code
stringlengths
101
5.91M
def replace_example_docstring(example_docstring): def docstring_decorator(fn): func_doc = fn.__doc__ lines = func_doc.split('\n') i = 0 while ((i < len(lines)) and (re.search('^\\s*Examples?:\\s*$', lines[i]) is None)): i += 1 if (i < len(lines)): lines[i] = example_docstring func_doc = '\n'.join(lines) else: raise ValueError(f'''The function {fn} should have an empty 'Examples:' in its docstring as placeholder, current docstring is: {func_doc}''') fn.__doc__ = func_doc return fn return docstring_decorator
def add_content_and_label(file_location, output_sentences, output_labels, label): with open(file_location) as file: content = ' '.join(file.readlines()).replace('\n', ' ').replace('\r', '') single_spaced_content = ' '.join(content.split()) output_sentences.write((single_spaced_content + '\n')) output_labels.write((str(label) + '\n'))
def get_learning_rate_schedules(specs): schedule_specs = specs['LearningRateSchedule'] schedules = [] for schedule_specs in schedule_specs: if (schedule_specs['Type'] == 'Step'): schedules.append(StepLearningRateSchedule(schedule_specs['Initial'], schedule_specs['Interval'], schedule_specs['Factor'])) elif (schedule_specs['Type'] == 'Warmup'): schedules.append(WarmupLearningRateSchedule(schedule_specs['Initial'], schedule_specs['Final'], schedule_specs['Length'])) elif (schedule_specs['Type'] == 'Constant'): schedules.append(ConstantLearningRateSchedule(schedule_specs['Value'])) else: raise Exception('no known learning rate schedule of type "{}"'.format(schedule_specs['Type'])) return schedules
class RandAugment(torch.nn.Module): def __init__(self, num_ops: int=2, magnitude: int=9, num_magnitude_bins: int=31, interpolation: InterpolationMode=InterpolationMode.NEAREST, fill: Optional[List[float]]=None) -> None: super().__init__() self.num_ops = num_ops self.magnitude = magnitude self.num_magnitude_bins = num_magnitude_bins self.interpolation = interpolation self.fill = fill def _augmentation_space(self, num_bins: int, image_size: List[int]) -> Dict[(str, Tuple[(Tensor, bool)])]: return {'Identity': (torch.tensor(0.0), False), 'Brightness': (torch.linspace(0.0, 0.9, num_bins), True), 'Color': (torch.linspace(0.0, 0.9, num_bins), True), 'Contrast': (torch.linspace(0.0, 0.9, num_bins), True), 'Sharpness': (torch.linspace(0.0, 0.9, num_bins), True), 'Posterize': ((8 - (torch.arange(num_bins) / ((num_bins - 1) / 4)).round().int()), False), 'Solarize': (torch.linspace(255.0, 0.0, num_bins), False), 'AutoContrast': (torch.tensor(0.0), False), 'Equalize': (torch.tensor(0.0), False)} def forward(self, img: Tensor) -> Tensor: fill = self.fill if isinstance(img, Tensor): if isinstance(fill, (int, float)): fill = ([float(fill)] * F.get_image_num_channels(img)) elif (fill is not None): fill = [float(f) for f in fill] for _ in range(self.num_ops): op_meta = self._augmentation_space(self.num_magnitude_bins, F.get_image_size(img)) op_index = int(torch.randint(len(op_meta), (1,)).item()) op_name = list(op_meta.keys())[op_index] (magnitudes, signed) = op_meta[op_name] magnitude = (float(magnitudes[self.magnitude].item()) if (magnitudes.ndim > 0) else 0.0) if (signed and torch.randint(2, (1,))): magnitude *= (- 1.0) img = _apply_op(img, op_name, magnitude, interpolation=self.interpolation, fill=fill) return img def __repr__(self) -> str: s = (self.__class__.__name__ + '(') s += 'num_ops={num_ops}' s += ', magnitude={magnitude}' s += ', num_magnitude_bins={num_magnitude_bins}' s += ', interpolation={interpolation}' s += ', fill={fill}' s += ')' return s.format(**self.__dict__)
def find_first_non_zero_pixel(points, instance_image): points = list(points) coord = points[0] for pixel in points: pixel = list(pixel) pixel[0] = np.clip(pixel[0], 0, (instance_image.shape[1] - 1)) pixel[1] = np.clip(pixel[1], 0, (instance_image.shape[0] - 1)) coord = pixel if (instance_image[(pixel[1], pixel[0])] > 0): break return coord
def sample_and_group_all(xyz, points, use_xyz=True): batch_size = tf.shape(xyz)[0] nsample = xyz.get_shape()[1].value new_xyz = tf.tile(np.array([0, 0, 0], dtype=np.float32).reshape((1, 1, 3)), (batch_size, 1, 1)) idx = tf.tile(np.array(range(nsample), dtype=np.float32).reshape((1, 1, nsample)), (batch_size, 1, 1)) grouped_xyz = tf.reshape(xyz, (batch_size, 1, nsample, 3)) if (points is not None): if use_xyz: new_points = tf.concat([xyz, points], axis=2) else: new_points = points new_points = tf.expand_dims(new_points, 1) else: new_points = grouped_xyz return (new_xyz, new_points, idx, grouped_xyz)
_task('cross_lingual_lm') class CrossLingualLMTask(FairseqTask): def add_args(parser): parser.add_argument('data', help='colon separated path to data directories list, will be iterated upon during epochs in round-robin manner') parser.add_argument('--tokens-per-sample', default=512, type=int, help='max number of total tokens over all segments per sample') parser.add_argument('--monolingual-langs', default='en', type=str, help='comma separated list of languages for which we want to train XLM on') parser.add_argument('--raw-text', default=False, action='store_true', help='load raw text dataset') parser.add_argument('--lazy-load', action='store_true', help='load the dataset lazily') parser.add_argument('--shuffle', action='store_true', help='shuffle each monolingual dataset while training') def __init__(self, args, dictionary): super().__init__(args) self.dictionary = dictionary self.seed = args.seed self.distributed_world_size = args.distributed_world_size self.langs2id = self._lang_to_id(args.monolingual_langs) def _lang_to_id(self, languages: str): lang2id = {} langs = [l.strip() for l in languages.split(',')] for (id, lang) in enumerate(langs): lang2id[lang] = id return lang2id def load_dictionary(cls, filename): return MaskedLMDictionary.load(filename) def build_dictionary(cls, filenames, workers=1, threshold=(- 1), nwords=(- 1), padding_factor=8): d = MaskedLMDictionary() for filename in filenames: Dictionary.add_file_to_dictionary(filename, d, tokenizer.tokenize_line, workers) d.finalize(threshold=threshold, nwords=nwords, padding_factor=padding_factor) return d def target_dictionary(self): return self.dictionary def setup_task(cls, args, **kwargs): dictionary = MaskedLMDictionary.load(os.path.join(args.data, 'dict.txt')) print('| dictionary: {} types'.format(len(dictionary))) return cls(args, dictionary) def _load_single_lang_dataset(self, split, epoch): loaded_datasets = [] paths = self.args.data.split(':') assert (len(paths) > 0) data_path = paths[(epoch % len(paths))] for k in itertools.count(): split_k = (split + (str(k) if (k > 0) else '')) path = os.path.join(data_path, split_k) ds = indexed_dataset.make_dataset(path, impl=self.args.dataset_impl, fix_lua_indexing=True, dictionary=self.dictionary) if (ds is None): if (k > 0): break else: raise FileNotFoundError('Dataset not found: {} ({})'.format(split, data_path)) loaded_datasets.append(TokenBlockDataset(ds, ds.sizes, (self.args.tokens_per_sample - 1), pad=self.dictionary.pad(), eos=self.dictionary.eos())) print('| {} {} {} examples'.format(data_path, split_k, len(loaded_datasets[(- 1)]))) if (len(loaded_datasets) == 1): dataset = loaded_datasets[0] sizes = dataset.sizes else: dataset = ConcatDataset(loaded_datasets) sizes = np.concatenate([ds.sizes for ds in loaded_datasets]) return (dataset, sizes) def load_dataset(self, split, epoch=0, combine=False, **kwargs): dataset_map = OrderedDict() for lang in self.langs2id.keys(): language_split = '{}.{}'.format(split, lang) (block_dataset, sizes) = self._load_single_lang_dataset(split=language_split, epoch=epoch) dataset_map[lang] = MaskedLMDataset(dataset=block_dataset, sizes=sizes, vocab=self.dictionary, pad_idx=self.dictionary.pad(), mask_idx=self.dictionary.mask(), classif_token_idx=self.dictionary.eos(), sep_token_idx=self.dictionary.eos(), shuffle=getattr(self.args, 'shuffle', False), has_pairs=False, segment_id=self.langs2id[lang], seed=self.seed) self.datasets[split] = MultiCorpusSampledDataset(dataset_map) print('| {} {} {} examples'.format(self.args.data.split(':')[epoch], split, len(self.datasets[split])))
def test_graph_gnm(): (n_v, n_e) = (100, 500) g = graph_Gnm(n_v, n_e) assert (g.num_v == n_v) assert (g.num_e == n_e)
('translate', inputs={'source_imgs': runway.image(description='input image to be translated'), 'Strokes': runway.number(min=100, max=700, default=100, description='number of strokes')}, outputs={'image': runway.image(description='output image containing the translated result')}) def translate(learn, inputs): os.makedirs('images', exist_ok=True) inputs['source_imgs'].save('images/temp.jpg') paths = os.path.join('images', 'temp.jpg') args.img_path = paths args.max_m_strokes = inputs['Strokes'] pt = ProgressivePainter(args=args) final_rendered_image = optimize_x(pt) formatted = ((final_rendered_image * 255) / np.max(final_rendered_image)).astype('uint8') img = Image.fromarray(formatted) return img
def perfect_plot(ax, xarr, yarr, label): if (len(xarr) != len(yarr)): ax.plot(xarr[:(- 1)], yarr, label=label) else: ax.plot(xarr, yarr, label=label)
class TestTensorboardXWriter(unittest.TestCase): def test_no_files_created(self) -> None: with tempfile.TemporaryDirectory() as tmp_dir: writer = TensorboardXWriter(tmp_dir) writer.close() self.assertFalse(os.listdir(tmp_dir)) def test_single_write(self) -> None: with tempfile.TemporaryDirectory() as tmp_dir: writer = TensorboardXWriter(tmp_dir) writer._writer.add_scalar('testing', 1, 1) writer.close() self.assertTrue(os.listdir(tmp_dir))
def attach_head_and_body(root): head = ET.Element('head') body = ET.Element('body') root.append(head) root.append(body) meta1 = ET.Element('meta') meta1.set('name', 'ocr-system') meta1.set('content', 'eperiodica_fulltext') meta2 = ET.Element('meta') meta2.set('name', 'ocr-capabilities') meta2.set('content', 'ocr_page ocr_author ocr_carea ocr_photo ocr_caption ocr_linear ocr_footer ocr_header ocr_pageno ocr_table ocr_section ocrx_block ocrx_word') head.append(meta1) head.append(meta2) return (head, body)
class PathBuffer(): def __init__(self, capacity_in_transitions): self._capacity = capacity_in_transitions self._transitions_stored = 0 self._first_idx_of_next_path = 0 self._path_segments = collections.deque() self._buffer = {} def add_path(self, path): for (key, buf_arr) in self._buffer.items(): path_array = path.get(key, None) if (path_array is None): raise ValueError('Key {} missing from path.'.format(key)) if ((len(path_array.shape) != 2) or (path_array.shape[1] != buf_arr.shape[1])): raise ValueError('Array {} has wrong shape.'.format(key)) path_len = self._get_path_length(path) (first_seg, second_seg) = self._next_path_segments(path_len) while (self._path_segments and self._segments_overlap(first_seg, self._path_segments[0][0])): self._path_segments.popleft() while (self._path_segments and self._segments_overlap(second_seg, self._path_segments[0][0])): self._path_segments.popleft() self._path_segments.append((first_seg, second_seg)) for (key, array) in path.items(): buf_arr = self._get_or_allocate_key(key, array) buf_arr[first_seg.start:first_seg.stop] = array[:len(first_seg)] buf_arr[second_seg.start:second_seg.stop] = array[len(first_seg):] if (second_seg.stop != 0): self._first_idx_of_next_path = second_seg.stop else: self._first_idx_of_next_path = first_seg.stop self._transitions_stored = min(self._capacity, (self._transitions_stored + path_len)) def sample_path(self): path_idx = np.random.randint(len(self._path_segments)) (first_seg, second_seg) = self._path_segments[path_idx] first_seg_indices = np.arange(first_seg.start, first_seg.stop) second_seg_indices = np.arange(second_seg.start, second_seg.stop) indices = np.concatenate([first_seg_indices, second_seg_indices]) path = {key: buf_arr[indices] for (key, buf_arr) in self._buffer.items()} return path def sample_transitions(self, batch_size): idx = np.random.choice(self._transitions_stored, batch_size) return {key: buf_arr[idx] for (key, buf_arr) in self._buffer.items()} def _next_path_segments(self, n_indices): if (n_indices > self._capacity): raise ValueError('Path is too long to store in buffer.') start = self._first_idx_of_next_path end = (start + n_indices) if (end > self._capacity): second_end = (end - self._capacity) return (range(start, self._capacity), range(0, second_end)) else: return (range(start, end), range(0, 0)) def _get_or_allocate_key(self, key, array): buf_arr = self._buffer.get(key, None) if (buf_arr is None): buf_arr = np.zeros((self._capacity, array.shape[1]), array.dtype) self._buffer[key] = buf_arr return buf_arr def clear(self): self._transitions_stored = 0 self._first_idx_of_next_path = 0 self._path_segments.clear() self._buffer.clear() def _get_path_length(path): length_key = None length = None for (key, value) in path.items(): if (length is None): length = len(value) length_key = key elif (len(value) != length): raise ValueError('path has inconsistent lengths between {!r} and {!r}.'.format(length_key, key)) if (not length): raise ValueError('Nothing in path') return length def _segments_overlap(seg_a, seg_b): if ((not seg_a) or (not seg_b)): return False first = seg_a second = seg_b if (seg_b.start < seg_a.start): (first, second) = (seg_b, seg_a) assert (first.start <= second.start) return (first.stop > second.start) def n_transitions_stored(self): return int(self._transitions_stored)
def load_obj_data(filename): v_list = [] vt_list = [] vc_list = [] vn_list = [] f_list = [] fn_list = [] ft_list = [] fp = open(filename, 'r') lines = fp.readlines() fp.close() for line in lines: if (len(line) < 2): continue line_data = line.strip().split(' ') if (line_data[0] == 'v'): v_list.append((float(line_data[1]), float(line_data[2]), float(line_data[3]))) if (len(line_data) == 7): vc_list.append((float(line_data[4]), float(line_data[5]), float(line_data[6]))) else: vc_list.append((0.5, 0.5, 0.5)) if (line_data[0] == 'vt'): vt_list.append((float(line_data[1]), float(line_data[2]))) if (line_data[0] == 'vn'): vn_list.append((float(line_data[1]), float(line_data[2]), float(line_data[3]))) if (line_data[0] == 'f'): def segElementData(ele_str): fv = None ft = None fn = None eles = ele_str.strip().split('/') if (len(eles) == 1): fv = (int(eles[0]) - 1) elif (len(eles) == 2): fv = (int(eles[0]) - 1) ft = (int(eles[1]) - 1) elif (len(eles) == 3): fv = (int(eles[0]) - 1) fn = (int(eles[2]) - 1) ft = (None if (eles[1] == '') else (int(eles[1]) - 1)) return (fv, ft, fn) (fv0, ft0, fn0) = segElementData(line_data[1]) (fv1, ft1, fn1) = segElementData(line_data[2]) (fv2, ft2, fn2) = segElementData(line_data[3]) f_list.append((fv0, fv1, fv2)) if ((ft0 is not None) and (ft1 is not None) and (ft2 is not None)): ft_list.append((ft0, ft1, ft2)) if ((fn0 is not None) and (fn1 is not None) and (fn2 is not None)): fn_list.append((fn0, fn1, fn2)) v_list = np.asarray(v_list) vn_list = np.asarray(vn_list) vt_list = np.asarray(vt_list) vc_list = np.asarray(vc_list) f_list = np.asarray(f_list) ft_list = np.asarray(ft_list) fn_list = np.asarray(fn_list) model = {'v': v_list, 'vt': vt_list, 'vc': vc_list, 'vn': vn_list, 'f': f_list, 'ft': ft_list, 'fn': fn_list} return model
def test_potential_paramunits_1d(): from galpy import potential from galpy.util import conversion (ro, vo) = (10.5, 195.0) pot = potential.KGPotential(amp=1.0, K=((40.0 * units.Msun) / (units.pc ** 2)), F=((0.02 * units.Msun) / (units.pc ** 3)), D=(200 * units.pc), ro=ro, vo=vo) pot_nounits = potential.KGPotential(amp=1.0, K=(((40.0 / conversion.surfdens_in_msolpc2(vo, ro)) * 2.0) * numpy.pi), F=(((0.02 / conversion.dens_in_msolpc3(vo, ro)) * 4.0) * numpy.pi), D=(0.2 / ro), ro=ro, vo=vo) assert (numpy.fabs((pot(1.5, use_physical=False) - pot_nounits(1.5, use_physical=False))) < (10.0 ** (- 8.0))), 'KGPotential w/ parameters w/ units does not behave as expected' pot = potential.KGPotential(amp=1.0, K=(((40.0 * units.Msun) / (units.pc ** 2)) * constants.G), F=(((0.02 * units.Msun) / (units.pc ** 3)) * constants.G), D=(200 * units.pc), ro=ro, vo=vo) pot_nounits = potential.KGPotential(amp=1.0, K=(40.0 / conversion.surfdens_in_msolpc2(vo, ro)), F=(0.02 / conversion.dens_in_msolpc3(vo, ro)), D=(0.2 / ro), ro=ro, vo=vo) assert (numpy.fabs((pot(1.5, use_physical=False) - pot_nounits(1.5, use_physical=False))) < (10.0 ** (- 8.0))), 'KGPotential w/ parameters w/ units does not behave as expected' pot = potential.IsothermalDiskPotential(amp=1.2, sigma=((30.0 * units.km) / units.s), ro=ro, vo=vo) pot_nounits = potential.IsothermalDiskPotential(amp=1.2, sigma=(30.0 / vo), ro=ro, vo=vo) assert (numpy.fabs((pot(1.5, use_physical=False) - pot_nounits(1.5, use_physical=False))) < (10.0 ** (- 8.0))), 'IsothermalDiskPotential w/ parameters w/ units does not behave as expected' return None
class TFRemBertPreTrainedModel(metaclass=DummyObject): _backends = ['tf'] def __init__(self, *args, **kwargs): requires_backends(self, ['tf'])
class DistEvalHook(EvalHook): def after_train_epoch(self, runner): if (not self.every_n_epochs(runner, self.interval)): return current_ckpt_path = osp.join(runner.work_dir, f'epoch_{(runner.epoch + 1)}.pth') json_path = osp.join(runner.work_dir, 'best.json') if (osp.exists(json_path) and (len(self.best_json) == 0)): self.best_json = mmcv.load(json_path) self.best_score = self.best_json['best_score'] self.best_ckpt = self.best_json['best_ckpt'] self.key_indicator = self.best_json['key_indicator'] from mmpose.apis import multi_gpu_test results = multi_gpu_test(runner.model, self.dataloader, tmpdir=osp.join(runner.work_dir, '.eval_hook'), gpu_collect=self.gpu_collect) if (runner.rank == 0): print('\n') key_score = self.evaluate(runner, results) if (self.save_best and self.compare_func(key_score, self.best_score)): self.best_score = key_score self.logger.info(f'Now best checkpoint is epoch_{(runner.epoch + 1)}.pth') self.best_json['best_score'] = self.best_score self.best_json['best_ckpt'] = current_ckpt_path self.best_json['key_indicator'] = self.key_indicator save_checkpoint(runner.model, osp.join(runner.work_dir, 'best.pth')) mmcv.dump(self.best_json, json_path)
def test_digits_cosine_lazy_init(): model = SumRedundancySelection(100, 'cosine', optimizer='lazy', initial_subset=digits_cosine_ranking[:5]) model.fit(X_digits) assert_array_equal(model.ranking[:(- 5)], digits_cosine_ranking[5:]) assert_array_almost_equal(model.gains[:(- 5)], digits_cosine_gains[5:], 4) assert_array_almost_equal(model.subset, X_digits[model.ranking])
class Critic(nn.Module): def __init__(self, state_dim, action_dim): super(Critic, self).__init__() self.l1 = nn.Linear(state_dim, 400) self.l2 = nn.Linear((400 + action_dim), 300) self.l3_additional = nn.Linear(300, 300) self.l3 = nn.Linear(300, 1) def forward(self, x, u): x = F.relu(self.l1(x)) x = F.relu(self.l2(torch.cat([x, u], 1))) x = self.l3_additional(x) x = self.l3(x) return x
def get_split_subset(args, ds, split): manual_seed(args.split_seed) indices = randperm(len(ds)) valid_size = round((len(ds) * args.split_ratio)) if (args.dataset == 'cifar10'): if (split == 'train'): ds = Subset(ds, indices[:(- valid_size)]) elif (split == 'val'): ds = Subset(ds, indices[(- valid_size):]) elif (args.dataset in ['imagenet', 'imagenet-mini']): if (split == 'train'): ds = Subset(ds, indices[:(- valid_size)]) elif (split == 'test'): ds = Subset(ds, indices[(- valid_size):]) return ds
class AsyncMultiGPUTrainer(MultiGPUTrainer, SingleCostFeedfreeTrainer, MultiPredictorTowerTrainer): def __init__(self, config, input_queue=None, average_gradient=True, predict_tower=None): if hasattr(config, 'dataset'): self._input_method = QueueInput(config.dataset, input_queue) else: self._input_method = config.data assert isinstance(self._input_method, QueueInput) super(AsyncMultiGPUTrainer, self).__init__(config) if (predict_tower is not None): logger.warn('[Deprecated] Argument `predict_tower` is deprecated for trainer. Use TrainConfig.predict_tower instead!') config.predict_tower = predict_tower self._setup_predictor_factory(config.predict_tower) self._average_gradient = average_gradient assert tf.test.is_gpu_available() def _setup(self): super(AsyncMultiGPUTrainer, self)._setup() grad_list = MultiGPUTrainer._multi_tower_grads(self.config.tower, (lambda : self._get_cost_and_grad()[1])) gradprocs = self.model.get_gradient_processor() if (self._average_gradient and (self.config.nr_tower > 1)): gradprocs.insert(0, ScaleGradient(('.*', (1.0 / self.config.nr_tower)), log=False)) grad_list = [apply_grad_processors(g, gradprocs) for g in grad_list] self.train_op = tf.group(self.config.optimizer.apply_gradients(grad_list[0], get_global_step_var()), summary_moving_average(), name='train_op') self._start_async_threads(grad_list) def _start_async_threads(self, grad_list): self.async_step_counter = itertools.count() self.training_threads = [] for k in range(1, len(self.config.tower)): train_op = self.config.optimizer.apply_gradients(grad_list[k]) def f(op=train_op): self.sess.run([op]) next(self.async_step_counter) th = LoopThread(f) th.pause() th.start() self.training_threads.append(th) self.async_running = False def run_step(self): if (not self.async_running): self.async_running = True for th in self.training_threads: th.resume() next(self.async_step_counter) self.sess.run(self.train_op) def _trigger_epoch(self): self.async_running = False for th in self.training_threads: th.pause() try: if (self.config.tower > 1): async_step_total_cnt = int(re.findall('[0-9]+', self.async_step_counter.__str__())[0]) self.write_scalar_summary('async_global_step', async_step_total_cnt) except: logger.exception('Cannot log async_global_step') super(AsyncMultiGPUTrainer, self)._trigger_epoch()
def stanford_pre(path_bf, path_af, path_root='/home/cc', data_name='nyt', flist='before-parse-problems-flist.txt'): files = os.listdir(path_bf) files = [os.path.join(path_bf, f) for f in files] num_flist = multiprocessing.cpu_count() slice = (len(files) // num_flist) for i in range(num_flist): if (i == (num_flist - 1)): tmp = files[(i * slice):] else: tmp = files[(i * slice):((i + 1) * slice)] with open(os.path.join(path_root, ((data_name + str(i)) + flist)), 'w') as fd: fd.write('\n'.join(tmp)) file_lists = [os.path.join(path_root, ((data_name + str(i)) + flist)) for i in range(num_flist)] inp = zip(file_lists, ([path_af] * len(file_lists))) pool = multiprocessing.Pool(processes=num_flist) pool.map(run_snlp, inp) pool.close() pool.join()
class TestBoxIOU(unittest.TestCase): def test_pairwise_iou(self): boxes1 = torch.tensor([[0.0, 0.0, 1.0, 1.0], [0.0, 0.0, 1.0, 1.0]]) boxes2 = torch.tensor([[0.0, 0.0, 1.0, 1.0], [0.0, 0.0, 0.5, 1.0], [0.0, 0.0, 1.0, 0.5], [0.0, 0.0, 0.5, 0.5], [0.5, 0.5, 1.0, 1.0], [0.5, 0.5, 1.5, 1.5]]) expected_ious = torch.tensor([[1.0, 0.5, 0.5, 0.25, 0.25, (0.25 / (2 - 0.25))], [1.0, 0.5, 0.5, 0.25, 0.25, (0.25 / (2 - 0.25))]]) ious = pairwise_iou(Boxes(boxes1), Boxes(boxes2)) assert torch.allclose(ious, expected_ious)
def regadv(): filename = sys.argv[1] config = {} config['jobs'] = [] model_list = ['ResNet18'] batch_size = [(512, 32)] for (i, modelname) in enumerate(model_list): job = {} job['eps'] = 0.047 job['alpha'] = 0.01 job['model'] = {} job['model']['name'] = modelname job['adv_train'] = {} job['adv_train']['attack'] = 'ml' job['logfilename'] = './log/SVHN/{}_adv.log'.format(modelname) job['savename'] = './models/SVHN/{}_adv.pth'.format(modelname) job['epoch'] = 100 job['epoch1'] = 60 job['epoch2'] = 80 job['lr'] = 0.1 job['momentum'] = 0.9 job['epoch1_lr'] = 0.01 job['epoch2_lr'] = 0.001 job['test_attack'] = ['untarg1', 'untarg2'] job['n_test_adv'] = 1000 job['train_batch_size'] = batch_size[i][0] job['test_batch_size'] = batch_size[i][1] config['jobs'].append(job) j = json.dumps(config, indent=4) with open(filename, 'w+') as f: f.write(j)
def unwrap_if_singleton(x): if (isinstance(x, dict) and (len(x) == 1) and ('single_element' in x)): return x['single_element'] return x
def train(train_loader, model, criterion, optimizer, epoch, args): batch_time = AverageMeter('Time', ':6.3f') data_time = AverageMeter('Data', ':6.3f') losses = AverageMeter('Loss', ':.4e') top1 = AverageMeter('', ':6.2f') top5 = AverageMeter('', ':6.2f') progress = ProgressMeter(len(train_loader), [batch_time, data_time, losses, top1, top5], prefix='Epoch: [{}]'.format(epoch)) model.train() end = time.time() for (i, (images, _)) in enumerate(train_loader): data_time.update((time.time() - end)) if (args.gpu is not None): images[0] = images[0].cuda(args.gpu, non_blocking=True) images[1] = images[1].cuda(args.gpu, non_blocking=True) if (args.method is None): (output, target) = model(im_q=images[0], im_k=images[1]) loss = criterion(output, target) if (args.method == 'ifm'): (output, output_adv, target) = model(im_q=images[0], im_k=images[1]) loss_orig = criterion(output, target) loss_adv = criterion(output_adv, target) loss = (loss_orig + (args.alpha * loss_adv)) loss /= (1 + args.alpha) if (args.method == 'ifm_only'): (output, output_adv, target) = model(im_q=images[0], im_k=images[1]) loss_adv = criterion(output_adv, target) loss = loss_adv (acc1, acc5) = accuracy(output, target, topk=(1, 5)) losses.update(loss.item(), images[0].size(0)) top1.update(acc1[0], images[0].size(0)) top5.update(acc5[0], images[0].size(0)) optimizer.zero_grad() loss.backward() optimizer.step() batch_time.update((time.time() - end)) end = time.time() if ((i % args.print_freq) == 0): progress.display(i)
def get_post_fmean(blm, X, Psi=None, w=None): if (Psi is None): Psi = blm.lik.linear.basis.get_basis(X) if (w is None): w = get_post_params_mean(blm) return (Psi.dot(w) + blm.lik.linear.bias)
_model def resnet12_wide(pretrained=False, **kwargs): model_args = dict(block=BasicBlock, layers=[1, 1, 1, 2], base_width=(64 * 2), our_ver=True, **kwargs) return _create_resnet('resnet12_wide', pretrained, **model_args)
def run_model_on_fold(name, max_len, embed_size, embed, bulid_fun): max_features = 50000 scores = {} scores.setdefault('fit_time', []) scores.setdefault('score_time', []) scores.setdefault('test_F1', []) scores.setdefault('test_Precision', []) scores.setdefault('test_Recall', []) scores.setdefault('test_Accuracy', []) scores.setdefault('test_Specificity', []) scores.setdefault('test_Sensitivity', []) for (fold_n, (train_index, valid_index)) in enumerate(folds.split(X, y)): print('Fold', fold_n, 'started at', time.ctime()) (X_train, X_valid) = (X[train_index], X[valid_index]) (y_train, y_valid) = (y[train_index], y[valid_index]) tk = Tokenizer(lower=True, filters='', num_words=max_features, oov_token=True) tk.fit_on_texts(X_train) train_tokenized = tk.texts_to_sequences(X_train) valid_tokenized = tk.texts_to_sequences(X_valid) X_train = pad_sequences(train_tokenized, maxlen=max_len) X_valid = pad_sequences(valid_tokenized, maxlen=max_len) embedding_matrix = create_embedding_matrix(embed, tk, max_features) model = bulid_fun(X_train, y_train, X_valid, y_valid, max_len, max_features, embed_size, embedding_matrix, lr=0.001, lr_d=0, spatial_dr=0.1, dense_units=128, conv_size=128, dr=0.1, patience=4, fold_id=fold_n) y_preds = [] for i in model.predict(X_valid): if (i[0] >= 0.5): y_preds.append(1) else: y_preds.append(0) print(accuracy_score(y_valid, y_preds)) scores['test_F1'].append(f1_score(y_valid, y_preds, average='macro')) scores['test_Precision'].append(precision_score(y_valid, y_preds, average='macro')) scores['test_Recall'].append(recall_score(y_valid, y_preds, average='macro')) scores['test_Accuracy'].append(accuracy_score(y_valid, y_preds)) scores['test_Specificity'].append(specificity(y_valid, y_preds)) scores['test_Sensitivity'].append(sensitivity(y_valid, y_preds)) print('{:<10} | {:<7} {:<7} {:<7} {:<7} {:<7} {:<7}'.format(str(name)[:7], str(('%.4f' % (sum(scores['test_F1']) / 10))), str(('%.4f' % (sum(scores['test_Precision']) / 10))), str(('%.4f' % (sum(scores['test_Recall']) / 10))), str(('%.4f' % (sum(scores['test_Accuracy']) / 10))), str(('%.4f' % (sum(scores['test_Specificity']) / 10))), str(('%.4f' % (sum(scores['test_Sensitivity']) / 10))))) f = open('setC.txt', 'a+') f.write(('{:<10} | {:<7} {:<7} {:<7} {:<7} {:<7} {:<7}'.format(str(name)[:7], str(('%.4f' % (sum(scores['test_F1']) / 10))), str(('%.4f' % (sum(scores['test_Precision']) / 10))), str(('%.4f' % (sum(scores['test_Recall']) / 10))), str(('%.4f' % (sum(scores['test_Accuracy']) / 10))), str(('%.4f' % (sum(scores['test_Specificity']) / 10))), str(('%.4f' % (sum(scores['test_Sensitivity']) / 10)))) + '\n')) f.close()
class Lossless(RateEstimator): def forward_help(self, z, _, parent=None): (batch_size, z_dim) = z.shape z_hat = z with closing(io.BytesIO()) as f: np.savez_compressed(f, to_numpy(z_hat)) bit_rate = ((f.getbuffer().nbytes * 8) / batch_size) nats_rate = (bit_rate * math.log(2)) rates = (nats_rate + (z.mean(dim=(- 1)) * 0)) logs = dict() other = dict() return (z_hat, rates, logs, other) def aux_parameters(self): return iter(())
class NLIDataset(torch.utils.data.Dataset): def __init__(self, premise, hypothesis, label, memory_key=None, memory_value=None, attention=None): assert (len(premise) == len(hypothesis)) self.premise = premise.astype(np.long) self.hypothesis = hypothesis.astype(np.long) if (label is not None): assert (len(hypothesis) == len(label)) self.label = label.astype(np.long) else: self.label = None self.memory_key = memory_key self.memory_value = memory_value self.attention = attention def __getitem__(self, index): prem = self.premise[index] hyp = self.hypothesis[index] if ((self.memory_key is not None) and (self.memory_value is not None)): mk = self.memory_key[index] mv = self.memory_value[index] else: mk = 0 mv = 0 if (self.attention is not None): att = self.attention[index] else: att = 0 cove_prem = 0 cove_hyp = 0 if (self.label is not None): lab = self.label[index] else: lab = 0 return (prem, hyp, mk, mv, att, cove_prem, cove_hyp, lab) def __len__(self): return len(self.hypothesis)
class MetadataField(Field[DataArray]): def __init__(self, metadata: Any) -> None: self.metadata = metadata def get_padding_lengths(self) -> Dict[(str, int)]: return {} def as_tensor(self, padding_lengths: Dict[(str, int)]) -> DataArray: return self.metadata def empty_field(self) -> 'MetadataField': return MetadataField(None) def batch_tensors(cls, tensor_list: List[DataArray]) -> DataArray: return tensor_list def __str__(self) -> str: return f'MetadataField (print field.metadata to see specific information).'
class First_order_chemical_synapse(SynapseModel): def __init__(self, conn, **kwargs): super(First_order_chemical_synapse, self).__init__(conn) from .Connections import FullConnection assert isinstance(conn, FullConnection) self._syn_tau_variables['tau[link]'] = kwargs.get('tau', 2.0) self._syn_variables['R[link]'] = np.zeros([1, conn.post_num]) self._syn_variables['WgtSum[link]'] = np.zeros([1, conn.post_num]) if (conn.post.model_name == 'complex'): self._syn_operations.append(['WgtSum[link]', 'mat_mult_weight_complex', (conn.pre_var_name + '[input][updated]'), 'weight[link]', 'complex_beta[post]']) else: self._syn_operations.append(['WgtSum[link]', 'mat_mult_weight', '[input]', 'weight[link]']) self._syn_operations.append(['R[link]', 'var_linear', 'tau[link]', 'R[link]', 'WgtSum[link][updated]']) self._syn_operations.append([(conn.post_var_name + '[post]'), 'assign', 'R[link][updated]'])
class WordSequence(nn.Module): def __init__(self, data): super(WordSequence, self).__init__() print(('build word sequence feature extractor: %s...' % data.word_feature_extractor)) self.gpu = data.HP_gpu self.use_char = data.use_char self.droplstm = nn.Dropout(data.HP_dropout) self.bilstm_flag = data.HP_bilstm self.lstm_layer = data.HP_lstm_layer self.wordrep = WordRep(data) self.input_size = data.word_emb_dim self.task_num = data.task_number self.domain_num = data.domain_number self.task_emb_size = data.task_emb_dim self.domain_emb_size = data.domain_emb_dim self.pretrain_task_emb = None self.pretrain_domain_emb = None self.model1_task_idx = data.task_alphabet.get_index(data.model1_task_name) self.model1_domain_idx = data.domain_alphabet.get_index(data.model1_domain_name) self.model2_task_idx = data.task_alphabet.get_index(data.model2_task_name) self.model2_domain_idx = data.domain_alphabet.get_index(data.model2_domain_name) self.model3_task_idx = data.task_alphabet.get_index(data.model3_task_name) self.model3_domain_idx = data.domain_alphabet.get_index(data.model3_domain_name) self.model4_task_idx = data.task_alphabet.get_index(data.model4_task_name) self.model4_domain_idx = data.domain_alphabet.get_index(data.model4_domain_name) if self.use_char: self.input_size += data.HP_char_hidden_dim if (data.char_feature_extractor == 'ALL'): self.input_size += data.HP_char_hidden_dim for idx in range(data.feature_num): self.input_size += data.feature_emb_dims[idx] if self.bilstm_flag: lstm_hidden = (data.HP_hidden_dim // 2) else: lstm_hidden = data.HP_hidden_dim self.word_feature_extractor = data.word_feature_extractor self.LSTM_param_generator = Network_param_generater(self.input_size, lstm_hidden, data) if (self.word_feature_extractor == 'GRU'): self.lstm = nn.GRU(self.input_size, lstm_hidden, num_layers=self.lstm_layer, batch_first=True, bidirectional=self.bilstm_flag) elif (self.word_feature_extractor == 'LSTM'): self.lstm = LSTM(self.input_size, lstm_hidden, num_layers=self.lstm_layer, batch_first=True, bidirectional=self.bilstm_flag) elif (self.word_feature_extractor == 'CNN'): self.word2cnn = nn.Linear(self.input_size, data.HP_hidden_dim) self.cnn_layer = data.HP_cnn_layer print('CNN layer: ', self.cnn_layer) self.cnn_list = nn.ModuleList() self.cnn_drop_list = nn.ModuleList() self.cnn_batchnorm_list = nn.ModuleList() kernel = 3 pad_size = ((kernel - 1) / 2) for idx in range(self.cnn_layer): self.cnn_list.append(nn.Conv1d(data.HP_hidden_dim, data.HP_hidden_dim, kernel_size=kernel, padding=pad_size)) self.cnn_drop_list.append(nn.Dropout(data.HP_dropout)) self.cnn_batchnorm_list.append(nn.BatchNorm1d(data.HP_hidden_dim)) if self.gpu: self.droplstm = self.droplstm.cuda() if (self.word_feature_extractor == 'CNN'): self.word2cnn = self.word2cnn.cuda() for idx in range(self.cnn_layer): self.cnn_list[idx] = self.cnn_list[idx].cuda() self.cnn_drop_list[idx] = self.cnn_drop_list[idx].cuda() self.cnn_batchnorm_list[idx] = self.cnn_batchnorm_list[idx].cuda() else: self.lstm = self.lstm.cuda() self.LSTM_param_generator = self.LSTM_param_generator.cuda() def forward(self, mode, word_inputs, feature_inputs, word_seq_lengths, char_inputs, char_seq_lengths, char_seq_recover): word_represent = self.wordrep(word_inputs, feature_inputs, word_seq_lengths, char_inputs, char_seq_lengths, char_seq_recover) if (self.word_feature_extractor == 'CNN'): word_in = F.tanh(self.word2cnn(word_represent)).transpose(2, 1).contiguous() for idx in range(self.cnn_layer): if (idx == 0): cnn_feature = F.relu(self.cnn_list[idx](word_in)) else: cnn_feature = F.relu(self.cnn_list[idx](cnn_feature)) cnn_feature = self.cnn_drop_list[idx](cnn_feature) cnn_feature = self.cnn_batchnorm_list[idx](cnn_feature) feature_out = cnn_feature.transpose(2, 1).contiguous() else: packed_words = pack_padded_sequence(word_represent, word_seq_lengths.cpu().numpy(), True) hidden = None if (mode == 'model1'): task_idx = self.model1_task_idx domain_idx = self.model1_domain_idx elif (mode == 'model2'): task_idx = self.model2_task_idx domain_idx = self.model2_domain_idx elif (mode == 'model3'): task_idx = self.model3_task_idx domain_idx = self.model3_domain_idx elif (mode == 'model4'): task_idx = self.model4_task_idx domain_idx = self.model4_domain_idx lstm_param = self.LSTM_param_generator(task_idx, domain_idx) (outputs_forward, outputs_backward, outputs) = (None, None, None) if ((mode == 'model1') or (mode == 'model3')): (lstm_out, hidden) = self.lstm(packed_words, lstm_param, hidden) (lstm_out, _) = pad_packed_sequence(lstm_out) feature_out = self.droplstm(lstm_out.transpose(1, 0)) (outputs_forward, outputs_backward) = feature_out.chunk(2, (- 1)) else: (lstm_out, hidden) = self.lstm(packed_words, lstm_param, hidden) (lstm_out, _) = pad_packed_sequence(lstm_out) outputs = self.droplstm(lstm_out.transpose(1, 0)) return (outputs_forward, outputs_backward, outputs)
class IMSATHeader(nn.Module): def __init__(self, output_k=10, num_sub_heads=5): super().__init__() self.output_k = output_k self.num_sub_heads = num_sub_heads self.heads = nn.ModuleList([nn.Sequential(nn.Linear(1200, self.output_k), nn.Softmax(dim=1)) for _ in range(self.num_sub_heads)]) def forward(self, input): results = [] for i in range(self.num_sub_heads): results.append(self.heads[i](input)) return results
def chunk_list(examples, chunk_size=2, pad_to_divisible=True): n_examples = len(examples) remainder = (n_examples % chunk_size) if (pad_to_divisible and (remainder > 0)): n_pad = (chunk_size - remainder) pad = random.choices(examples, k=n_pad) examples = (examples + pad) n_examples = len(examples) remainder = 0 chunked_examples = [] n_chunks = int((n_examples / chunk_size)) n_chunks = ((n_chunks + 1) if (remainder > 0) else n_chunks) for i in range(n_chunks): chunked_examples.append(examples[(i * chunk_size):((i + 1) * chunk_size)]) return chunked_examples
def get_arrow_hex_str(batched_data, names): import pyarrow as pa sink = pa.BufferOutputStream() pred_arrow = pa.record_batch(batched_data, names=names) with pa.ipc.new_stream(sink, pred_arrow.schema) as writer: writer.write_batch(pred_arrow) pred_arrow = sink.getvalue().hex() pred_arrow = pred_arrow.decode('utf-8') sink.close() return pred_arrow
class MetricLogger(object): def __init__(self, delimiter='\t'): self.meters = defaultdict(SmoothedValue) self.delimiter = delimiter def update(self, **kwargs): for (k, v) in kwargs.items(): if (v is None): continue if isinstance(v, torch.Tensor): v = v.item() assert isinstance(v, (float, int)) self.meters[k].update(v) def __getattr__(self, attr): if (attr in self.meters): return self.meters[attr] if (attr in self.__dict__): return self.__dict__[attr] raise AttributeError("'{}' object has no attribute '{}'".format(type(self).__name__, attr)) def __str__(self): loss_str = [] for (name, meter) in self.meters.items(): loss_str.append('{}: {}'.format(name, str(meter))) return self.delimiter.join(loss_str) def synchronize_between_processes(self): for meter in self.meters.values(): meter.synchronize_between_processes() def add_meter(self, name, meter): self.meters[name] = meter def log_every(self, iterable, print_freq, header=None): i = 0 if (not header): header = '' start_time = time.time() end = time.time() iter_time = SmoothedValue(fmt='{avg:.4f} ({min:.4f} -- {max:.4f})') data_time = SmoothedValue(fmt='{avg:.4f} ({min:.4f} -- {max:.4f})') space_fmt = ((':' + str(len(str(len(iterable))))) + 'd') log_msg = [header, (('[{0' + space_fmt) + '}/{1}]'), 'eta: {eta}', '{meters}', 'time: {time}', 'data: {data}'] if torch.cuda.is_available(): log_msg.append('max mem: {memory:.0f}') log_msg = self.delimiter.join(log_msg) MB = (1024.0 * 1024.0) for obj in iterable: data_time.update((time.time() - end)) (yield obj) iter_time.update((time.time() - end)) if (((i % print_freq) == 0) or (i == (len(iterable) - 1))): eta_seconds = (iter_time.global_avg * (len(iterable) - i)) eta_string = str(datetime.timedelta(seconds=int(eta_seconds))) if torch.cuda.is_available(): print(log_msg.format(i, len(iterable), eta=eta_string, meters=str(self), time=str(iter_time), data=str(data_time), memory=(torch.cuda.max_memory_allocated() / MB))) else: print(log_msg.format(i, len(iterable), eta=eta_string, meters=str(self), time=str(iter_time), data=str(data_time))) i += 1 end = time.time() total_time = (time.time() - start_time) total_time_str = str(datetime.timedelta(seconds=int(total_time))) print('{} Total time: {} ({:.4f} s / it)'.format(header, total_time_str, (total_time / len(iterable))))
class ResNetPerfCallback(MainCallback): def before_val_epoch(self, runner): if (runner.config['ipex'] and runner.config['int8'] and runner.config['calibration']): print('running int8 calibration step\n') import intel_extension_for_pytorch as ipex from torch.ao.quantization import MinMaxObserver, PerChannelMinMaxObserver, QConfig qconfig = QConfig(activation=MinMaxObserver.with_args(qscheme=torch.per_tensor_symmetric, dtype=torch.qint8), weight=PerChannelMinMaxObserver.with_args(dtype=torch.qint8, qscheme=torch.per_channel_symmetric)) x = torch.randn(1, 3, 224, 224) prepared_model = ipex.quantization.prepare(runner.model, qconfig, x, inplace=True) with torch.no_grad(): for (i, (images, target)) in enumerate(runner.val_loader): images = images.contiguous(memory_format=torch.channels_last) prepared_model(images) if (i == 4): print(i) break prepared_model.save_qconf_summary(runner.config['configure_dir']) print('calibration step done') if (not runner.num_steps): runner.num_steps = len(runner.val_loader) invalidInputError((runner.num_steps > runner.config['warmup_iterations']), 'total steps should be larger than warmup iterations') if runner.config['dummy']: images = torch.randn(runner.config['batch'], 3, 224, 224) target = torch.arange(1, (runner.config['batch'] + 1)).long() if runner.config['ipex']: images = images.contiguous(memory_format=torch.channels_last) if runner.config['bf16']: images = images.to(torch.bfloat16) runner.put('images', images) runner.put('target', target) if (runner.config['warmup_iterations'] > 0): print('running warmup iterations') def on_val_forward(self, runner): if runner.config['dummy']: images = runner.get('images') target = runner.get('target') else: (images, target) = next(iter(runner.val_loader)) if runner.config['ipex']: images = images.contiguous(memory_format=torch.channels_last) if runner.config['bf16']: images = images.to(torch.bfloat16) runner.batch = (images, target) if (runner.batch_idx < runner.config['warmup_iterations']): (output, target, loss) = self.forward(runner, images, target, warmup=True) else: (output, target, loss) = self.forward(runner, images, target, warmup=False) runner.output = output runner.loss = loss runner.target = target def forward(self, runner, images, target, warmup=False): if warmup: if ((not runner.config['jit']) and runner.config['bf16']): with torch.cpu.amp.autocast(): output = runner.model(images) else: output = runner.model(images) else: with runner.timers.record('non_warmup_eval_fwd'): if ((not runner.config['jit']) and runner.config['bf16']): with torch.cpu.amp.autocast(): output = runner.model(images) else: output = runner.model(images) if runner.config['bf16']: output = output.to(torch.float32) loss = runner.criterion(output, target) return (output, target, loss)
def _add_category(context, name: str=None, color: Tuple[float]=None) -> None: if (name in context.scene.categories.keys()): log.warning(f'Skipping duplicate category {name}.') return if (color is None): color = zpy.color.random_color(output_style='frgb') log.info(f'Choosing random color for category {name}: {color}') new_category = context.scene.categories.add() new_category.name = name new_category.color = color
class Interpolation(): def interpolate_position(dataframe: NumTrajDF, sampling_rate: float, ip_type: Optional[Text]='linear', class_label_col: Optional[Text]=''): df = dataframe.reset_index() df_chunks = helper._df_split_helper(df) processes = ([None] * len(df_chunks)) manager = mlp.Manager() return_list = manager.list() ip_type = ip_type.lower().strip() if (ip_type == 'linear'): for i in range(len(processes)): processes[i] = mlp.Process(target=Interpolation._linear_ip, args=(df_chunks[i], sampling_rate, return_list, class_label_col)) processes[i].start() for j in range(len(processes)): processes[j].join() elif (ip_type == 'cubic'): for i in range(len(processes)): processes[i] = mlp.Process(target=Interpolation._cubic_ip, args=(df_chunks[i], sampling_rate, return_list, class_label_col)) processes[i].start() for j in range(len(processes)): processes[j].join() elif (ip_type == 'kinematic'): for i in range(len(processes)): processes[i] = mlp.Process(target=Interpolation._kinematic_ip, args=(df_chunks[i], sampling_rate, return_list, class_label_col)) processes[i].start() for j in range(len(processes)): processes[j].join() elif (ip_type == 'random-walk'): for i in range(len(processes)): processes[i] = mlp.Process(target=Interpolation._random_walk_ip, args=(df_chunks[i], sampling_rate, return_list, class_label_col)) processes[i].start() for j in range(len(processes)): processes[j].join() else: raise ValueError(f'Interpolation type: {ip_type} specified does not exist. Please check theinterpolation type specified and type again.') return NumTrajDF(pd.concat(return_list).reset_index(), const.LAT, const.LONG, const.DateTime, const.TRAJECTORY_ID) def _linear_ip(dataframe: Union[(pd.DataFrame, NumTrajDF)], sampling_rate: float, return_list: list, class_label_col): if (class_label_col == ''): dataframe = dataframe.reset_index()[[const.DateTime, const.TRAJECTORY_ID, const.LAT, const.LONG]].set_index(const.DateTime) else: dataframe = dataframe.reset_index()[[const.DateTime, const.TRAJECTORY_ID, const.LAT, const.LONG, class_label_col]].set_index(const.DateTime) ids_ = list(dataframe[const.TRAJECTORY_ID].value_counts().keys()) df_chunks = [dataframe.loc[(dataframe[const.TRAJECTORY_ID] == ids_[i])] for i in range(len(ids_))] small_pool = mlp.Pool(NUM_CPU) final = small_pool.starmap(helper.linear_help, zip(df_chunks, ids_, itertools.repeat(sampling_rate), itertools.repeat(class_label_col))) small_pool.close() small_pool.join() return_list.append(pd.concat(final)) def _cubic_ip(dataframe: Union[(pd.DataFrame, NumTrajDF)], sampling_rate: float, return_list: list, class_label_col): try: if (class_label_col == ''): dataframe = dataframe.reset_index()[[const.DateTime, const.TRAJECTORY_ID, const.LAT, const.LONG]].set_index(const.DateTime) else: dataframe = dataframe.reset_index()[[const.DateTime, const.TRAJECTORY_ID, const.LAT, const.LONG, class_label_col]].set_index(const.DateTime) ids_ = list(dataframe[const.TRAJECTORY_ID].value_counts().keys()) df_chunks = [dataframe.loc[(dataframe[const.TRAJECTORY_ID] == ids_[i])] for i in range(len(ids_))] small_pool = mlp.Pool(NUM_CPU) final = small_pool.starmap(helper.cubic_help, zip(df_chunks, ids_, itertools.repeat(sampling_rate), itertools.repeat(class_label_col))) small_pool.close() small_pool.join() return_list.append(pd.concat(final)) except ValueError: raise ValueError def _kinematic_ip(dataframe: Union[(pd.DataFrame, NumTrajDF)], sampling_rate, return_list, class_label_col): if (class_label_col == ''): dataframe = dataframe.reset_index()[[const.DateTime, const.TRAJECTORY_ID, const.LAT, const.LONG]].set_index(const.DateTime) else: dataframe = dataframe.reset_index()[[const.DateTime, const.TRAJECTORY_ID, const.LAT, const.LONG, class_label_col]].set_index(const.DateTime) ids_ = list(dataframe[const.TRAJECTORY_ID].value_counts().keys()) df_chunks = [dataframe.loc[(dataframe[const.TRAJECTORY_ID] == ids_[i])] for i in range(len(ids_))] small_pool = mlp.Pool(NUM_CPU) final = small_pool.starmap(helper.kinematic_help, zip(df_chunks, ids_, itertools.repeat(sampling_rate), itertools.repeat(class_label_col))) small_pool.close() small_pool.join() return_list.append(pd.concat(final)) def _random_walk_ip(dataframe: Union[(pd.DataFrame, NumTrajDF)], sampling_rate, return_list, class_label_col): if (class_label_col == ''): dataframe = dataframe.reset_index()[[const.DateTime, const.TRAJECTORY_ID, const.LAT, const.LONG]].set_index(const.DateTime) else: dataframe = dataframe.reset_index()[[const.DateTime, const.TRAJECTORY_ID, const.LAT, const.LONG, class_label_col]].set_index(const.DateTime) ids_ = list(dataframe[const.TRAJECTORY_ID].value_counts().keys()) df_chunks = [dataframe.loc[(dataframe[const.TRAJECTORY_ID] == ids_[i])] for i in range(len(ids_))] small_pool = mlp.Pool(NUM_CPU) final = small_pool.starmap(helper.random_walk_help, zip(df_chunks, ids_, itertools.repeat(sampling_rate), itertools.repeat(class_label_col))) small_pool.close() small_pool.join() return_list.append(pd.concat(final))
def sample_vectors(samples, num): (num_samples, device) = (samples.shape[0], samples.device) if (num_samples >= num): indices = torch.randperm(num_samples, device=device)[:num] else: indices = torch.randint(0, num_samples, (num,), device=device) return samples[indices]
def get_figure_properties(fig): return {'figwidth': fig.get_figwidth(), 'figheight': fig.get_figheight(), 'dpi': fig.dpi}
def evaluate(Net, config, load_data, train, test, optim_func): file_name = config['exp_name'] for trial in range(config['num_trials_eval']): csv_name = (((file_name + '_t') + str(trial)) + '.csv') model_name = (((file_name + '_t') + str(trial)) + '.pt') num_epochs = config['num_epochs_eval'] set_all_seeds((config['seed'] + trial)) df_train_loss = pd.DataFrame() df_test_acc = pd.DataFrame(columns=['epoch', 'test_acc', 'train_time']) df_lr = pd.DataFrame() net = Net(config) device = 'cpu' if torch.cuda.is_available(): device = 'cuda' net.to(device) criterion = SF.mse_count_loss(correct_rate=config['correct_rate'], incorrect_rate=config['incorrect_rate']) (optimizer, scheduler, loss_dependent) = optim_func(net, config) if config['early_stopping']: early_stopping = EarlyStopping_acc(patience=config['patience'], verbose=True, path=model_name) early_stopping.early_stop = False early_stopping.best_score = None (trainset, testset) = load_data(config) config['dataset_length'] = len(trainset) trainloader = DataLoader(trainset, batch_size=int(config['batch_size']), shuffle=True) testloader = DataLoader(testset, batch_size=int(config['batch_size']), shuffle=False) if loss_dependent: old_loss_hist = float('inf') print(f"=======Trial: {trial}, Batch: {config['batch_size']}, beta: {config['beta']:.3f}, threshold: {config['threshold']:.2f}, slope: {config['slope']}, lr: {config['lr']:.3e}======") for epoch in range(num_epochs): start_time = time.time() (loss_list, lr_list) = train(config, net, trainloader, criterion, optimizer, device, scheduler) epoch_time = (time.time() - start_time) if loss_dependent: avg_loss_hist = (sum(loss_list) / len(loss_list)) if (avg_loss_hist > old_loss_hist): for param_group in optimizer.param_groups: param_group['lr'] = (param_group['lr'] * 0.5) else: old_loss_hist = avg_loss_hist test_accuracy = test(config, net, testloader, device) print(f'Epoch: {epoch} Test Accuracy: {test_accuracy}') df_lr = df_lr.append(lr_list, ignore_index=True) df_train_loss = df_train_loss.append(loss_list, ignore_index=True) df_test_acc = df_test_acc.append({'epoch': epoch, 'test_acc': test_accuracy, 'train_time': epoch_time}, ignore_index=True) if config['save_csv']: df_train_loss.to_csv(('loss_' + csv_name), index=False) df_test_acc.to_csv(('acc_' + csv_name), index=False) df_lr.to_csv(('lr_' + csv_name), index=False) if config['early_stopping']: early_stopping(test_accuracy, net) if early_stopping.early_stop: print('Early stopping') early_stopping.early_stop = False early_stopping.best_score = None break
def replaceRule(expr, repls): for (k, m) in repls.items(): expr = expr.replace(k, m, map=False, simultaneous=True, exact=False) return expr
class Walker2dFullObsEnv(mujoco_env.MujocoEnv, utils.EzPickle): def __init__(self): asset_path = os.path.join(os.path.dirname(__file__), 'assets/walker2d.xml') mujoco_env.MujocoEnv.__init__(self, asset_path, 4) utils.EzPickle.__init__(self) def step(self, a): posbefore = self.sim.data.qpos[0] self.do_simulation(a, self.frame_skip) (posafter, height, ang) = self.sim.data.qpos[0:3] alive_bonus = 1.0 reward = ((posafter - posbefore) / self.dt) reward += alive_bonus reward -= (0.001 * np.square(a).sum()) done = (not ((height > 0.8) and (height < 2.0) and (ang > (- 1.0)) and (ang < 1.0))) ob = self._get_obs() return (ob, reward, done, {}) def _get_obs(self): qpos = self.sim.data.qpos qvel = self.sim.data.qvel return np.concatenate([qpos, qvel]).ravel() def reset_model(self): self.set_state((self.init_qpos + self.np_random.uniform(low=(- 0.005), high=0.005, size=self.model.nq)), (self.init_qvel + self.np_random.uniform(low=(- 0.005), high=0.005, size=self.model.nv))) return self._get_obs() def viewer_setup(self): self.viewer.cam.trackbodyid = 2 self.viewer.cam.distance = (self.model.stat.extent * 0.5) self.viewer.cam.lookat[2] = 1.15 self.viewer.cam.elevation = (- 20)
class TupleTensorOutputModel(nn.Module): def __init__(self): super().__init__() self.layer_1 = nn.Linear((28 * 28), 12) self.layer_2 = nn.Linear((28 * 28), 12) self.layer_3 = nn.Linear(24, 1) def forward(self, x1, x2): x1 = self.layer_1(x1) x2 = self.layer_2(x2) x = torch.cat([x1, x2], axis=1) output = self.layer_3(x) return (output, (x1, x2, x))
_criterion('cross_entropy') class CrossEntropyCriterion(FairseqCriterion): def __init__(self, args, task): super().__init__(args, task) def forward(self, model, sample, reduce=True): net_output = model(**sample['net_input']) lprobs = model.get_normalized_probs(net_output, log_probs=True) lprobs = lprobs.view((- 1), lprobs.size((- 1))) target = model.get_targets(sample, net_output).view((- 1)) loss = F.nll_loss(lprobs, target, size_average=False, ignore_index=self.padding_idx, reduce=reduce) sample_size = (sample['target'].size(0) if self.args.sentence_avg else sample['ntokens']) logging_output = {'loss': (utils.item(loss.data) if reduce else loss.data), 'ntokens': sample['ntokens'], 'nsentences': sample['target'].size(0), 'sample_size': sample_size} if (not reduce): logging_output['model_out'] = net_output return (loss, sample_size, logging_output) def aggregate_logging_outputs(logging_outputs): loss_sum = sum((log.get('loss', 0) for log in logging_outputs)) ntokens = sum((log.get('ntokens', 0) for log in logging_outputs)) nsentences = sum((log.get('nsentences', 0) for log in logging_outputs)) sample_size = sum((log.get('sample_size', 0) for log in logging_outputs)) agg_output = {'loss': ((loss_sum / sample_size) / math.log(2)), 'ntokens': ntokens, 'nsentences': nsentences, 'sample_size': sample_size} if (sample_size != ntokens): agg_output['nll_loss'] = ((loss_sum / ntokens) / math.log(2)) return agg_output
def create_dir(dir_path): if (not osp.exists(dir_path)): os.makedirs(dir_path) return dir_path
class SpatialTemporalEnsemble(nn.Module): def __init__(self, is_temporal_ensemble=False): super().__init__() self.is_temporal_ensemble = is_temporal_ensemble def _transform(self, imgs, mode): is_single_image = False if (imgs.ndim == 4): if self.is_temporal_ensemble: raise ValueError('"is_temporal_ensemble" must be False if the input is an image.') is_single_image = True imgs = imgs.unsqueeze(1) if (mode == 'vertical'): imgs = imgs.flip(4).clone() elif (mode == 'horizontal'): imgs = imgs.flip(3).clone() elif (mode == 'transpose'): imgs = imgs.permute(0, 1, 2, 4, 3).clone() if is_single_image: imgs = imgs.squeeze(1) return imgs def spatial_ensemble(self, imgs, model): img_list = [imgs.cpu()] for mode in ['vertical', 'horizontal', 'transpose']: img_list.extend([self._transform(t, mode) for t in img_list]) output_list = [model(t.to(imgs.device)).cpu() for t in img_list] for i in range(len(output_list)): if (i > 3): output_list[i] = self._transform(output_list[i], 'transpose') if ((i % 4) > 1): output_list[i] = self._transform(output_list[i], 'horizontal') if (((i % 4) % 2) == 1): output_list[i] = self._transform(output_list[i], 'vertical') outputs = torch.stack(output_list, dim=0) outputs = outputs.mean(dim=0, keepdim=False) return outputs.to(imgs.device) def forward(self, imgs, model): outputs = self.spatial_ensemble(imgs, model) if self.is_temporal_ensemble: outputs += self.spatial_ensemble(imgs.flip(1), model).flip(1) outputs *= 0.5 return outputs
class HiLAMParallel(BaseHiGraphModel): def __init__(self, args): super().__init__(args) total_edge_index_list = ((list(self.m2m_edge_index) + list(self.mesh_up_edge_index)) + list(self.mesh_down_edge_index)) total_edge_index = torch.cat(total_edge_index_list, dim=1) self.edge_split_sections = [ei.shape[1] for ei in total_edge_index_list] if (args.processor_layers == 0): self.processor = (lambda x, edge_attr: (x, edge_attr)) else: processor_nets = [InteractionNet(total_edge_index, args.hidden_dim, hidden_layers=args.hidden_layers, edge_chunk_sizes=self.edge_split_sections, aggr_chunk_sizes=self.N_mesh_levels) for _ in range(args.processor_layers)] self.processor = pyg.nn.Sequential('mesh_rep, edge_rep', [(net, 'mesh_rep, mesh_rep, edge_rep -> mesh_rep, edge_rep') for net in processor_nets]) def hi_processor_step(self, mesh_rep_levels, mesh_same_rep, mesh_up_rep, mesh_down_rep): mesh_rep = torch.cat(mesh_rep_levels, dim=1) mesh_edge_rep = torch.cat(((mesh_same_rep + mesh_up_rep) + mesh_down_rep), axis=1) (mesh_rep, mesh_edge_rep) = self.processor(mesh_rep, mesh_edge_rep) mesh_rep_levels = list(torch.split(mesh_rep, self.N_mesh_levels, dim=1)) mesh_edge_rep_sections = torch.split(mesh_edge_rep, self.edge_split_sections, dim=1) mesh_same_rep = mesh_edge_rep_sections[:self.N_levels] mesh_up_rep = mesh_edge_rep_sections[self.N_levels:(self.N_levels + (self.N_levels - 1))] mesh_down_rep = mesh_edge_rep_sections[(self.N_levels + (self.N_levels - 1)):] return (mesh_rep_levels, mesh_same_rep, mesh_up_rep, mesh_down_rep)
def init_weights(net, init_type='kaiming', scale=1.0, std=0.02): print('initialization method [{:s}]'.format(init_type)) if (init_type == 'normal'): weights_init_normal_ = functools.partial(weights_init_normal, std=std) net.apply(weights_init_normal_) elif (init_type == 'kaiming'): weights_init_kaiming_ = functools.partial(weights_init_kaiming, scale=scale) net.apply(weights_init_kaiming_) elif (init_type == 'orthogonal'): net.apply(weights_init_orthogonal) else: raise NotImplementedError('initialization method [{:s}] not implemented'.format(init_type))
class World(object): def __init__(self, name, args, timeout): self.client = None self.name = name self.args = args self.timeout = timeout self.server_fps = 0.0 self.simulation_time = 0 self.server_clock = pygame.time.Clock() self.world = None self.town_map = None self.actors_with_transforms = [] self._hud = None self._input = None self.surface_size = [0, 0] self.prev_scaled_size = 0 self.scaled_size = 0 self.hero_actor = None self.spawned_hero = None self.hero_transform = None self.scale_offset = [0, 0] self.vehicle_id_surface = None self.result_surface = None self.traffic_light_surfaces = TrafficLightSurfaces() self.affected_traffic_light = None self.map_image = None self.border_round_surface = None self.original_surface_size = None self.hero_surface = None self.actors_surface = None def _get_data_from_carla(self): try: self.client = carla.Client(self.args.host, self.args.port) self.client.set_timeout(self.timeout) if (self.args.map is None): world = self.client.get_world() else: world = self.client.load_world(self.args.map) town_map = world.get_map() return (world, town_map) except RuntimeError as ex: logging.error(ex) exit_game() def start(self, hud, input_control): (self.world, self.town_map) = self._get_data_from_carla() settings = self.world.get_settings() settings.no_rendering_mode = self.args.no_rendering self.world.apply_settings(settings) self.map_image = MapImage(carla_world=self.world, carla_map=self.town_map, pixels_per_meter=PIXELS_PER_METER, show_triggers=self.args.show_triggers, show_connections=self.args.show_connections, show_spawn_points=self.args.show_spawn_points) self._hud = hud self._input = input_control self.original_surface_size = min(self._hud.dim[0], self._hud.dim[1]) self.surface_size = self.map_image.big_map_surface.get_width() self.scaled_size = int(self.surface_size) self.prev_scaled_size = int(self.surface_size) self.actors_surface = pygame.Surface((self.map_image.surface.get_width(), self.map_image.surface.get_height())) self.actors_surface.set_colorkey(COLOR_BLACK) self.vehicle_id_surface = pygame.Surface((self.surface_size, self.surface_size)).convert() self.vehicle_id_surface.set_colorkey(COLOR_BLACK) self.border_round_surface = pygame.Surface(self._hud.dim, pygame.SRCALPHA).convert() self.border_round_surface.set_colorkey(COLOR_WHITE) self.border_round_surface.fill(COLOR_BLACK) center_offset = (int((self._hud.dim[0] / 2)), int((self._hud.dim[1] / 2))) pygame.draw.circle(self.border_round_surface, COLOR_ALUMINIUM_1, center_offset, int((self._hud.dim[1] / 2))) pygame.draw.circle(self.border_round_surface, COLOR_WHITE, center_offset, int(((self._hud.dim[1] - 8) / 2))) scaled_original_size = (self.original_surface_size * (1.0 / 0.9)) self.hero_surface = pygame.Surface((scaled_original_size, scaled_original_size)).convert() self.result_surface = pygame.Surface((self.surface_size, self.surface_size)).convert() self.result_surface.set_colorkey(COLOR_BLACK) self.select_hero_actor() self.hero_actor.set_autopilot(False) self._input.wheel_offset = HERO_DEFAULT_SCALE self._input.control = carla.VehicleControl() weak_self = weakref.ref(self) self.world.on_tick((lambda timestamp: World.on_world_tick(weak_self, timestamp))) def select_hero_actor(self): hero_vehicles = [actor for actor in self.world.get_actors() if (('vehicle' in actor.type_id) and (actor.attributes['role_name'] == 'hero'))] if (len(hero_vehicles) > 0): self.hero_actor = random.choice(hero_vehicles) self.hero_transform = self.hero_actor.get_transform() else: self._spawn_hero() def _spawn_hero(self): blueprint = random.choice(self.world.get_blueprint_library().filter(self.args.filter)) blueprint.set_attribute('role_name', 'hero') if blueprint.has_attribute('color'): color = random.choice(blueprint.get_attribute('color').recommended_values) blueprint.set_attribute('color', color) while (self.hero_actor is None): spawn_points = self.world.get_map().get_spawn_points() spawn_point = (random.choice(spawn_points) if spawn_points else carla.Transform()) self.hero_actor = self.world.try_spawn_actor(blueprint, spawn_point) self.hero_transform = self.hero_actor.get_transform() self.spawned_hero = self.hero_actor def tick(self, clock): actors = self.world.get_actors() self.actors_with_transforms = [(actor, actor.get_transform()) for actor in actors] if (self.hero_actor is not None): self.hero_transform = self.hero_actor.get_transform() self.update_hud_info(clock) def update_hud_info(self, clock): hero_mode_text = [] if (self.hero_actor is not None): hero_speed = self.hero_actor.get_velocity() hero_speed_text = (3.6 * math.sqrt((((hero_speed.x ** 2) + (hero_speed.y ** 2)) + (hero_speed.z ** 2)))) affected_traffic_light_text = 'None' if (self.affected_traffic_light is not None): state = self.affected_traffic_light.state if (state == carla.TrafficLightState.Green): affected_traffic_light_text = 'GREEN' elif (state == carla.TrafficLightState.Yellow): affected_traffic_light_text = 'YELLOW' else: affected_traffic_light_text = 'RED' affected_speed_limit_text = self.hero_actor.get_speed_limit() if math.isnan(affected_speed_limit_text): affected_speed_limit_text = 0.0 hero_mode_text = ['Hero Mode: ON', ('Hero ID: %7d' % self.hero_actor.id), ('Hero Vehicle: %14s' % get_actor_display_name(self.hero_actor, truncate=14)), ('Hero Speed: %3d km/h' % hero_speed_text), 'Hero Affected by:', (' Traffic Light: %12s' % affected_traffic_light_text), (' Speed Limit: %3d km/h' % affected_speed_limit_text)] else: hero_mode_text = ['Hero Mode: OFF'] self.server_fps = self.server_clock.get_fps() self.server_fps = ('inf' if (self.server_fps == float('inf')) else round(self.server_fps)) info_text = [('Server: % 16s FPS' % self.server_fps), ('Client: % 16s FPS' % round(clock.get_fps())), ('Simulation Time: % 12s' % datetime.timedelta(seconds=int(self.simulation_time))), ('Map Name: %10s' % self.town_map.name)] self._hud.add_info(self.name, info_text) self._hud.add_info('HERO', hero_mode_text) def on_world_tick(weak_self, timestamp): self = weak_self() if (not self): return self.server_clock.tick() self.server_fps = self.server_clock.get_fps() self.simulation_time = timestamp.elapsed_seconds def _show_nearby_vehicles(self, vehicles): info_text = [] if ((self.hero_actor is not None) and (len(vehicles) > 1)): location = self.hero_transform.location vehicle_list = [x[0] for x in vehicles if (x[0].id != self.hero_actor.id)] def distance(v): return location.distance(v.get_location()) for (n, vehicle) in enumerate(sorted(vehicle_list, key=distance)): if (n > 15): break vehicle_type = get_actor_display_name(vehicle, truncate=22) info_text.append(('% 5d %s' % (vehicle.id, vehicle_type))) self._hud.add_info('NEARBY VEHICLES', info_text) def _split_actors(self): vehicles = [] traffic_lights = [] speed_limits = [] walkers = [] for actor_with_transform in self.actors_with_transforms: actor = actor_with_transform[0] if ('vehicle' in actor.type_id): vehicles.append(actor_with_transform) elif ('traffic_light' in actor.type_id): traffic_lights.append(actor_with_transform) elif ('speed_limit' in actor.type_id): speed_limits.append(actor_with_transform) elif ('walker.pedestrian' in actor.type_id): walkers.append(actor_with_transform) return (vehicles, traffic_lights, speed_limits, walkers) def _render_traffic_lights(self, surface, list_tl, world_to_pixel): self.affected_traffic_light = None for tl in list_tl: world_pos = tl.get_location() pos = world_to_pixel(world_pos) if self.args.show_triggers: corners = Util.get_bounding_box(tl) corners = [world_to_pixel(p) for p in corners] pygame.draw.lines(surface, COLOR_BUTTER_1, True, corners, 2) if (self.hero_actor is not None): corners = Util.get_bounding_box(tl) corners = [world_to_pixel(p) for p in corners] tl_t = tl.get_transform() transformed_tv = tl_t.transform(tl.trigger_volume.location) hero_location = self.hero_actor.get_location() d = hero_location.distance(transformed_tv) s = (Util.length(tl.trigger_volume.extent) + Util.length(self.hero_actor.bounding_box.extent)) if (d <= s): self.affected_traffic_light = tl srf = self.traffic_light_surfaces.surfaces['h'] surface.blit(srf, srf.get_rect(center=pos)) srf = self.traffic_light_surfaces.surfaces[tl.state] surface.blit(srf, srf.get_rect(center=pos)) def _render_speed_limits(self, surface, list_sl, world_to_pixel, world_to_pixel_width): font_size = world_to_pixel_width(2) radius = world_to_pixel_width(2) font = pygame.font.SysFont('Arial', font_size) for sl in list_sl: (x, y) = world_to_pixel(sl.get_location()) white_circle_radius = int((radius * 0.75)) pygame.draw.circle(surface, COLOR_SCARLET_RED_1, (x, y), radius) pygame.draw.circle(surface, COLOR_ALUMINIUM_0, (x, y), white_circle_radius) limit = sl.type_id.split('.')[2] font_surface = font.render(limit, True, COLOR_ALUMINIUM_5) if self.args.show_triggers: corners = Util.get_bounding_box(sl) corners = [world_to_pixel(p) for p in corners] pygame.draw.lines(surface, COLOR_PLUM_2, True, corners, 2) if (self.hero_actor is not None): angle = ((- self.hero_transform.rotation.yaw) - 90.0) font_surface = pygame.transform.rotate(font_surface, angle) offset = font_surface.get_rect(center=(x, y)) surface.blit(font_surface, offset) else: surface.blit(font_surface, ((x - (radius / 2)), (y - (radius / 2)))) def _render_walkers(self, surface, list_w, world_to_pixel): for w in list_w: color = COLOR_PLUM_0 bb = w[0].bounding_box.extent corners = [carla.Location(x=(- bb.x), y=(- bb.y)), carla.Location(x=bb.x, y=(- bb.y)), carla.Location(x=bb.x, y=bb.y), carla.Location(x=(- bb.x), y=bb.y)] w[1].transform(corners) corners = [world_to_pixel(p) for p in corners] pygame.draw.polygon(surface, color, corners) def _render_vehicles(self, surface, list_v, world_to_pixel): for v in list_v: color = COLOR_SKY_BLUE_0 if (int(v[0].attributes['number_of_wheels']) == 2): color = COLOR_CHOCOLATE_1 if (v[0].attributes['role_name'] == 'hero'): color = COLOR_CHAMELEON_0 bb = v[0].bounding_box.extent corners = [carla.Location(x=(- bb.x), y=(- bb.y)), carla.Location(x=(bb.x - 0.8), y=(- bb.y)), carla.Location(x=bb.x, y=0), carla.Location(x=(bb.x - 0.8), y=bb.y), carla.Location(x=(- bb.x), y=bb.y), carla.Location(x=(- bb.x), y=(- bb.y))] v[1].transform(corners) corners = [world_to_pixel(p) for p in corners] pygame.draw.lines(surface, color, False, corners, int(math.ceil((4.0 * self.map_image.scale)))) def render_actors(self, surface, vehicles, traffic_lights, speed_limits, walkers): self._render_traffic_lights(surface, [tl[0] for tl in traffic_lights], self.map_image.world_to_pixel) self._render_speed_limits(surface, [sl[0] for sl in speed_limits], self.map_image.world_to_pixel, self.map_image.world_to_pixel_width) self._render_vehicles(surface, vehicles, self.map_image.world_to_pixel) self._render_walkers(surface, walkers, self.map_image.world_to_pixel) def clip_surfaces(self, clipping_rect): self.actors_surface.set_clip(clipping_rect) self.vehicle_id_surface.set_clip(clipping_rect) self.result_surface.set_clip(clipping_rect) def _compute_scale(self, scale_factor): m = self._input.mouse_pos px = ((m[0] - self.scale_offset[0]) / float(self.prev_scaled_size)) py = ((m[1] - self.scale_offset[1]) / float(self.prev_scaled_size)) diff_between_scales = (((float(self.prev_scaled_size) * px) - (float(self.scaled_size) * px)), ((float(self.prev_scaled_size) * py) - (float(self.scaled_size) * py))) self.scale_offset = ((self.scale_offset[0] + diff_between_scales[0]), (self.scale_offset[1] + diff_between_scales[1])) self.prev_scaled_size = self.scaled_size self.map_image.scale_map(scale_factor) def render(self, display): if (self.actors_with_transforms is None): return self.result_surface.fill(COLOR_BLACK) (vehicles, traffic_lights, speed_limits, walkers) = self._split_actors() scale_factor = self._input.wheel_offset self.scaled_size = int((self.map_image.width * scale_factor)) if (self.scaled_size != self.prev_scaled_size): self._compute_scale(scale_factor) self.actors_surface.fill(COLOR_BLACK) self.render_actors(self.actors_surface, vehicles, traffic_lights, speed_limits, walkers) self._hud.render_vehicles_ids(self.vehicle_id_surface, vehicles, self.map_image.world_to_pixel, self.hero_actor, self.hero_transform) self._show_nearby_vehicles(vehicles) surfaces = ((self.map_image.surface, (0, 0)), (self.actors_surface, (0, 0)), (self.vehicle_id_surface, (0, 0))) angle = (0.0 if (self.hero_actor is None) else (self.hero_transform.rotation.yaw + 90.0)) self.traffic_light_surfaces.rotozoom((- angle), self.map_image.scale) center_offset = (0, 0) if (self.hero_actor is not None): hero_location_screen = self.map_image.world_to_pixel(self.hero_transform.location) hero_front = self.hero_transform.get_forward_vector() translation_offset = (((hero_location_screen[0] - (self.hero_surface.get_width() / 2)) + (hero_front.x * PIXELS_AHEAD_VEHICLE)), ((hero_location_screen[1] - (self.hero_surface.get_height() / 2)) + (hero_front.y * PIXELS_AHEAD_VEHICLE))) clipping_rect = pygame.Rect(translation_offset[0], translation_offset[1], self.hero_surface.get_width(), self.hero_surface.get_height()) self.clip_surfaces(clipping_rect) Util.blits(self.result_surface, surfaces) self.border_round_surface.set_clip(clipping_rect) self.hero_surface.fill(COLOR_ALUMINIUM_4) self.hero_surface.blit(self.result_surface, ((- translation_offset[0]), (- translation_offset[1]))) rotated_result_surface = pygame.transform.rotozoom(self.hero_surface, angle, 0.9).convert() center = ((display.get_width() / 2), (display.get_height() / 2)) rotation_pivot = rotated_result_surface.get_rect(center=center) display.blit(rotated_result_surface, rotation_pivot) display.blit(self.border_round_surface, (0, 0)) else: translation_offset = (((self._input.mouse_offset[0] * scale_factor) + self.scale_offset[0]), ((self._input.mouse_offset[1] * scale_factor) + self.scale_offset[1])) center_offset = (((abs((display.get_width() - self.surface_size)) / 2) * scale_factor), 0) clipping_rect = pygame.Rect(((- translation_offset[0]) - center_offset[0]), (- translation_offset[1]), self._hud.dim[0], self._hud.dim[1]) self.clip_surfaces(clipping_rect) Util.blits(self.result_surface, surfaces) display.blit(self.result_surface, ((translation_offset[0] + center_offset[0]), translation_offset[1])) def destroy(self): if (self.spawned_hero is not None): self.spawned_hero.destroy()
class TestGraphMatMulFusion(unittest.TestCase): def setUpClass(self): build_fake_yaml() self.op_wise_sequences = TensorflowQuery(local_config_file=os.path.join(os.path.dirname(neural_compressor.__file__), 'adaptor/tensorflow.yaml')).get_eightbit_patterns() def tearDownClass(self): os.remove('fake_yaml.yaml') _random() def test_matmul_biasadd_relu_requantize_fusion(self): g = tf.Graph() with g.as_default(): x_data = np.array([[0.1, 0.2], [0.2, 0.3]]) y_data = np.array([[1, 2], [3, 4]], dtype=np.float32) x = tf.placeholder(tf.float32, shape=[2, 2], name='x') y = tf.constant(y_data, dtype=tf.float32, shape=[2, 2]) z = tf.matmul(x, y) z = tf.nn.bias_add(z, [1, 2]) z = tf.nn.relu(z, name='op_to_store') found_quantized_matmul = False with tf.Session() as sess: sess.run(z, feed_dict={x: x_data, y: y_data}) float_graph_def = sess.graph.as_graph_def() from neural_compressor.experimental import Quantization, common quantizer = Quantization('fake_yaml.yaml') dataset = quantizer.dataset('dummy', shape=(2, 2), label=True) quantizer.calib_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.eval_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.model = float_graph_def output_graph = quantizer.fit() for i in output_graph.graph_def.node: if (i.op == 'QuantizedMatMulWithBiasAndReluAndRequantize'): found_quantized_matmul = True break self.assertEqual(found_quantized_matmul, True) _random() def test_first_matmul_biasadd_relu_fusion(self): x_data = np.array([[0.1, 0.2], [0.2, 0.3]]) y_data = np.array([[1, 2], [3, 4]], dtype=np.float32) x = tf.placeholder(tf.float32, shape=[2, 2], name='x') y = tf.constant(y_data, dtype=tf.float32, shape=[2, 2]) z = tf.matmul(x, y) z = tf.nn.bias_add(z, [1, 2]) z = tf.nn.relu(z, name='op_to_store') with tf.Session() as sess: sess.run(z, feed_dict={x: x_data, y: y_data}) float_graph_def = sess.graph.as_graph_def() from neural_compressor.experimental import Quantization, common quantizer = Quantization('fake_yaml.yaml') dataset = quantizer.dataset('dummy', shape=(2, 2), label=True) quantizer.calib_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.eval_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.model = float_graph_def output_graph = quantizer.fit() found_quantized_matmul = False for i in output_graph.graph_def.node: if ((i.op == 'QuantizeV2') and (i.name == 'MatMul_eightbit_quantize_x') and (i.attr['T'].type == dtypes.quint8)): found_quantized_matmul = True break self.assertEqual(found_quantized_matmul, True) _random() def test_matmul_biasadd_requantize_dequantize_fusion(self): g = tf.Graph() with g.as_default(): x_data = np.array([[0.1, 0.2], [0.2, 0.3]]) y_data = np.array([[1, 2], [3, 4]], dtype=np.float32) x = tf.placeholder(tf.float32, shape=[2, 2], name='x') y = tf.constant(y_data, dtype=tf.float32, shape=[2, 2]) z = tf.matmul(x, y) z = tf.nn.bias_add(z, [1, 2]) z = tf.identity(z, name='op_to_store') found_quantized_matmul = False if (tf.version.VERSION < '2.2.0'): found_quantized_matmul = True else: with tf.Session() as sess: sess.run(z, feed_dict={x: x_data, y: y_data}) float_graph_def = sess.graph.as_graph_def() from neural_compressor.experimental import Quantization, common quantizer = Quantization('fake_yaml.yaml') dataset = quantizer.dataset('dummy', shape=(2, 2), label=True) quantizer.calib_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.eval_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.model = float_graph_def output_graph = quantizer.fit() for i in output_graph.graph_def.node: if (i.op == 'QuantizedMatMulWithBiasAndDequantize'): found_quantized_matmul = True break self.assertEqual(found_quantized_matmul, True) _random() def test_matmul_biasadd_requantize_dequantize_last_fusion(self): g = tf.Graph() with g.as_default(): x_data = np.array([[0.1, 0.2], [0.2, 0.3]]) y_data = np.array([[1, 2], [3, 4]], dtype=np.float32) x = tf.placeholder(tf.float32, shape=[2, 2], name='x') y = tf.constant(y_data, dtype=tf.float32, shape=[2, 2]) z = tf.matmul(x, y) z = tf.nn.bias_add(z, [1, 2], name='op_to_store') found_quantized_matmul = False if (tf.version.VERSION < '2.2.0'): found_quantized_matmul = True else: with tf.Session() as sess: sess.run(z, feed_dict={x: x_data, y: y_data}) float_graph_def = sess.graph.as_graph_def() from neural_compressor.experimental import Quantization, common quantizer = Quantization('fake_yaml.yaml') dataset = quantizer.dataset('dummy', shape=(2, 2), label=True) quantizer.calib_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.eval_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.model = float_graph_def output_graph = quantizer.fit() for i in output_graph.graph_def.node: if ((i.op == 'QuantizedMatMulWithBiasAndDequantize') and (i.name == 'op_to_store')): found_quantized_matmul = True break self.assertEqual(found_quantized_matmul, True) _random() def test_disable_matmul_fusion(self): g = tf.Graph() with g.as_default(): x_data = np.array([[0.1, 0.2], [0.2, 0.3]]) y_data = np.array([[1, 2], [3, 4]], dtype=np.float32) x = tf.placeholder(tf.float32, shape=[2, 2], name='x') y = tf.constant(y_data, dtype=tf.float32, shape=[2, 2]) z = tf.matmul(x, y, name='no_quant_matmul') z = tf.nn.relu6(z, name='op_to_store') found_quantized_matmul = False with tf.Session() as sess: sess.run(z, feed_dict={x: x_data, y: y_data}) float_graph_def = sess.graph.as_graph_def() from neural_compressor.experimental import Quantization, common quantizer = Quantization('fake_yaml.yaml') dataset = quantizer.dataset('dummy', shape=(2, 2), label=True) quantizer.calib_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.eval_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.model = float_graph_def output_graph = quantizer.fit() for i in output_graph.graph_def.node: if ((i.op == 'QuantizedMatMulWithBiasAndDequantize') and (i.name == 'op_to_store')): found_quantized_matmul = True break self.assertEqual(found_quantized_matmul, False) _random() def test_disable_matmul_fusion_with_transpose_b_true(self): g = tf.Graph() with g.as_default(): x_data = np.array([[0.1, 0.2], [0.2, 0.3]]) y_data = np.array([[1, 2], [3, 4]], dtype=np.float32) x = tf.placeholder(tf.float32, shape=[2, 2], name='x') y = tf.constant(y_data, dtype=tf.float32, shape=[2, 2]) z = tf.matmul(x, y, name='no_quant_matmul', transpose_b=True) z = tf.nn.relu6(z, name='op_to_store') found_quantized_matmul = False with tf.Session() as sess: sess.run(z, feed_dict={x: x_data, y: y_data}) float_graph_def = sess.graph.as_graph_def() from neural_compressor.experimental import Quantization, common quantizer = Quantization('fake_yaml.yaml') dataset = quantizer.dataset('dummy', shape=(2, 2), label=True) quantizer.calib_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.eval_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.model = float_graph_def output_graph = quantizer.fit() for i in output_graph.graph_def.node: if ((i.op == 'QuantizedMatMulWithBiasAndDequantize') and (i.name == 'op_to_store')): found_quantized_matmul = True break self.assertEqual(found_quantized_matmul, False) _random() ((float(tf.__version__[:3]) > 2.7), 'only tf lower than 2.8 enable dummy biasadd') def test_matmul_with_dummy_biasadd(self): g = tf.Graph() with g.as_default(): x_data = np.array([[0.1, 0.2], [0.2, 0.3]]) y_data = np.array([[1, 2], [3, 4]], dtype=np.float32) x = tf.placeholder(tf.float32, shape=[2, 2], name='x') y = tf.constant(y_data, dtype=tf.float32, shape=[2, 2]) z = tf.matmul(x, y, name='no_quant_matmul') z = tf.identity(z, name='op_to_store') found_quantized_matmul = True with tf.Session() as sess: sess.run(z, feed_dict={x: x_data, y: y_data}) float_graph_def = sess.graph.as_graph_def() from neural_compressor.experimental import Quantization, common quantizer = Quantization('fake_yaml.yaml') dataset = quantizer.dataset('dummy', shape=(2, 2), label=True) quantizer.calib_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.eval_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.model = float_graph_def output_graph = quantizer.fit() for i in output_graph.graph_def.node: if (i.op == 'MatMul'): found_quantized_matmul = False break self.assertEqual(found_quantized_matmul, True) _random() ((float(tf.__version__[:3]) > 2.7), 'only tf lower than 2.8 enable dummy biasadd') def test_matmul_with_nan(self): g = tf.Graph() with g.as_default(): x_data = np.array([[0.1, 0.2], [0.2, 0.3]]) nan_array = np.empty((2, 2), dtype=np.float32) nan_array[:] = np.NaN x = tf.placeholder(tf.float32, shape=[2, 2], name='x') z = tf.matmul(x, nan_array, name='no_quant_matmul') z = tf.identity(z, name='op_to_store') found_quantized_matmul = True with tf.Session() as sess: sess.run(z, feed_dict={x: x_data}) float_graph_def = sess.graph.as_graph_def() from neural_compressor.experimental import Quantization, common quantizer = Quantization('fake_yaml.yaml') dataset = quantizer.dataset('dummy', shape=(2, 2), label=True) quantizer.calib_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.eval_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.model = float_graph_def output_graph = quantizer.fit() for i in output_graph.graph_def.node: if (i.op == 'MatMul'): found_quantized_matmul = False break self.assertEqual(found_quantized_matmul, True) _random() def test_matmul_with_reshape_transpose(self): g = tf.Graph() with g.as_default(): x_data = np.array([[0.1, 0.2], [0.2, 0.3]]) y_data = np.array([[1, 2], [3, 4]], dtype=np.float32) x = tf.placeholder(tf.float32, shape=[2, 2], name='x') y = tf.constant(y_data, dtype=tf.float32, shape=[2, 2]) transpose = tf.transpose(y, perm=[1, 0]) reshape = tf.reshape(transpose, [2, 2]) z = tf.matmul(x, reshape, name='no_quant_matmul') z = tf.nn.bias_add(z, [1, 2], name='op_to_store') found_quantized_matmul = True with tf.Session() as sess: sess.run(z, feed_dict={x: x_data, y: y_data}) float_graph_def = sess.graph.as_graph_def() from neural_compressor.experimental import Quantization, common quantizer = Quantization('fake_yaml.yaml') dataset = quantizer.dataset('dummy', shape=(2, 2), label=True) quantizer.calib_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.eval_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.model = float_graph_def output_graph = quantizer.fit() for i in output_graph.graph_def.node: if (i.op == 'MatMul'): found_quantized_matmul = False break self.assertEqual(found_quantized_matmul, True) _random() def test_matmul_with_add(self): g = tf.Graph() with g.as_default(): x_data = np.array([[0.1, 0.2], [0.2, 0.3]]) y_data = np.array([[1, 2], [3, 4]], dtype=np.float32) x = tf.placeholder(tf.float32, shape=[2, 2], name='x') y = tf.constant(y_data, dtype=tf.float32, shape=[2, 2]) transpose = tf.transpose(y, perm=[1, 0]) reshape = tf.reshape(transpose, [2, 2]) z = tf.matmul(x, reshape, name='no_quant_matmul') z = tf.math.add(z, [1, 2], name='op_to_store') found_quantized_matmul = True with tf.Session() as sess: sess.run(z, feed_dict={x: x_data, y: y_data}) float_graph_def = sess.graph.as_graph_def() from neural_compressor.experimental import Quantization, common quantizer = Quantization('fake_yaml.yaml') dataset = quantizer.dataset('dummy', shape=(2, 2), label=True) quantizer.calib_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.eval_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.model = float_graph_def output_graph = quantizer.fit() for i in output_graph.graph_def.node: if (i.op == 'MatMul'): found_quantized_matmul = False break self.assertEqual(found_quantized_matmul, True) _random() def test_matmul_biasadd_requantize_dequantize_fusion_with_softmax(self): g = tf.Graph() with g.as_default(): x_data = np.array([[0.1, 0.2], [0.2, 0.3]]) y_data = np.array([[1, 2], [3, 4]], dtype=np.float32) x = tf.placeholder(tf.float32, shape=[2, 2], name='x') y = tf.constant(y_data, dtype=tf.float32, shape=[2, 2]) z = tf.matmul(x, y) biasadd = tf.nn.bias_add(z, [1, 2]) biasadd1 = tf.nn.bias_add(biasadd, [1, 1]) y1 = tf.constant(x_data, dtype=tf.float32, shape=[2, 2]) matmul1 = tf.matmul(biasadd1, y1) biasadd2 = tf.nn.bias_add(matmul1, [1, 1]) z = tf.nn.softmax(biasadd2, name='op_to_store') found_quantized_matmul = False if (tf.version.VERSION < '2.2.0'): found_quantized_matmul = False else: with tf.Session() as sess: sess.run(z, feed_dict={x: x_data, y: y_data}) float_graph_def = sess.graph.as_graph_def() from neural_compressor.experimental import Quantization, common quantizer = Quantization('fake_yaml.yaml') dataset = quantizer.dataset('dummy', shape=(2, 2), label=True) quantizer.calib_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.eval_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.model = float_graph_def output_graph = quantizer.fit() count = 0 for i in output_graph.model.as_graph_def().node: if (i.op == 'QuantizedMatMulWithBiasAndDequantize'): count += 1 found_quantized_matmul = bool((count > 1)) if (tf.__version__ < '2.6.0'): self.assertEqual(found_quantized_matmul, False) else: self.assertEqual(found_quantized_matmul, True) def test_matmul_biasadd_relu_non_const_weight(self): g = tf.Graph() with g.as_default(): x_data = np.array([[0.1, 0.2], [0.2, 0.3]]) x = tf.placeholder(tf.float32, shape=[2, 2], name='x') y = tf.matmul(x, x, name='no_quant_matmul') biasadd = tf.nn.bias_add(y, [1, 2]) z = tf.nn.relu(biasadd) found_quantized_matmul = True with tf.Session() as sess: sess.run(z, feed_dict={x: x_data}) float_graph_def = sess.graph.as_graph_def() from neural_compressor.experimental import Quantization, common quantizer = Quantization('fake_yaml.yaml') dataset = quantizer.dataset('dummy', shape=(2, 2), label=True) quantizer.calib_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.eval_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.model = float_graph_def output_graph = quantizer.fit() for i in output_graph.graph_def.node: if (i.op == 'MatMul'): found_quantized_matmul = False break self.assertEqual(found_quantized_matmul, False) def test_matmul_biasadd_non_const_weight(self): g = tf.Graph() with g.as_default(): x_data = np.array([[0.1, 0.2], [0.2, 0.3]]) x = tf.placeholder(tf.float32, shape=[2, 2], name='x') y = tf.matmul(x, x, name='no_quant_matmul') z = tf.nn.bias_add(y, [1, 2]) found_quantized_matmul = True with tf.Session() as sess: sess.run(z, feed_dict={x: x_data}) float_graph_def = sess.graph.as_graph_def() from neural_compressor.experimental import Quantization, common quantizer = Quantization('fake_yaml.yaml') dataset = quantizer.dataset('dummy', shape=(2, 2), label=True) quantizer.calib_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.eval_dataloader = common.DataLoader(dataset, batch_size=2) quantizer.model = float_graph_def output_graph = quantizer.fit() for i in output_graph.graph_def.node: if (i.op == 'MatMul'): found_quantized_matmul = False break self.assertEqual(found_quantized_matmul, False)
def reveal_fog_of_war(top_down_map: np.ndarray, current_fog_of_war_mask: np.ndarray, current_point: np.ndarray, current_angle: float, fov: float=90, max_line_len: float=100) -> np.ndarray: fov = np.deg2rad(fov) angles = np.arange(((- fov) / 2), (fov / 2), step=(1.0 / max_line_len), dtype=np.float32) fog_of_war_mask = current_fog_of_war_mask.copy() _draw_loop(top_down_map, fog_of_war_mask, current_point, current_angle, max_line_len, angles) return fog_of_war_mask
def save_episode_result(environment, test_result): res_dict = environment.get_final_result() date = environment.day idx = environment.episode_idx test_result.loc[((date + '_') + str(idx))] = [res_dict['pnl'], res_dict['nd_pnl'], res_dict['avg_abs_position'], res_dict['profit_ratio'], res_dict['volume']]
def main(): atheris.Setup(sys.argv, TestOneInput, enable_python_coverage=True) atheris.Fuzz()
def world_info_from_env(): local_rank = 0 for v in ('LOCAL_RANK', 'MPI_LOCALRANKID', 'SLURM_LOCALID', 'OMPI_COMM_WORLD_LOCAL_RANK'): if (v in os.environ): local_rank = int(os.environ[v]) break global_rank = 0 for v in ('RANK', 'PMI_RANK', 'SLURM_PROCID', 'OMPI_COMM_WORLD_RANK'): if (v in os.environ): global_rank = int(os.environ[v]) break world_size = 1 for v in ('WORLD_SIZE', 'PMI_SIZE', 'SLURM_NTASKS', 'OMPI_COMM_WORLD_SIZE'): if (v in os.environ): world_size = int(os.environ[v]) break return (local_rank, global_rank, world_size)
class SkipProofDatasetCreator(DatasetCreator): def __init__(self, fp): super().__init__(fp) self.seen = set() def process_dp(self, dp): (result, proof_term) = get_skip_proof_datapoint(dp) guard = (lambda : ('PREDICT' in result)) if guard(): if (not ((result, proof_term) in self.seen)): self.seen.add((result, proof_term)) result = {'skip_proof': result, 'proof_term': proof_term} result_msg = json.dumps(result) self.fp.write((result_msg + '\n'))
def convert(model, qconfig_mapping): for ((op_name, op_type), qconfig) in qconfig_mapping.items(): if (qconfig.weight_dtype not in FP8_DTYPE): continue module = fetch_module(model, op_name) if (module is None): logger.info(f'{op_name} is not found in model.') continue if (qconfig.approach != 'dynamic'): _remove_observer(module, qconfig) module = _replace_module(module, qconfig) set_module(model, op_name, module) htcore.mark_step() return model
def get_sumo_binary(): sumo_binary = 'sumo' if Settings.USE_GUI: if (Settings.SYSTEM == 'Windows'): sumo_binary = 'sumo-gui.exe' elif (Settings.SYSTEM == 'Linux'): sumo_binary = 'sumo-gui' elif (Settings.SYSTEM == 'Windows'): sumo_binary = 'sumo.exe' elif (Settings.SYSTEM == 'Linux'): sumo_binary = 'sumo' return sumo_binary
def insert_receptors(path_db, name, receptors, max_cdr3_length=32): labels = set() for quantities in receptors.values(): labels.update(quantities.keys()) labels = sorted(list(labels)) dtype_receptor = ([('tra_vgene', 'S16'), ('tra_cdr3', ('S' + str(max_cdr3_length))), ('tra_jgene', 'S16'), ('trb_vgene', 'S16'), ('trb_cdr3', ('S' + str(max_cdr3_length))), ('trb_jgene', 'S16')] + [(('frequency_' + label), 'f8') for label in labels]) rs = np.zeros(len(receptors), dtype=dtype_receptor) for (i, (receptor, quantities)) in enumerate(receptors.items()): (tra_vgene, tra_cdr3, tra_jgene, trb_vgene, trb_cdr3, trb_jgene) = receptor.split(':') rs[i]['tra_vgene'] = tra_vgene rs[i]['tra_cdr3'] = tra_cdr3 rs[i]['tra_jgene'] = tra_jgene rs[i]['trb_vgene'] = trb_vgene rs[i]['trb_cdr3'] = trb_cdr3 rs[i]['trb_jgene'] = trb_jgene for label in quantities.keys(): rs[i][('frequency_' + label)] = quantities[label] flag = ('r+' if os.path.isfile(path_db) else 'w') with h5py.File(path_db, flag) as db: rs_db = db.create_dataset(name, (rs.size,), dtype_receptor) rs_db[:] = rs
def read_non_scored_words(non_scored_words_file): for line in non_scored_words_file.readlines(): parts = line.split() if (not (len(parts) == 1)): raise RuntimeError('segment_ctm_edits.py: bad line in non-scored-words file {0}: {1}'.format(non_scored_words_file, line)) _global_non_scored_words.add(parts[0]) non_scored_words_file.close()
def _at_least_version(actual_version, required_version): actual = [int(v) for v in actual_version.split('.')] required = [int(v) for v in required_version.split('.')] return (actual >= required)
def conv2d_bn(x, filters, kernel_size, strides=1, padding='same', activation='relu', use_bias=False, name=None): x = Conv2D(filters, kernel_size, strides=strides, padding=padding, use_bias=use_bias, name=name)(x) if (not use_bias): bn_axis = (1 if (K.image_data_format() == 'channels_first') else 3) bn_name = (None if (name is None) else (name + '_bn')) x = BatchNormalization(axis=bn_axis, scale=False, name=bn_name)(x) if (activation is not None): ac_name = (None if (name is None) else (name + '_ac')) x = Activation(activation, name=ac_name)(x) return x
class LeakyClamp(torch.autograd.Function): def forward(ctx: Any, x: torch.Tensor, min: float, max: float) -> torch.Tensor: ctx.save_for_backward((x.ge(min) * x.le(max))) return torch.clamp(x, min=min, max=max) def backward(ctx: Any, grad_output: torch.Tensor) -> Tuple[(torch.Tensor, None, None)]: (mask,) = ctx.saved_tensors mask = mask.type_as(grad_output) return (((grad_output * mask) + ((grad_output * (1 - mask)) * eps)), None, None)
class AvgStatistic(Statistic): decay: bool = False debias: bool = False def new_step(self): (self.val, self.count) = (0.0, 0) def accumulate(self, val): self.count += 1 self.val += self._get_val1(val) def _get_val1(self, val): return val.mean() def _get_val2(self, state, val, param): return (state.add_((1 - param), val) if self.decay else state.add_(val)) def _get_val3(self, state, val, param): v = val.view(val.size(0), (- 1)).mean(1) return (state.add_((1 - param), v) if self.decay else state.add_(v)) def update(self, state, param, val=None, step=None): if (self.scope == StatScope.Weight): res = self._get_val2(state.mul_(param), val, param) elif (self.scope == StatScope.Channel): res = self._get_val3(state.mul_(param), val, param) elif (self.scope == StatScope.Layer): res = ((state * param) + (self._get_val1(val) * ((1 - param) if self.decay else 1.0))) elif (self.count != 0): res = ((state * param) + ((self.val / self.count) * ((1 - param) if self.decay else 1.0))) else: return state if (self.debias and (step is not None)): res /= (1 - (param ** step)) return res
class LinearFilter(nn.Module): def __init__(self, filter_size, filter_initializer, filter_optimizer=None, feature_extractor=None): super().__init__() self.filter_size = filter_size self.filter_initializer = filter_initializer self.filter_optimizer = filter_optimizer self.feature_extractor = feature_extractor for m in self.feature_extractor.modules(): if isinstance(m, nn.Conv2d): n = ((m.kernel_size[0] * m.kernel_size[1]) * m.out_channels) m.weight.data.normal_(0, math.sqrt((2.0 / n))) if (m.bias is not None): m.bias.data.zero_() elif isinstance(m, nn.BatchNorm2d): m.weight.data.fill_(1) m.bias.data.zero_() def forward(self, train_feat, test_feat, train_bb, *args, **kwargs): assert (train_bb.dim() == 3) num_sequences = train_bb.shape[1] if (train_feat.dim() == 5): train_feat = train_feat.view((- 1), *train_feat.shape[(- 3):]) if (test_feat.dim() == 5): test_feat = test_feat.view((- 1), *test_feat.shape[(- 3):]) train_feat = self.extract_classification_feat(train_feat, num_sequences) test_feat = self.extract_classification_feat(test_feat, num_sequences) (filter, filter_iter, losses) = self.get_filter(train_feat, train_bb, *args, **kwargs) test_scores = [self.classify(f, test_feat) for f in filter_iter] return test_scores def extract_classification_feat(self, feat, num_sequences=None): if (self.feature_extractor is None): return feat if (num_sequences is None): return self.feature_extractor(feat) output = self.feature_extractor(feat) return output.view((- 1), num_sequences, *output.shape[(- 3):]) def classify(self, weights, feat): scores = filter_layer.apply_filter(feat, weights) return scores def get_filter(self, feat, bb, *args, **kwargs): weights = self.filter_initializer(feat, bb) if (self.filter_optimizer is not None): (weights, weights_iter, losses) = self.filter_optimizer(weights, *args, feat=feat, bb=bb, **kwargs) else: weights_iter = [weights] losses = None return (weights, weights_iter, losses)
class PhotoAIAPIRouter(APIRouter): def __init__(self) -> None: super().__init__() self.chatbot = None def set_chatbot(self, chatbot, use_deepspeed, world_size, host, port) -> None: self.chatbot = chatbot self.use_deepspeed = use_deepspeed self.world_size = world_size self.host = host self.port = port def get_chatbot(self): if (self.chatbot is None): logger.error('Chatbot instance is not found.') raise RuntimeError('Chatbot instance has not been set.') return self.chatbot async def handle_voice_chat_request(self, prompt: str, audio_output_path: Optional[str]=None) -> str: chatbot = self.get_chatbot() try: plugins['tts']['enable'] = True plugins['retrieval']['enable'] = False config = GenerationConfig(audio_output_path=audio_output_path) (result, link) = chatbot.chat_stream(query=prompt, config=config) def audio_file_generate(result): for path in result: with open(path, mode='rb') as file: bytes = file.read() data = base64.b64encode(bytes) (yield f'''data: {data} ''') (yield f'''data: [DONE] ''') return StreamingResponse(audio_file_generate(result), media_type='text/event-stream') except Exception as e: raise Exception(e)
class SchedulerBaseTests(unittest.TestCase): def test_save_load_from_different_config(self): obj = SchedulerObject() setattr(diffusers, 'SchedulerObject', SchedulerObject) logger = logging.get_logger('diffusers.configuration_utils') with tempfile.TemporaryDirectory() as tmpdirname: obj.save_config(tmpdirname) with CaptureLogger(logger) as cap_logger_1: config = SchedulerObject2.load_config(tmpdirname) new_obj_1 = SchedulerObject2.from_config(config) with open(os.path.join(tmpdirname, SchedulerObject.config_name), 'r') as f: data = json.load(f) data['unexpected'] = True with open(os.path.join(tmpdirname, SchedulerObject.config_name), 'w') as f: json.dump(data, f) with CaptureLogger(logger) as cap_logger_2: config = SchedulerObject.load_config(tmpdirname) new_obj_2 = SchedulerObject.from_config(config) with CaptureLogger(logger) as cap_logger_3: config = SchedulerObject2.load_config(tmpdirname) new_obj_3 = SchedulerObject2.from_config(config) assert (new_obj_1.__class__ == SchedulerObject2) assert (new_obj_2.__class__ == SchedulerObject) assert (new_obj_3.__class__ == SchedulerObject2) assert (cap_logger_1.out == '') assert (cap_logger_2.out == "The config attributes {'unexpected': True} were passed to SchedulerObject, but are not expected and will be ignored. Please verify your config.json configuration file.\n") assert (cap_logger_2.out.replace('SchedulerObject', 'SchedulerObject2') == cap_logger_3.out) def test_save_load_compatible_schedulers(self): SchedulerObject2._compatibles = ['SchedulerObject'] SchedulerObject._compatibles = ['SchedulerObject2'] obj = SchedulerObject() setattr(diffusers, 'SchedulerObject', SchedulerObject) setattr(diffusers, 'SchedulerObject2', SchedulerObject2) logger = logging.get_logger('diffusers.configuration_utils') with tempfile.TemporaryDirectory() as tmpdirname: obj.save_config(tmpdirname) with open(os.path.join(tmpdirname, SchedulerObject.config_name), 'r') as f: data = json.load(f) data['f'] = [0, 0] data['unexpected'] = True with open(os.path.join(tmpdirname, SchedulerObject.config_name), 'w') as f: json.dump(data, f) with CaptureLogger(logger) as cap_logger: config = SchedulerObject.load_config(tmpdirname) new_obj = SchedulerObject.from_config(config) assert (new_obj.__class__ == SchedulerObject) assert (cap_logger.out == "The config attributes {'unexpected': True} were passed to SchedulerObject, but are not expected and will be ignored. Please verify your config.json configuration file.\n") def test_save_load_from_different_config_comp_schedulers(self): SchedulerObject3._compatibles = ['SchedulerObject', 'SchedulerObject2'] SchedulerObject2._compatibles = ['SchedulerObject', 'SchedulerObject3'] SchedulerObject._compatibles = ['SchedulerObject2', 'SchedulerObject3'] obj = SchedulerObject() setattr(diffusers, 'SchedulerObject', SchedulerObject) setattr(diffusers, 'SchedulerObject2', SchedulerObject2) setattr(diffusers, 'SchedulerObject3', SchedulerObject3) logger = logging.get_logger('diffusers.configuration_utils') logger.setLevel(diffusers.logging.INFO) with tempfile.TemporaryDirectory() as tmpdirname: obj.save_config(tmpdirname) with CaptureLogger(logger) as cap_logger_1: config = SchedulerObject.load_config(tmpdirname) new_obj_1 = SchedulerObject.from_config(config) with CaptureLogger(logger) as cap_logger_2: config = SchedulerObject2.load_config(tmpdirname) new_obj_2 = SchedulerObject2.from_config(config) with CaptureLogger(logger) as cap_logger_3: config = SchedulerObject3.load_config(tmpdirname) new_obj_3 = SchedulerObject3.from_config(config) assert (new_obj_1.__class__ == SchedulerObject) assert (new_obj_2.__class__ == SchedulerObject2) assert (new_obj_3.__class__ == SchedulerObject3) assert (cap_logger_1.out == '') assert (cap_logger_2.out == "{'f'} was not found in config. Values will be initialized to default values.\n") assert (cap_logger_3.out == "{'f'} was not found in config. Values will be initialized to default values.\n") def test_default_arguments_not_in_config(self): pipe = DiffusionPipeline.from_pretrained('hf-internal-testing/tiny-stable-diffusion-pipe', torch_dtype=torch.float16) assert (pipe.scheduler.__class__ == DDIMScheduler) assert (pipe.scheduler.config.timestep_spacing == 'leading') pipe.scheduler = EulerDiscreteScheduler.from_config(pipe.scheduler.config) assert (pipe.scheduler.config.timestep_spacing == 'linspace') pipe.scheduler = EulerDiscreteScheduler.from_config(pipe.scheduler.config, timestep_spacing='trailing') assert (pipe.scheduler.config.timestep_spacing == 'trailing') pipe.scheduler = LMSDiscreteScheduler.from_config(pipe.scheduler.config) assert (pipe.scheduler.config.timestep_spacing == 'trailing') pipe.scheduler = LMSDiscreteScheduler.from_config(pipe.scheduler.config) assert (pipe.scheduler.config.timestep_spacing == 'trailing') def test_default_solver_type_after_switch(self): pipe = DiffusionPipeline.from_pretrained('hf-internal-testing/tiny-stable-diffusion-pipe', torch_dtype=torch.float16) assert (pipe.scheduler.__class__ == DDIMScheduler) pipe.scheduler = DEISMultistepScheduler.from_config(pipe.scheduler.config) assert (pipe.scheduler.config.solver_type == 'logrho') pipe.scheduler = UniPCMultistepScheduler.from_config(pipe.scheduler.config) assert (pipe.scheduler.config.solver_type == 'bh2')
def load_dataset_config(cfg_path): cfg = OmegaConf.load(cfg_path).datasets cfg = cfg[list(cfg.keys())[0]] return cfg
class DummyModel(EztorchBaseModule, ABC): def __init__(self, input_shape: int, transform: Optional[DictConfig]=None) -> None: super().__init__() self.transform = (hydra.utils.instantiate(transform) if (transform is not None) else None) self.save_hyperparameters() input_dim = math.prod(input_shape) self.layer = nn.Linear(input_dim, 1) def configure_optimizers(self) -> Dict[(Any, Any)]: return torch.optim.Adam(self.parameters(), 0.0001) def forward(self, x: Tensor) -> Tensor: x = torch.flatten(x, 1) x = self.layer(x) return x def training_step(self, batch: Iterable[Any], batch_idx: int): x = batch['input'] if (self.transform is not None): with torch.no_grad(): with torch.cuda.amp.autocast(enabled=False): x = self.transform(x) if (type(x) is list): pred = self.forward(x[0]) else: pred = self.forward(x) return torch.nn.functional.binary_cross_entropy_with_logits(pred, torch.ones_like(pred))
def flavour_compair(d1, d2): diffs1 = {x: [] for x in 'tp fp fn fl None'.split()} diffs2 = {x: [] for x in 'tp fp fn fl None'.split()} for arc in d1['tp']: if (arc in d2['tp']): diffs1['tp'].append(arc) elif (arc in d2['fp']): diffs1['fp'].append(arc) elif (arc in d2['fn']): diffs1['fn'].append(arc) elif (arc in d2['fl']): diffs1['fl'].append(arc) else: diffs1['None'].append(arc) for arc in d2['tp']: if (arc in d1['tp']): diffs2['tp'].append(arc) elif (arc in d1['fp']): diffs2['fp'].append(arc) elif (arc in d1['fn']): diffs2['fn'].append(arc) elif (arc in d1['fl']): diffs2['fl'].append(arc) else: diffs2['None'].append(arc) print([(x, len(y)) for (x, y) in diffs1.items()]) print([(x, len(y)) for (x, y) in diffs2.items()]) return (diffs1, diffs2)
class NormsMethods(Generic[T_co], ExtensionMethods): l0: Callable[(..., T_co)] = norms.l0 l1: Callable[(..., T_co)] = norms.l1 l2: Callable[(..., T_co)] = norms.l2 linf: Callable[(..., T_co)] = norms.linf lp: Callable[(..., T_co)] = norms.lp
def gradients_collection(ys, xs, grad_ys=None, **kwargs): return gradients(ys, xs, grad_ys, checkpoints='collection', **kwargs)
class RPNModule(torch.nn.Module): def __init__(self, cfg, in_channels): super(RPNModule, self).__init__() self.cfg = cfg.clone() anchor_generator = make_anchor_generator(cfg) rpn_head = registry.RPN_HEADS[cfg.MODEL.RPN.RPN_HEAD] head = rpn_head(cfg, in_channels, anchor_generator.num_anchors_per_location()[0]) rpn_box_coder = BoxCoder(weights=(1.0, 1.0, 1.0, 1.0)) self.rpn_box_coder = rpn_box_coder box_selector_train = make_rpn_postprocessor(cfg, rpn_box_coder, is_train=True) box_selector_test = make_rpn_postprocessor(cfg, rpn_box_coder, is_train=False) loss_evaluator = make_rpn_loss_evaluator(cfg, rpn_box_coder) self.anchor_generator = anchor_generator self.head = head self.box_selector_train = box_selector_train self.box_selector_test = box_selector_test self.loss_evaluator = loss_evaluator resolution = cfg.MODEL.ROI_BOX_HEAD.POOLER_RESOLUTION scales = cfg.MODEL.ROI_BOX_HEAD.POOLER_SCALES sampling_ratio = cfg.MODEL.ROI_BOX_HEAD.POOLER_SAMPLING_RATIO pooler = Pooler(output_size=(resolution, resolution), scales=scales, sampling_ratio=sampling_ratio) self.pooler = pooler def forward(self, images, features, atten_logits, atten_map, targets=None): image_size = (images.tensors.shape[(- 2)], images.tensors.shape[(- 1)]) result = [] anchors = self.anchor_generator(images, features, atten_logits) for (anchor, atten_logit) in zip(anchors, atten_logits.mean(1)): boxes = anchor[0].bbox boxlist = BoxList(boxes, features[0].shape[(- 2):], mode='xyxy') boxlist = boxlist.clip_to_image(remove_empty=False) boxlist = boxlist.resize(anchor[0].size) boxlist = boxlist.clip_to_image(remove_empty=False) boxlist = remove_small_boxes(boxlist, 20) objectness = self.pooler([atten_logit.unsqueeze(0).unsqueeze(1)], [boxlist]).mean(3).mean(2).squeeze().sigmoid() boxlist.add_field('objectness', objectness) boxlist = boxlist_nms(boxlist, 0.7, max_proposals=2000, score_field='objectness') result.append(boxlist) return (result, {}) if self.training: return self._forward_train(anchors, objectness, rpn_box_regression, targets) else: return self._forward_test(anchors, objectness, rpn_box_regression) def _forward_train(self, anchors, objectness, rpn_box_regression, targets): if self.cfg.MODEL.RPN_ONLY: boxes = anchors else: with torch.no_grad(): boxes = self.box_selector_train(anchors, objectness, rpn_box_regression, targets) losses = {} return (boxes, losses) def _forward_test(self, anchors, objectness, rpn_box_regression): boxes = self.box_selector_test(anchors, objectness, rpn_box_regression) if self.cfg.MODEL.RPN_ONLY: inds = [box.get_field('objectness').sort(descending=True)[1] for box in boxes] boxes = [box[ind] for (box, ind) in zip(boxes, inds)] return (boxes, {})
def main(): args = parser.parse_args() print(args) if (args.seed is not None): random.seed(args.seed) torch.manual_seed(args.seed) cudnn.deterministic = True warnings.warn('You have chosen to seed training. This will turn on the CUDNN deterministic setting, which can slow down your training considerably! You may see unexpected behavior when restarting from checkpoints.') if (args.gpu is not None): warnings.warn('You have chosen a specific GPU. This will completely disable data parallelism.') if ((args.dist_url == 'env://') and (args.world_size == (- 1))): args.world_size = int(os.environ['WORLD_SIZE']) args.distributed = ((args.world_size > 1) or args.multiprocessing_distributed) ngpus_per_node = torch.cuda.device_count() if args.multiprocessing_distributed: args.world_size = (ngpus_per_node * args.world_size) mp.spawn(main_worker, nprocs=ngpus_per_node, args=(ngpus_per_node, args)) else: main_worker(args.gpu, ngpus_per_node, args)
class TanhTransformedDistribution(tfd.TransformedDistribution): def __init__(self, distribution: tfd.Distribution, validate_args: bool=False): super().__init__(distribution=distribution, bijector=tfb.Tanh(), validate_args=validate_args) def mode(self) -> jnp.ndarray: return self.bijector.forward(self.distribution.mode()) def _parameter_properties(cls, dtype: Optional[Any], num_classes=None): td_properties = super()._parameter_properties(dtype, num_classes=num_classes) del td_properties['bijector'] return td_properties
class Cell(nn.Module): def __init__(self, genotype, C_prev_prev, C_prev, C, reduction, reduction_prev): super(Cell, self).__init__() print(C_prev_prev, C_prev, C) if reduction_prev: self.preprocess0 = FactorizedReduce(C_prev_prev, C) else: self.preprocess0 = ReLUConvBN(C_prev_prev, C, (1, 1), (1, 1), (0, 0)) self.preprocess1 = ReLUConvBN(C_prev, C, (1, 1), (1, 1), (0, 0)) if reduction: (op_names, indices) = zip(*genotype.reduce) concat = genotype.reduce_concat else: (op_names, indices) = zip(*genotype.normal) concat = genotype.normal_concat self._compile(C, op_names, indices, concat, reduction) def _compile(self, C, op_names, indices, concat, reduction): assert (len(op_names) == len(indices)) self._steps = (len(op_names) // 2) self._concat = concat self.multiplier = len(concat) self._ops = nn.ModuleList() for (name, index) in zip(op_names, indices): stride = (2 if (reduction and (index < 2)) else 1) op = OPS[name](C, stride, True) self._ops += [op] self._indices = indices def forward(self, s0, s1, drop_prob): s0 = self.preprocess0(s0) s1 = self.preprocess1(s1) states = [s0, s1] for i in range(self._steps): h1 = states[self._indices[(2 * i)]] h2 = states[self._indices[((2 * i) + 1)]] op1 = self._ops[(2 * i)] op2 = self._ops[((2 * i) + 1)] h1 = op1(h1) h2 = op2(h2) if (self.training and (drop_prob > 0.0)): if (not isinstance(op1, Identity)): h1 = drop_path(h1, drop_prob) if (not isinstance(op2, Identity)): h2 = drop_path(h2, drop_prob) s = (h1 + h2) states += [s] return torch.cat([states[i] for i in self._concat], dim=1)
class TimeLimit(EnvWrapper): def __init__(self, env, duration): super().__init__(env) self._duration = duration self._step = None def step(self, action): assert (self._step is not None), 'Must reset environment.' (obs, reward, done, info) = self.env.step(action) self._step += 1 if (self._step >= self._duration): if isinstance(info, EnvInfo): info = EnvInfo(info.discount, info.game_score, True) self._step = None return EnvStep(obs, reward, done, info) def reset(self): self._step = 0 return self.env.reset()
def start_server(args): scorer = Scorer(args) app = web.Application([('/result', ResultHandler, dict(scorer=scorer)), ('/src', SourceHandler, dict(scorer=scorer)), ('/hypo', HypothesisHandler, dict(scorer=scorer)), ('/', EvalSessionHandler, dict(scorer=scorer))], debug=False) app.listen(args.port, max_buffer_size=(1024 ** 3)) logger.info(f'Evaluation Server Started (process id {os.getpid()}). Listening to port {args.port} ') ioloop.IOLoop.current().start()
class DatasetSplit(Dataset): def __init__(self, dataset, idxs, Y=None): self.dataset = dataset self.idxs = [int(i) for i in idxs] self.mal = False if (Y is not None): self.mal = True self.mal_Y = Y def __len__(self): return len(self.idxs) def __getitem__(self, item): (image, label) = self.dataset[self.idxs[item]] if (self.mal == True): label_mal = self.mal_Y[item] return (torch.tensor(image), torch.tensor(label_mal), torch.tensor(label)) return (torch.tensor(image), torch.tensor(label))
def setup(rank, world_size, port): os.environ['MASTER_ADDR'] = 'localhost' os.environ['MASTER_PORT'] = str(port) dist.init_process_group('nccl', rank=rank, world_size=world_size, init_method='env://')
class Txt2ImgIterableBaseDataset(IterableDataset): def __init__(self, num_records=0, valid_ids=None, size=256): super().__init__() self.num_records = num_records self.valid_ids = valid_ids self.sample_ids = valid_ids self.size = size print(f'{self.__class__.__name__} dataset contains {self.__len__()} examples.') def __len__(self): return self.num_records def __iter__(self): pass
def update(config, args): config['model_dir'] = get_value(config['model_dir'], args.model_dir) config['training_opt']['batch_size'] = get_value(config['training_opt']['batch_size'], args.batch_size) return config
def get_metric(y_true_aspect, y_predict_aspect, y_true_opinion, y_predict_opinion, y_true_sentiment, y_predict_sentiment, mask, train_op): (f_a, f_o) = (0, 0) (true_aspect, true_sentiment) = convert_to_list(y_true_aspect, y_true_sentiment, mask) (predict_aspect, predict_sentiment) = convert_to_list(y_predict_aspect, y_predict_sentiment, mask) (true_opinion, _) = convert_to_list(y_true_opinion, y_true_sentiment, mask) (predict_opinion, _) = convert_to_list(y_predict_opinion, y_predict_sentiment, mask) (f_aspect, acc_s, f_s, f_absa) = score(true_aspect, predict_aspect, true_sentiment, predict_sentiment, 0) if train_op: (f_opinion, _, _, _) = score(true_opinion, predict_opinion, true_sentiment, predict_sentiment, 1) return (f_aspect, f_opinion, acc_s, f_s, f_absa)
class KIEDecoderTRIE(nn.Module): def __init__(self, ocr_dict, entity_dict, use_crf=False, ins_lvl_mean=False, d_model=(- 1), lstm_args=None): super(KIEDecoderTRIE, self).__init__() self.debug_mode = False self.ins_lvl_mean = ins_lvl_mean self.ocr_dict = ocr_dict self.rev_ocr_dict = dict() for (key, val) in ocr_dict.items(): self.rev_ocr_dict[val] = key self.entity_dict = entity_dict bilstm_kwargs = lstm_args.get('bilstm_kwargs', dict()) mlp_kwargs = lstm_args.get('mlp_kwargs', dict()) bilstm_kwargs.update(input_size=(2 * d_model)) if bilstm_kwargs['bidirectional']: mlp_kwargs.update(in_dim=(2 * bilstm_kwargs['hidden_size'])) else: mlp_kwargs.update(in_dim=bilstm_kwargs['hidden_size']) mlp_kwargs.update(out_dim=len(self.entity_dict)) self.bilstm_layer = BiLSTMLayer(bilstm_kwargs, mlp_kwargs, pad_val=self.ocr_dict['<PAD>'], apply_norm=lstm_args.get('apply_norm', False)) self.union_layer = UnionLayer(debug_mode=self.debug_mode) self.soft_max_func = nn.Softmax(dim=(- 1)) self.use_crf = use_crf if use_crf: self.crf = CRF(len(entity_dict), batch_first=True) else: self.criterion = nn.CrossEntropyLoss(ignore_index=self.entity_dict['<PAD>'], reduction='mean') def forward(self, ocr_logits, multi_modal_context, texts=None, tags=None, logits_logger=None, sorted_idx=None): if self.training: (B, N, L) = texts.shape C = ocr_logits.shape[(- 1)] multi_modal_context = multi_modal_context.reshape((B * N), C).unsqueeze(1).expand((B * N), L, C) if (not self.debug_mode): (new_kie_logits, new_mask, new_tags, mask) = self.union_layer(torch.cat((ocr_logits, multi_modal_context), dim=(- 1)).reshape(B, N, L, (- 1)), texts, self.ocr_dict['<END>'], tags_=tags, tag_pad=self.entity_dict['<PAD>'], sorted_idx=sorted_idx) else: raise NotImplementedError (new_kie_logits, new_mask, new_tags, mask, doc_texts) = self.union_layer(kie_logits.reshape(B, N, L, (- 1)), texts, self.ocr_dict['<END>'], tags_=tags, tag_pad=self.entity_dict['<PAD>'], sorted_idx=sorted_idx) docs = [] for i in range(B): cur_doc = '' for j in doc_texts[i]: cur_idx = j.item() if (cur_idx == self.ocr_dict['<END>']): break cur_doc += self.rev_ocr_dict[cur_idx] docs.append(cur_doc) new_kie_logits = self.bilstm_layer(new_kie_logits, new_mask.sum(dim=(- 1)), (None, None)) if self.use_crf: if self.ins_lvl_mean: log_likelihood = self.crf(new_kie_logits, new_tags, mask=new_mask) if sorted_idx: total_num_box = 0 for idx_set in sorted_idx: total_num_box += len(idx_set) log_likelihood /= total_num_box else: raise ValueError(f'sorted_idx is required for calculating the num of boxes') else: log_likelihood = self.crf(new_kie_logits, new_tags, mask=new_mask) log_likelihood /= new_mask.sum() if ('CRF' not in logits_logger): logits_logger['CRF'] = [(- log_likelihood.reshape(1))] else: logits_logger['CRF'].append((- log_likelihood.reshape(1))) elif ('CRF' not in logits_logger): logits_logger['CRF'] = [self.criterion(new_kie_logits.reshape((- 1), new_kie_logits.shape[(- 1)]), new_tags.reshape((- 1)))] else: logits_logger['CRF'].append(self.criterion(new_kie_logits.reshape((- 1), new_kie_logits.shape[(- 1)]), new_tags.reshape((- 1)))) else: logits_logger['ocr_logits'] = ocr_logits logits_logger['multi_modal_context'] = multi_modal_context def crf_decode(self, logits_logger, shape_, sorted_idx, gt_texts=None): if self.training: raise RuntimeError('crf_decode should be inference only') (B, N) = shape_ (L, C) = logits_logger['ocr_logits'].shape[1:] multi_modal_context = logits_logger['multi_modal_context'].reshape((B * N), C).unsqueeze(1).expand((B * N), L, C) ocr_logits = logits_logger['ocr_logits'] if (not isinstance(gt_texts, type(None))): pred_texts = gt_texts else: (_, pred_texts) = torch.max(self.soft_max_func(logits_logger['REC'][(- 1)]), dim=(- 1)) (new_kie_logits, new_mask, mask) = self.union_layer(torch.cat((ocr_logits, multi_modal_context), dim=(- 1)).reshape(B, N, L, (- 1)), pred_texts.reshape(B, N, (- 1)), self.ocr_dict['<END>'], sorted_idx=sorted_idx) new_kie_logits = self.bilstm_layer(new_kie_logits, new_mask.sum(dim=(- 1)), (None, None)) if self.use_crf: best_paths = self.crf.decode(new_kie_logits, new_mask) ins_lvl_tags = self.union_layer.split_out(best_paths, mask, sorted_idx=sorted_idx) if ('CRF' not in logits_logger): logits_logger['CRF'] = [ins_lvl_tags] else: logits_logger['CRF'].append(ins_lvl_tags) else: ins_lvl_logits = self.union_layer.split_logits_out(new_kie_logits, mask, sorted_idx=sorted_idx) if ('CRF' not in logits_logger): logits_logger['CRF'] = [ins_lvl_logits] else: logits_logger['CRF'].append(ins_lvl_logits)
class SharedMLP(nn.Sequential): def __init__(self, args: List[int], *, bn: bool=False, activation=nn.ReLU(inplace=True), preact: bool=False, first: bool=False, name: str=''): super().__init__() for i in range((len(args) - 1)): self.add_module((name + 'layer{}'.format(i)), Conv2d(args[i], args[(i + 1)], bn=(((not first) or (not preact) or (i != 0)) and bn), activation=(activation if ((not first) or (not preact) or (i != 0)) else None), preact=preact))
def parse(opt_path, root_path, is_train=True, debug=False): with open(opt_path, mode='r') as f: (Loader, _) = ordered_yaml() opt = yaml.load(f, Loader=Loader) if (debug and (not opt['name'].startswith('debug'))): opt['name'] = ('debug_' + opt['name']) opt['is_train'] = is_train if (opt['num_gpu'] == 'auto'): opt['num_gpu'] = torch.cuda.device_count() for (phase, dataset) in opt['datasets'].items(): phase = phase.split('_')[0] dataset['phase'] = phase if ('scale' in opt): dataset['scale'] = opt['scale'] if (dataset.get('dataroot_gt') is not None): dataset['dataroot_gt'] = osp.expanduser(dataset['dataroot_gt']) if (dataset.get('dataroot_lq') is not None): dataset['dataroot_lq'] = osp.expanduser(dataset['dataroot_lq']) if (opt.get('folder_suffix') == None): opt['folder_suffix'] = '' for (key, val) in opt['path'].items(): if ((val is not None) and (('resume_state' in key) or ('pretrain_network' in key))): opt['path'][key] = osp.expanduser(val) if is_train: experiments_root = osp.join(root_path, 'experiments', opt['folder_suffix'], opt['name']) opt['path']['experiments_root'] = experiments_root opt['path']['models'] = osp.join(experiments_root, 'models') opt['path']['training_states'] = osp.join(experiments_root, 'training_states') opt['path']['log'] = experiments_root opt['path']['visualization'] = osp.join(experiments_root, 'visualization') if ('debug' in opt['name']): if ('val' in opt): opt['val']['val_freq'] = 8 opt['logger']['print_freq'] = 1 opt['logger']['save_checkpoint_freq'] = 8 else: results_root = osp.join(root_path, 'results', opt['folder_suffix'], opt['name']) opt['path']['results_root'] = results_root opt['path']['log'] = results_root opt['path']['visualization'] = osp.join(results_root, 'visualization') return opt
class Model(nn.Module): def __init__(self, nclasses): super().__init__() self.features = [6, 100, 100, nclasses] self.bandwidths = [64, 16, 10] assert (len(self.bandwidths) == (len(self.features) - 1)) sequence = [] grid = s2_equatorial_grid(max_beta=0, n_alpha=(2 * self.bandwidths[0]), n_beta=1) sequence.append(S2Convolution(self.features[0], self.features[1], self.bandwidths[0], self.bandwidths[1], grid)) for l in range(1, (len(self.features) - 2)): nfeature_in = self.features[l] nfeature_out = self.features[(l + 1)] b_in = self.bandwidths[l] b_out = self.bandwidths[(l + 1)] sequence.append(nn.BatchNorm3d(nfeature_in, affine=True)) sequence.append(nn.ReLU()) grid = so3_equatorial_grid(max_beta=0, max_gamma=0, n_alpha=(2 * b_in), n_beta=1, n_gamma=1) sequence.append(SO3Convolution(nfeature_in, nfeature_out, b_in, b_out, grid)) sequence.append(nn.BatchNorm3d(self.features[(- 2)], affine=True)) sequence.append(nn.ReLU()) self.sequential = nn.Sequential(*sequence) output_features = self.features[(- 2)] self.out_layer = nn.Linear(output_features, self.features[(- 1)]) def forward(self, x): x = self.sequential(x) x = so3_integrate(x) x = self.out_layer(x) return F.log_softmax(x, dim=1)
def test_fs_observer_resource_event(dir_obs, sample_run, tmpfile): (basedir, obs) = dir_obs _id = obs.started_event(**sample_run) run_dir = basedir.join(_id) obs.resource_event(tmpfile.name) res_dir = basedir.join('_resources') assert res_dir.exists() assert (len(res_dir.listdir()) == 1) assert (res_dir.listdir()[0].read() == tmpfile.content) run = json.loads(run_dir.join('run.json').read()) assert (len(run['resources']) == 1) assert (run['resources'][0] == [tmpfile.name, res_dir.listdir()[0].strpath])
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): 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() 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 create_oracles_with_ilp(dataname, path_read, path_wt_distributed): process_one_example_with_ilp(path_read, path_wt_distributed, '6b43f5e79b3cdf1c4a6debad958d5d70358f4269', 30, data_name) process_one_example_with_ilp(path_read, path_wt_distributed, 'e1dc607107cc484c4bc1515cfa414a45088d4048', 30, data_name) process_one_example_with_ilp(path_read, path_wt_distributed, 'b3f9fa06492bed4394ca521100b10e87431d0ddb', 30, data_name) process_one_example_with_ilp(path_read, path_wt_distributed, '8e03f1cc2061dbfb947d4e790395bddbf070db53', 30, data_name)
def data_reader(): data_root = 'data/Casia_maxpy_clean' images_data = [] images_label = [] batch_size = 128 count_value = 0 class_file_list = os.listdir(data_root) for i in tqdm.tqdm(range(len(class_file_list)), ncols=80): images_path = os.listdir(os.path.join(data_root, class_file_list[i])) for image_path in images_path: images_data.append(vision.image_load(((((data_root + '/') + class_file_list[i]) + '/') + image_path), backend='cv2')) images_label.append(i) count_value += 1 print(count_value) assert (len(images_data) == len(images_label)) def reader(): for i in range((int((len(images_data) / batch_size)) - 1)): data = images_data[(i * batch_size):((i + 1) * batch_size)] data = np.array(data) data = np.transpose(data, (0, 3, 1, 2)) data = ((data - 127.5) / 127.5) (yield data) return reader
class LoadGinConfigOperator(bpy.types.Operator): bl_idname = 'scene.zpy_load_gin_config' bl_label = 'Load gin config from file.' bl_description = 'Load gin config from file.' bl_category = 'ZPY' bl_options = {'REGISTER'} DEFAULT_TEXT_NAME = 'config' def execute(self, context): zpy.blender.load_text_from_file(bpy.path.abspath(context.scene.zpy_gin_config_path), text_name=self.DEFAULT_TEXT_NAME) return {'FINISHED'}
def complete_device(device): if (not torch.cuda.is_available()): return torch.device('cpu') if (type(device) == str): device = torch.device(device) if ((device.type == 'cuda') and (device.index is None)): return torch.device(device.type, torch.cuda.current_device()) return device
class EvolutionStrategyEmitter(EmitterBase): def __init__(self, archive, *, x0, sigma0, ranker='2imp', es='cma_es', es_kwargs=None, selection_rule='filter', restart_rule='no_improvement', bounds=None, batch_size=None, seed=None): EmitterBase.__init__(self, archive, solution_dim=archive.solution_dim, bounds=bounds) seed_sequence = (seed if isinstance(seed, np.random.SeedSequence) else np.random.SeedSequence(seed)) (opt_seed, ranker_seed) = seed_sequence.spawn(2) self._x0 = np.array(x0, dtype=archive.dtype) check_shape(self._x0, 'x0', archive.solution_dim, 'archive.solution_dim') self._sigma0 = sigma0 if (selection_rule not in ['mu', 'filter']): raise ValueError(f'Invalid selection_rule {selection_rule}') self._selection_rule = selection_rule self._restart_rule = restart_rule self._restarts = 0 self._itrs = 0 _ = self._check_restart(0) self._opt = _get_es(es, sigma0=sigma0, batch_size=batch_size, solution_dim=self._solution_dim, seed=opt_seed, dtype=self.archive.dtype, lower_bounds=self.lower_bounds, upper_bounds=self.upper_bounds, **(es_kwargs if (es_kwargs is not None) else {})) self._opt.reset(self._x0) self._ranker = _get_ranker(ranker, ranker_seed) self._ranker.reset(self, archive) self._batch_size = self._opt.batch_size def x0(self): return self._x0 def batch_size(self): return self._batch_size def restarts(self): return self._restarts def itrs(self): return self._itrs def ask(self): return self._opt.ask() def _check_restart(self, num_parents): if isinstance(self._restart_rule, numbers.Integral): return ((self._itrs % self._restart_rule) == 0) if (self._restart_rule == 'no_improvement'): return (num_parents == 0) if (self._restart_rule == 'basic'): return False raise ValueError(f'Invalid restart_rule {self._restart_rule}') def tell(self, solution, objective, measures, add_info, **fields): (data, add_info) = validate_batch(self.archive, {'solution': solution, 'objective': objective, 'measures': measures, **fields}, add_info) self._itrs += 1 new_sols = add_info['status'].astype(bool).sum() (indices, ranking_values) = self._ranker.rank(self, self.archive, data, add_info) num_parents = (new_sols if (self._selection_rule == 'filter') else (self._batch_size // 2)) self._opt.tell(indices, ranking_values, num_parents) if (self._opt.check_stop(ranking_values[indices]) or self._check_restart(new_sols)): new_x0 = self.archive.sample_elites(1)['solution'][0] self._opt.reset(new_x0) self._ranker.reset(self, self.archive) self._restarts += 1
def integer_mixed_cell(dim, nbr, idx, verbose=True): from ast import literal_eval from phcpy.phcpy2c3 import py2c_intcelcon_get_inner_normal as getnormal from phcpy.phcpy2c3 import py2c_intcelcon_mixed_volume as mixvol from phcpy.phcpy2c3 import py2c_intcelcon_number_of_points_in_cell as npts from phcpy.phcpy2c3 import py2c_intcelcon_get_point_in_cell as getpoint normal = literal_eval(getnormal(dim, idx)) mv = mixvol(idx) lenpts = literal_eval(npts(idx, nbr)) if verbose: print('inner normal :', normal) print('mixed volume :', mv) print('number of points :', lenpts) supp = [[] for _ in range(nbr)] for i in range(nbr): if verbose: print('points in support', (i + 1), ':') for j in range(lenpts[i]): point = getpoint(dim, idx, (i + 1), (j + 1)) if verbose: print(eval(point)) supp[i].append(eval(point)) return (normal, mv, tuple(supp))
class MasterServicer(elastic_training_pb2_grpc.MasterServicer): def __init__(self, task_manager, job_manager, speed_monitor: SpeedMonitor, rdzv_managers: Dict[(str, RendezvousManager)], job_metric_collector=None, elastic_ps_service=None, sync_service=None): self._task_manager: TaskManager = task_manager self._job_manager: JobManager = job_manager self._speed_monitor = speed_monitor self._rdzv_managers = rdzv_managers self._kv_store = KVStoreService() self._job_metric_collector: JobMetricCollector = job_metric_collector self._elastic_ps_service: ElasticPsService = elastic_ps_service self._sync_service: SyncService = sync_service self._lock = threading.Lock() self._version = 0 self._start_training_time = 0 self._start_autoscale = False def get(self, request, _): node_type = request.node_type node_id = request.node_id req_message = grpc.deserialize_message(request.data) response = elastic_training_pb2.Message() if (not req_message): return response message = None if isinstance(req_message, grpc.TaskRequest): message = self._get_task(node_type, node_id, req_message) elif isinstance(req_message, grpc.ShardCheckpointRequest): message = self._get_shard_checkpoint(req_message) elif isinstance(req_message, grpc.ClusterVersionRequest): message = self._get_cluster_version(req_message) elif isinstance(req_message, grpc.RunningNodesRequest): message = self._get_running_nodes() elif isinstance(req_message, grpc.JoinRendezvousRequest): message = self._join_rendezvous(req_message) elif isinstance(req_message, grpc.WaitingNodeNumRequest): message = self._num_nodes_waiting(req_message.rdzv_name) elif isinstance(req_message, grpc.NetworkReadyRequest): message = self._check_fault_node() elif isinstance(req_message, grpc.StragglerExistRequest): message = self._check_straggler() elif isinstance(req_message, grpc.JoinRendezvousRequest): message = self._join_rendezvous(req_message) elif isinstance(req_message, grpc.CommWorldRequest): message = self._get_comm_world(req_message) elif isinstance(req_message, grpc.KeyValuePair): message = self._kv_store_get(req_message) elif isinstance(req_message, grpc.PsNodesRequest): message = self._query_ps_nodes() elif isinstance(req_message, grpc.TrainingStatusRequest): message = self._get_training_status() elif isinstance(req_message, grpc.ParallelConfigRequest): message = self._get_paral_config() elif isinstance(req_message, grpc.CheckHardwareResetRequest): message = self._need_to_restart_training(node_type, node_id) if message: response.data = message.serialize() return response def _get_task(self, node_type, node_id, request: grpc.TaskRequest): if (not self._start_training_time): self._start_training_time = int(time.time()) shard = grpc.Shard() res = grpc.Task(shard=shard) ds_name = request.dataset_name dataset = self._task_manager.get_dataset(ds_name) if (not dataset): return res task = self._task_manager.get_dataset_task(node_type, node_id, ds_name) if task: res.task_id = task.task_id res.type = task.task_type res.shard.name = task.shard.name res.shard.start = task.shard.start res.shard.end = task.shard.end if task.shard.record_indices: res.shard.indices = task.shard.record_indices elif (not dataset.completed()): res.type = elastic_training_pb2.WAIT with self._lock: self._task_manager.reset_worker_start_task_time(node_id) return res def _get_shard_checkpoint(self, request: grpc.ShardCheckpointRequest): response = grpc.ShardCheckpoint() dataset = self._task_manager.get_dataset(request.dataset_name) checkpoint = dataset.checkpoint() if checkpoint: response.content = checkpoint.to_json() return response def _get_cluster_version(self, request: grpc.ClusterVersionRequest): message = grpc.ClusterVersion() if (not self._elastic_ps_service): return message if (request.task_type == NodeType.WORKER): message.version = self._elastic_ps_service.get_worker_version(request.version_type, request.task_id) elif (request.task_type == NodeType.PS): message.version = self._elastic_ps_service.get_ps_version(request.version_type, request.task_id) return message def _query_ps_nodes(self): res = grpc.PsNodes(nodes=[]) training_ps: List[Node] = self._job_manager.get_next_cluster_ps() ready = self._job_manager.ready_for_new_ps_cluster() ps_failure = self._job_manager.has_ps_failure() for ps in training_ps: ps_meta = grpc.NodeMeta() ps_meta.type = NodeType.PS ps_meta.addr = ps.service_addr ps_meta.cpu = ps.config_resource.cpu ps_meta.memory = int(ps.config_resource.memory) res.nodes.append(ps_meta) logger.info('PS nodes : %s', res) res.new_ps_ready = ready res.ps_failure = ps_failure return res def _get_running_nodes(self): res = grpc.RunningNodes(nodes=[]) nodes: List[Node] = self._job_manager.get_running_nodes() for node in nodes: meta = grpc.NodeMeta() meta.type = node.type meta.addr = node.service_addr meta.cpu = node.config_resource.cpu meta.memory = node.config_resource.memory if node.config_resource.gpu_type: meta.gpu_type = node.config_resource.gpu_type meta.gpu = node.config_resource.gpu_num res.nodes.append(meta) return res def _get_training_status(self): res = grpc.TrainingStatus() if self._task_manager.training_started(): res.status = TrainingLoopStatus.START else: res.status = TrainingLoopStatus.PENDING return res def _check_fault_node(self): rdzv_manager: NetworkCheckRendezvousManager = self._rdzv_managers[RendezvousName.NETWORK_CHECK] (nodes, reason) = rdzv_manager.check_fault_node() res = grpc.NetworkCheckResult(nodes=nodes, reason=reason) return res def _check_straggler(self): rdzv_manager: NetworkCheckRendezvousManager = self._rdzv_managers[RendezvousName.NETWORK_CHECK] (nodes, reason) = rdzv_manager.get_straggler() res = grpc.NetworkCheckResult(nodes=nodes, reason=reason) return res def _join_rendezvous(self, request: grpc.JoinRendezvousRequest): rdzv_manager = self._rdzv_managers[request.rdzv_name] round = rdzv_manager.join_rendezvous(request.node_id, request.local_world_size) if (request.rdzv_name == RendezvousName.NETWORK_CHECK): training_manager = self._rdzv_managers[RendezvousName.ELASTIC_TRAINING] training_manager.clear_waiting_nodes() res = grpc.RendezvousState(round=round) return res def _num_nodes_waiting(self, rdzv_name): waiting_num = self._rdzv_managers[rdzv_name].num_nodes_waiting() res = grpc.RendezvousState(waiting_num=waiting_num) return res def _get_comm_world(self, request: grpc.CommWorldRequest): rdzv_manager = self._rdzv_managers[request.rdzv_name] (rdzv_round, group, nodes) = rdzv_manager.get_comm_world(request.node_id) res = grpc.RendezvousState(world={}) res.group = group res.round = rdzv_round for (rank_id, worker_num) in nodes.items(): res.world[rank_id] = worker_num return res def _kv_store_get(self, request: grpc.KeyValuePair): value = self._kv_store.get(request.key) res = grpc.KeyValuePair(request.key, value) return res def _get_paral_config(self): res = self._job_manager.get_opt_strategy() if (not res): res = grpc.ParallelConfig() return res def _need_to_restart_training(self, node_type, node_id): restart = self._job_manager.verify_restarting_worker_training(node_type, node_id) res = grpc.ParallelConfig() res.restart = restart return res def report(self, request, _): node_type = request.node_type node_id = request.node_id message = grpc.deserialize_message(request.data) response = elastic_training_pb2.Response() if (not message): return response success = False if isinstance(message, grpc.DatasetShardParams): success = self._collect_dataset_shard_params(message) elif isinstance(message, grpc.ResourceStats): success = self._update_node_resource_usage(node_type, node_id, message) elif isinstance(message, grpc.ModelInfo): success = self._collect_model_info(message) elif isinstance(message, grpc.GlobalStep): success = self._collect_global_step(message) elif isinstance(message, grpc.ShardCheckpoint): success = self._restore_shard_checkpoint(message) elif isinstance(message, grpc.TaskResult): success = self._report_task_result(message) elif isinstance(message, grpc.ClusterVersion): success = self._update_cluster_version(message) elif isinstance(message, grpc.NodeAddress): success = self._update_node_address(message) elif isinstance(message, grpc.NetworkStatus): success = self._update_node_status(message) elif isinstance(message, grpc.NodeEvent): success = self._update_node_event(message) elif isinstance(message, grpc.SyncJoin): success = self._join_sync(node_type, node_id, message) elif isinstance(message, grpc.SyncFinish): success = self._sync_finished(message) elif isinstance(message, grpc.SyncBarrier): success = self._barrier(message) elif isinstance(message, grpc.NodeFailure): success = self._report_failure(node_type, node_id, message) elif isinstance(message, grpc.RendezvousParams): success = self._report_rdzv_params(message) elif isinstance(message, grpc.PsReady): success = self._ready_for_ps_relaunch() elif isinstance(message, grpc.KeyValuePair): success = self._kv_store_set(message) elif isinstance(message, grpc.ParallelConfig): success = self._report_paral_config(node_type, node_id, message) response.success = success return response def _ready_for_ps_relaunch(self): self._job_manager.post_ps_ready() return True def _collect_dataset_shard_params(self, metrics: grpc.DatasetShardParams): num_minibatches_per_task = (metrics.num_minibatches_per_shard or _DEFAULT_NUM_MINIBATCHES_PER_SHARD) shard_size = (metrics.batch_size * num_minibatches_per_task) splitter = new_dataset_splitter(metrics.shuffle, shard_size, metrics.dataset_size, metrics.num_epochs, metrics.dataset_name, metrics.storage_type) self._task_manager.new_dataset(metrics.batch_size, metrics.dataset_size, metrics.dataset_name, splitter, metrics.task_type) if self._job_metric_collector: self._job_metric_collector.collect_dataset_metric(metrics.dataset_name, metrics.dataset_size, metrics.storage_type) if (metrics.task_type == elastic_training_pb2.TRAINING): self._job_metric_collector.collect_training_hyper_params(metrics.num_epochs, metrics.batch_size) return True def _update_node_resource_usage(self, node_type, node_id, metrics: grpc.ResourceStats): logger.debug(f'Update resource usage for {node_type}-{node_id},cpu={metrics.cpu}, memory={metrics.memory},gpu_stats={metrics.gpu_stats}') if self._job_manager: self._job_manager.update_node_resource_usage(node_type, node_id, metrics.cpu, metrics.memory, metrics.gpu_stats) return True def _collect_model_info(self, metrics: grpc.ModelInfo): if self._job_metric_collector: self._job_metric_collector.collect_model_metric(metrics) return True def _collect_global_step(self, metrics: grpc.GlobalStep): self._speed_monitor.collect_global_step(metrics.step, metrics.timestamp) self._collect_runtime_stats() self._check_start_auto_scale_worker() return True def _restore_shard_checkpoint(self, message: grpc.ShardCheckpoint): success = self._task_manager.restore_dataset_from_checkpoint(message.content) return success def _collect_runtime_stats(self): if (self._job_metric_collector and self._job_manager): nodes = self._job_manager.get_running_nodes() self._job_metric_collector.collect_runtime_stats(self._speed_monitor, nodes) def _report_task_result(self, request: grpc.TaskResult): success = True if request.err_message: logger.warning(('Worker reported error: ' + request.err_message)) success = False (task, _) = self._task_manager.report_dataset_task(request, success) if ((not self._start_autoscale) and self._job_manager and (self._speed_monitor.completed_global_step == 0) and ((int(time.time()) - self._start_training_time) > _dlrover_context.seconds_to_autoscale_worker)): logger.info('Start autoscale for non-training jobs') self._job_manager.start_auto_scaling() self._start_autoscale = True if (self._job_metric_collector and task and (task.task_type == elastic_training_pb2.PREDICTION)): self._collect_runtime_stats() self._check_start_auto_scale_worker() return success def _check_start_auto_scale_worker(self): sample_count = self._speed_monitor.get_sample_count() if ((not self._start_autoscale) and (sample_count >= _dlrover_context.sample_count_to_adjust_worker)): logger.info('Start autoscale with %s stats samples', sample_count) self._job_manager.start_auto_scaling() self._start_autoscale = True def _update_cluster_version(self, message: grpc.ClusterVersion): if (not self._elastic_ps_service): return False if (message.task_type == NodeType.WORKER): self._elastic_ps_service.update_worker_version(message.task_id, message.version_type, message.version) elif (message.task_type == NodeType.PS): self._elastic_ps_service.update_ps_version(message.task_id, message.version_type, message.version) return True def _update_node_address(self, message: grpc.NodeAddress): self._job_manager.update_node_service_addr(node_type=message.type, node_id=message.id, service_addr=message.addr) return True def _update_node_status(self, message: grpc.NetworkStatus): net_rdzv_manager = self._rdzv_managers.get(RendezvousName.NETWORK_CHECK, None) if net_rdzv_manager: succeed = (message.status == NodeStatus.SUCCEEDED) net_rdzv_manager.report_network_check_result(message.rank, succeed, message.elasped_time) return True def _update_node_event(self, message: grpc.NodeEvent): node = Node(message.event_type, message.node.id) event = NodeEvent('exit', node) ray_event_queue.put(event) return True def _join_sync(self, node_type, node_id, message: grpc.SyncJoin): success = False if self._sync_service: success = self._sync_service.join_sync(message.sync_name, node_type, node_id) return success def _sync_finished(self, message: grpc.SyncFinish): success = False if self._sync_service: success = self._sync_service.sync_finished(message.sync_name) return success def _barrier(self, message: grpc.SyncBarrier): if (not self._sync_service): return False if message.notify: success = self._sync_service.notify_barrier(message.barrier_name) else: success = self._sync_service.barrier(message.barrier_name) return success def _report_rdzv_params(self, message: grpc.RendezvousParams): for manager in self._rdzv_managers.values(): manager.update_rdzv_params(min_nodes=message.min_nodes, max_ndoes=message.max_nodes, waiting_timeout=message.waiting_timeout, node_unit=message.node_unit) return True def _report_failure(self, node_type, node_id, message: grpc.NodeFailure): self._job_manager.handle_training_failure(node_type, node_id, message.restart_count, message.error_data, message.level) return True def _kv_store_set(self, message: grpc.KeyValuePair): self._kv_store.set(message.key, message.value) return True def _report_paral_config(self, node_type, node_id, message: grpc.ParallelConfig): if self._job_manager: logger.debug('Update parallel config for %s-%s: %s', node_type, node_id, message) self._job_manager.update_node_paral_config(node_type, node_id, message) return True
class _UpConv(nn.Module): def __init__(self, in_ch, out_ch, momentum=0.1): super(_UpConv, self).__init__() self.up = nn.Sequential(nn.Upsample(scale_factor=2), nn.Conv2d(in_ch, out_ch, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False), nn.BatchNorm2d(out_ch, momentum=momentum), nn.ReLU(inplace=True)) def forward(self, x): x = self.up(x) return x