code
stringlengths
101
5.91M
def c(dfs, numL, numR): r = Rule.fromDFS(dfs) assert (r.numLeftComponents == numL) assert (r.numRightComponents == numR) commonChecks(r)
def bottle3(f, x_tuple): x_sizes = tuple(map((lambda x: x.size()), x_tuple)) y = f(*map((lambda x: x[0].view(((x[1][0] * x[1][1]) * x[1][2]), *x[1][3:])), zip(x_tuple, x_sizes))) y_size = y.size() return y.view(x_sizes[0][0], x_sizes[0][1], x_sizes[0][2], *y_size[1:])
class BridgeTowerForImageAndTextRetrieval(metaclass=DummyObject): _backends = ['torch'] def __init__(self, *args, **kwargs): requires_backends(self, ['torch'])
def polarity(importtext): text = word_tokenize(importtext) tokens = nltk.pos_tag(text) polarity = TextBlob(importtext).sentiment[0] sentiment = TextBlob(importtext).sentiment[1] polaritylist = list() for i in range(0, len(tokens), 3): if (i <= (len(tokens) - 3)): words = ((((text[i] + ' ') + text[(i + 1)]) + ' ') + text[(i + 2)]) polaritylist.append(TextBlob(words).sentiment[0]) else: pass avgpolarity = np.mean(polaritylist) stdpolarity = np.std(polaritylist) varpolarity = np.var(polaritylist) return [float(avgpolarity), float(stdpolarity), float(varpolarity)]
class PointnetSAModule(PointnetSAModuleMSG): def __init__(self, *, mlp: List[int], npoint: int=None, radius: float=None, nsample: int=None, bn: bool=True, use_xyz: bool=True, pool_method='max_pool'): super().__init__(mlps=[mlp], npoint=npoint, radii=[radius], nsamples=[nsample], bn=bn, use_xyz=use_xyz, pool_method=pool_method)
def train(train_dataloader, ae, optimizer, optimizer_step, cuda_details: gnn_utils.CudaDetails, tb_logger, lambda_value, property_pred_factor): loss_meter = gnn_utils.AverageMeter() time_meter = gnn_utils.AverageMeter() time_on_calc = gnn_utils.AverageMeter() prediction_mse_meter = gnn_utils.AverageMeter() pre_time = time.time() with tqdm.tqdm(train_dataloader, total=len(train_dataloader)) as t: for (i, (padded_seq, lengths, order, properties)) in enumerate(t): packed_seq = rnn.pack_padded_sequence(padded_seq, lengths, batch_first=True) packed_seq = cuda_details.return_cudafied(packed_seq) properties = cuda_details.return_cudafied(properties) pre_calc_time = time.time() if ((i % 100) == 0): ae.encoder.shallow_dist._tb_logger = tb_logger ae_obj = ae(packed_seq, lambda_=lambda_value).mean() prediction_of_property = ae.prop_predictor_(ae._last_z_sample_on_obj) prop_loss = F.mse_loss(input=prediction_of_property.squeeze(), target=properties.squeeze()) loss = (- ae_obj) loss += (property_pred_factor * prop_loss) optimizer.zero_grad() loss.backward() optimizer_step() if ((i % 100) == 0): ae.encoder.shallow_dist._tb_logger = None loss_meter.update(loss.item(), n=lengths.shape[0]) time_meter.update((time.time() - pre_time)) time_on_calc.update((time.time() - pre_calc_time)) prediction_mse_meter.update(prop_loss.item(), n=lengths.shape[0]) pre_time = time.time() if (tb_logger is not None): tb_logger.add_scalar('property_mse', prop_loss.item()) t.set_postfix(avg_epoch_loss=f'{loss_meter.avg:.4E}', total_time=f'{time_meter.avg:.3E}', calc_time=f'{time_on_calc.avg:.3E}', prop_mse=f'{prediction_mse_meter.avg:.3E}')
_model_architecture('s2ut_conformer', 's2ut_conformer') def s2ut_conformer_architecture_base(args): args.attn_type = getattr(args, 'attn_type', None) args.pos_enc_type = getattr(args, 'pos_enc_type', 'abs') args.input_feat_per_channel = getattr(args, 'input_feat_per_channel', 80) args.input_channels = getattr(args, 'input_channels', 1) args.max_source_positions = getattr(args, 'max_source_positions', 6000) args.encoder_embed_dim = getattr(args, 'encoder_embed_dim', 256) args.encoder_ffn_embed_dim = getattr(args, 'encoder_ffn_embed_dim', 2048) args.encoder_attention_heads = getattr(args, 'encoder_attention_heads', 4) args.dropout = getattr(args, 'dropout', 0.1) args.encoder_layers = getattr(args, 'encoder_layers', 16) args.depthwise_conv_kernel_size = getattr(args, 'depthwise_conv_kernel_size', 31) s2ut_architecture_base(args)
class LEVIRCDPlus(torch.utils.data.Dataset): splits = ['train', 'test'] def __init__(self, root: str='.data/levircd_plus', split: str='train', transform: Compose=Compose([ToTensor()])): assert (split in self.splits) self.root = root self.transform = transform self.files = self.load_files(root, split) def load_files(root: str, split: str): files = [] images = glob(os.path.join(root, split, 'A', '*.png')) images = sorted([os.path.basename(image) for image in images]) for image in images: image1 = os.path.join(root, split, 'A', image) image2 = os.path.join(root, split, 'B', image) mask = os.path.join(root, split, 'label', image) files.append(dict(image1=image1, image2=image2, mask=mask)) return files def __len__(self) -> int: return len(self.files) def __getitem__(self, idx: int) -> Dict: files = self.files[idx] mask = np.array(Image.open(files['mask'])) mask = np.clip(mask, 0, 1) image1 = np.array(Image.open(files['image1'])) image2 = np.array(Image.open(files['image2'])) (image1, image2, mask) = self.transform([image1, image2, mask]) x = torch.stack([image1, image2], dim=0) return dict(x=x, mask=mask)
def create_FDS_train_subset(args): print('Creating FDS statistics updating subset...') frame = pd.read_csv(os.path.join(args.data_dir, 'nyu2_train.csv'), header=None) select_id = np.load(os.path.join(args.data_dir, 'FDS_train_subset_id.npy')) frame.iloc[select_id].to_csv(os.path.join(args.data_dir, 'nyu2_train_FDS_subset.csv'), index=False, header=False)
def parse_arguments(): parser = argparse.ArgumentParser() parser.add_argument('--input_model', type=str, required=False, default='zfnet512-12.onnx') parser.add_argument('--output_model', type=str, required=True) return parser.parse_args()
class TestHPO(unittest.TestCase): search_space = {'learning_rate': SearchSpace((0.0001, 0.001)), 'num_train_epochs': SearchSpace(bound=(20, 100), interval=1), 'weight_decay': SearchSpace((0.0001, 0.001)), 'cooldown_epochs': SearchSpace(bound=(0, 10), interval=1), 'sparsity_warm_epochs': SearchSpace(bound=(0, 5), interval=1), 'per_device_train_batch_size': SearchSpace((5, 20), 1)} def test_searcher(self): hpo_config = HPOConfig({'num_train_epochs': self.search_space['num_train_epochs'], 'cooldown_epochs': self.search_space['cooldown_epochs']}, searcher='grid') searcher = GridSearcher({'num_train_epochs': self.search_space['num_train_epochs'], 'cooldown_epochs': self.search_space['cooldown_epochs']}) conf_searcher = prepare_hpo(hpo_config) self.assertEqual(searcher.__class__, conf_searcher.__class__) for _ in range(5): self.assertEqual(searcher.suggest(), conf_searcher.suggest()) hpo_config = HPOConfig(self.search_space, 'random') searcher = prepare_hpo(hpo_config) for _ in range(5): searcher.suggest() hpo_config = HPOConfig(self.search_space, 'bo') searcher = prepare_hpo(hpo_config) params = [] for _ in range(5): searcher.suggest() searcher.get_feedback(np.random.random()) for _ in range(5): param = searcher.suggest() if (param in params): continue params.append(param) searcher.feedback(param, np.random.random()) hpo_config = HPOConfig(self.search_space, 'xgb', higher_is_better=True, min_train_samples=3) searcher = prepare_hpo(hpo_config) for _ in range(5): searcher.suggest() searcher.get_feedback(np.random.random()) for _ in range(5): param = searcher.suggest() searcher.feedback(param, np.random.random()) def test_search_space(self): ds = DiscreteSearchSpace(bound=[0, 10]) get_ds = SearchSpace(bound=[0, 10], interval=1) self.assertEqual(ds.__class__, get_ds.__class__) self.assertEqual(ds.index(1), ds.get_nth_value(1)) ds = DiscreteSearchSpace(value=[1, 2, 3, 4]) self.assertEqual(ds.get_all(), [1, 2, 3, 4]) ds = DiscreteSearchSpace(bound=[0.01, 0.1]) self.assertEqual(ds.interval, 0.01) self.assertIn(ds.get_value(), ds.get_all()) self.assertEqual(ds.get_value(2), ds.get_nth_value(2)) cs = ContinuousSearchSpace(bound=[0.01, 0.1]) self.assertTrue((cs.get_value() >= 0.01)) self.assertTrue((cs.get_value() < 0.1)) def test_sa(self): def f(x): return np.mean(np.log((x ** 2)), axis=1) points = np.random.randn(5, 6) optimizer = SimulatedAnnealingOptimizer(T0=100, Tf=0, alpha=0.9, higher_is_better=True) result = optimizer.gen_next_params(f, points) optimizer = SimulatedAnnealingOptimizer(T0=1, Tf=0.01, alpha=None, higher_is_better=False) result2 = optimizer.gen_next_params(f, points) self.assertTrue((len(result) == len(result2)))
def make_agent(id, **kwargs): if isinstance(id, str): wargs = dict(**_agent_registry[id]) del wargs['agent'] wargs.update(kwargs) instance = _agent_registry[id]['agent'](name=id, **wargs) return instance else: return id(**kwargs)
def get_extensions(): this_dir = os.path.dirname(os.path.abspath(__file__)) extensions_dir = this_dir main_file = glob.glob(os.path.join(extensions_dir, '*.cpp')) source_cpu = glob.glob(os.path.join(extensions_dir, 'cpu', '*.cpp')) source_cuda = glob.glob(os.path.join(extensions_dir, 'cuda', '*.cu')) sources = (main_file + source_cpu) extension = CppExtension extra_compile_args = {'cxx': []} define_macros = [] if (torch.cuda.is_available() and (CUDA_HOME is not None)): extension = CUDAExtension sources += source_cuda define_macros += [('WITH_CUDA', None)] extra_compile_args['nvcc'] = ['-DCUDA_HAS_FP16=1', '-D__CUDA_NO_HALF_OPERATORS__', '-D__CUDA_NO_HALF_CONVERSIONS__', '-D__CUDA_NO_HALF2_OPERATORS__'] sources = [os.path.join(extensions_dir, s) for s in sources] include_dirs = [extensions_dir] ext_modules = [extension('C_ROIPooling', sources, include_dirs=include_dirs, define_macros=define_macros, extra_compile_args=extra_compile_args)] return ext_modules
class HfApi(): def __init__(self, endpoint=None): self.endpoint = (endpoint if (endpoint is not None) else ENDPOINT) def login(self, username: str, password: str) -> str: path = '{}/api/login'.format(self.endpoint) r = requests.post(path, json={'username': username, 'password': password}) r.raise_for_status() d = r.json() return d['token'] def whoami(self, token: str) -> Tuple[(str, List[str])]: path = '{}/api/whoami'.format(self.endpoint) r = requests.get(path, headers={'authorization': 'Bearer {}'.format(token)}) r.raise_for_status() d = r.json() return (d['user'], d['orgs']) def logout(self, token: str) -> None: path = '{}/api/logout'.format(self.endpoint) r = requests.post(path, headers={'authorization': 'Bearer {}'.format(token)}) r.raise_for_status() def presign(self, token: str, filename: str, organization: Optional[str]=None) -> PresignedUrl: path = '{}/api/presign'.format(self.endpoint) r = requests.post(path, headers={'authorization': 'Bearer {}'.format(token)}, json={'filename': filename, 'organization': organization}) r.raise_for_status() d = r.json() return PresignedUrl(**d) def presign_and_upload(self, token: str, filename: str, filepath: str, organization: Optional[str]=None) -> str: urls = self.presign(token, filename=filename, organization=organization) with open(filepath, 'rb') as f: pf = TqdmProgressFileReader(f) data = (f if (pf.total_size > 0) else '') r = requests.put(urls.write, data=data, headers={'content-type': urls.type}) r.raise_for_status() pf.close() return urls.access def list_objs(self, token: str, organization: Optional[str]=None) -> List[S3Obj]: path = '{}/api/listObjs'.format(self.endpoint) params = ({'organization': organization} if (organization is not None) else None) r = requests.get(path, params=params, headers={'authorization': 'Bearer {}'.format(token)}) r.raise_for_status() d = r.json() return [S3Obj(**x) for x in d] def delete_obj(self, token: str, filename: str, organization: Optional[str]=None): path = '{}/api/deleteObj'.format(self.endpoint) r = requests.delete(path, headers={'authorization': 'Bearer {}'.format(token)}, json={'filename': filename, 'organization': organization}) r.raise_for_status() def model_list(self) -> List[ModelInfo]: path = '{}/api/models'.format(self.endpoint) r = requests.get(path) r.raise_for_status() d = r.json() return [ModelInfo(**x) for x in d]
def load_dataset(n_jobs, use_gpu, pin_memory, ascending, corpus, audio, text): print('Prepare dataloader for training/validation') (audio_transform, feat_dim) = create_transform(audio.copy()) tokenizer = load_text_encoder(**text) (tr_set, dv_set, tr_loader_bs, dv_loader_bs, mode, data_msg) = create_dataset(tokenizer, ascending, **corpus) collect_tr = partial(collect_audio_batch, audio_transform=audio_transform, mode=mode) collect_dv = partial(collect_audio_batch, audio_transform=audio_transform, mode='test') shuffle = ((mode == 'train') and (not ascending)) drop_last = shuffle tr_set = DataLoader(tr_set, batch_size=tr_loader_bs, shuffle=shuffle, drop_last=drop_last, collate_fn=collect_tr, num_workers=n_jobs, pin_memory=use_gpu) dv_set = DataLoader(dv_set, batch_size=dv_loader_bs, shuffle=False, drop_last=False, collate_fn=collect_dv, num_workers=n_jobs, pin_memory=pin_memory) data_msg.append('I/O spec. | Audio feature = {}\t| feature dim = {}\t| Token type = {}\t| Vocab size = {}'.format(audio['feat_type'], feat_dim, tokenizer.token_type, tokenizer.vocab_size)) return (tr_set, dv_set, feat_dim, tokenizer.vocab_size, tokenizer, data_msg)
class ModelCriterionConfig(FairseqDataclass): loss_weights: Dict[(str, float)] = field(default_factory=dict, metadata={'help': 'weights for the loss terms'}) log_keys: List[str] = field(default_factory=list, metadata={'help': 'additional output keys to log'})
def track_parallel_progress(func, tasks, nproc, initializer=None, initargs=None, bar_width=50, chunksize=1, skip_first=False, keep_order=True): if isinstance(tasks, tuple): assert (len(tasks) == 2) assert isinstance(tasks[0], collections_abc.Iterable) assert isinstance(tasks[1], int) task_num = tasks[1] tasks = tasks[0] elif isinstance(tasks, collections_abc.Iterable): task_num = len(tasks) else: raise TypeError('"tasks" must be an iterable object or a (iterator, int) tuple') pool = init_pool(nproc, initializer, initargs) start = (not skip_first) task_num -= ((nproc * chunksize) * int(skip_first)) prog_bar = ProgressBar(task_num, bar_width, start) results = [] if keep_order: gen = pool.imap(func, tasks, chunksize) else: gen = pool.imap_unordered(func, tasks, chunksize) for result in gen: results.append(result) if skip_first: if (len(results) < (nproc * chunksize)): continue elif (len(results) == (nproc * chunksize)): prog_bar.start() continue prog_bar.update() sys.stdout.write('\n') pool.close() pool.join() return results
class ContinuousInverseModel(nn.Module): def __init__(self, state_size, action_size, log_std_low=(- 10.0), log_std_high=2.0, hidden_size=256, dist_impl='pyd'): super().__init__() assert (dist_impl in ['pyd', 'beta']) self.fc1 = nn.Linear((state_size * 2), hidden_size) self.fc2 = nn.Linear(hidden_size, hidden_size) self.fc3 = nn.Linear(hidden_size, (2 * action_size)) self.log_std_low = log_std_low self.log_std_high = log_std_high self.apply(weight_init) self.dist_impl = dist_impl def forward(self, state, next_state): inp = torch.cat((state, next_state), dim=(- 1)) x = F.relu(self.fc1(inp)) x = F.relu(self.fc2(x)) out = self.fc3(x) if (self.dist_impl == 'pyd'): return distributions.create_tanh_normal(out, self.log_std_low, self.log_std_high) elif (self.dist_impl == 'beta'): return distributions.create_beta(out)
def test_glorot_normal_receptive_field(): from lasagne.init import GlorotNormal sample = GlorotNormal().sample((50, 50, 2)) assert ((- 0.01) < sample.mean() < 0.01) assert (0.09 < sample.std() < 0.11)
class Bottleneck(nn.Module): expansion = 4 def __init__(self, inplanes, planes, stride=1, downsample=None): super(Bottleneck, self).__init__() self.conv1 = nn.Conv2d(inplanes, planes, kernel_size=1, bias=False) self.bn1 = nn.BatchNorm2d(planes) self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=stride, padding=1, bias=False) self.bn2 = nn.BatchNorm2d(planes) self.conv3 = nn.Conv2d(planes, (planes * 4), kernel_size=1, bias=False) self.bn3 = nn.BatchNorm2d((planes * 4)) self.relu = nn.ReLU(inplace=True) self.downsample = downsample self.stride = stride def forward(self, x): residual = x out = self.conv1(x) out = self.bn1(out) out = self.relu(out) out = self.conv2(out) out = self.bn2(out) out = self.relu(out) out = self.conv3(out) out = self.bn3(out) if (self.downsample is not None): residual = self.downsample(x) out += residual out = self.relu(out) return out
def loss_G_fn(P, D, options, images, gen_images): d_gen = D(P.augment_fn(gen_images)) if (options['loss'] == 'nonsat'): g_loss = F.softplus((- d_gen)).mean() elif (options['loss'] == 'lsgan'): g_loss = (0.5 * ((d_gen - 1.0) ** 2).mean()) else: g_loss = (- d_gen.mean()) return g_loss
def initialize_double_double_artificial_homotopy(target, start, homogeneous=False, vrblvl=0): if (vrblvl > 0): print('in initialize_double_double_artificial_homotopy', end='') print(', homogeneous :', homogeneous) print('the target system :') for pol in target: print(pol) print('the start system :') for pol in start: print(pol) set_double_double_target_system(target, vrblvl) set_double_double_start_system(start, vrblvl) phc = get_phcfun() aprc = pointer(c_int32(1)) bpars = (c_int32 * 2)() bpars[0] = c_int32(vrblvl) bpars[1] = c_int32(int(homogeneous)) bpar = pointer(bpars) ccc = pointer(c_double(0.0)) vrb = c_int32(vrblvl) if (vrblvl > 0): print('-> initialize_double_double_artificial_homotopy calls phc', end='') retval = phc(860, aprc, bpar, ccc, vrb) if (vrblvl > 0): print(', return value :', retval) return retval
class MAP(): def __init__(self, length=20): self.length = length def init(self, train): return def reset(self): self.test = 0 self.pos = 0 def skip(self, for_item=0, session=(- 1)): pass def add_multiple(self, result, next_items, for_item=0, session=0, position=None): last_recall = 0 res = 0 for i in range(self.length): recall = self.recall(result[:i].index, next_items) precision = self.precision(result[:i].index, next_items) res += (precision * (recall - last_recall)) last_recall = recall self.pos += res self.test += 1 def add(self, result, next_item, for_item=0, session=0, pop_bin=None, position=None): sum = 0 for i in range(self.length): sum += self.mrr(result, next_item, (i + 1)) self.pos += (sum / self.length) self.test += 1 def recall(self, result, next_items): return (len((set(next_items) & set(result))) / len(next_items)) def precision(self, result, next_items): return (len((set(next_items) & set(result))) / self.length) def mrr(self, result, next_item, n): res = result[:n] if (next_item in res.index): rank = (res.index.get_loc(next_item) + 1) return (1.0 / rank) else: return 0 def add_batch(self, result, next_item): i = 0 for (part, series) in result.iteritems(): result.sort_values(part, ascending=False, inplace=True) self.add(series, next_item[i]) i += 1 def result(self): return ((('' + str(self.length)) + ': '), (self.pos / self.test))
class CollectLayerHistogram(unittest.TestCase): def setUp(self): model = BuildFakeModel(width_mult=1) (layer_tensor, include_layer) = (OrderedDict(), OrderedDict()) i = 0 for (key, value) in model.state_dict().items(): if (not value.ndim): value = np.expand_dims(value, axis=0) if (i > 200): pass else: include_layer[key] = np.array(value, dtype=np.float32) layer_tensor[key] = np.array(value, dtype=np.float32) i += 1 self.layer_histogram_collector = LayerHistogramCollector(num_bins=8001, layer_tensor=layer_tensor, include_layer=include_layer, logger=logger) def test_layer_histogram(self): self.layer_histogram_collector.collect() self.assertEqual((self.layer_histogram_collector.layer_tensor.keys() & self.layer_histogram_collector.include_layer.keys()), self.layer_histogram_collector.hist_dict.keys())
def load_tensorrt_plugin(): global plugin_is_loaded lib_path = get_tensorrt_op_path() if ((not plugin_is_loaded) and os.path.exists(lib_path)): ctypes.CDLL(lib_path) plugin_is_loaded = True
class KPFCNN(nn.Module): def __init__(self, config): super(KPFCNN, self).__init__() self.encoder = KPCNN(config) self.config = config self.blocks = nn.ModuleDict() start_i = 0 for (block_i, block) in enumerate(config.architecture): if ('upsample' in block): start_i = block_i break layer = (config.num_layers - 1) r = ((config.first_subsampling_dl * config.density_parameter) * (2 ** layer)) in_fdim = (config.first_features_dim * (2 ** layer)) out_fdim = in_fdim block_in_layer = 0 for (block_i, block) in enumerate(config.architecture[start_i:]): is_strided = ('strided' in block) self.blocks[f'layer{layer}/{block}'] = get_block(block, config, int((1.5 * in_fdim)), out_fdim, radius=r, strided=is_strided) in_fdim = out_fdim block_in_layer += 1 if ('upsample' in block): out_fdim = (out_fdim // 2) r *= 0.5 layer -= 1 block_in_layer = 0 self.blocks['segmentation_head'] = nn.Sequential(nn.Linear(out_fdim, config.first_features_dim), nn.BatchNorm1d(config.first_features_dim, momentum=config.batch_norm_momentum, eps=1e-06), nn.LeakyReLU(negative_slope=0.2), nn.Linear(config.first_features_dim, config.num_classes)) def forward(self, inputs): features = self.feature_extraction(inputs) logits = self.segmentation_head(features) return logits def feature_extraction(self, inputs): F = self.encoder.feature_extraction(inputs) features = F[(- 1)] layer = (self.config.num_layers - 1) r = ((self.config.first_subsampling_dl * self.config.density_parameter) * (2 ** layer)) fdim = (self.config.first_features_dim * (2 ** layer)) start_i = 0 for (block_i, block) in enumerate(self.config.architecture): if ('upsample' in block): start_i = block_i break for (block_i, block) in enumerate(self.config.architecture[start_i:]): block_ops = self.blocks[f'layer{layer}/{block}'] if ('upsample' in block): if (block == 'nearest_upsample'): upsample_indices = inputs['upsamples'][(layer - 1)] else: raise ValueError(f'Unknown block type. {block}') features = block_ops(upsample_indices, features) else: if (block in ['unary', 'simple', 'resnet', 'resnetb']): query_points = inputs['points'][layer] support_points = inputs['points'][layer] neighbors_indices = inputs['neighbors'][layer] elif (block in ['simple_strided', 'resnetb_strided', 'resnetb_deformable_strided']): query_points = inputs['points'][(layer + 1)] support_points = inputs['points'][layer] neighbors_indices = inputs['pools'][layer] else: raise ValueError(f'Unknown block type. {block}') features = block_ops(query_points, support_points, neighbors_indices, features) if ('upsample' in block): layer -= 1 r *= 0.5 fdim = (fdim // 2) features = torch.cat((features, F[layer]), dim=1) return features def segmentation_head(self, features): logits = self.blocks['segmentation_head'](features) return logits
def InceptionV3(include_top=True, weights='imagenet', input_tensor=None, input_shape=None, pooling=None, classes=1000, **kwargs): global backend, layers, models, keras_utils (backend, layers, models, keras_utils) = get_submodules_from_kwargs(kwargs) if (not ((weights in {'imagenet', None}) or os.path.exists(weights))): raise ValueError('The `weights` argument should be either `None` (random initialization), `imagenet` (pre-training on ImageNet), or the path to the weights file to be loaded.') if ((weights == 'imagenet') and include_top and (classes != 1000)): raise ValueError('If using `weights` as `"imagenet"` with `include_top` as true, `classes` should be 1000') if (input_tensor is None): img_input = layers.Input(shape=input_shape) elif (not backend.is_keras_tensor(input_tensor)): img_input = layers.Input(tensor=input_tensor, shape=input_shape) else: img_input = input_tensor if (backend.image_data_format() == 'channels_first'): channel_axis = 1 else: channel_axis = 4 x = conv3d_bn(img_input, 32, 3, 3, 3, strides=(2, 2, 2), padding='valid') x = conv3d_bn(x, 32, 3, 3, 3, padding='valid') x = conv3d_bn(x, 64, 3, 3, 3) x = layers.MaxPooling3D((3, 3, 3), strides=(2, 2, 2))(x) x = conv3d_bn(x, 80, 1, 1, 1, padding='valid') x = conv3d_bn(x, 192, 3, 3, 3, padding='valid') x = layers.MaxPooling3D((3, 3, 3), strides=(2, 2, 2))(x) branch1x1 = conv3d_bn(x, 64, 1, 1, 1) branch5x5 = conv3d_bn(x, 48, 1, 1, 1) branch5x5 = conv3d_bn(branch5x5, 64, 5, 5, 5) branch3x3dbl = conv3d_bn(x, 64, 1, 1, 1) branch3x3dbl = conv3d_bn(branch3x3dbl, 96, 3, 3, 3) branch3x3dbl = conv3d_bn(branch3x3dbl, 96, 3, 3, 3) branch_pool = layers.AveragePooling3D((3, 3, 3), strides=(1, 1, 1), padding='same')(x) branch_pool = conv3d_bn(branch_pool, 32, 1, 1, 1) x = layers.concatenate([branch1x1, branch5x5, branch3x3dbl, branch_pool], axis=channel_axis, name='mixed0') branch1x1 = conv3d_bn(x, 64, 1, 1, 1) branch5x5 = conv3d_bn(x, 48, 1, 1, 1) branch5x5 = conv3d_bn(branch5x5, 64, 5, 5, 5) branch3x3dbl = conv3d_bn(x, 64, 1, 1, 1) branch3x3dbl = conv3d_bn(branch3x3dbl, 96, 3, 3, 3) branch3x3dbl = conv3d_bn(branch3x3dbl, 96, 3, 3, 3) branch_pool = layers.AveragePooling3D((3, 3, 3), strides=(1, 1, 1), padding='same')(x) branch_pool = conv3d_bn(branch_pool, 64, 1, 1, 1) x = layers.concatenate([branch1x1, branch5x5, branch3x3dbl, branch_pool], axis=channel_axis, name='mixed1') branch1x1 = conv3d_bn(x, 64, 1, 1, 1) branch5x5 = conv3d_bn(x, 48, 1, 1, 1) branch5x5 = conv3d_bn(branch5x5, 64, 5, 5, 5) branch3x3dbl = conv3d_bn(x, 64, 1, 1, 1) branch3x3dbl = conv3d_bn(branch3x3dbl, 96, 3, 3, 3) branch3x3dbl = conv3d_bn(branch3x3dbl, 96, 3, 3, 3) branch_pool = layers.AveragePooling3D((3, 3, 3), strides=(1, 1, 1), padding='same')(x) branch_pool = conv3d_bn(branch_pool, 64, 1, 1, 1) x = layers.concatenate([branch1x1, branch5x5, branch3x3dbl, branch_pool], axis=channel_axis, name='mixed2') branch3x3 = conv3d_bn(x, 384, 3, 3, 3, strides=(2, 2, 2), padding='valid') branch3x3dbl = conv3d_bn(x, 64, 1, 1, 1) branch3x3dbl = conv3d_bn(branch3x3dbl, 96, 3, 3, 3) branch3x3dbl = conv3d_bn(branch3x3dbl, 96, 3, 3, 3, strides=(2, 2, 2), padding='valid') branch_pool = layers.MaxPooling3D((3, 3, 3), strides=(2, 2, 2))(x) x = layers.concatenate([branch3x3, branch3x3dbl, branch_pool], axis=channel_axis, name='mixed3') branch1x1 = conv3d_bn(x, 192, 1, 1, 1) branch7x7 = conv3d_bn(x, 128, 1, 1, 1) branch7x7 = conv3d_bn(branch7x7, 128, 1, 7, 1) branch7x7 = conv3d_bn(branch7x7, 192, 7, 1, 1) branch7x7dbl = conv3d_bn(x, 128, 1, 1, 1) branch7x7dbl = conv3d_bn(branch7x7dbl, 128, 7, 1, 1) branch7x7dbl = conv3d_bn(branch7x7dbl, 128, 1, 7, 1) branch7x7dbl = conv3d_bn(branch7x7dbl, 128, 7, 1, 1) branch7x7dbl = conv3d_bn(branch7x7dbl, 192, 1, 7, 1) branch_pool = layers.AveragePooling3D((3, 3, 3), strides=(1, 1, 1), padding='same')(x) branch_pool = conv3d_bn(branch_pool, 192, 1, 1, 1) x = layers.concatenate([branch1x1, branch7x7, branch7x7dbl, branch_pool], axis=channel_axis, name='mixed4') for i in range(2): branch1x1 = conv3d_bn(x, 192, 1, 1, 1) branch7x7 = conv3d_bn(x, 160, 1, 1, 1) branch7x7 = conv3d_bn(branch7x7, 160, 1, 7, 1) branch7x7 = conv3d_bn(branch7x7, 192, 7, 1, 1) branch7x7dbl = conv3d_bn(x, 160, 1, 1, 1) branch7x7dbl = conv3d_bn(branch7x7dbl, 160, 7, 1, 1) branch7x7dbl = conv3d_bn(branch7x7dbl, 160, 1, 7, 1) branch7x7dbl = conv3d_bn(branch7x7dbl, 160, 7, 1, 1) branch7x7dbl = conv3d_bn(branch7x7dbl, 192, 1, 7, 1) branch_pool = layers.AveragePooling3D((3, 3, 3), strides=(1, 1, 1), padding='same')(x) branch_pool = conv3d_bn(branch_pool, 192, 1, 1, 1) x = layers.concatenate([branch1x1, branch7x7, branch7x7dbl, branch_pool], axis=channel_axis, name=('mixed' + str((5 + i)))) branch1x1 = conv3d_bn(x, 192, 1, 1, 1) branch7x7 = conv3d_bn(x, 192, 1, 1, 1) branch7x7 = conv3d_bn(branch7x7, 192, 1, 7, 1) branch7x7 = conv3d_bn(branch7x7, 192, 7, 1, 1) branch7x7dbl = conv3d_bn(x, 192, 1, 1, 1) branch7x7dbl = conv3d_bn(branch7x7dbl, 192, 7, 1, 1) branch7x7dbl = conv3d_bn(branch7x7dbl, 192, 1, 7, 1) branch7x7dbl = conv3d_bn(branch7x7dbl, 192, 7, 1, 1) branch7x7dbl = conv3d_bn(branch7x7dbl, 192, 1, 7, 1) branch_pool = layers.AveragePooling3D((3, 3, 3), strides=(1, 1, 1), padding='same')(x) branch_pool = conv3d_bn(branch_pool, 192, 1, 1, 1) x = layers.concatenate([branch1x1, branch7x7, branch7x7dbl, branch_pool], axis=channel_axis, name='mixed7') branch3x3 = conv3d_bn(x, 192, 1, 1, 1) branch3x3 = conv3d_bn(branch3x3, 320, 3, 3, 3, strides=(2, 2, 2), padding='valid') branch7x7x3 = conv3d_bn(x, 192, 1, 1, 1) branch7x7x3 = conv3d_bn(branch7x7x3, 192, 1, 7, 1) branch7x7x3 = conv3d_bn(branch7x7x3, 192, 7, 1, 1) branch7x7x3 = conv3d_bn(branch7x7x3, 192, 3, 3, 3, strides=(2, 2, 2), padding='valid') branch_pool = layers.MaxPooling3D((3, 3, 3), strides=(2, 2, 2))(x) x = layers.concatenate([branch3x3, branch7x7x3, branch_pool], axis=channel_axis, name='mixed8') for i in range(2): branch1x1 = conv3d_bn(x, 320, 1, 1, 1) branch3x3 = conv3d_bn(x, 384, 1, 1, 1) branch3x3_1 = conv3d_bn(branch3x3, 384, 1, 3, 1) branch3x3_2 = conv3d_bn(branch3x3, 384, 3, 1, 1) branch3x3 = layers.concatenate([branch3x3_1, branch3x3_2], axis=channel_axis, name=('mixed9_' + str(i))) branch3x3dbl = conv3d_bn(x, 448, 1, 1, 1) branch3x3dbl = conv3d_bn(branch3x3dbl, 384, 3, 3, 3) branch3x3dbl_1 = conv3d_bn(branch3x3dbl, 384, 1, 3, 1) branch3x3dbl_2 = conv3d_bn(branch3x3dbl, 384, 3, 1, 1) branch3x3dbl = layers.concatenate([branch3x3dbl_1, branch3x3dbl_2], axis=channel_axis) branch_pool = layers.AveragePooling3D((3, 3, 3), strides=(1, 1, 1), padding='same')(x) branch_pool = conv3d_bn(branch_pool, 192, 1, 1, 1) x = layers.concatenate([branch1x1, branch3x3, branch3x3dbl, branch_pool], axis=channel_axis, name=('mixed' + str((9 + i)))) if include_top: x = layers.GlobalAveragePooling3D(name='avg_pool')(x) x = layers.Dense(classes, activation='softmax', name='predictions')(x) elif (pooling == 'avg'): x = layers.GlobalAveragePooling3D()(x) elif (pooling == 'max'): x = layers.GlobalMaxPooling3D()(x) if (input_tensor is not None): inputs = keras_utils.get_source_inputs(input_tensor) else: inputs = img_input model = models.Model(inputs, x, name='inception_v3') if (weights == 'imagenet'): if include_top: weights_path = keras_utils.get_file('inception_v3_weights_tf_dim_ordering_tf_kernels.h5', WEIGHTS_PATH, cache_subdir='models', file_hash='9a0d58056eeedaa3f26cb7ebd46da564') else: weights_path = keras_utils.get_file('inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5', WEIGHTS_PATH_NO_TOP, cache_subdir='models', file_hash='bcbd6486424b2319ff4ef7d526e38f63') model.load_weights(weights_path) elif (weights is not None): model.load_weights(weights) return model
(version='2.0') class Optimizers(object): def __init__(self, framework): assert (framework in ('tensorflow', 'pytorch', 'pytorch_fx')), 'framework support tensorflow pytorch' self.optimizers = framework_optimizers[framework]().optimizers def __getitem__(self, optimizer_type): assert (optimizer_type in self.optimizers.keys()), 'only support optimizers in {}'.format(self.optimizers.keys()) return self.optimizers[optimizer_type] def register(self, name, optimizer_cls): assert (name not in self.optimizers.keys()), 'registered optimizer name already exists.' self.optimizers.update({name: optimizer_cls})
class Permute(torch.nn.Module): def __init__(self, dims): super().__init__() self.dims = dims def forward(self, input: Tensor) -> Tensor: return input.permute(self.dims).contiguous()
def test_modelcheckpoint_get_state(): fpath = 'tests/test_model_functioning/modelcheckpoint/' model_checkpoint = ModelCheckpoint(filepath='/'.join([fpath, 'weights_out']), monitor='val_loss') trainer = Trainer(model, objective='binary', callbacks=[model_checkpoint], verbose=0) trainer.fit(X_wide=X_wide, X_tab=X_tab, target=target, n_epochs=1, batch_size=16) with open('/'.join([fpath, 'checkpoint.p']), 'wb') as f: pickle.dump(model_checkpoint, f) with open('/'.join([fpath, 'checkpoint.p']), 'rb') as f: model_checkpoint = pickle.load(f) self_dict_keys = model_checkpoint.__dict__.keys() no_trainer = ('trainer' not in self_dict_keys) no_model = ('model' not in self_dict_keys) shutil.rmtree('tests/test_model_functioning/modelcheckpoint/') assert (no_trainer and no_model)
def train(model, loader, optimizer, device, weights): model.train() total_loss = total_examples = 0 total_correct = total_examples = 0 for data in loader: data = data.to(device) if (data.train_mask.sum() == 0): continue optimizer.zero_grad() out = model(data.x, data.edge_index)[data.train_mask] y = data.y.squeeze(1)[data.train_mask] loss = F.nll_loss(out, y, weight=weights) loss.backward() optimizer.step() num_examples = data.train_mask.sum().item() total_loss += (loss.item() * num_examples) total_examples += num_examples total_correct += out.argmax(dim=(- 1)).eq(y).sum().item() total_examples += y.size(0) return ((total_loss / total_examples), (total_correct / total_examples))
def tune_model(model_type, X_tune, y_tune, X_val, y_val, tree_type=None, scoring='nll', bagging_frac=1.0, gridsearch=True, cond_mean_type='base', n_stopping_rounds=25, in_dir=None, logger=None, verbose=0, n_jobs=1): start = time.time() model = get_model(model_type=model_type, tree_type=tree_type, scoring=scoring, bagging_frac=bagging_frac) model_val = None tune_dict = {'base_model': model} if (in_dir is not None): if logger: logger.info(f''' loading saved validation model from {in_dir}/...''') result = np.load(os.path.join(in_dir, 'results.npy'), allow_pickle=True)[()] model_val = util.load_model(model_type=tree_type, fp=result['saved_models']['model_val']) model_test = util.load_model(model_type=tree_type, fp=result['saved_models']['model_test']) tune_dict['model_val'] = model_val tune_dict['model_test'] = model_test tune_dict['tune_time_model'] = result['timing']['tune_model'] else: param_grid = get_params(model_type=model_type, tree_type=tree_type, n_train=len(X_tune)) if ((model_type == 'ibug') and (tree_type == 'lgb') and (args.custom_dir == 'ibug_bart')): param_grid['n_estimators'] = [10, 50, 100, 200] if ((model_type in ['constant', 'ibug', 'pgbm', 'bart', 'cbu', 'knn']) and gridsearch): if logger: logger.info('\nmodel: {}, param_grid: {}'.format(model_type, param_grid)) cv_results = [] best_score = None best_model = None best_params = None param_dicts = list(util.product_dict(**param_grid)) for (i, param_dict) in enumerate(param_dicts): temp_model = clone(model).set_params(**param_dict).fit(X_tune, y_tune) y_val_hat = temp_model.predict(X_val) param_dict['score'] = mean_squared_error(y_val, y_val_hat) cv_results.append(param_dict) if logger: logger.info(f"[{(i + 1):,}/{len(param_dicts):,}] {param_dict}, cum. time: {(time.time() - start):.3f}s, score: {param_dict['score']:.3f}") if ((best_score is None) or (param_dict['score'] < best_score)): best_score = param_dict['score'] best_model = temp_model best_params = param_dict df = pd.DataFrame(cv_results).sort_values('score', ascending=True) del best_params['score'] if logger: logger.info(f''' gridsearch results: {df}''') assert (best_model is not None) model_val = best_model elif (model_type in ['constant', 'ibug', 'knn']): assert (tree_type in ['lgb', 'xgb', 'cb', 'ngboost', 'pgbm']) if (tree_type == 'lgb'): model_val = clone(model).fit(X_tune, y_tune, eval_set=[(X_val, y_val)], eval_metric='mse', early_stopping_rounds=n_stopping_rounds) best_n_estimators = model_val.best_iteration_ elif (tree_type == 'xgb'): model_val = clone(model).fit(X_tune, y_tune, eval_set=[(X_val, y_val)], early_stopping_rounds=n_stopping_rounds) best_n_estimators = model_val.best_ntree_limit elif (tree_type == 'cb'): model_val = clone(model).fit(X_tune, y_tune, eval_set=[(X_val, y_val)], early_stopping_rounds=n_stopping_rounds) best_n_estimators = model_val.tree_count_ elif (tree_type == 'ngboost'): model_val = clone(model).fit(X_tune, y_tune, X_val=X_val, Y_val=y_val, early_stopping_rounds=n_stopping_rounds) if (model_val.best_val_loss_itr is None): best_n_estimators = model_val.n_estimators else: best_n_estimators = (model_val.best_val_loss_itr + 1) elif (tree_type == 'pgbm'): model_val = clone(model).fit(X_tune, y_tune, eval_set=(X_val, y_val), early_stopping_rounds=n_stopping_rounds) best_n_estimators = model_val.learner_.best_iteration else: raise ValueError(f'Unknown tree type {tree_type}') best_n_estimators = max(best_n_estimators, MIN_NUM_TREES) best_params = {'n_estimators': best_n_estimators} elif (model_type == 'bart'): model_val = clone(model).fit(X_tune, y_tune) best_params = {} elif (model_type == 'cbu'): model_val = clone(model).fit(X_tune, y_tune) best_params = {} elif (model_type == 'pgbm'): model_val = clone(model).fit(X_tune, y_tune, eval_set=(X_val, y_val), early_stopping_rounds=n_stopping_rounds) best_n_estimators = model_val.learner_.best_iteration best_n_estimators = max(best_n_estimators, MIN_NUM_TREES) best_params = {'n_estimators': best_n_estimators} else: assert (model_type == 'ngboost') model_val = clone(model).fit(X_tune, y_tune, X_val=X_val, Y_val=y_val, early_stopping_rounds=n_stopping_rounds) if (model_val.best_val_loss_itr is None): best_n_estimators = model_val.n_estimators else: best_n_estimators = (model_val.best_val_loss_itr + 1) best_n_estimators = max(best_n_estimators, MIN_NUM_TREES) best_params = {'n_estimators': best_n_estimators} tune_dict['model_val'] = model_val tune_dict['best_params'] = best_params tune_dict['tune_time_model'] = (time.time() - start) if logger: logger.info(f''' best params: {tune_dict['best_params']}''') logger.info(f"tune time (model): {tune_dict['tune_time_model']:.3f}s") tune_dict['tune_time_extra'] = 0 if (model_type in ['ibug', 'knn']): best_params_wrapper = {} WrapperClass = (IBUGWrapper if (model_type == 'ibug') else KNNWrapper) model_val_wrapper = WrapperClass(scoring=scoring, variance_calibration=False, cond_mean_type=cond_mean_type, verbose=verbose, n_jobs=n_jobs, logger=logger) if logger: logger.info('\nTuning k and min. scale...') start = time.time() model_val_wrapper = model_val_wrapper.fit(model_val, X_tune, y_tune, X_val=X_val, y_val=y_val) best_params_wrapper = {'k': model_val_wrapper.k_, 'min_scale': model_val_wrapper.min_scale_, 'cond_mean_type': model_val_wrapper.cond_mean_type} if (model_type == 'knn'): best_params_wrapper['max_feat'] = model_val_wrapper.max_feat_ tune_dict['model_val_wrapper'] = model_val_wrapper tune_dict['best_params_wrapper'] = best_params_wrapper tune_dict['WrapperClass'] = WrapperClass tune_dict['tune_time_extra'] = (time.time() - start) if logger: logger.info(f'best params (wrapper): {best_params_wrapper}') logger.info(f"tune time (extra): {tune_dict['tune_time_extra']:.3f}s") if (model_type in ['ibug', 'knn']): (loc_val, scale_val) = (model_val_wrapper.loc_val_, model_val_wrapper.scale_val_) elif (model_type in ['constant', 'ngboost', 'pgbm', 'bart', 'cbu']): (loc_val, scale_val) = get_loc_scale(model_val, model_type, X=X_val, y_train=y_tune) tune_dict['loc_val'] = loc_val tune_dict['scale_val'] = scale_val return tune_dict
class AutoEncoder(nn.Module): def __init__(self): super(AutoEncoder, self).__init__() self.autoencoder = nn.Sequential(nn.Conv2d(4, 8, kernel_size=3, stride=1, padding=1), nn.LeakyReLU(inplace=True), nn.AvgPool2d(kernel_size=2, stride=2), nn.Conv2d(8, 12, kernel_size=3, stride=1, padding=1), nn.LeakyReLU(inplace=True), nn.AvgPool2d(kernel_size=2, stride=2), nn.Conv2d(12, 256, kernel_size=3, stride=1, padding=1), nn.LeakyReLU(inplace=True), nn.Upsample(scale_factor=2, mode='nearest'), nn.Conv2d(256, 128, kernel_size=3, stride=1, padding=1), nn.LeakyReLU(inplace=True), nn.Upsample(scale_factor=2, mode='nearest')) self.protocombiner = nn.Sequential(nn.Conv2d(131, 1, kernel_size=1, stride=1, padding=0), nn.Sigmoid()) def forward(self, imgs, same_proto, oppo_proto): x = torch.cat([imgs, same_proto], dim=1) x = self.autoencoder(x) rec_same = torch.cat([x, same_proto], dim=1) rec_oppo = torch.cat([x, oppo_proto], dim=1) return (self.protocombiner(rec_same), self.protocombiner(rec_oppo))
class SpeedController(PID): def __init__(self, params: Optional[PIDParam]=None): params = (SpeedControllerParam() if (params is None) else params) super(SpeedController, self).__init__(params) def from_vehicle_params(cls, model_param: ModelParameters) -> 'SpeedController': params = SpeedControllerParam(setpoint_minmax=model_param.vx_limits, output_minmax=model_param.acc_limits) return SpeedController(params)
def load_predictions(pred_path, gt_path, w2i_path): raw_preds = load_json(pred_path) gt_data = load_json(gt_path) word2idx = load_json(w2i_path) idx2word = {i: w for (w, i) in word2idx.items()} qid2ans = {int(e['qid']): int(e['answer_idx']) for e in gt_data} qid2bbox = {int(e['qid']): e['bbox'] for e in gt_data} bbox_preds = dict() for e in raw_preds['raw_bbox']: qid = None for i in range(5): if (len(e[str(i)]) > 0): qid = e[str(i)][0]['qid'] assert (qid is not None) ans_idx = qid2ans[int(qid)] cur_gt_bbox = qid2bbox[int(qid)] cur_correct_bbox_preds = e[str(ans_idx)] key_template = '{vid_name}_{qid}_{img_idx:05d}' for p in cur_correct_bbox_preds: annotated_word_ids = [(word2idx[clean_label(b['label'])] if (clean_label(b['label']) in word2idx) else word2idx['<unk>']) for b in cur_gt_bbox[str(p['img_idx'])]] collected_bbox = [] for (idx, b) in enumerate(p['bbox']): if (p['word'] in annotated_word_ids): collected_bbox.append([idx2word[p['word']], float(p['pred'][idx]), b]) key_str = key_template.format(vid_name=p['vid_name'], qid=qid, img_idx=p['img_idx']) if (key_str not in bbox_preds): bbox_preds[key_str] = [] bbox_preds[key_str].extend(collected_bbox) preds = dict(ts_answer=raw_preds['ts_answer'], bbox=bbox_preds) return preds
def tryLoad(name, default=None): try: import user_config except: return None if hasattr(user_config, name): return getattr(user_config, name) return default
class MaskedSeparableConv2D(_MaskedConv): def __init__(self, filters, kernel_size, strides=(1, 1), padding='valid', data_format='channels_last', dilation_rate=(1, 1), depth_multiplier=1, activation=None, use_bias=True, depthwise_initializer='global_uniform', pointwise_initializer='global_uniform', depthwise_regularizer=None, pointwise_regularizer=None, bias_initializer=init_ops.zeros_initializer(), bias_regularizer=None, activity_regularizer=None, trainable=True, name=None, task_id=1, **kwargs): super(MaskedSeparableConv2D, self).__init__(rank=2, filters=filters, kernel_size=kernel_size, strides=strides, padding=padding, data_format=data_format, dilation_rate=dilation_rate, activation=activation, use_bias=use_bias, bias_initializer=bias_initializer, bias_regularizer=bias_regularizer, activity_regularizer=activity_regularizer, trainable=trainable, name=name, task_id=task_id, **kwargs) self.depth_multiplier = depth_multiplier self.depthwise_initializer = depthwise_initializer self.pointwise_initializer = pointwise_initializer self.depthwise_regularizer = depthwise_regularizer self.pointwise_regularizer = pointwise_regularizer def build(self, input_shape): input_shape = tensor_shape.TensorShape(input_shape) channel_axis = (1 if (self.data_format == 'channels_first') else (- 1)) if (input_shape[channel_axis].value is None): raise ValueError('The channel dimension of the inputs should be defined. Found `None`.') input_dim = input_shape[channel_axis].value depthwise_kernel_shape = (self.kernel_size + (input_dim, self.depth_multiplier)) pointwise_kernel_shape = (((1,) * self.rank) + ((self.depth_multiplier * input_dim), self.filters)) self.depthwise_mask = self.add_variable(name='depthwise_mask', shape=depthwise_kernel_shape, initializer=init_ops.zeros_initializer(), trainable=False, dtype=tf.int8) self.depthwise_kernel = self.add_variable(name='depthwise_kernel', shape=depthwise_kernel_shape, initializer=self.depthwise_initializer, regularizer=self.depthwise_regularizer, trainable=True, dtype=self.dtype) self.depthwise_threshold = self.add_variable(name='depthwise_threshold', shape=[], initializer=init_ops.zeros_initializer(), trainable=False, dtype=self.dtype) self.pointwise_mask = self.add_variable(name='pointwise_mask', shape=pointwise_kernel_shape, initializer=init_ops.zeros_initializer(), trainable=False, dtype=tf.int8) self.pointwise_kernel = self.add_variable(name='pointwise_kernel', shape=pointwise_kernel_shape, initializer=self.pointwise_initializer, regularizer=self.pointwise_regularizer, trainable=True, dtype=self.dtype) self.pointwise_threshold = self.add_variable(name='pointwise_threshold', shape=[], initializer=init_ops.zeros_initializer(), trainable=False, dtype=self.dtype) if (FLAGS.reset_weights_in_new_locations and FLAGS.open_ratio): depthwise_conditional_op = change_mask_and_weight(self.depthwise_mask, self.depthwise_kernel, self.task_id, FLAGS.open_ratio, FLAGS.cell_scope_to_be_assigned_current_task_id) elif FLAGS.open_ratio: depthwise_conditional_op = change_mask(self.depthwise_mask, self.task_id, FLAGS.open_ratio, FLAGS.cell_scope_to_be_assigned_current_task_id) else: depthwise_conditional_op = control_flow_ops.no_op() if (FLAGS.reset_weights_in_new_locations and FLAGS.open_ratio): pointwise_conditional_op = change_mask_and_weight(self.pointwise_mask, self.pointwise_kernel, self.task_id, FLAGS.open_ratio, FLAGS.cell_scope_to_be_assigned_current_task_id) elif FLAGS.open_ratio: pointwise_conditional_op = change_mask(self.pointwise_mask, self.task_id, FLAGS.open_ratio, FLAGS.cell_scope_to_be_assigned_current_task_id) else: pointwise_conditional_op = control_flow_ops.no_op() with tf.control_dependencies([depthwise_conditional_op, pointwise_conditional_op]): if FLAGS.share_only_task_1: depthwise_boolean_mask = tf.cast(tf.logical_or(tf.equal(tf.identity(self.depthwise_mask), 1), tf.equal(tf.identity(self.depthwise_mask), self.task_id)), dtype=tf.float32) pointwise_boolean_mask = tf.cast(tf.logical_or(tf.equal(tf.identity(self.pointwise_mask), 1), tf.equal(tf.identity(self.pointwise_mask), self.task_id)), dtype=tf.float32) else: depthwise_boolean_mask = tf.cast(tf.logical_and(tf.greater_equal(tf.identity(self.depthwise_mask), 1), tf.less_equal(tf.identity(self.depthwise_mask), self.task_id)), dtype=tf.float32) pointwise_boolean_mask = tf.cast(tf.logical_and(tf.greater_equal(tf.identity(self.pointwise_mask), 1), tf.less_equal(tf.identity(self.pointwise_mask), self.task_id)), dtype=tf.float32) self.masked_depthwise_kernel = math_ops.multiply(depthwise_boolean_mask, self.depthwise_kernel, MASKED_WEIGHT_NAME) self.masked_pointwise_kernel = math_ops.multiply(pointwise_boolean_mask, self.pointwise_kernel, MASKED_WEIGHT_NAME) if (self.depthwise_mask not in ops.get_collection_ref(MASK_COLLECTION)): ops.add_to_collection(MASK_COLLECTION, self.depthwise_mask) ops.add_to_collection(MASK_COLLECTION, self.pointwise_mask) ops.add_to_collection(MASKED_WEIGHT_COLLECTION, self.masked_depthwise_kernel) ops.add_to_collection(MASKED_WEIGHT_COLLECTION, self.masked_pointwise_kernel) ops.add_to_collection(THRESHOLD_COLLECTION, self.depthwise_threshold) ops.add_to_collection(THRESHOLD_COLLECTION, self.pointwise_threshold) ops.add_to_collection(WEIGHT_COLLECTION, self.depthwise_kernel) ops.add_to_collection(WEIGHT_COLLECTION, self.pointwise_kernel) if self.use_bias: original_scope = self._scope with tf.variable_scope('task_{}'.format(self.task_id)) as scope: self._scope = scope self.bias = self.add_variable(name='bias', shape=(self.filters,), initializer=self.bias_initializer, regularizer=self.bias_regularizer, trainable=True, dtype=self.dtype) self._scope = original_scope else: self.bias = None self.input_spec = base.InputSpec(ndim=(self.rank + 2), axes={channel_axis: input_dim}) self.built = True def call(self, inputs): if (self.data_format == 'channels_last'): strides = (((1,) + self.strides) + (1,)) else: strides = ((1, 1) + self.strides) outputs = nn.separable_conv2d(inputs, self.masked_depthwise_kernel, self.masked_pointwise_kernel, strides=strides, padding=self.padding.upper(), rate=self.dilation_rate, data_format=utils.convert_data_format(self.data_format, ndim=4)) if (self.bias is not None): if (self.data_format == 'channels_first'): if (self.rank == 1): bias = array_ops.reshape(self.bias, (1, self.filters, 1)) outputs += bias if (self.rank == 2): outputs = nn.bias_add(outputs, self.bias, data_format='NCHW') if (self.rank == 3): outputs_shape = outputs.shape.as_list() outputs_4d = array_ops.reshape(outputs, [outputs_shape[0], outputs_shape[1], (outputs_shape[2] * outputs_shape[3]), outputs_shape[4]]) outputs_4d = nn.bias_add(outputs_4d, self.bias, data_format='NCHW') outputs = array_ops.reshape(outputs_4d, outputs_shape) else: outputs = nn.bias_add(outputs, self.bias, data_format='NHWC') if (self.activation is not None): return self.activation(outputs) return outputs
def default_install_dir(target_abi): install_dir = '/tmp/mace_run' if ((target_abi == 'armeabi-v7a') or (target_abi == 'arm64-v8a')): install_dir = '/data/local/tmp/mace_run' return install_dir
class PruningCallbacks(BaseCallbacks): def __init__(self, conf=None, model=None): super(PruningCallbacks, self).__init__(conf=conf, model=model) self.pruners_info = process_config(self.conf) self.pruners = [] self._generate_pruners() self.generate_hooks() def on_train_end(self): for on_train_end_hook in self.hooks_dict['on_train_end']: on_train_end_hook() if ((self.conf.framework == 'pytorch') and isinstance(self.model.model, torch.nn.Module)): get_sparsity_ratio(self.pruners, self.model) elif ((self.conf.framework == 'keras') and isinstance(self.model.model, tf.keras.Model)): get_sparsity_ratio_tf(self.pruners, self.model) def __repr__(self): return 'Pruning Callbacks' def generate_hooks(self): for pruner in self.pruners: for key in self.hooks.keys(): if hasattr(pruner, key): self.register_hook(key, getattr(pruner, key)) def _generate_pruners(self): if ((self.conf.framework == 'pytorch') and isinstance(self.model.model, torch.nn.Module)): from .pruner.model_slim.pattern_analyzer import SelfMHASearcher for info in self.pruners_info: if ('mha' in info['pattern']): pa_obj = SelfMHASearcher(self.model.model) (modules, _) = pa_obj.search(split_qkv_ffn=False) modules = pa_obj.obtain_mha_module(modules) modules = pa_obj.from_layer_name_to_object(modules) if (len(modules) == 0): logger.warning('one pruner hooks no mha modules, please have a check') self.pruners.append(get_pruner(info, modules)) else: modules = parse_to_prune(info, self.model.model) if (modules == {}): logger.warning('one pruner hooks no layers, please have a check') self.pruners.append(get_pruner(info, modules)) info['modules'] = [key for key in modules.keys()] info['len_of_modules'] = len(info['modules']) logger.info(info) elif ((self.conf.framework == 'keras') and isinstance(self.model.model, tf.keras.Model)): from tensorflow.python.ops.numpy_ops import np_config np_config.enable_numpy_behavior() for info in self.pruners_info: modules = parse_to_prune_tf(info, self.model.model) if (modules == {}): logger.warning('one pruner hooks no layers, please have a check') self.pruners.append(get_pruner(info, modules, 'keras')) info['modules'] = [key for key in modules.keys()] info['len_of_modules'] = len(info['modules']) logger.info(info) else: assert False, 'now only support {}'.format(PRUNERS.keys())
def witness_set_of_hypersurface(nvar, hpol, precision='d'): if (precision == 'd'): from phcpy.phcpy2c3 import py2c_standard_witset_of_hypersurface from phcpy.interface import load_standard_system from phcpy.interface import load_standard_solutions py2c_standard_witset_of_hypersurface(nvar, len(hpol), hpol) return (load_standard_system(), load_standard_solutions()) elif (precision == 'dd'): from phcpy.phcpy2c3 import py2c_dobldobl_witset_of_hypersurface from phcpy.interface import load_dobldobl_system from phcpy.interface import load_dobldobl_solutions py2c_dobldobl_witset_of_hypersurface(nvar, len(hpol), hpol) return (load_dobldobl_system(), load_dobldobl_solutions()) elif (precision == 'qd'): from phcpy.phcpy2c3 import py2c_quaddobl_witset_of_hypersurface from phcpy.interface import load_quaddobl_system from phcpy.interface import load_quaddobl_solutions py2c_quaddobl_witset_of_hypersurface(nvar, len(hpol), hpol) return (load_quaddobl_system(), load_quaddobl_solutions()) else: print('wrong argument for precision') return None
class MeshRenderer(nn.Module): def __init__(self, rasterize_fov, znear=0.1, zfar=10, rasterize_size=224): super(MeshRenderer, self).__init__() x = (np.tan(np.deg2rad((rasterize_fov * 0.5))) * znear) self.ndc_proj = torch.tensor(ndc_projection(x=x, n=znear, f=zfar)).matmul(torch.diag(torch.tensor([1.0, (- 1), (- 1), 1]))) self.rasterize_size = rasterize_size self.glctx = None def forward(self, vertex, tri, feat=None): device = vertex.device rsize = int(self.rasterize_size) ndc_proj = self.ndc_proj.to(device) if (vertex.shape[(- 1)] == 3): vertex = torch.cat([vertex, torch.ones([*vertex.shape[:2], 1]).to(device)], dim=(- 1)) vertex[(..., 1)] = (- vertex[(..., 1)]) vertex_ndc = (vertex ndc_proj.t()) if (self.glctx is None): self.glctx = dr.RasterizeGLContext(device=device) print(('create glctx on device cuda:%d' % device.index)) ranges = None if (isinstance(tri, List) or (len(tri.shape) == 3)): vum = vertex_ndc.shape[1] fnum = torch.tensor([f.shape[0] for f in tri]).unsqueeze(1).to(device) fstartidx = (torch.cumsum(fnum, dim=0) - fnum) ranges = torch.cat([fstartidx, fnum], axis=1).type(torch.int32).cpu() for i in range(tri.shape[0]): tri[i] = (tri[i] + (i * vum)) vertex_ndc = torch.cat(vertex_ndc, dim=0) tri = torch.cat(tri, dim=0) tri = tri.type(torch.int32).contiguous() (rast_out, _) = dr.rasterize(self.glctx, vertex_ndc.contiguous(), tri, resolution=[rsize, rsize], ranges=ranges) (depth, _) = dr.interpolate(vertex.reshape([(- 1), 4])[(..., 2)].unsqueeze(1).contiguous(), rast_out, tri) depth = depth.permute(0, 3, 1, 2) mask = (rast_out[(..., 3)] > 0).float().unsqueeze(1) depth = (mask * depth) image = None if (feat is not None): (image, _) = dr.interpolate(feat, rast_out, tri) image = image.permute(0, 3, 1, 2) image = (mask * image) return (mask, depth, image)
def test_func(x): print('Running test_func') p = mp.current_process() y = ((x * x) if (p.runner is None) else x) print(y)
def blh2xyz(latitude, longitude, height): latitude = math.radians(latitude) longitude = math.radians(longitude) e = math.sqrt((1 - ((B ** 2) / (A ** 2)))) N = (A / math.sqrt((1 - ((e ** 2) * (math.sin(latitude) ** 2))))) X = (((N + height) * math.cos(latitude)) * math.cos(longitude)) Y = (((N + height) * math.cos(latitude)) * math.sin(longitude)) Z = (((N * (1 - (e ** 2))) + height) * math.sin(latitude)) return (X, Y, Z)
def interpolate_img(img, coords, order=3, mode='nearest', cval=0.0): return map_coordinates(img, coords, order=order, mode=mode, cval=cval)
class ResNet(nn.Module): __factory = {18: torchvision.models.resnet18, 34: torchvision.models.resnet34, 50: torchvision.models.resnet50, 101: torchvision.models.resnet101, 152: torchvision.models.resnet152} def __init__(self, depth, pretrained=True, cut_at_pooling=False, num_features=0, norm=False, dropout=0, num_classes=0, FCN=False, T=1, dim=256, num_parts=6): super(ResNet, self).__init__() self.depth = depth self.pretrained = pretrained self.cut_at_pooling = cut_at_pooling self.FCN = FCN self.T = T self.reduce_dim = dim self.num_parts = num_parts if (depth not in ResNet.__factory): raise KeyError('Unsupported depth:', depth) self.base = ResNet.__factory[depth](pretrained=pretrained) if self.FCN: self.base.layer4[0].conv2.stride = (1, 1) self.base.layer4[0].downsample[0].stride = (1, 1) self.num_features = num_features self.num_classes = num_classes self.dropout = dropout self.instance = nn.ModuleList() for i in range((self.num_parts + 1)): local_conv = nn.Linear(2048, self.num_features, bias=False) init.kaiming_normal_(local_conv.weight, mode='fan_out') local_bn = nn.BatchNorm1d(self.num_features) init.constant_(local_bn.weight, 1) init.constant_(local_bn.bias, 0) fc = AngleLinear(self.num_features, self.num_classes) self.instance.append(nn.Sequential(nn.Dropout(self.dropout), local_conv, local_bn, fc)) self.drop = nn.Dropout(self.dropout) self.local_mask = nn.Conv2d(self.reduce_dim, self.num_parts, kernel_size=1, bias=True) init.kaiming_normal_(self.local_mask.weight, mode='fan_out') init.constant_(self.local_mask.bias, 0) elif (not self.cut_at_pooling): self.num_features = num_features self.norm = norm self.dropout = dropout self.has_embedding = (num_features > 0) self.num_classes = num_classes out_planes = self.base.fc.in_features if self.has_embedding: self.feat = nn.Linear(out_planes, self.num_features, bias=False) self.feat_bn = nn.BatchNorm1d(self.num_features) init.kaiming_normal_(self.feat.weight, mode='fan_out') init.constant_(self.feat_bn.weight, 1) init.constant_(self.feat_bn.bias, 0) else: self.num_features = out_planes if (self.dropout > 0): self.drop = nn.Dropout(self.dropout) if (self.num_classes > 0): self.classifier = nn.Linear(self.num_features, self.num_classes) init.normal_(self.classifier.weight, std=0.001) init.constant_(self.classifier.bias, 0) if (not self.pretrained): self.reset_params() def forward(self, inputs, part_labels=None): x = inputs if (part_labels is None): tmp = torch.FloatTensor(range(1, 25)) tmp = ((tmp - 0.1) / 4).int() part_labels = tmp.unsqueeze(0).expand(inputs.size(0), tmp.size(0)) part_labels = torch.autograd.Variable(part_labels.cuda()) for (name, module) in self.base._modules.items(): if (name == 'avgpool'): break x = module(x) if self.cut_at_pooling: return x if self.FCN: T = self.T y = self.drop(x).unsqueeze(1) stride = (2048 // self.reduce_dim) y = F.avg_pool3d(y, kernel_size=(stride, 1, 8), stride=(stride, 1, 8)).squeeze(1) x_global = F.avg_pool2d(x, (24, 8)) local_score = self.local_mask(y) local_score = local_score.squeeze(3) score = F.softmax((1 * local_score.detach()), 1) pscore = score.sum(2) score = (score / pscore.unsqueeze(2).expand_as(score)) (bb, cc, hh, ww) = x.size() feat = (x.unsqueeze(2).expand(bb, cc, self.num_parts, hh, ww) * score.unsqueeze(1).unsqueeze(4).expand(bb, cc, self.num_parts, hh, ww)) feat = feat.sum(4).sum(3).unsqueeze(3) x = feat out0 = x.view(x.size(0), (- 1)) out0 = (x / torch.clamp(x.norm(2, 1).unsqueeze(1).expand_as(x), min=1e-12)) x_list = list(x.chunk(x.size(2), 2)) x_list.append(x_global) c = [] for (tensor, branch) in zip(x_list, self.instance): tensor = tensor.contiguous().view(tensor.size(0), (- 1)) c.append(branch(tensor)) ps = local_score return (out0, c, ps, pscore) x = F.avg_pool2d(x, x.size()[2:]) x = x.view(x.size(0), (- 1)) out1 = x out1 = (x / x.norm(2, 1).unsqueeze(1).expand_as(x)) if self.has_embedding: x = self.feat(x) x = self.feat_bn(x) out2 = (x / x.norm(2, 1).unsqueeze(1).expand_as(x)) if self.norm: x = (x / x.norm(2, 1).unsqueeze(1).expand_as(x)) if (self.dropout > 0): x = self.drop(x) if (self.num_classes > 0): x = self.classifier(x) return (out2, x) def reset_params(self): for m in self.modules(): if isinstance(m, nn.Conv2d): init.kaiming_normal_(m.weight, mode='fan_out') if (m.bias is not None): init.constant_(m.bias, 0) elif isinstance(m, nn.BatchNorm2d): init.constant_(m.weight, 1) init.constant_(m.bias, 0) elif isinstance(m, nn.Linear): init.normal_(m.weight, std=0.001) if (m.bias is not None): init.constant_(m.bias, 0)
def load_results(sent_lst, tokenizer): full_result_dict = {} failed_instances = [] found_idx = [] sent_lst_lst = list(sent_lst.items()) for (idx, (key, val)) in enumerate(sent_lst_lst): if (idx in full_result_dict.keys()): continue word_lst1 = [x.text for x in tokenizer(val['obs1'])] word_lst2 = [x.text for x in tokenizer(val['obs2'])] target_file = f'{INPUT_PATH}_*_{SPLIT}_{idx}.json' file_lst = glob.glob(target_file) try: assert (len(file_lst) == 1) except: print('the file must have existed in a batched version') target_file = f'{INPUT_PATH}_*_{idx}.json' file_lst = glob.glob(target_file) print(file_lst, target_file) print(file_lst) target_file = file_lst[0] if ('x128' in target_file): infill_lst = [] with open(target_file, 'r') as f: for line in f: example = json.loads(line)[0] infill_ = example.split()[len(word_lst1):(- len(word_lst2))] infill_ = ' '.join(infill_) infill_lst.append(infill_) result_dict = {'pred_samples': infill_lst, 'sample': None, 'obs1': val['obs1'], 'obs2': val['obs2']} full_result_dict[idx] = result_dict else: with open(target_file, 'r') as f: for line in f: example = ast.literal_eval(line.strip()) (index, template) = list(example.keys())[0] print(index, idx) if (int(index) < int(idx)): continue assert (int(index) == int(idx)) found_idx.append(idx) example = list(example.values())[0] (kk, val) = sent_lst_lst[idx] word_lst1 = [x.text for x in tokenizer(val['obs1'])] word_lst2 = [x.text for x in tokenizer(val['obs2'])] infill_lst = [' '.join(xx.split()[len(word_lst1):(- len(word_lst2))]) for xx in example] result_dict = {'pred_samples': infill_lst, 'sample': None, 'obs1': val['obs1'], 'obs2': val['obs2']} full_result_dict[idx] = result_dict idx += 1 with open('full_diff_test_outputs_aug.json', 'w') as f: json.dump(full_result_dict, f) return full_result_dict
def mood(sentence, **kwargs): if isinstance(sentence, str): try: from pattern.en import parse, Sentence sentence = Sentence(parse(sentence)) except ImportError: pass if imperative(sentence, **kwargs): return IMPERATIVE if conditional(sentence, **kwargs): return CONDITIONAL if subjunctive(sentence, **kwargs): return SUBJUNCTIVE else: return INDICATIVE
class Attribute(JsonSerializer): def __init__(self, name: str, attribute_type: str, value: Any): super().__init__() self.name = name self.attribute_type = attribute_type self.value = value
class Generic_MIL_Dataset(Generic_WSI_Classification_Dataset): def __init__(self, data_dir, **kwargs): super().__init__(**kwargs) self.data_dir = data_dir self.use_h5 = False def load_from_h5(self, toggle): self.use_h5 = toggle def __getitem__(self, idx): import h5py slide_id = self.slide_data['slide'][idx] label = self.slide_data['label'][idx] if (type(self.data_dir) == dict): source = self.slide_data['source'][idx] data_dir = self.data_dir[source] else: data_dir = self.data_dir if (not self.use_h5): if self.data_dir: full_path = os.path.join(data_dir, f'{slide_id}.pt') features = torch.load(full_path) if self.lasthalf: features = torch.split(features, 1024, dim=1)[1] return (features, label) else: return (slide_id, label) else: full_path = os.path.join(data_dir, 'h5_files', f'{slide_id}.h5') with h5py.File(full_path, 'r') as hdf5_file: features = hdf5_file['features'][:] coords = hdf5_file['coords'][:] if self.lasthalf: features = torch.from_numpy(features.split(2)[1]) else: features = torch.from_numpy(features) return (features, label, coords)
class PowerTransformerComponent(Rescaling, AutotabularPreprocessingAlgorithm): def __init__(self, random_state: Optional[np.random.RandomState]=None): from sklearn.preprocessing import PowerTransformer self.preprocessor = PowerTransformer(copy=False) def get_properties(dataset_properties: Optional[DATASET_PROPERTIES_TYPE]=None) -> Dict[(str, Optional[Union[(str, int, bool, Tuple)]])]: return {'shortname': 'PowerTransformer', 'name': 'PowerTransformer', 'handles_missing_values': False, 'handles_nominal_values': False, 'handles_numerical_features': True, 'prefers_data_scaled': False, 'prefers_data_normalized': False, 'handles_regression': True, 'handles_classification': True, 'handles_multiclass': True, 'handles_multilabel': True, 'handles_multioutput': True, 'is_deterministic': True, 'handles_sparse': False, 'handles_dense': True, 'input': (DENSE, UNSIGNED_DATA), 'output': (INPUT,), 'preferred_dtype': None}
class Timer(): def __init__(self, keys): self.keys = keys self.n = {} self.running_time = {} self.total_time = {} self.reset() def start(self, key): self.running_time[key] = time.time() return self def stop(self, key): self.total_time[key] = (time.time() - self.running_time[key]) self.n[key] += 1 self.running_time[key] = None return self def reset(self): for k in self.keys: self.total_time[k] = 0 self.running_time[k] = None self.n[k] = 0 return self def value(self): vals = {} for k in self.keys: if (self.n[k] == 0): raise ValueError('Trying to divide by zero in TimeMeter') else: vals[k] = (self.total_time[k] / self.n[k]) return vals
def _loads(s, *, fix_imports=True, encoding='ASCII', errors='strict', buffers=None): if isinstance(s, str): raise TypeError("Can't load pickle from unicode string") file = io.BytesIO(s) return _Unpickler(file, fix_imports=fix_imports, buffers=buffers, encoding=encoding, errors=errors).load()
def main(): parser = argparse.ArgumentParser(description='Diffuser Pipeline for processing images with prompts.') parser.add_argument('--prompt_path', type=str, default='dataset/animal.json', help='Path to the JSON file containing prompts.') parser.add_argument('--save_path', type=str, default='train_set/animal/', help='Path to save processed images.') args = parser.parse_args() pipe = DiffusionPipeline.from_pretrained('stabilityai/stable-diffusion-xl-base-1.0', torch_dtype=torch.float16, use_safetensors=True, variant='fp16') pipe.to('cuda') with open(args.prompt_path, 'r') as f: data = json.load(f) for info in data: img = info['image'].split('/')[1] prompt = info['prompt'] images = pipe(prompt=prompt).images[0] images.save((args.save_path + img))
class Task(NamedTuple): video_name: str video_path: str out_path: str min_frame: int max_frame: int target_fps: float target_num_frames: int width: int height: int max_height: int
def evaluate(args): with open(args.data, 'rb') as f: test_dataset: SNLIDataset = pickle.load(f) word_vocab = test_dataset.word_vocab label_vocab = test_dataset.label_vocab model = SNLIModel(num_classes=len(label_vocab), num_words=len(word_vocab), word_dim=args.word_dim, hidden_dim=args.hidden_dim, clf_hidden_dim=args.clf_hidden_dim, clf_num_layers=args.clf_num_layers, use_leaf_rnn=args.leaf_rnn, intra_attention=args.intra_attention, use_batchnorm=args.batchnorm, dropout_prob=args.dropout, bidirectional=args.bidirectional) num_params = sum((np.prod(p.size()) for p in model.parameters())) num_embedding_params = np.prod(model.word_embedding.weight.size()) print(f'# of parameters: {num_params}') print(f'# of word embedding parameters: {num_embedding_params}') print(f'# of parameters (excluding word embeddings): {(num_params - num_embedding_params)}') model.load_state_dict(torch.load(args.model, map_location='cpu')) model.eval() model.to(args.device) torch.set_grad_enabled(False) test_data_loader = DataLoader(dataset=test_dataset, batch_size=args.batch_size, collate_fn=test_dataset.collate) num_correct = 0 num_data = len(test_dataset) for batch in test_data_loader: pre = batch['pre'].to(args.device) hyp = batch['hyp'].to(args.device) pre_length = batch['pre_length'].to(args.device) hyp_length = batch['hyp_length'].to(args.device) label = batch['label'].to(args.device) logits = model(pre=pre, pre_length=pre_length, hyp=hyp, hyp_length=hyp_length) label_pred = logits.max(1)[1] num_correct_batch = torch.eq(label, label_pred).long().sum() num_correct_batch = num_correct_batch.item() num_correct += num_correct_batch print(f'# data: {num_data}') print(f'# correct: {num_correct}') print(f'Accuracy: {(num_correct / num_data):.4f}')
_model('s2t_transformer_w2v2') class S2TTransformerModelW2V2(FairseqEncoderDecoderModel): def __init__(self, encoder, decoder): super().__init__(encoder, decoder) def add_args(parser): parser.add_argument('--conv-kernel-sizes', type=str, metavar='N', help='kernel sizes of Conv1d subsampling layers') parser.add_argument('--conv-channels', type=int, metavar='N', help='# of channels in Conv1d subsampling layers') parser.add_argument('--activation-fn', type=str, default='relu', choices=utils.get_available_activation_fns(), help='activation function to use') parser.add_argument('--dropout', type=float, metavar='D', help='dropout probability') parser.add_argument('--attention-dropout', type=float, metavar='D', help='dropout probability for attention weights') parser.add_argument('--activation-dropout', '--relu-dropout', type=float, metavar='D', help='dropout probability after activation in FFN.') parser.add_argument('--encoder-embed-dim', type=int, metavar='N', help='encoder embedding dimension') parser.add_argument('--encoder-ffn-embed-dim', type=int, metavar='N', help='encoder embedding dimension for FFN') parser.add_argument('--translation-encoder-layers', type=int, metavar='N', help='num translation encoder layers') parser.add_argument('--encoder-attention-heads', type=int, metavar='N', help='num encoder attention heads') parser.add_argument('--encoder-normalize-before', action='store_true', help='apply layernorm before each encoder block') parser.add_argument('--decoder-embed-dim', type=int, metavar='N', help='decoder embedding dimension') parser.add_argument('--decoder-ffn-embed-dim', type=int, metavar='N', help='decoder embedding dimension for FFN') parser.add_argument('--decoder-layers', type=int, metavar='N', help='num decoder layers') parser.add_argument('--decoder-attention-heads', type=int, metavar='N', help='num decoder attention heads') parser.add_argument('--decoder-normalize-before', action='store_true', help='apply layernorm before each decoder block') parser.add_argument('--share-decoder-input-output-embed', action='store_true', help='share decoder input and output embeddings') parser.add_argument('--layernorm-embedding', action='store_true', help='add layernorm to embedding') parser.add_argument('--no-scale-embedding', action='store_true', help='if True, dont scale embeddings') parser.add_argument('--load-pretrained-mt-encoder-decoder-from', type=str, metavar='STR', help='model to take mt encoder/decoder weights from (for initialization)') parser.add_argument('--encoder-freezing-updates', type=int, metavar='N', help='freeze encoder for first N updates') parser.add_argument('--mixup', action='store_true', help='if mix input of translation encoder') parser.add_argument('--mixup-arguments', type=str, metavar='STR', help='arguments for adjusting the probability p of mixup') parser.add_argument('--w2v2-model-path', type=str, metavar='STR', help='path/to/wav2vec/model') parser.add_argument('--freeze-w2v', action='store_true', help='if we want to freeze the w2v features') def build_encoder(cls, args, task=None, embed_tokens=None): return S2TTransformerEncoderW2V2(args, task.target_dictionary, embed_tokens) def build_decoder(cls, args, task, embed_tokens): return TransformerDecoderScriptable(args, task.target_dictionary, embed_tokens) def build_model(cls, args, task): base_architecture(args) def build_embedding(dictionary, embed_dim): num_embeddings = len(dictionary) padding_idx = dictionary.pad() return Embedding(num_embeddings, embed_dim, padding_idx) decoder_embed_tokens = build_embedding(task.target_dictionary, args.decoder_embed_dim) encoder_embed_tokens = decoder_embed_tokens encoder = cls.build_encoder(args, task, encoder_embed_tokens) decoder = cls.build_decoder(args, task, decoder_embed_tokens) mt_pretraining_path = getattr(args, 'load_pretrained_mt_encoder_decoder_from', None) if ((mt_pretraining_path is not None) and Path(mt_pretraining_path).exists()): mt_state = checkpoint_utils.load_checkpoint_to_cpu(mt_pretraining_path) mt_encoder_state_dict = OrderedDict() mt_decoder_state_dict = OrderedDict() for key in mt_state['model'].keys(): if key.startswith('encoder'): subkey = key[(len('encoder') + 1):] mt_encoder_state_dict[subkey] = mt_state['model'][key] if key.startswith('decoder'): subkey = key[(len('decoder') + 1):] mt_decoder_state_dict[subkey] = mt_state['model'][key] encoder.load_state_dict(mt_encoder_state_dict, strict=False) decoder.load_state_dict(mt_decoder_state_dict, strict=False) logger.info(f'loaded pretrained mt encoder and decoder from: {mt_pretraining_path}') return cls(encoder, decoder) def get_normalized_probs(self, net_output: Tuple[(Tensor, Optional[Dict[(str, List[Optional[Tensor]])]])], log_probs: bool, sample: Optional[Dict[(str, Tensor)]]=None): lprobs = self.get_normalized_probs_scriptable(net_output, log_probs, sample) lprobs.batch_first = True return lprobs def forward(self, audio, audio_lengths, source, source_lengths, prev_output_tokens, align_pad, align_lengths): encoder_out = self.encoder(audio, audio_lengths, source, source_lengths, align_pad, align_lengths) decoder_out = self.decoder(prev_output_tokens=prev_output_tokens, encoder_out=encoder_out) return decoder_out
class SegNet(nn.Module): def __init__(self, input_nbr=3, label_nbr=22): super(SegNet, self).__init__() batchNorm_momentum = 0.1 self.conv11 = nn.Conv2d(input_nbr, 64, kernel_size=3, padding=1) self.bn11 = nn.BatchNorm2d(64, momentum=batchNorm_momentum) self.conv12 = nn.Conv2d(64, 64, kernel_size=3, padding=1) self.bn12 = nn.BatchNorm2d(64, momentum=batchNorm_momentum) self.conv21 = nn.Conv2d(64, 128, kernel_size=3, padding=1) self.bn21 = nn.BatchNorm2d(128, momentum=batchNorm_momentum) self.conv22 = nn.Conv2d(128, 128, kernel_size=3, padding=1) self.bn22 = nn.BatchNorm2d(128, momentum=batchNorm_momentum) self.conv31 = nn.Conv2d(128, 256, kernel_size=3, padding=1) self.bn31 = nn.BatchNorm2d(256, momentum=batchNorm_momentum) self.conv32 = nn.Conv2d(256, 256, kernel_size=3, padding=1) self.bn32 = nn.BatchNorm2d(256, momentum=batchNorm_momentum) self.conv33 = nn.Conv2d(256, 256, kernel_size=3, padding=1) self.bn33 = nn.BatchNorm2d(256, momentum=batchNorm_momentum) self.conv41 = nn.Conv2d(256, 512, kernel_size=3, padding=1) self.bn41 = nn.BatchNorm2d(512, momentum=batchNorm_momentum) self.conv42 = nn.Conv2d(512, 512, kernel_size=3, padding=1) self.bn42 = nn.BatchNorm2d(512, momentum=batchNorm_momentum) self.conv43 = nn.Conv2d(512, 512, kernel_size=3, padding=1) self.bn43 = nn.BatchNorm2d(512, momentum=batchNorm_momentum) self.conv51 = nn.Conv2d(512, 512, kernel_size=3, padding=1) self.bn51 = nn.BatchNorm2d(512, momentum=batchNorm_momentum) self.conv52 = nn.Conv2d(512, 512, kernel_size=3, padding=1) self.bn52 = nn.BatchNorm2d(512, momentum=batchNorm_momentum) self.conv53 = nn.Conv2d(512, 512, kernel_size=3, padding=1) self.bn53 = nn.BatchNorm2d(512, momentum=batchNorm_momentum) self.conv53d = nn.Conv2d(512, 512, kernel_size=3, padding=1) self.bn53d = nn.BatchNorm2d(512, momentum=batchNorm_momentum) self.conv52d = nn.Conv2d(512, 512, kernel_size=3, padding=1) self.bn52d = nn.BatchNorm2d(512, momentum=batchNorm_momentum) self.conv51d = nn.Conv2d(512, 512, kernel_size=3, padding=1) self.bn51d = nn.BatchNorm2d(512, momentum=batchNorm_momentum) self.conv43d = nn.Conv2d(512, 512, kernel_size=3, padding=1) self.bn43d = nn.BatchNorm2d(512, momentum=batchNorm_momentum) self.conv42d = nn.Conv2d(512, 512, kernel_size=3, padding=1) self.bn42d = nn.BatchNorm2d(512, momentum=batchNorm_momentum) self.conv41d = nn.Conv2d(512, 256, kernel_size=3, padding=1) self.bn41d = nn.BatchNorm2d(256, momentum=batchNorm_momentum) self.conv33d = nn.Conv2d(256, 256, kernel_size=3, padding=1) self.bn33d = nn.BatchNorm2d(256, momentum=batchNorm_momentum) self.conv32d = nn.Conv2d(256, 256, kernel_size=3, padding=1) self.bn32d = nn.BatchNorm2d(256, momentum=batchNorm_momentum) self.conv31d = nn.Conv2d(256, 128, kernel_size=3, padding=1) self.bn31d = nn.BatchNorm2d(128, momentum=batchNorm_momentum) self.conv22d = nn.Conv2d(128, 128, kernel_size=3, padding=1) self.bn22d = nn.BatchNorm2d(128, momentum=batchNorm_momentum) self.conv21d = nn.Conv2d(128, 64, kernel_size=3, padding=1) self.bn21d = nn.BatchNorm2d(64, momentum=batchNorm_momentum) self.conv12d = nn.Conv2d(64, 64, kernel_size=3, padding=1) self.bn12d = nn.BatchNorm2d(64, momentum=batchNorm_momentum) self.conv11d = nn.Conv2d(64, label_nbr, kernel_size=3, padding=1) def forward(self, x): x11 = F.relu(self.bn11(self.conv11(x))) x12 = F.relu(self.bn12(self.conv12(x11))) (x1p, id1) = F.max_pool2d(x12, kernel_size=2, stride=2, return_indices=True) x21 = F.relu(self.bn21(self.conv21(x1p))) x22 = F.relu(self.bn22(self.conv22(x21))) (x2p, id2) = F.max_pool2d(x22, kernel_size=2, stride=2, return_indices=True) x31 = F.relu(self.bn31(self.conv31(x2p))) x32 = F.relu(self.bn32(self.conv32(x31))) x33 = F.relu(self.bn33(self.conv33(x32))) (x3p, id3) = F.max_pool2d(x33, kernel_size=2, stride=2, return_indices=True) x41 = F.relu(self.bn41(self.conv41(x3p))) x42 = F.relu(self.bn42(self.conv42(x41))) x43 = F.relu(self.bn43(self.conv43(x42))) (x4p, id4) = F.max_pool2d(x43, kernel_size=2, stride=2, return_indices=True) x51 = F.relu(self.bn51(self.conv51(x4p))) x52 = F.relu(self.bn52(self.conv52(x51))) x53 = F.relu(self.bn53(self.conv53(x52))) (x5p, id5) = F.max_pool2d(x53, kernel_size=2, stride=2, return_indices=True) x5d = F.max_unpool2d(x5p, id5, kernel_size=2, stride=2) x53d = F.relu(self.bn53d(self.conv53d(x5d))) x52d = F.relu(self.bn52d(self.conv52d(x53d))) x51d = F.relu(self.bn51d(self.conv51d(x52d))) x4d = F.max_unpool2d(x51d, id4, kernel_size=2, stride=2) x43d = F.relu(self.bn43d(self.conv43d(x4d))) x42d = F.relu(self.bn42d(self.conv42d(x43d))) x41d = F.relu(self.bn41d(self.conv41d(x42d))) x3d = F.max_unpool2d(x41d, id3, kernel_size=2, stride=2) x33d = F.relu(self.bn33d(self.conv33d(x3d))) x32d = F.relu(self.bn32d(self.conv32d(x33d))) x31d = F.relu(self.bn31d(self.conv31d(x32d))) x2d = F.max_unpool2d(x31d, id2, kernel_size=2, stride=2) x22d = F.relu(self.bn22d(self.conv22d(x2d))) x21d = F.relu(self.bn21d(self.conv21d(x22d))) x1d = F.max_unpool2d(x21d, id1, kernel_size=2, stride=2) x12d = F.relu(self.bn12d(self.conv12d(x1d))) x11d = self.conv11d(x12d) return x11d
class DataCfg(TypedDict): type: str mode: str size: Sequence[int] supp_idxs: Optional[Sequence[int]] use_depth: Optional[bool] use_hints: Optional[bool] use_benchmark: Optional[bool] use_strong_aug: Optional[bool] as_torch: Optional[bool] use_aug: Optional[bool] log_time: Optional[bool] train: Optional['DataCfg'] val: Optional['DataCfg'] test: Optional['DataCfg']
def load_local_or_remote_file(filepath, file_type=None): local_path = local_path_from_s3_or_local_path(filepath) if (file_type is None): extension = local_path.split('.')[(- 1)] if (extension == 'npy'): file_type = NUMPY else: file_type = PICKLE else: file_type = PICKLE if (file_type == NUMPY): object = np.load(open(local_path, 'rb')) elif (file_type == JOBLIB): object = joblib.load(local_path) else: object = pickle.load(open(local_path, 'rb')) print('loaded', local_path) return object
class FuncNonContiguousArgs(): def forward(self, input_ids, some_other_args, token_type_ids, attention_mask): return None
def set_severity(args): if (args.dataset != 'kitti'): args.severity = args.robustness_severities[args.severity_idx] return True if (args.task == 'initial'): args.severity = '' return True if (args.severity_idx < len(globals.KITTI_SEVERITIES[args.task])): args.severity = globals.KITTI_SEVERITIES[args.task][args.severity_idx] return True return False
def inverse_dict(d): assert (len(d.keys()) == len(set(d.keys()))) return {v: k for (k, v) in d.items()}
def get_model_list(): _start_prompt = ' Transformers currently provides the following architectures' _end_prompt = '1. Want to contribute a new model?' with open(os.path.join(REPO_PATH, 'README.md'), 'r', encoding='utf-8', newline='\n') as f: lines = f.readlines() start_index = 0 while (not lines[start_index].startswith(_start_prompt)): start_index += 1 start_index += 1 result = [] current_line = '' end_index = start_index while (not lines[end_index].startswith(_end_prompt)): if lines[end_index].startswith('1.'): if (len(current_line) > 1): result.append(current_line) current_line = lines[end_index] elif (len(lines[end_index]) > 1): current_line = f'{current_line[:(- 1)]} {lines[end_index].lstrip()}' end_index += 1 if (len(current_line) > 1): result.append(current_line) return ''.join(result)
class EagerModeCtx(): def __init__(self, eagerly: bool) -> None: assert isinstance(eagerly, bool), f'argument eagerly should not be {eagerly.__class__}. It must be a boolean.' self.eagerly = eagerly def __enter__(self) -> None: self.old_mode = tf.config.functions_run_eagerly() tf.config.run_functions_eagerly(self.eagerly) def __exit__(self, exc_type, exc_val, exc_tb) -> None: tf.config.run_functions_eagerly(self.old_mode)
def parse_args(): parser = argparse.ArgumentParser() parser.add_argument('--data_name', default='gsm8k', type=str) parser.add_argument('--model_name_or_path', default='gpt-4', type=str) parser.add_argument('--prompt_type', default='pal', type=str) parser.add_argument('--split', default='test', type=str) parser.add_argument('--num_test_sample', default=(- 1), type=int) parser.add_argument('--seed', default=0, type=int) parser.add_argument('--start', default=0, type=int) parser.add_argument('--end', default=(- 1), type=int) parser.add_argument('--temperature', default=0, type=float) parser.add_argument('--n_sampling', default=1, type=int) parser.add_argument('--top_p', default=0.95, type=float) parser.add_argument('--shuffle', action='store_true') parser.add_argument('--use_train_prompt_format', action='store_true') parser.add_argument('--code_concat', action='store_true') parser.add_argument('--code_exec_warning', action='store_true') parser.add_argument('--max_func_call', default=4, type=int) parser.add_argument('--max_code_fix_retries', default=4, type=int) parser.add_argument('--verbose', action='store_true') args = parser.parse_args() args.top_p = (1 if (args.temperature == 0) else args.top_p) args.max_code_fix_retries = min(args.max_code_fix_retries, int((args.max_func_call / 2))) if (args.prompt_type in ['cr']): args.max_func_call = max(args.max_func_call, 10) return args
class AlexDagRnn(nn.Module): def __init__(self, embedding_size, hidden_size): super(AlexDagRnn, self).__init__() self.embedding_size = embedding_size self.hidden_size = hidden_size (sym_dict, _) = ut.get_sym_dict() self.max_tok = (len(sym_dict) + 1) embedding = nn.Embedding(self.max_tok, self.embedding_size) initrange = (0.5 / self.embedding_size) embedding.weight.data.uniform_((- initrange), initrange) self.final_embeddings = embedding self.linear_token_1 = nn.Linear(self.embedding_size, self.hidden_size) self.linear_token_2 = nn.Linear((3 * self.hidden_size), self.hidden_size) self.lstm_ins = nn.LSTM(self.hidden_size, self.hidden_size) def embed_instr(self, instr): op = self.final_embeddings(instr.opcode) srcs = ([self.final_embeddings(min(src, self.max_tok)) for src in instr.srcs] + [torch.zeros(self.embedding_size)]) def forward(self, datum): embeddings = {}
def get_tag_device(args: object) -> str: tag = '' if torch.cuda.is_available(): txt = subprocess.run(['nvidia-smi', '--list-gpus'], stdout=subprocess.PIPE).stdout.decode('utf-8').split('\n') try: cudaids = args.cudaid.split(',') tag = 'CUDA devices: \n' for cid in cudaids: tag += 'ID: {} - {} \n'.format(cid, txt[int(cid)]) except IndexError: tag = 'CUDA devices: lost.' return tag
_model def ese_vovnet19b_dw(pretrained=False, **kwargs): return _create_vovnet('ese_vovnet19b_dw', pretrained=pretrained, **kwargs)
def get_user_config(usr_config_path, default_config=None): if (default_config is None): config = get_default_config() else: config = default_config usr_config = get_config_from_file(usr_config_path) config.merge_from_other_cfg(usr_config) config = get_conditional_config(config) return config
def calibrate(prompt_model: PromptForClassification, dataloader: PromptDataLoader) -> torch.Tensor: all_logits = [] prompt_model.eval() for batch in tqdm(dataloader, desc='ContextCali'): batch = batch.to(prompt_model.device) logits = prompt_model.forward_without_verbalize(batch) all_logits.append(logits.detach()) all_logits = torch.cat(all_logits, dim=0) return all_logits.mean(dim=0)
def create_lmsm_network(outfname_train, outfname_deploy, source_train, source_test, softmax_weight, use_OLE, batch_size_train, lambda_, num_classes=10): template_train = open('model/cifar_network.prototxt', 'r').read() template_deploy = open('model/cifar_network.prototxt', 'r').read() if use_OLE: template_train = template_train.replace('_OLE_COMMENT_', '') else: template_train = template_train.replace('_OLE_COMMENT_', '#') template_train = template_train.replace('_BATCH_SIZE_TRAIN_', str(batch_size_train)) template_train = template_train.replace('_SOURCE_TRAIN_', source_train) template_train = template_train.replace('_SOURCE_TEST_', source_test) template_train = template_train.replace('_SOFTMAX_WEIGHT_', str(softmax_weight)) template_train = template_train.replace('_LAMBDA_', str(lambda_)) write_to_file(outfname_train, template_train) write_to_file(outfname_deploy, template_deploy)
def pca_sub_df(df, task, ref_depth): data_dict = pkl.load(open(scores_path, 'rb')) accs = [get_acc(data_dict, probe_task, seed, layer=ref_depth, dims=0, run='average') for seed in REF_SEEDS] acc_dict = dict(zip(REF_SEEDS, accs)) best_seed = max(acc_dict, key=acc_dict.get) sub_df = df[(((df.layer1 == ref_depth) & (df.layer2 == ref_depth)) & ((df.seed1 == best_seed) | (df.seed2 == best_seed)))] return sub_df
def register_images(output_map, grapher, prefix='train'): if ((args.distributed_rank == 0) and (grapher is not None)): for (k, v) in output_map.items(): if isinstance(v, dict): register_images(output_map[k], grapher, prefix=prefix) if (('img' in k) or ('imgs' in k)): key_name = '-'.join(k.split('_')[0:(- 1)]) img = torchvision.utils.make_grid(v, normalize=True, scale_each=True) grapher.add_image('{}_{}'.format(prefix, key_name), img.detach(), global_step=0)
class BidirectionalLSTM(nn.Module): def __init__(self, nIn, nHidden, nOut, dropout=0.2): super(BidirectionalLSTM, self).__init__() self.rnn = nn.LSTM(nIn, nHidden, bidirectional=True, batch_first=True) self.embedding = nn.Linear((nHidden * 2), nOut) self.dropout = nn.Dropout(p=dropout) def forward(self, input, lengths=None): (hidden, _) = self.rnn(input) hidden = self.dropout(hidden) output = self.embedding(hidden) return output
_module() class ResNeXt2(ResNet2): arch_settings = {50: (Bottleneck, (3, 4, 6, 3)), 101: (Bottleneck, (3, 4, 23, 3)), 152: (Bottleneck, (3, 8, 36, 3))} def __init__(self, groups=1, base_width=4, patch_path=None, **kwargs): self.groups = groups self.base_width = base_width super(ResNeXt2, self).__init__(**kwargs) self.patch_path = patch_path for p in self.parameters(): p.requires_grad = False self.patch = get_patch_tensor(self.patch_path, self.training) def make_res_layer(self, **kwargs): return ResLayer(groups=self.groups, base_width=self.base_width, base_channels=self.base_channels, **kwargs)
def search_similar(s1, dlist=DATASET_IDS, MAX_SIMILARS=10): similars = {s2: similarity(s1, s2) for s2 in dlist if similarity(s1, s2)} top_match = Counter(similars).most_common((MAX_SIMILARS + 1)) return top_match
def _ada_boost_hp_space(name_func, base_estimator=None, n_estimators=None, learning_rate=None, random_state=None): hp_space = dict(base_estimator=base_estimator, n_estimators=(_boosting_n_estimators(name_func('n_estimators')) if (n_estimators is None) else n_estimators), learning_rate=(_ada_boost_learning_rate(name_func('learning_rate')) if (learning_rate is None) else learning_rate), random_state=_random_state(name_func('rstate'), random_state)) return hp_space
def pageinate(page, maxPage, n): pages = [page] min = (page - 1) max = (page + 1) while (len(pages) < n): if (((((2 * page) - min) <= max) or (max > maxPage)) and (min > 0)): pages.append(min) min -= 1 elif (max <= maxPage): pages.append(max) max += 1 else: break pages = sorted(pages) if ((maxPage > n) and (n > 5)): if (pages[0] != 1): pages[0] = 1 pages[1] = (- 1) if (pages[(- 1)] != maxPage): pages[(- 1)] = maxPage pages[(- 2)] = (- 1) return pages
def get_device(): device = ('cuda' if torch.cuda.is_available() else 'cpu') if (torch.backends.mps.is_available() and torch.backends.mps.is_built()): device = 'mps' if (device == 'mps'): print("WARNING: MPS currently doesn't seem to work, and messes up backpropagation without any visible torch errors. I recommend using CUDA on a colab notebook or CPU instead if you're facing inexplicable issues with generations.") return device
def ignore_buffers_decorator(func): def wrapper_ignore_buffers_decorator(self, raw_state): raw_state = raw_state[(- 1)] if (len(raw_state['parsed_mkt_data']) == 0): pass else: raw_state['parsed_mkt_data'] = raw_state['parsed_mkt_data'][(- 1)] if raw_state['parsed_volume_data']: raw_state['parsed_volume_data'] = raw_state['parsed_volume_data'][(- 1)] return func(self, raw_state) return wrapper_ignore_buffers_decorator
class Layer(object): def __init__(self): raise NotImplementedError((str(type(self)) + ' does not implement this method')) def get_output_shape(self): raise NotImplementedError((str(type(self)) + ' does not implement this method')) def output(self): raise NotImplementedError((str(type(self)) + ' does not implement this method')) def reset_params(self): raise NotImplementedError((str(type(self)) + ' does not implement this method'))
def get_vehicle_corners_from_dict(state_dict): x = state_dict['center-x'] y = state_dict['center-y'] psi = state_dict['heading'] body_shape = state_dict['corners'] center = np.array([x, y]) R = np.array([[np.cos(psi), (- np.sin(psi))], [np.sin(psi), np.cos(psi)]]) corners = (R body_shape.T).T return (corners + center)
class TrainOptions(BaseOptions): def initialize(self): BaseOptions.initialize(self) self.parser.add_argument('--display_freq', type=int, default=100, help='frequency of showing training results on screen') self.parser.add_argument('--print_freq', type=int, default=100, help='frequency of showing training results on console') self.parser.add_argument('--save_latest_freq', type=int, default=1000, help='frequency of saving the latest results') self.parser.add_argument('--save_epoch_freq', type=int, default=1, help='frequency of saving checkpoints at the end of epochs') self.parser.add_argument('--continue_train', action='store_true', help='continue training: load the latest model') self.parser.add_argument('--phase', type=str, default='train', help='train, val, test, etc') self.parser.add_argument('--which_epoch', type=str, default='latest', help='which epoch to load? set to latest to use latest cached model') self.parser.add_argument('--niter', type=int, default=200, help='# of iter at starting learning rate') self.parser.add_argument('--niter_decay', type=int, default=200, help='# of iter to linearly decay learning rate to zero') self.parser.add_argument('--beta1', type=float, default=0.5, help='momentum term of adam') self.parser.add_argument('--lr', type=float, default=0.0002, help='initial learning rate for adam') self.parser.add_argument('--TTUR', action='store_true', help='Use TTUR training scheme') self.parser.add_argument('--gan_mode', type=str, default='ls', help='(ls|original|hinge)') self.parser.add_argument('--pool_size', type=int, default=1, help='the size of image buffer that stores previously generated images') self.parser.add_argument('--no_html', action='store_true', help='do not save intermediate training results to [opt.checkpoints_dir]/[opt.name]/web/') self.isTrain = True
def main(): cuda = torch.cuda.is_available() device = torch.device(('cuda' if cuda else 'cpu')) print('Used device:', device) encoder = CNN_4Layer(in_channels=3) encoder = encoder.to(device) save_path = 'prototransfer/checkpoints/random_init_conv4' print('Save path is:', save_path) def save_checkpoint(state): filename = (save_path + '.pth.tar') torch.save(state, filename) save_checkpoint({'epoch': 0, 'n_no_improvement': 0, 'model': encoder.state_dict(), 'optimizer': None, 'scheduler': None, 'loss': np.inf, 'best_loss': np.inf, 'best_accuracy': 0, 'accuracy': 0, 'setup': None})
def ReadFile(tthread, batchInterval): (w, h) = (6, 6) y = [[0 for x in range(w)] for y in range(h)] y_sum = [0 for x in range(w)] inputEvents = (tthread * batchInterval) gs_path = (FILE_FOLER + '/GS/threads = {}/totalEvents = {}'.format(tthread, inputEvents)) lines = open(gs_path).readlines() idx = locateIdx(lines) for line in lines[idx:]: breakdown_value = line.split('\t') print(breakdown_value) for i in range(0, 6): y[i][0] += float(breakdown_value[(i + 1)]) y_sum[0] += float(breakdown_value[(i + 1)]) bfs_path = (FILE_FOLER + '/BFS/threads = {}/totalEvents = {}'.format(tthread, inputEvents)) lines = open(bfs_path).readlines() idx = locateIdx(lines) for line in lines[idx:]: breakdown_value = line.split('\t') print(breakdown_value) for i in range(0, 6): y[i][1] += float(breakdown_value[(i + 1)]) y_sum[1] += float(breakdown_value[(i + 1)]) dfs_path = (FILE_FOLER + '/DFS/threads = {}/totalEvents = {}'.format(tthread, inputEvents)) lines = open(dfs_path).readlines() idx = locateIdx(lines) for line in lines[idx:]: breakdown_value = line.split('\t') print(breakdown_value) for i in range(0, 6): y[i][2] += float(breakdown_value[(i + 1)]) y_sum[2] += float(breakdown_value[(i + 1)]) op_gs_path = (FILE_FOLER + '/OPGS/threads = {}/totalEvents = {}'.format(tthread, inputEvents)) lines = open(op_gs_path).readlines() idx = locateIdx(lines) for line in lines[idx:]: breakdown_value = line.split('\t') print(breakdown_value) for i in range(0, 6): y[i][3] += float(breakdown_value[(i + 1)]) y_sum[3] += float(breakdown_value[(i + 1)]) op_bfs_path = (FILE_FOLER + '/OPBFS/threads = {}/totalEvents = {}'.format(tthread, inputEvents)) lines = open(op_bfs_path).readlines() idx = locateIdx(lines) for line in lines[idx:]: breakdown_value = line.split('\t') print(breakdown_value) for i in range(0, 6): y[i][4] += float(breakdown_value[(i + 1)]) y_sum[4] += float(breakdown_value[(i + 1)]) op_dfs_path = (FILE_FOLER + '/OPDFS/threads = {}/totalEvents = {}'.format(tthread, inputEvents)) lines = open(op_dfs_path).readlines() idx = locateIdx(lines) for line in lines[idx:]: breakdown_value = line.split('\t') print(breakdown_value) for i in range(0, 6): y[i][5] += float(breakdown_value[(i + 1)]) y_sum[5] += float(breakdown_value[(i + 1)]) for i in range(h): for j in range(w): if (y_sum[j] != 0): y[i][j] = ((y[i][j] / y_sum[j]) * 100) print(y) return y
class WideResNet(nn.Module): def __init__(self, depth, num_classes, widen_factor=1, bn_aff=True, shortcut=True, dropRate=0.0): super(WideResNet, self).__init__() nChannels = [16, (16 * widen_factor), (32 * widen_factor), (64 * widen_factor)] assert (((depth - 4) % 6) == 0) n = ((depth - 4) / 6) block = BasicBlock self.bn_aff = bn_aff self.shortcut = shortcut self.conv1 = nn.Conv2d(3, nChannels[0], kernel_size=3, stride=1, padding=1, bias=False) self.block1 = NetworkBlock(n, nChannels[0], nChannels[1], block, 1, self.bn_aff, self.shortcut, dropRate) self.block2 = NetworkBlock(n, nChannels[1], nChannels[2], block, 2, self.bn_aff, self.shortcut, dropRate) self.block3 = NetworkBlock(n, nChannels[2], nChannels[3], block, 2, self.bn_aff, self.shortcut, dropRate) self.bn1 = nn.BatchNorm2d(nChannels[3], affine=self.bn_aff) self.relu = nn.ReLU(inplace=True) self.linear = nn.Linear(nChannels[3], num_classes) self.nChannels = nChannels for m in self.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))) elif isinstance(m, nn.BatchNorm2d): m.weight.data.fill_(1) m.bias.data.zero_() elif isinstance(m, nn.Linear): m.bias.data.zero_() def forward(self, x): out = self.conv1(x) out = self.block1(out) out = self.block2(out) out = self.block3(out) out = self.relu(self.bn1(out)) out = F.avg_pool2d(out, 8) out = out.view((- 1), self.nChannels[3]) return self.linear(out) def get_bn_before_relu(self): bn1 = self.block2.layer[0].bn1 bn2 = self.block3.layer[0].bn1 bn3 = self.bn1 return [bn1, bn2, bn3] def get_channel_num(self): return self.nChannels[1:] def extract_feature(self, x, preReLU=False): out = self.conv1(x) feat1 = self.block1(out) feat2 = self.block2(feat1) feat3 = self.block3(feat2) out = self.relu(self.bn1(feat3)) out = F.avg_pool2d(out, 8) out = out.view((- 1), self.nChannels[3]) out = self.linear(out) if preReLU: feat1 = self.block2.layer[0].bn1(feat1) feat2 = self.block3.layer[0].bn1(feat2) feat3 = self.bn1(feat3) return ([feat1, feat2, feat3], out)
def test(): global best_acc net.eval() test_loss = 0 correct = 0 total = 0 with torch.no_grad(): for (batch_idx, (inputs, targets)) in enumerate(testloader): (inputs, targets) = (inputs.to(device), targets.to(device)) outputs = net(inputs) loss = criterion(outputs, targets) test_loss += loss.item() (_, predicted) = outputs.max(1) total += targets.size(0) correct += predicted.eq(targets).sum().item() print('Test acc: ', (correct / total)) if ((correct / total) > best_acc): best_acc = (correct / total) print('Saving..') state = {'net': net.state_dict(), 'acc': best_acc} if (not os.path.isdir('checkpoint')): os.mkdir('checkpoint') torch.save(state, './checkpoint/best_ckpt.pth') print('Best acc: ', best_acc) net.train()
class OnnxrtModel(Model): def __init__(self, path: str) -> None: super().__init__(path) self._nc_model_instance: Optional[ONNXModel] = None def domain(self) -> Domain: try: input_node_names = {node.name for node in self.nc_model_instance.graph().input} node_names = {node.name for node in self.nc_model_instance.nodes()} boundary_nodes = [node.name for node in self.nc_model_instance.graph().input] boundary_nodes.extend([node.name for node in self.nc_model_instance.graph().output]) op_names = {node.op_type for node in self.nc_model_instance.nodes()} except Exception: return Domain() nlp_domain = self._check_nlp_domain(input_node_names) if (nlp_domain is not None): return nlp_domain object_detection_domain = self._check_object_detection_domain(node_names, boundary_nodes, op_names) if (object_detection_domain is not None): return object_detection_domain image_recognition_domain = self._check_image_recognition_domain(op_names) if (image_recognition_domain is not None): return image_recognition_domain return Domain() def _has_any_name_parts(nodes: set, name_parts: List[str]) -> bool: matching_names = [] for node in nodes: for partial_name in name_parts: search = re.match(partial_name, node) if search: matching_names.append(node) return bool(matching_names) def _check_nlp_domain(self, input_node_names: set) -> Optional[Domain]: if self._has_any_name_parts(input_node_names, ['input_ids', 'input_mask', 'segment_ids']): return Domain(domain=Domains.NLP.value, domain_flavour=DomainFlavours.NONE.value) return None def _check_object_detection_domain(self, node_names: set, boundary_nodes: list, op_names: set) -> Optional[Domain]: if (self._has_any_name_parts(node_names, ['NonMaxSuppression', 'box', 'score', 'class', 'detection']) or self._has_any_name_parts(op_names, ['NonMaxSuppression']) or self._has_any_name_parts(set(boundary_nodes), ['NonMaxSuppression', 'box', 'score', 'class', 'detection'])): return Domain(domain=Domains.OBJECT_DETECTION.value, domain_flavour=DomainFlavours.NONE.value) return None def _check_image_recognition_domain(self, op_names: set) -> Optional[Domain]: if self._has_any_name_parts(op_names, ['Conv', 'Relu/Clip']): return Domain(domain=Domains.IMAGE_RECOGNITION.value, domain_flavour=DomainFlavours.NONE.value) return None def input_shape(self) -> Shape: from google.protobuf.json_format import MessageToDict shape = None trusted = False if (self.domain.domain == Domains.IMAGE_RECOGNITION.value): shape_list = [] for input_node in self.filtered_input_nodes: node_dict = MessageToDict(input_node) dimensions = node_dict.get('type', {}).get('tensorType', {}).get('shape', {}).get('dim', []) input_shape = [] for dim in dimensions: if (dim.get('dimValue', None) is not None): input_shape.append(dim['dimValue']) shape_list.append(','.join(input_shape)) if (len(shape_list) == 1): shape = shape_list[0] shape = remove_number_of_samples_from_shape(shape) trusted = True if (len(shape_list) > 1): adjusted_shapes_list = [] for sub_shape in shape_list: adjusted_shapes_list.append(remove_number_of_samples_from_shape(sub_shape)) shape = ','.join([(('[' + sub_shape) + ']') for sub_shape in adjusted_shapes_list]) trusted = True return Shape(shape=shape, trusted=trusted) def shape_elements_order(self) -> List[str]: return ['channels', 'height', 'width'] def get_framework_name() -> str: return Frameworks.ONNX.value def get_model_graph(self) -> Graph: graph_reader = OnnxrtReader(self) return graph_reader.read() def nc_model_instance(self) -> ONNXModel: self._ensure_nc_model_instance() return self._nc_model_instance def _ensure_nc_model_instance(self) -> None: if (self._nc_model_instance is not None): return model_name = os.path.splitext(os.path.basename(self.path))[0] Logger().get_logger().setLevel(log.level) self._nc_model_instance = NCModel(self.path) self._nc_model_instance.name = model_name def supports_path(path: str) -> bool: return ('onnx' == get_file_extension(path)) def guard_requirements_installed(self) -> None: check_module('onnx') check_module('onnxruntime') if (sys.version_info < (3, 11)): check_module('onnxruntime_extensions') def filtered_input_nodes(self) -> List[Any]: input_nodes = self.nc_model_instance.graph().input name_to_input = {} for input in input_nodes: name_to_input[input.name] = input for initializer in self.nc_model_instance.graph().initializer: if (initializer.name in name_to_input): input_nodes.remove(name_to_input[initializer.name]) return input_nodes
def merge_dict(user, default): if (isinstance(user, dict) and isinstance(default, dict)): for (k, v) in default.items(): if (k not in user): user[k] = v else: user[k] = merge_dict(user[k], v) return user
class TestDummy(unittest.TestCase): def test_encode(self): commands = [['python', 'encode.py', '--users', '--items'], ['python', 'lr.py', 'data/dummy/X-ui.npz'], ['python', 'lr.py', '--folds', 'strong', 'data/dummy/X-ui.npz'], ['python', 'fm.py', 'data/dummy/X-ui.npz'], ['python', 'fm.py', '--folds', 'weak', 'data/dummy/X-ui.npz'], ['python', 'sktm.py', '--model', 'irt'], ['python', 'sktm.py', '--model', 'pfa'], ['python', 'sktm.py', '--model', 'iswf']] for command in commands: p = check_output(command)
_args('v', 'is', 'is', 'is', 'is') def im2col(g, input, kernel_size, dilation, padding, stride): input_h = size(g, input, g.op('Constant', value_t=torch.tensor(2))) input_w = size(g, input, g.op('Constant', value_t=torch.tensor(3))) (stride_h, stride_w) = (stride[0], stride[1]) (padding_h, padding_w) = (padding[0], padding[1]) (dilation_h, dilation_w) = (dilation[0], dilation[1]) (kernel_h, kernel_w) = (kernel_size[0], kernel_size[1]) blocks_row_indices = _get_im2col_indices_along_dim(g, input_h, kernel_h, dilation_h, padding_h, stride_h) blocks_col_indices = _get_im2col_indices_along_dim(g, input_w, kernel_w, dilation_w, padding_w, stride_w) output_shape = _get_im2col_output_shape(g, input, kernel_h, kernel_w) padded_input = _get_im2col_padded_input(g, input, padding_h, padding_w) output = g.op('Gather', padded_input, blocks_row_indices, axis_i=2) output = g.op('Gather', output, blocks_col_indices, axis_i=4) output = g.op('Transpose', output, perm_i=[0, 1, 2, 4, 3, 5]) return g.op('Reshape', output, output_shape)
def plot(json_fname, results_fname, store_plots='', plots_to_latex=''): name = '.'.join(json_fname.split('.')[:(- 1)]) data = json.loads(open(json_fname, 'r', encoding='utf-8').read()) fi = open(results_fname, 'r', encoding='utf-8') truely_detected_and_truely_corrected = [[], []] truely_detected_and_falsely_corrected = [[], []] falsely_detected = [[], []] count_of_absence_of_correct_chars = [0, 0] w3 = open(f'{name}_falsely_detected.txt', 'w', encoding='utf-8') w4 = open(f'{name}_falsely_corrected.txt', 'w', encoding='utf-8') for (line, entry) in zip(fi, data): (origin_num, wrong_sent, correct_sent, predict_sent, num) = line.strip().split('\t') pos_to_error = dict([(e['error_position'], e) for e in entry['errors']]) for (pos, (w, c, p)) in enumerate(zip(wrong_sent, correct_sent, predict_sent)): if ((w != c) and (w != p)): e = pos_to_error[pos] assert (e['corrected_to'] == p) if (c != p): candidatas = dict(sorted(list(e['candidates'].items()), reverse=True, key=(lambda it: it[1]))[:5]) absent = 'no' if (c not in candidatas): count_of_absence_of_correct_chars[0] += 1 absent = 'yes' truely_detected_and_falsely_corrected[0].append(e['confidence']) truely_detected_and_falsely_corrected[1].append(e['similarity']) w4.write(('\t'.join([wrong_sent, f'pos={pos}', f'w={w}', f'c={c}', f'p={p}', f"sim={e['similarity']}", f'absent={absent}']) + '\n')) else: truely_detected_and_truely_corrected[0].append(e['confidence']) truely_detected_and_truely_corrected[1].append(e['similarity']) elif ((w == c) and (w != p)): e = pos_to_error[pos] candidates = dict(sorted(list(e['candidates'].items()), reverse=True, key=(lambda it: it[1]))[:5]) absent = 'no' if (c not in candidates): count_of_absence_of_correct_chars[1] += 1 absent = 'yes' falsely_detected[0].append(e['confidence']) falsely_detected[1].append(e['similarity']) w3.write(('\t'.join([wrong_sent, f'pos={pos}', f'w={w}', f'c={c}', f'p={p}', f"sim={e['similarity']}", f'absent={absent}']) + '\n')) print(f'In {len(truely_detected_and_falsely_corrected[0])} falsely corrected characters, {count_of_absence_of_correct_chars[0]} are because of absent correct candidates.') print(f'In {len(falsely_detected[0])} falsely detected characters, {count_of_absence_of_correct_chars[1]} are because of absent correct candidates.') plt.plot(truely_detected_and_truely_corrected[0], truely_detected_and_truely_corrected[1], 'ro', truely_detected_and_falsely_corrected[0], truely_detected_and_falsely_corrected[1], 'bo', falsely_detected[0], falsely_detected[1], 'x') plt.axis([0.0, 1.0, 0.0, 1.0]) plt.show() if plots_to_latex: produce_latex(truely_detected_and_truely_corrected, truely_detected_and_falsely_corrected, falsely_detected, os.path.join(plots_to_latex, f'{name}_latex.txt')) if store_plots: axes = plt.gca() plt.savefig(os.path.join(store_plots, f'{name}.png')) axes.set_xlim([0.95, 1]) axes.set_ylim([0.0, 0.6]) plt.savefig(os.path.join(store_plots, f'{name}2.png'))
def process_single_fragment(fragment_id, color_files, depth_files, n_files, n_fragments, config): if config['path_intrinsic']: intrinsic = o3d.io.read_pinhole_camera_intrinsic(config['path_intrinsic']) else: intrinsic = o3d.camera.PinholeCameraIntrinsic(o3d.camera.PinholeCameraIntrinsicParameters.PrimeSenseDefault) sid = (fragment_id * config['n_frames_per_fragment']) eid = min((sid + config['n_frames_per_fragment']), n_files) make_posegraph_for_fragment(config['path_dataset'], sid, eid, color_files, depth_files, fragment_id, n_fragments, intrinsic, with_opencv, config) optimize_posegraph_for_fragment(config['path_dataset'], fragment_id, config) make_pointcloud_for_fragment(config['path_dataset'], color_files, depth_files, fragment_id, n_fragments, intrinsic, config)
class AdvWeightPerturb(object): def __init__(self, model, proxy, proxy_optim, gamma): super(AdvWeightPerturb, self).__init__() self.model = model self.proxy = proxy self.proxy_optim = proxy_optim self.gamma = gamma def calc_awp(self, inputs_adv, targets): self.proxy.load_state_dict(self.model.state_dict()) self.proxy.train() loss = (- F.cross_entropy(self.proxy(inputs_adv), targets)) self.proxy_optim.zero_grad() loss.backward() self.proxy_optim.step() diff = diff_in_weights(self.model, self.proxy) return diff def perturb(self, diff): add_into_weights(self.model, diff, coeff=(1.0 * self.gamma)) def restore(self, diff): add_into_weights(self.model, diff, coeff=((- 1.0) * self.gamma))
class ParetoSetModel(torch.nn.Module): def __init__(self, n_dim, n_obj): super(ParetoSetModel, self).__init__() self.n_dim = n_dim self.n_obj = n_obj self.fc1 = nn.Linear(self.n_obj, 256) self.fc2 = nn.Linear(256, 256) self.fc3 = nn.Linear(256, self.n_dim) def forward(self, pref): x = torch.relu(self.fc1(pref)) x = torch.relu(self.fc2(x)) x = self.fc3(x) x = torch.sigmoid(x) return x.to(torch.float64)
class Warmup(Scheduler): def __init__(self, delta: float) -> None: from bigdl.dllib.optim.optimizer import Warmup as BWarmup self.scheduler = BWarmup(delta) def get_scheduler(self) -> 'optimizer.Warmup': return self.scheduler
def line_bounding_2D_activation(x_minus, x_plus, y_minus, y_plus, tanh=True): (kl, bl, ku, bu) = getConvenientGeneralActivationBound(y_minus, y_plus, 'sigmoid') if tanh: X_l = torch.tanh(x_minus) X_u = torch.tanh(x_plus) else: X_l = x_minus X_u = x_plus I_l = (X_l >= 0).float() I_u = (X_u >= 0).float() alpha_l = torch.zeros(x_minus.shape, device=x_minus.device) beta_l = (((I_l * X_l) * kl) + (((1 - I_l) * X_l) * ku)) gamma_l = (((I_l * X_l) * bl) + (((1 - I_l) * X_l) * bu)) alpha_u = torch.zeros(x_plus.shape, device=x_minus.device) beta_u = (((I_u * X_u) * ku) + (((1 - I_u) * X_u) * kl)) gamma_u = (((I_u * X_u) * bu) + (((1 - I_u) * X_u) * bl)) return (alpha_l, beta_l, gamma_l, alpha_u, beta_u, gamma_u)
def test_text_envs(): env = gym.make('FrozenLake-v0') video = VideoRecorder(env) try: env.reset() video.capture_frame() video.close() finally: os.remove(video.path)
def data(ndata=100, baseline=1, freq=10, sigma=1.0, **kwargs): t = (baseline * np.sort(np.random.rand(ndata))) y = transit_model(t, freq, **kwargs) dy = (sigma * np.ones_like(t)) y += (dy * np.random.randn(len(t))) return (t, y, dy)
def save_tokenizer(tokenizer, path): with open(path, 'wb') as f: pickle.dump(tokenizer, f) print('tokenizer saved in {}'.format(path))