Is this the first document?'] elif (Estimator.__name__ == 'DictVectorizer'): X = [{'foo': 1, 'bar': 2}, {'foo': 3, 'baz': 1}] y = None else: (X, y) = make_classification(n_samples=20, n_features=3, n_redundant=0, n_classes=2, random_state=2) y = _enforce_estimator_tags_y(est, y) X = _enforce_estimator_tags_X(est, X) if ('1dlabels' in est._get_tags()['X_types']): est.fit(y) elif ('2dlabels' in est._get_tags()['X_types']): est.fit(np.c_[(y, y)]) elif ('3darray' in est._get_tags()['X_types']): est.fit(X[(np.newaxis, ...)], y) else: est.fit(X, y) for attr in attributes: if (attr.name in skipped_attributes): continue desc = ' '.join(attr.desc).lower() if ('only ' in desc): continue with ignore_warnings(category=FutureWarning): assert hasattr(est, attr.name) fit_attr = _get_all_fitted_attributes(est) fit_attr_names = [attr.name for attr in attributes] undocumented_attrs = set(fit_attr).difference(fit_attr_names) undocumented_attrs = set(undocumented_attrs).difference(skipped_attributes) if undocumented_attrs: raise AssertionError(f'Undocumented attributes for {Estimator.__name__}: {undocumented_attrs}')
class Decoder(nn.Module): def __init__(self, x_dim, z_dim): super(Decoder, self).__init__() self.model = nn.Sequential(nn.Linear(z_dim, 512), nn.ReLU(), nn.Linear(512, x_dim)) def forward(self, z): img = self.model(z) return img
class ImageGPTConfig(PretrainedConfig): model_type = 'imagegpt' keys_to_ignore_at_inference = ['past_key_values'] attribute_map = {'hidden_size': 'n_embd', 'max_position_embeddings': 'n_positions', 'num_attention_heads': 'n_head', 'num_hidden_layers': 'n_layer'} def __init__(self, vocab_size=(512 + 1), n_positions=(32 * 32), n_embd=512, n_layer=24, n_head=8, n_inner=None, activation_function='quick_gelu', resid_pdrop=0.1, embd_pdrop=0.1, attn_pdrop=0.1, layer_norm_epsilon=1e-05, initializer_range=0.02, scale_attn_weights=True, use_cache=True, tie_word_embeddings=False, scale_attn_by_inverse_layer_idx=False, reorder_and_upcast_attn=False, **kwargs): self.vocab_size = vocab_size self.n_positions = n_positions self.n_embd = n_embd self.n_layer = n_layer self.n_head = n_head self.n_inner = n_inner self.activation_function = activation_function self.resid_pdrop = resid_pdrop self.embd_pdrop = embd_pdrop self.attn_pdrop = attn_pdrop self.layer_norm_epsilon = layer_norm_epsilon self.initializer_range = initializer_range self.scale_attn_weights = scale_attn_weights self.use_cache = use_cache self.scale_attn_by_inverse_layer_idx = scale_attn_by_inverse_layer_idx self.reorder_and_upcast_attn = reorder_and_upcast_attn self.tie_word_embeddings = tie_word_embeddings super().__init__(tie_word_embeddings=tie_word_embeddings, **kwargs)
class _fasterRCNN(nn.Module): def __init__(self, classes, class_agnostic): super(_fasterRCNN, self).__init__() self.classes = classes self.n_classes = len(classes) self.class_agnostic = class_agnostic self.RCNN_loss_cls = 0 self.RCNN_loss_bbox = 0 self.RCNN_rpn = _RPN(self.dout_base_model) self.RCNN_proposal_target = _ProposalTargetLayer(self.n_classes) self.RCNN_roi_pool = ROIPool((cfg.POOLING_SIZE, cfg.POOLING_SIZE), (1.0 / 16.0)) self.RCNN_roi_align = ROIAlign((cfg.POOLING_SIZE, cfg.POOLING_SIZE), (1.0 / 16.0), 0) def forward(self, im_data, im_info, gt_boxes, num_boxes): batch_size = im_data.size(0) im_info = im_info.data gt_boxes = gt_boxes.data num_boxes = num_boxes.data base_feat = self.RCNN_base(im_data) (rois, rpn_loss_cls, rpn_loss_bbox) = self.RCNN_rpn(base_feat, im_info, gt_boxes, num_boxes) if self.training: roi_data = self.RCNN_proposal_target(rois, gt_boxes, num_boxes) (rois, rois_label, rois_target, rois_inside_ws, rois_outside_ws) = roi_data rois_label = Variable(rois_label.view((- 1)).long()) rois_target = Variable(rois_target.view((- 1), rois_target.size(2))) rois_inside_ws = Variable(rois_inside_ws.view((- 1), rois_inside_ws.size(2))) rois_outside_ws = Variable(rois_outside_ws.view((- 1), rois_outside_ws.size(2))) else: rois_label = None rois_target = None rois_inside_ws = None rois_outside_ws = None rpn_loss_cls = 0 rpn_loss_bbox = 0 rois = Variable(rois) if (cfg.POOLING_MODE == 'align'): pooled_feat = self.RCNN_roi_align(base_feat, rois.view((- 1), 5)) elif (cfg.POOLING_MODE == 'pool'): pooled_feat = self.RCNN_roi_pool(base_feat, rois.view((- 1), 5)) pooled_feat = self._head_to_tail(pooled_feat) bbox_pred = self.RCNN_bbox_pred(pooled_feat) if (self.training and (not self.class_agnostic)): bbox_pred_view = bbox_pred.view(bbox_pred.size(0), int((bbox_pred.size(1) / 4)), 4) bbox_pred_select = torch.gather(bbox_pred_view, 1, rois_label.view(rois_label.size(0), 1, 1).expand(rois_label.size(0), 1, 4)) bbox_pred = bbox_pred_select.squeeze(1) cls_score = self.RCNN_cls_score(pooled_feat) cls_prob = F.softmax(cls_score, 1) RCNN_loss_cls = 0 RCNN_loss_bbox = 0 if self.training: RCNN_loss_cls = F.cross_entropy(cls_score, rois_label) RCNN_loss_bbox = _smooth_l1_loss(bbox_pred, rois_target, rois_inside_ws, rois_outside_ws) cls_prob = cls_prob.view(batch_size, rois.size(1), (- 1)) bbox_pred = bbox_pred.view(batch_size, rois.size(1), (- 1)) return (rois, cls_prob, bbox_pred, rpn_loss_cls, rpn_loss_bbox, RCNN_loss_cls, RCNN_loss_bbox, rois_label) def _init_weights(self): def normal_init(m, mean, stddev, truncated=False): if truncated: m.weight.data.normal_().fmod_(2).mul_(stddev).add_(mean) else: m.weight.data.normal_(mean, stddev) m.bias.data.zero_() normal_init(self.RCNN_rpn.RPN_Conv, 0, 0.01, cfg.TRAIN.TRUNCATED) normal_init(self.RCNN_rpn.RPN_cls_score, 0, 0.01, cfg.TRAIN.TRUNCATED) normal_init(self.RCNN_rpn.RPN_bbox_pred, 0, 0.01, cfg.TRAIN.TRUNCATED) normal_init(self.RCNN_cls_score, 0, 0.01, cfg.TRAIN.TRUNCATED) normal_init(self.RCNN_bbox_pred, 0, 0.001, cfg.TRAIN.TRUNCATED) def create_architecture(self): self._init_modules() self._init_weights()
class SawyerDoorEnvV2(SawyerXYZEnv): def __init__(self): hand_low = ((- 0.5), 0.4, 0.05) hand_high = (0.5, 1, 0.5) obj_low = (0.0, 0.85, 0.15) obj_high = (0.1, 0.95, 0.15) goal_low = ((- 0.3), 0.4, 0.1499) goal_high = ((- 0.2), 0.5, 0.1501) super().__init__(self.model_name, hand_low=hand_low, hand_high=hand_high) self.init_config = {'obj_init_angle': np.array([0.3]), 'obj_init_pos': np.array([0.1, 0.95, 0.15]), 'hand_init_pos': np.array([0, 0.6, 0.2])} self.goal = np.array([(- 0.2), 0.7, 0.15]) self.obj_init_pos = self.init_config['obj_init_pos'] self.obj_init_angle = self.init_config['obj_init_angle'] self.hand_init_pos = self.init_config['hand_init_pos'] self.max_path_length = 150 self._random_reset_space = Box(np.array(obj_low), np.array(obj_high)) self.goal_space = Box(np.array(goal_low), np.array(goal_high)) self.door_angle_idx = self.model.get_joint_qpos_addr('doorjoint') def model_name(self): return full_v2_path_for('sawyer_xyz/sawyer_door_pull.xml') _assert_task_is_set def step(self, action): ob = super().step(action) (reward, reachDist, pullDist) = self.compute_reward(action, ob) self.curr_path_length += 1 info = {'reachDist': reachDist, 'goalDist': pullDist, 'epRew': reward, 'pickRew': None, 'success': float((pullDist <= 0.08))} return (ob, reward, False, info) def _get_pos_objects(self): return self.data.get_geom_xpos('handle').copy() def _set_obj_xyz(self, pos): qpos = self.data.qpos.copy() qvel = self.data.qvel.copy() qpos[self.door_angle_idx] = pos qvel[self.door_angle_idx] = 0 self.set_state(qpos.flatten(), qvel.flatten()) def reset_model(self): self._reset_hand() self.objHeight = self.data.get_geom_xpos('handle')[2] self.obj_init_pos = (self._get_state_rand_vec() if self.random_init else self.init_config['obj_init_pos']) self._target_pos = (self.obj_init_pos + np.array([(- 0.3), (- 0.45), 0.0])) self.sim.model.body_pos[self.model.body_name2id('door')] = self.obj_init_pos self.sim.model.site_pos[self.model.site_name2id('goal')] = self._target_pos self._set_obj_xyz(0) self.maxPullDist = np.linalg.norm((self.data.get_geom_xpos('handle')[:(- 1)] - self._target_pos[:(- 1)])) self.target_reward = ((1000 * self.maxPullDist) + (1000 * 2)) return self._get_obs() def _reset_hand(self): super()._reset_hand() (rightFinger, leftFinger) = (self._get_site_pos('rightEndEffector'), self._get_site_pos('leftEndEffector')) self.init_fingerCOM = ((rightFinger + leftFinger) / 2) self.reachCompleted = False def compute_reward(self, actions, obs): del actions objPos = obs[3:6] (rightFinger, leftFinger) = (self._get_site_pos('rightEndEffector'), self._get_site_pos('leftEndEffector')) fingerCOM = ((rightFinger + leftFinger) / 2) pullGoal = self._target_pos pullDist = np.linalg.norm((objPos[:(- 1)] - pullGoal[:(- 1)])) reachDist = np.linalg.norm((objPos - fingerCOM)) reachRew = (- reachDist) self.reachCompleted = (reachDist < 0.05) def pullReward(): c1 = 1000 c2 = 0.01 c3 = 0.001 if self.reachCompleted: pullRew = ((1000 * (self.maxPullDist - pullDist)) + (c1 * (np.exp(((- (pullDist ** 2)) / c2)) + np.exp(((- (pullDist ** 2)) / c3))))) pullRew = max(pullRew, 0) return pullRew else: return 0 pullRew = pullReward() reward = (reachRew + pullRew) return [reward, reachDist, pullDist]
def hide_rename_eval_setting(eval_setting_name): setting = m_repo.get_evaluation_setting(name=eval_setting_name, load_evaluations=True) for e in m_repo.get_evaluations([x.uuid for x in setting.evaluations]): m_repo.hide_evaluation(e.uuid) new_name = (eval_setting_name + f'_hidden_{random.randint(0, 10000)}') m_repo.rename_evaluation_setting(setting.uuid, new_name) m_repo.hide_evaluation_setting(setting.uuid)
def find_all(a_str, sub): start = 0 while True: start = a_str.find(sub, start) if (start == (- 1)): return (yield start) start += len(sub)
def build_voxel_generator(voxel_config): voxel_generator = VoxelGenerator(voxel_size=voxel_config.VOXEL_SIZE, point_cloud_range=voxel_config.RANGE, max_num_points=voxel_config.MAX_POINTS_NUM_PER_VOXEL, max_voxels=20000) return voxel_generator
class TestFromCTypes(object): def check(ctype, dtype): dtype = np.dtype(dtype) assert_equal(np.dtype(ctype), dtype) assert_equal(np.dtype(ctype()), dtype) def test_array(self): c8 = ctypes.c_uint8 self.check((3 * c8), (np.uint8, (3,))) self.check((1 * c8), (np.uint8, (1,))) self.check((0 * c8), (np.uint8, (0,))) self.check((1 * (3 * c8)), ((np.uint8, (3,)), (1,))) self.check((3 * (1 * c8)), ((np.uint8, (1,)), (3,))) def test_padded_structure(self): class PaddedStruct(ctypes.Structure): _fields_ = [('a', ctypes.c_uint8), ('b', ctypes.c_uint16)] expected = np.dtype([('a', np.uint8), ('b', np.uint16)], align=True) self.check(PaddedStruct, expected) def test_bit_fields(self): class BitfieldStruct(ctypes.Structure): _fields_ = [('a', ctypes.c_uint8, 7), ('b', ctypes.c_uint8, 1)] assert_raises(TypeError, np.dtype, BitfieldStruct) assert_raises(TypeError, np.dtype, BitfieldStruct()) def test_pointer(self): p_uint8 = ctypes.POINTER(ctypes.c_uint8) assert_raises(TypeError, np.dtype, p_uint8) def test_void_pointer(self): self.check(ctypes.c_void_p, np.uintp) def test_union(self): class Union(ctypes.Union): _fields_ = [('a', ctypes.c_uint8), ('b', ctypes.c_uint16)] expected = np.dtype(dict(names=['a', 'b'], formats=[np.uint8, np.uint16], offsets=[0, 0], itemsize=2)) self.check(Union, expected) def test_union_with_struct_packed(self): class Struct(ctypes.Structure): _pack_ = 1 _fields_ = [('one', ctypes.c_uint8), ('two', ctypes.c_uint32)] class Union(ctypes.Union): _fields_ = [('a', ctypes.c_uint8), ('b', ctypes.c_uint16), ('c', ctypes.c_uint32), ('d', Struct)] expected = np.dtype(dict(names=['a', 'b', 'c', 'd'], formats=['u1', np.uint16, np.uint32, [('one', 'u1'), ('two', np.uint32)]], offsets=[0, 0, 0, 0], itemsize=ctypes.sizeof(Union))) self.check(Union, expected) def test_union_packed(self): class Struct(ctypes.Structure): _fields_ = [('one', ctypes.c_uint8), ('two', ctypes.c_uint32)] _pack_ = 1 class Union(ctypes.Union): _pack_ = 1 _fields_ = [('a', ctypes.c_uint8), ('b', ctypes.c_uint16), ('c', ctypes.c_uint32), ('d', Struct)] expected = np.dtype(dict(names=['a', 'b', 'c', 'd'], formats=['u1', np.uint16, np.uint32, [('one', 'u1'), ('two', np.uint32)]], offsets=[0, 0, 0, 0], itemsize=ctypes.sizeof(Union))) self.check(Union, expected) def test_packed_structure(self): class PackedStructure(ctypes.Structure): _pack_ = 1 _fields_ = [('a', ctypes.c_uint8), ('b', ctypes.c_uint16)] expected = np.dtype([('a', np.uint8), ('b', np.uint16)]) self.check(PackedStructure, expected) def test_large_packed_structure(self): class PackedStructure(ctypes.Structure): _pack_ = 2 _fields_ = [('a', ctypes.c_uint8), ('b', ctypes.c_uint16), ('c', ctypes.c_uint8), ('d', ctypes.c_uint16), ('e', ctypes.c_uint32), ('f', ctypes.c_uint32), ('g', ctypes.c_uint8)] expected = np.dtype(dict(formats=[np.uint8, np.uint16, np.uint8, np.uint16, np.uint32, np.uint32, np.uint8], offsets=[0, 2, 4, 6, 8, 12, 16], names=['a', 'b', 'c', 'd', 'e', 'f', 'g'], itemsize=18)) self.check(PackedStructure, expected) def test_big_endian_structure_packed(self): class BigEndStruct(ctypes.BigEndianStructure): _fields_ = [('one', ctypes.c_uint8), ('two', ctypes.c_uint32)] _pack_ = 1 expected = np.dtype([('one', 'u1'), ('two', '>u4')]) self.check(BigEndStruct, expected) def test_little_endian_structure_packed(self): class LittleEndStruct(ctypes.LittleEndianStructure): _fields_ = [('one', ctypes.c_uint8), ('two', ctypes.c_uint32)] _pack_ = 1 expected = np.dtype([('one', 'u1'), ('two', '<u4')]) self.check(LittleEndStruct, expected) def test_little_endian_structure(self): class PaddedStruct(ctypes.LittleEndianStructure): _fields_ = [('a', ctypes.c_uint8), ('b', ctypes.c_uint16)] expected = np.dtype([('a', '<B'), ('b', '<H')], align=True) self.check(PaddedStruct, expected) def test_big_endian_structure(self): class PaddedStruct(ctypes.BigEndianStructure): _fields_ = [('a', ctypes.c_uint8), ('b', ctypes.c_uint16)] expected = np.dtype([('a', '>B'), ('b', '>H')], align=True) self.check(PaddedStruct, expected) def test_simple_endian_types(self): self.check(ctypes.c_uint16.__ctype_le__, np.dtype('<u2')) self.check(ctypes.c_uint16.__ctype_be__, np.dtype('>u2')) self.check(ctypes.c_uint8.__ctype_le__, np.dtype('u1')) self.check(ctypes.c_uint8.__ctype_be__, np.dtype('u1')) all_types = set(np.typecodes['All']) all_pairs = permutations(all_types, 2) .parametrize('pair', all_pairs) def test_pairs(self, pair): pair_type = np.dtype('{},{}'.format(*pair)) expected = np.dtype([('f0', pair[0]), ('f1', pair[1])]) assert_equal(pair_type, expected)
def _named_idx(idx): if ((idx < 0) or (idx > 2)): raise ValueError(('idx must be between 0 and 2, got %d' % idx)) return ('x', 'y', 'z')[idx]
def build_transform(is_train, args): resize_im = (args.input_size > 32) if is_train: transform = create_transform(input_size=args.input_size, is_training=True, color_jitter=args.color_jitter, auto_augment=args.aa, interpolation=args.train_interpolation, re_prob=args.reprob, re_mode=args.remode, re_count=args.recount) if (not resize_im): transform.transforms[0] = transforms.RandomCrop(args.input_size, padding=4) return transform t = [] if resize_im: size = int(((256 / 224) * args.input_size)) t.append(transforms.Resize(size, interpolation=3)) t.append(transforms.CenterCrop(args.input_size)) t.append(transforms.ToTensor()) t.append(transforms.Normalize(IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD)) return transforms.Compose(t)
class LinformerEncoder(RobertaEncoder): def __init__(self, args, dictionary): super().__init__(args, dictionary) self.register_buffer('version', torch.tensor(2)) def build_encoder(self, args, dictionary, embed_tokens): encoder = LinformerTransformerEncoder(args, dictionary, embed_tokens) encoder.apply(init_bert_params) return encoder def upgrade_state_dict_named(self, state_dict, name): super().upgrade_state_dict_named(state_dict, name) prefix = ((name + '.') if (name != '') else '') if (utils.item(state_dict.get(f'{prefix}version', torch.tensor(1))) < 2): state_dict[f'{prefix}version'] = torch.tensor(1) if (not torch.allclose(state_dict[f'{prefix}sentence_encoder.embed_tokens.weight'], state_dict[f'{prefix}lm_head.weight'])): self.lm_head = self.build_lm_head(embed_dim=self.args.encoder_embed_dim, output_dim=len(self.dictionary), activation_fn=self.args.activation_fn, weight=None)
def write_predictions_extended(all_examples, all_features, all_results, n_best_size, max_answer_length, output_prediction_file, output_nbest_file, output_null_log_odds_file, orig_data_file, start_n_top, end_n_top, version_2_with_negative, tokenizer, verbose_logging): _PrelimPrediction = collections.namedtuple('PrelimPrediction', ['feature_index', 'start_index', 'end_index', 'start_log_prob', 'end_log_prob']) _NbestPrediction = collections.namedtuple('NbestPrediction', ['text', 'start_log_prob', 'end_log_prob']) logger.info('Writing predictions to: %s', output_prediction_file) example_index_to_features = collections.defaultdict(list) for feature in all_features: example_index_to_features[feature.example_index].append(feature) unique_id_to_result = {} for result in all_results: unique_id_to_result[result.unique_id] = result all_predictions = collections.OrderedDict() all_nbest_json = collections.OrderedDict() scores_diff_json = collections.OrderedDict() for (example_index, example) in enumerate(all_examples): features = example_index_to_features[example_index] prelim_predictions = [] score_null = 1000000 for (feature_index, feature) in enumerate(features): result = unique_id_to_result[feature.unique_id] cur_null_score = result.cls_logits score_null = min(score_null, cur_null_score) for i in range(start_n_top): for j in range(end_n_top): start_log_prob = result.start_top_log_probs[i] start_index = result.start_top_index[i] j_index = ((i * end_n_top) + j) end_log_prob = result.end_top_log_probs[j_index] end_index = result.end_top_index[j_index] if (start_index >= (feature.paragraph_len - 1)): continue if (end_index >= (feature.paragraph_len - 1)): continue if (not feature.token_is_max_context.get(start_index, False)): continue if (end_index < start_index): continue length = ((end_index - start_index) + 1) if (length > max_answer_length): continue prelim_predictions.append(_PrelimPrediction(feature_index=feature_index, start_index=start_index, end_index=end_index, start_log_prob=start_log_prob, end_log_prob=end_log_prob)) prelim_predictions = sorted(prelim_predictions, key=(lambda x: (x.start_log_prob + x.end_log_prob)), reverse=True) seen_predictions = {} nbest = [] for pred in prelim_predictions: if (len(nbest) >= n_best_size): break feature = features[pred.feature_index] tok_tokens = feature.tokens[pred.start_index:(pred.end_index + 1)] orig_doc_start = feature.token_to_orig_map[pred.start_index] orig_doc_end = feature.token_to_orig_map[pred.end_index] orig_tokens = example.doc_tokens[orig_doc_start:(orig_doc_end + 1)] tok_text = tokenizer.convert_tokens_to_string(tok_tokens) tok_text = tok_text.strip() tok_text = ' '.join(tok_text.split()) orig_text = ' '.join(orig_tokens) final_text = get_final_text(tok_text, orig_text, tokenizer.do_lower_case, verbose_logging) if (final_text in seen_predictions): continue seen_predictions[final_text] = True nbest.append(_NbestPrediction(text=final_text, start_log_prob=pred.start_log_prob, end_log_prob=pred.end_log_prob)) if (not nbest): nbest.append(_NbestPrediction(text='', start_log_prob=(- 1000000.0), end_log_prob=(- 1000000.0))) total_scores = [] best_non_null_entry = None for entry in nbest: total_scores.append((entry.start_log_prob + entry.end_log_prob)) if (not best_non_null_entry): best_non_null_entry = entry probs = _compute_softmax(total_scores) nbest_json = [] for (i, entry) in enumerate(nbest): output = collections.OrderedDict() output['text'] = entry.text output['probability'] = probs[i] output['start_log_prob'] = entry.start_log_prob output['end_log_prob'] = entry.end_log_prob nbest_json.append(output) assert (len(nbest_json) >= 1) assert (best_non_null_entry is not None) score_diff = score_null scores_diff_json[example.qas_id] = score_diff all_predictions[example.qas_id] = best_non_null_entry.text all_nbest_json[example.qas_id] = nbest_json with open(output_prediction_file, 'w') as writer: writer.write((json.dumps(all_predictions, indent=4) + '\n')) with open(output_nbest_file, 'w') as writer: writer.write((json.dumps(all_nbest_json, indent=4) + '\n')) if version_2_with_negative: with open(output_null_log_odds_file, 'w') as writer: writer.write((json.dumps(scores_diff_json, indent=4) + '\n')) with open(orig_data_file, 'r', encoding='utf-8') as reader: orig_data = json.load(reader)['data'] qid_to_has_ans = make_qid_to_has_ans(orig_data) has_ans_qids = [k for (k, v) in qid_to_has_ans.items() if v] no_ans_qids = [k for (k, v) in qid_to_has_ans.items() if (not v)] (exact_raw, f1_raw) = get_raw_scores(orig_data, all_predictions) out_eval = {} find_all_best_thresh_v2(out_eval, all_predictions, exact_raw, f1_raw, scores_diff_json, qid_to_has_ans) return out_eval
def get_statistics(args, datasource): scaler = sklearn.preprocessing.StandardScaler() pbar = tqdm.tqdm(range(len(datasource.mus.tracks))) for ind in pbar: x = datasource.mus.tracks[ind].audio.T audio = nn.NdArray.from_numpy_array(x[(None, ...)]) target_spec = get_spectogram(*get_stft(audio, n_fft=args.nfft, n_hop=args.nhop), mono=True) pbar.set_description('Compute dataset statistics') scaler.partial_fit(np.squeeze(target_spec.data)) std = np.maximum(scaler.scale_, (0.0001 * np.max(scaler.scale_))) return (scaler.mean_, std)
def main(args, config): utils.init_distributed_mode(args) device = torch.device(args.device) seed = (args.seed + utils.get_rank()) torch.manual_seed(seed) np.random.seed(seed) random.seed(seed) cudnn.benchmark = True start_epoch = 0 max_epoch = config['schedular']['epochs'] warmup_epochs = config['schedular']['warmup_epochs'] print('Creating vqa datasets') datasets = create_dataset('vqa', config) if args.distributed: num_tasks = utils.get_world_size() global_rank = utils.get_rank() samplers = create_sampler(datasets, [True, False, False], num_tasks, global_rank) else: samplers = [None, None, None] (train_loader, val_loader, test_loader) = create_loader(datasets, samplers, batch_size=[config['batch_size_train'], config['batch_size_test'], config['batch_size_test']], num_workers=[4, 4, 4], is_trains=[True, False, False], collate_fns=[vqa_collate_fn, vqa_collate_fn, None]) tokenizer = BertTokenizer.from_pretrained(args.encoder, bos_token='[CLS]', eos_token='[SEP]', add_single_sep=False) print('Creating model') model = DaVinciVQA(config=config, encoder=args.encoder, text_decoder=args.text_decoder, tokenizer=tokenizer) model = model.to(device) arg_opt = utils.AttrDict(config['optimizer']) optimizer = create_optimizer(arg_opt, model) arg_sche = utils.AttrDict(config['schedular']) step_per_epoch = len(train_loader) arg_sche['num_warmup_steps'] = (arg_sche['warmup_epochs'] * step_per_epoch) arg_sche['num_training_steps'] = (arg_sche['epochs'] * step_per_epoch) (lr_scheduler, _) = create_scheduler(arg_sche, optimizer) if args.checkpoint: if args.evaluate: checkpoint = torch.load(args.checkpoint, map_location='cpu') state_dict = checkpoint['model'] msg = model.load_state_dict(state_dict, strict=False) print(('load checkpoint from %s' % args.checkpoint)) print(msg) else: checkpoint = torch.load(args.checkpoint, map_location='cpu') state_dict = checkpoint['model'] for key in list(state_dict.keys())[:]: new_key = ('davinci.' + key) state_dict[new_key] = state_dict[key] del state_dict[key] pos_embed_reshaped = interpolate_pos_embed(state_dict['davinci.visual_encoder.pos_embed'], model.davinci.visual_encoder) state_dict['davinci.visual_encoder.pos_embed'] = pos_embed_reshaped msg = model.load_state_dict(state_dict, strict=False) print(('load checkpoint from %s' % args.checkpoint)) print(msg) model_without_ddp = model (model, optimizer) = amp.initialize(model, optimizer, opt_level='O1') if args.distributed: model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[args.gpu], find_unused_parameters=True) model_without_ddp = model.module print('Start training') start_time = time.time() for epoch in range(start_epoch, max_epoch): if (not args.evaluate): if args.distributed: train_loader.sampler.set_epoch(epoch) train_stats = train(model, train_loader, optimizer, tokenizer, epoch, warmup_epochs, device, lr_scheduler, config) vqa_test_result = test(model, test_loader, tokenizer, device, config) result_file = save_result(vqa_test_result, args.result_dir, ('vqa_test_result_epoch%d' % epoch)) if args.evaluate: break if utils.is_main_process(): log_stats = {**{f'train_{k}': v for (k, v) in train_stats.items()}, 'epoch': epoch} with open(os.path.join(args.output_dir, 'log.txt'), 'a') as f: f.write((json.dumps(log_stats) + '\n')) save_obj = {'model': model_without_ddp.state_dict(), 'optimizer': optimizer.state_dict(), 'lr_scheduler': lr_scheduler.state_dict(), 'config': config, 'epoch': epoch} torch.save(save_obj, os.path.join(args.output_dir, ('checkpoint_%02d.pth' % epoch))) dist.barrier() total_time = (time.time() - start_time) total_time_str = str(datetime.timedelta(seconds=int(total_time))) print('Training time {}'.format(total_time_str))
() def dataset(test_dataset): if (test_dataset == ''): return None elif (test_dataset not in soundata.DATASETS): raise ValueError('{} is not a dataset in soundata'.format(test_dataset)) data_home = os.path.join('tests/resources/sound_datasets_full', test_dataset) return soundata.initialize(test_dataset, data_home)
class BandGapsConf(Struct): def __init__(self, filename, approx, region_selects, mat_pars, options, evp_options, eigenmomenta_options, band_gaps_options, coefs_save_name='coefs', corrs_save_names=None, incwd=None, output_dir=None, **kwargs): Struct.__init__(self, approx=approx, region_selects=region_selects, mat_pars=mat_pars, options=options, evp_options=evp_options, eigenmomenta_options=eigenmomenta_options, band_gaps_options=band_gaps_options, **kwargs) self.incwd = get_default(incwd, (lambda x: x)) self.conf = Struct() self.conf.filename_mesh = self.incwd(filename) output_dir = get_default(output_dir, self.incwd('output')) default = {'evp': 'evp', 'corrs_rs': 'corrs_rs'} self.corrs_save_names = get_default(corrs_save_names, default) io = MeshIO.any_from_filename(self.conf.filename_mesh) (self.bbox, self.dim) = io.read_bounding_box(ret_dim=True) rpc_axes = (nm.eye(self.dim, dtype=nm.float64) * (self.bbox[1] - self.bbox[0])) self.conf.options = options self.conf.options.update({'output_dir': output_dir, 'volume': {'value': get_lattice_volume(rpc_axes)}, 'coefs': 'coefs', 'requirements': 'requirements', 'coefs_filename': coefs_save_name}) self.conf.mat_pars = mat_pars self.conf.solvers = self.define_solvers() self.conf.regions = self.define_regions() self.conf.materials = self.define_materials() self.conf.fields = self.define_fields() self.conf.variables = self.define_variables() (self.conf.ebcs, self.conf.epbcs, self.conf.lcbcs, self.all_periodic) = self.define_bcs() self.conf.functions = self.define_functions() self.conf.integrals = self.define_integrals() (self.equations, self.expr_coefs) = self.define_equations() self.conf.coefs = self.define_coefs() self.conf.requirements = self.define_requirements() def __call__(self): return ProblemConf.from_dict(self.conf.__dict__, import_file(__file__)) def define_solvers(self): solvers = {'ls_d': ('ls.auto_direct', {'use_presolve': True}), 'ls_i': ('ls.scipy_iterative', {'method': 'cg', 'i_max': 1000, 'eps_a': 1e-12}), 'newton': ('nls.newton', {'i_max': 1, 'eps_a': 0.0001})} return solvers def define_regions(self): regions = {'Y': 'all', 'Y_m': self.region_selects.matrix, 'Y_c': self.region_selects.inclusion, 'Gamma_mc': ('r.Y_m *v r.Y_c', 'facet')} regions.update(define_box_regions(self.dim, self.bbox[0], self.bbox[1], 1e-05)) return regions def define_materials(self): materials = {'m': ({'D_m': self.mat_pars.D_m, 'density_m': self.mat_pars.density_m, 'D_c': self.mat_pars.D_c, 'density_c': self.mat_pars.density_c}, None, None, {'special_constant': True})} return materials def define_fields(self): fields = {'vector_Y_m': ('real', self.dim, 'Y_m', self.approx), 'vector_Y_c': ('real', self.dim, 'Y_c', self.approx), 'scalar_Y': ('real', 1, 'Y', 1)} return fields def define_variables(self): variables = {'u_m': ('unknown field', 'vector_Y_m'), 'v_m': ('test field', 'vector_Y_m', 'u_m'), 'Pi': ('parameter field', 'vector_Y_m', '(set-to-None)'), 'u1_m': ('parameter field', 'vector_Y_m', '(set-to-None)'), 'u2_m': ('parameter field', 'vector_Y_m', '(set-to-None)'), 'u_c': ('unknown field', 'vector_Y_c'), 'v_c': ('test field', 'vector_Y_c', 'u_c'), 'aux': ('parameter field', 'scalar_Y', '(set-to-None)')} return variables def define_bcs(self): ebcs = {'fixed_corners': ('Corners', {'u_m.all': 0.0}), 'fixed_gamma_mc': ('Gamma_mc', {'u_c.all': 0.0})} epbcs = {} all_periodic = [] for vn in ['u_m']: val = {('%s.all' % vn): ('%s.all' % vn)} epbcs.update({('periodic_%s_x' % vn): (['Left', 'Right'], val, 'match_y_line'), ('periodic_%s_y' % vn): (['Top', 'Bottom'], val, 'match_x_line')}) all_periodic.extend([('periodic_%s_x' % vn), ('periodic_%s_y' % vn)]) lcbcs = {} return (ebcs, epbcs, lcbcs, all_periodic) def define_functions(self): functions = {'match_x_line': (per.match_x_line,), 'match_y_line': (per.match_y_line,)} return functions def define_integrals(self): integrals = {'i': 2} return integrals def define_equations(self): equations = {} equations['corrs_rs'] = {'balance_of_forces': 'dw_lin_elastic.i.Y_m( m.D_m, v_m, u_m )\n = - dw_lin_elastic.i.Y_m( m.D_m, v_m, Pi )'} equations['evp'] = {'lhs': 'dw_lin_elastic.i.Y_c( m.D_c, v_c, u_c )', 'rhs': 'dw_dot.i.Y_c( m.density_c, v_c, u_c )'} expr_coefs = {'D': 'dw_lin_elastic.i.Y_m( m.D_m, u1_m, u2_m )', 'VF': 'ev_volume.i.%s(aux)', 'ema': 'ev_integrate.i.Y_c( m.density_c, u_c )'} return (equations, expr_coefs) def define_coefs(self): from copy import copy ema_options = copy(self.eigenmomenta_options) ema_options.update({'var_name': 'u_c'}) dispersion_options = copy(self.band_gaps_options) dispersion_options.update({'log_save_name': 'dispersion.log'}) coefs = {'VF': {'regions': ['Y_m', 'Y_c'], 'expression': self.expr_coefs['VF'], 'class': cb.VolumeFractions}, 'dv_info': {'requires': ['c.VF'], 'region_to_material': {'Y_m': ('m', 'density_m'), 'Y_c': ('m', 'density_c')}, 'class': cp.DensityVolumeInfo}, 'eigenmomenta': {'requires': ['evp', 'c.dv_info'], 'expression': self.expr_coefs['ema'], 'options': ema_options, 'class': cp.Eigenmomenta}, 'M': {'requires': ['evp', 'c.dv_info', 'c.eigenmomenta'], 'class': cp.AcousticMassTensor}, 'band_gaps': {'requires': ['evp', 'c.eigenmomenta', 'c.M'], 'options': self.band_gaps_options, 'class': cp.BandGaps}, 'D': {'requires': ['pis', 'corrs_rs'], 'expression': self.expr_coefs['D'], 'set_variables': set_coef_d, 'class': cb.CoefSymSym}, 'Gamma': {'requires': ['c.D'], 'options': {'mode': 'simple', 'incident_wave_dir': None}, 'class': cp.ChristoffelAcousticTensor}, 'dispersion': {'requires': ['evp', 'c.eigenmomenta', 'c.M', 'c.Gamma'], 'options': dispersion_options, 'class': cp.BandGaps}, 'polarization_angles': {'requires': ['c.dispersion'], 'options': {'incident_wave_dir': None}, 'class': cp.PolarizationAngles}, 'phase_velocity': {'requires': ['c.dv_info', 'c.Gamma'], 'options': {'eigensolver': 'eig.sgscipy'}, 'class': cp.PhaseVelocity}, 'filenames': {}} return coefs def define_requirements(self): requirements = {'evp': {'ebcs': ['fixed_gamma_mc'], 'epbcs': None, 'equations': self.equations['evp'], 'save_name': self.corrs_save_names['evp'], 'options': self.evp_options, 'class': cp.SimpleEVP}, 'pis': {'variables': ['u_m'], 'class': cb.ShapeDimDim}, 'corrs_rs': {'requires': ['pis'], 'ebcs': ['fixed_corners'], 'epbcs': self.all_periodic, 'equations': self.equations['corrs_rs'], 'set_variables': [('Pi', 'pis', 'u_m')], 'save_name': self.corrs_save_names['corrs_rs'], 'is_linear': True, 'class': cb.CorrDimDim, 'is_linear': True}} return requirements
class GAN(object): def __init__(self, z_dim, crop_image_size, resized_image_size, batch_size, data_dir): celebA_dataset = celebA.read_dataset(data_dir) self.z_dim = z_dim self.crop_image_size = crop_image_size self.resized_image_size = resized_image_size self.batch_size = batch_size filename_queue = tf.train.string_input_producer(celebA_dataset.train_images) self.images = self._read_input_queue(filename_queue) def _read_input(self, filename_queue): class DataRecord(object): pass reader = tf.WholeFileReader() (key, value) = reader.read(filename_queue) record = DataRecord() decoded_image = tf.image.decode_jpeg(value, channels=3) cropped_image = tf.cast(tf.image.crop_to_bounding_box(decoded_image, 55, 35, self.crop_image_size, self.crop_image_size), tf.float32) decoded_image_4d = tf.expand_dims(cropped_image, 0) resized_image = tf.image.resize_bilinear(decoded_image_4d, [self.resized_image_size, self.resized_image_size]) record.input_image = tf.squeeze(resized_image, squeeze_dims=[0]) return record def _read_input_queue(self, filename_queue): print('Setting up image reader...') read_input = self._read_input(filename_queue) num_preprocess_threads = 4 num_examples_per_epoch = 800 min_queue_examples = int((0.1 * num_examples_per_epoch)) print('Shuffling') input_image = tf.train.batch([read_input.input_image], batch_size=self.batch_size, num_threads=num_preprocess_threads, capacity=(min_queue_examples + (2 * self.batch_size))) input_image = utils.process_image(input_image, 127.5, 127.5) return input_image def _generator(self, z, dims, train_phase, activation=tf.nn.relu, scope_name='generator'): N = len(dims) image_size = (self.resized_image_size // (2 ** (N - 1))) with tf.variable_scope(scope_name) as scope: W_z = utils.weight_variable([self.z_dim, ((dims[0] * image_size) * image_size)], name='W_z') b_z = utils.bias_variable([((dims[0] * image_size) * image_size)], name='b_z') h_z = (tf.matmul(z, W_z) + b_z) h_z = tf.reshape(h_z, [(- 1), image_size, image_size, dims[0]]) h_bnz = utils.batch_norm(h_z, dims[0], train_phase, scope='gen_bnz') h = activation(h_bnz, name='h_z') utils.add_activation_summary(h) for index in range((N - 2)): image_size *= 2 W = utils.weight_variable([5, 5, dims[(index + 1)], dims[index]], name=('W_%d' % index)) b = utils.bias_variable([dims[(index + 1)]], name=('b_%d' % index)) deconv_shape = tf.pack([tf.shape(h)[0], image_size, image_size, dims[(index + 1)]]) h_conv_t = utils.conv2d_transpose_strided(h, W, b, output_shape=deconv_shape) h_bn = utils.batch_norm(h_conv_t, dims[(index + 1)], train_phase, scope=('gen_bn%d' % index)) h = activation(h_bn, name=('h_%d' % index)) utils.add_activation_summary(h) image_size *= 2 W_pred = utils.weight_variable([5, 5, dims[(- 1)], dims[(- 2)]], name='W_pred') b_pred = utils.bias_variable([dims[(- 1)]], name='b_pred') deconv_shape = tf.pack([tf.shape(h)[0], image_size, image_size, dims[(- 1)]]) h_conv_t = utils.conv2d_transpose_strided(h, W_pred, b_pred, output_shape=deconv_shape) pred_image = tf.nn.tanh(h_conv_t, name='pred_image') utils.add_activation_summary(pred_image) return pred_image def _discriminator(self, input_images, dims, train_phase, activation=tf.nn.relu, scope_name='discriminator', scope_reuse=False): N = len(dims) with tf.variable_scope(scope_name) as scope: if scope_reuse: scope.reuse_variables() h = input_images skip_bn = True for index in range((N - 2)): W = utils.weight_variable([5, 5, dims[index], dims[(index + 1)]], name=('W_%d' % index)) b = utils.bias_variable([dims[(index + 1)]], name=('b_%d' % index)) h_conv = utils.conv2d_strided(h, W, b) if skip_bn: h_bn = h_conv skip_bn = False else: h_bn = utils.batch_norm(h_conv, dims[(index + 1)], train_phase, scope=('disc_bn%d' % index)) h = activation(h_bn, name=('h_%d' % index)) utils.add_activation_summary(h) shape = h.get_shape().as_list() image_size = (self.resized_image_size // (2 ** (N - 2))) h_reshaped = tf.reshape(h, [self.batch_size, ((image_size * image_size) * shape[3])]) W_pred = utils.weight_variable([((image_size * image_size) * shape[3]), dims[(- 1)]], name='W_pred') b_pred = utils.bias_variable([dims[(- 1)]], name='b_pred') h_pred = (tf.matmul(h_reshaped, W_pred) + b_pred) return (tf.nn.sigmoid(h_pred), h_pred, h) def _cross_entropy_loss(self, logits, labels, name='x_entropy'): xentropy = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(logits, labels)) tf.scalar_summary(name, xentropy) return xentropy def _get_optimizer(self, optimizer_name, learning_rate, optimizer_param): self.learning_rate = learning_rate if (optimizer_name == 'Adam'): return tf.train.AdamOptimizer(learning_rate, beta1=optimizer_param) elif (optimizer_name == 'RMSProp'): return tf.train.RMSPropOptimizer(learning_rate, decay=optimizer_param) else: raise ValueError(('Unknown optimizer %s' % optimizer_name)) def _train(self, loss_val, var_list, optimizer): grads = optimizer.compute_gradients(loss_val, var_list=var_list) for (grad, var) in grads: utils.add_gradient_summary(grad, var) return optimizer.apply_gradients(grads) def _setup_placeholder(self): self.train_phase = tf.placeholder(tf.bool) self.z_vec = tf.placeholder(tf.float32, [self.batch_size, self.z_dim], name='z') def _gan_loss(self, logits_real, logits_fake, feature_real, feature_fake, use_features=False): discriminator_loss_real = self._cross_entropy_loss(logits_real, tf.ones_like(logits_real), name='disc_real_loss') discriminator_loss_fake = self._cross_entropy_loss(logits_fake, tf.zeros_like(logits_fake), name='disc_fake_loss') self.discriminator_loss = (discriminator_loss_fake + discriminator_loss_real) gen_loss_disc = self._cross_entropy_loss(logits_fake, tf.ones_like(logits_fake), name='gen_disc_loss') if use_features: gen_loss_features = (tf.reduce_mean(tf.nn.l2_loss((feature_real - feature_fake))) / (self.crop_image_size ** 2)) else: gen_loss_features = 0 self.gen_loss = (gen_loss_disc + (0.1 * gen_loss_features)) tf.scalar_summary('Discriminator_loss', self.discriminator_loss) tf.scalar_summary('Generator_loss', self.gen_loss) def create_network(self, generator_dims, discriminator_dims, optimizer='Adam', learning_rate=0.0002, optimizer_param=0.9, improved_gan_loss=True): print('Setting up model...') self._setup_placeholder() tf.histogram_summary('z', self.z_vec) self.gen_images = self._generator(self.z_vec, generator_dims, self.train_phase, scope_name='generator') tf.image_summary('image_real', self.images, max_images=2) tf.image_summary('image_generated', self.gen_images, max_images=2) def leaky_relu(x, name='leaky_relu'): return utils.leaky_relu(x, alpha=0.2, name=name) (discriminator_real_prob, logits_real, feature_real) = self._discriminator(self.images, discriminator_dims, self.train_phase, activation=leaky_relu, scope_name='discriminator', scope_reuse=False) (discriminator_fake_prob, logits_fake, feature_fake) = self._discriminator(self.gen_images, discriminator_dims, self.train_phase, activation=leaky_relu, scope_name='discriminator', scope_reuse=True) self._gan_loss(logits_real, logits_fake, feature_real, feature_fake, use_features=improved_gan_loss) train_variables = tf.trainable_variables() for v in train_variables: utils.add_to_regularization_and_summary(var=v) self.generator_variables = [v for v in train_variables if v.name.startswith('generator')] self.discriminator_variables = [v for v in train_variables if v.name.startswith('discriminator')] optim = self._get_optimizer(optimizer, learning_rate, optimizer_param) self.generator_train_op = self._train(self.gen_loss, self.generator_variables, optim) self.discriminator_train_op = self._train(self.discriminator_loss, self.discriminator_variables, optim) def initialize_network(self, logs_dir): print('Initializing network...') self.logs_dir = logs_dir self.sess = tf.Session() self.summary_op = tf.merge_all_summaries() self.saver = tf.train.Saver() self.summary_writer = tf.train.SummaryWriter(self.logs_dir, self.sess.graph) self.sess.run(tf.initialize_all_variables()) ckpt = tf.train.get_checkpoint_state(self.logs_dir) if (ckpt and ckpt.model_checkpoint_path): self.saver.restore(self.sess, ckpt.model_checkpoint_path) print('Model restored...') self.coord = tf.train.Coordinator() self.threads = tf.train.start_queue_runners(self.sess, self.coord) def train_model(self, max_iterations): try: print('Training model...') for itr in xrange(1, max_iterations): batch_z = np.random.uniform((- 1.0), 1.0, size=[self.batch_size, self.z_dim]).astype(np.float32) feed_dict = {self.z_vec: batch_z, self.train_phase: True} self.sess.run(self.discriminator_train_op, feed_dict=feed_dict) self.sess.run(self.generator_train_op, feed_dict=feed_dict) if ((itr % 10) == 0): (g_loss_val, d_loss_val, summary_str) = self.sess.run([self.gen_loss, self.discriminator_loss, self.summary_op], feed_dict=feed_dict) print(('Step: %d, generator loss: %g, discriminator_loss: %g' % (itr, g_loss_val, d_loss_val))) self.summary_writer.add_summary(summary_str, itr) if ((itr % 2000) == 0): self.saver.save(self.sess, (self.logs_dir + 'model.ckpt'), global_step=itr) except tf.errors.OutOfRangeError: print('Done training -- epoch limit reached') except KeyboardInterrupt: print('Ending Training...') finally: self.coord.request_stop() self.coord.join(self.threads) def visualize_model(self): print('Sampling images from model...') batch_z = np.random.uniform((- 1.0), 1.0, size=[self.batch_size, self.z_dim]).astype(np.float32) feed_dict = {self.z_vec: batch_z, self.train_phase: False} images = self.sess.run(self.gen_images, feed_dict=feed_dict) images = utils.unprocess_image(images, 127.5, 127.5).astype(np.uint8) shape = [4, (self.batch_size // 4)] utils.save_imshow_grid(images, self.logs_dir, 'generated.png', shape=shape)
def RatVal(a, b, ctx=None): if z3_debug(): _z3_assert((_is_int(a) or isinstance(a, str)), 'First argument cannot be converted into an integer') _z3_assert((_is_int(b) or isinstance(b, str)), 'Second argument cannot be converted into an integer') return simplify((RealVal(a, ctx) / RealVal(b, ctx)))
class BasicBlock(nn.Module): expansion = 1 def __init__(self, inplanes, planes, stride=1, downsample=None, dilation=(1, 1), residual=True): super(BasicBlock, self).__init__() self.conv1 = conv3x3(inplanes, planes, stride, padding=dilation[0], dilation=dilation[0]) self.bn1 = BatchNorm(planes) self.relu = nn.ReLU(inplace=True) self.conv2 = conv3x3(planes, planes, padding=dilation[1], dilation=dilation[1]) self.bn2 = BatchNorm(planes) self.downsample = downsample self.stride = stride self.residual = residual def forward(self, x): residual = x out = self.conv1(x) out = self.bn1(out) out = self.relu(out) out = self.conv2(out) out = self.bn2(out) if (self.downsample is not None): residual = self.downsample(x) if self.residual: out += residual out = self.relu(out) return out
def fused_bn(x, mean, var, gain=None, bias=None, eps=1e-05): scale = torch.rsqrt((var + eps)) if (gain is not None): scale = (scale * gain) shift = (mean * scale) if (bias is not None): shift = (shift - bias) return ((x * scale) - shift)
class _DecreasingVarianceModel(QuadraticMeanAndRBFKernel, TrainableProbabilisticModel): def __init__(self, data: Dataset): super().__init__() self._data = data _check_shapes def predict(self, query_points: TensorType) -> tuple[(TensorType, TensorType)]: (mean, var) = super().predict(query_points) return (mean, (var / len(self._data))) def update(self, dataset: Dataset) -> None: self._data = dataset def optimize(self, dataset: Dataset) -> None: pass
class UniqueSinglingOutQueries(): def __init__(self): self._set: Set[str] = set() self._list: List[str] = [] def check_and_append(self, query: str, df: pd.DataFrame): sorted_query = ''.join(sorted(query)) if (sorted_query not in self._set): counts = safe_query_counts(query=query, df=df) if ((counts is not None) and (counts == 1)): self._set.add(sorted_query) self._list.append(query) def __len__(self): return len(self._list) def queries(self) -> List[str]: return self._list
def import_fsspec(name: str) -> ModuleType: try: import fsspec except ModuleNotFoundError as err: raise ImportError(f'''to use {name}, you must install fsspec: pip install fsspec or conda install -c conda-forge fsspec ''') from err import_pyarrow_parquet(name) return fsspec
def transform_params(params, params_tf, num_classes): params['root_block']['conv_root']['kernel'] = params_tf['resnet/root_block/standardized_conv2d/kernel'] for block in ['block1', 'block2', 'block3', 'block4']: units = set([re.findall('unit\\d+', p)[0] for p in params_tf.keys() if (p.find(block) >= 0)]) for unit in units: for (i, group) in enumerate(['a', 'b', 'c']): params[block][unit][f'conv{(i + 1)}']['kernel'] = params_tf[f'resnet/{block}/{unit}/{group}/standardized_conv2d/kernel'] params[block][unit][f'gn{(i + 1)}']['bias'] = params_tf[f'resnet/{block}/{unit}/{group}/group_norm/beta'][(None, None, None)] params[block][unit][f'gn{(i + 1)}']['scale'] = params_tf[f'resnet/{block}/{unit}/{group}/group_norm/gamma'][(None, None, None)] projs = [p for p in params_tf.keys() if (p.find(f'{block}/{unit}/a/proj') >= 0)] assert (len(projs) <= 1) if projs: params[block][unit]['conv_proj']['kernel'] = params_tf[projs[0]] params['norm-pre-head']['bias'] = params_tf['resnet/group_norm/beta'][(None, None, None)] params['norm-pre-head']['scale'] = params_tf['resnet/group_norm/gamma'][(None, None, None)] params['conv_head']['kernel'] = np.zeros((params['conv_head']['kernel'].shape[0], num_classes), dtype=np.float32) params['conv_head']['bias'] = np.zeros(num_classes, dtype=np.float32)
class ArgMaxParameter(message.Message): __metaclass__ = reflection.GeneratedProtocolMessageType DESCRIPTOR = _ARGMAXPARAMETER
def _fake_quantize_per_tensor_affine_grad_reference(dY, X, scale, zero_point, quant_min, quant_max): Xq = torch.round(((X * (1.0 / scale)) + zero_point)) mask = ((Xq >= quant_min) * (Xq <= quant_max)) res = torch.zeros_like(dY) res[mask] = dY[mask] return res
def make_square(img_size=(64, 64), num_points_per_cluster=8, cluster_radius=1): is_square = False while (not is_square): point_1_x = random.randint((0 + cluster_radius), (img_size[0] - cluster_radius)) point_1_y = random.randint((0 + cluster_radius), (img_size[1] - cluster_radius)) point_2_x = random.randint((0 + cluster_radius), (img_size[0] - cluster_radius)) point_2_y = random.randint((0 + cluster_radius), (img_size[1] - cluster_radius)) (point_3_x, point_3_y, point_4_x, point_4_y) = get_point_square(point_1_x, point_1_y, point_2_x, point_2_y) if (((point_3_x + cluster_radius) > img_size[0]) or ((point_3_y + cluster_radius) > img_size[1]) or ((point_3_x - cluster_radius) < 0) or ((point_3_y - cluster_radius) < 0)): continue if (((point_4_x + cluster_radius) > img_size[0]) or ((point_4_y + cluster_radius) > img_size[1]) or ((point_4_x - cluster_radius) < 0) or ((point_4_y - cluster_radius) < 0)): continue points = [] points = get_cluster_points(num_points_per_cluster, point_1_x, point_1_y, points, cluster_radius) points = get_cluster_points(num_points_per_cluster, point_2_x, point_2_y, points, cluster_radius) points = get_cluster_points(num_points_per_cluster, point_3_x, point_3_y, points, cluster_radius) points = get_cluster_points(num_points_per_cluster, point_4_x, point_4_y, points, cluster_radius) image = np.zeros((img_size[0], img_size[1], 1)) for p in points: image = cv2.circle(image, p, radius=2, color=255, thickness=(- 1)) is_square = True return image