code
stringlengths
101
5.91M
class UNetMidBlock2D(nn.Module): def __init__(self, in_channels: int, temb_channels: int, dropout: float=0.0, num_layers: int=1, resnet_eps: float=1e-06, resnet_time_scale_shift: str='default', resnet_act_fn: str='swish', resnet_groups: int=32, attn_groups: Optional[int]=None, resnet_pre_norm: bool=True, add_attention: bool=True, attention_head_dim: int=1, output_scale_factor: float=1.0): super().__init__() resnet_groups = (resnet_groups if (resnet_groups is not None) else min((in_channels // 4), 32)) self.add_attention = add_attention if (attn_groups is None): attn_groups = (resnet_groups if (resnet_time_scale_shift == 'default') else None) resnets = [ResnetBlock2D(in_channels=in_channels, out_channels=in_channels, temb_channels=temb_channels, eps=resnet_eps, groups=resnet_groups, dropout=dropout, time_embedding_norm=resnet_time_scale_shift, non_linearity=resnet_act_fn, output_scale_factor=output_scale_factor, pre_norm=resnet_pre_norm)] attentions = [] if (attention_head_dim is None): logger.warn(f'It is not recommend to pass `attention_head_dim=None`. Defaulting `attention_head_dim` to `in_channels`: {in_channels}.') attention_head_dim = in_channels for _ in range(num_layers): if self.add_attention: attentions.append(Attention(in_channels, heads=(in_channels // attention_head_dim), dim_head=attention_head_dim, rescale_output_factor=output_scale_factor, eps=resnet_eps, norm_num_groups=attn_groups, spatial_norm_dim=(temb_channels if (resnet_time_scale_shift == 'spatial') else None), residual_connection=True, bias=True, upcast_softmax=True, _from_deprecated_attn_block=True)) else: attentions.append(None) resnets.append(ResnetBlock2D(in_channels=in_channels, out_channels=in_channels, temb_channels=temb_channels, eps=resnet_eps, groups=resnet_groups, dropout=dropout, time_embedding_norm=resnet_time_scale_shift, non_linearity=resnet_act_fn, output_scale_factor=output_scale_factor, pre_norm=resnet_pre_norm)) self.attentions = nn.ModuleList(attentions) self.resnets = nn.ModuleList(resnets) def forward(self, hidden_states: torch.FloatTensor, temb: Optional[torch.FloatTensor]=None) -> torch.FloatTensor: hidden_states = self.resnets[0](hidden_states, temb) for (attn, resnet) in zip(self.attentions, self.resnets[1:]): if (attn is not None): hidden_states = attn(hidden_states, temb=temb) hidden_states = resnet(hidden_states, temb) return hidden_states
def normalize(x, stats): if (stats is None): return x return ((x - stats.mean) / stats.std)
def get_file_size(filename): size_in_mb = (os.path.getsize(filename) / float((1024 ** 2))) return size_in_mb
def main(): import argparse parser = argparse.ArgumentParser() parser.add_argument('dataset_loader', type=str) parser.add_argument('dataset_path', type=str) parser.add_argument('--video_names', type=str, nargs='+', help='Only generate training data for a subset of videos. If not set, will include all videos in dataset_path.') parser.add_argument('--num_train_samples', type=int, help='How many training samples to generate') parser.add_argument('--num_eval_samples', type=int, help='How many evaluation samples to generate') parser.add_argument('--temp_dir', type=str, help='Where to store temporary intermediate results') parser.add_argument('--work_dir', type=str, help='Root folder for all experiments') parser.add_argument('--num_process', type=int, help='How many worker processes') parser.add_argument('--random_seed', type=int, help='Optional random seed for deterministic results') add_resizing_arguments(parser) add_image_processing_arguments(parser) args = parser.parse_args() last_progress = None for progress in generate(**vars(args)): prog_percent = int((progress * 100)) if (prog_percent != last_progress): logger.info(f'Generating training data: {prog_percent}% done') last_progress = prog_percent
def format_str_one(v, float_prec=6, int_pad=1): if (isinstance(v, torch.Tensor) and (v.numel() == 1)): v = v.item() if isinstance(v, float): return (('{:.' + str(float_prec)) + 'f}').format(v) if (isinstance(v, int) and int_pad): return (('{:0' + str(int_pad)) + 'd}').format(v) return str(v)
def create_vocabulary_from_data(datasets: DatasetDict, word_delimiter_token: Optional[str]=None, unk_token: Optional[str]=None, pad_token: Optional[str]=None): def extract_all_chars(batch): all_text = ' '.join(batch['target_text']) vocab = list(set(all_text)) return {'vocab': [vocab], 'all_text': [all_text]} vocabs = datasets.map(extract_all_chars, batched=True, batch_size=(- 1), keep_in_memory=True, remove_columns=datasets['train'].column_names) vocab_set = functools.reduce((lambda vocab_1, vocab_2: (set(vocab_1['vocab'][0]) | set(vocab_2['vocab'][0]))), vocabs.values()) vocab_dict = {v: k for (k, v) in enumerate(sorted(list(vocab_set)))} if (word_delimiter_token is not None): vocab_dict[word_delimiter_token] = vocab_dict[' '] del vocab_dict[' '] if (unk_token is not None): vocab_dict[unk_token] = len(vocab_dict) if (pad_token is not None): vocab_dict[pad_token] = len(vocab_dict) return vocab_dict
def train_val_test_generate(dataframe, model_params): (train_val_test_x, train_val_test_y, len_x_samples, len_before_x_samples) = pad_all_cases(dataframe, dataframe['NO3'].values, model_params, model_params['min_before'], model_params['max_before'], model_params['min_after'], model_params['max_after'], model_params['output_length']) train_val_test_y = np.expand_dims(train_val_test_y, axis=2) return (train_val_test_x, train_val_test_y, len_x_samples, len_before_x_samples)
def main(): parser = argparse.ArgumentParser(description='Tool to average the params of input checkpoints to produce a new checkpoint') parser.add_argument('--inputs', required=True, nargs='+', help='Input checkpoint file paths.') parser.add_argument('--output', required=True, metavar='FILE', help='Write the new checkpoint containing the averaged weights to this path.') num_group = parser.add_mutually_exclusive_group() num_group.add_argument('--num-epoch-checkpoints', type=int, help='if set, will try to find checkpoints with names checkpoint_xx.pt in the path specified by input, and average last this many of them.') num_group.add_argument('--num-update-checkpoints', type=int, help='if set, will try to find checkpoints with names checkpoint_ee_xx.pt in the path specified by input, and average last this many of them.') parser.add_argument('--checkpoint-upper-bound', type=int, help='when using --num-epoch-checkpoints, this will set an upper bound on which epoch to use, when using --num-update-checkpoints, this will set an upper bound on which update to usee.g., with --num-epoch-checkpoints=10 --checkpoint-upper-bound=50, checkpoints 41-50 would be averaged.e.g., with --num-update-checkpoints=10 --checkpoint-upper-bound=50000, checkpoints 40500-50000 would be averaged assuming --save-interval-updates 500') args = parser.parse_args() print(args) num = None is_update_based = False if (args.num_update_checkpoints is not None): num = args.num_update_checkpoints is_update_based = True elif (args.num_epoch_checkpoints is not None): num = args.num_epoch_checkpoints assert ((args.checkpoint_upper_bound is None) or ((args.num_epoch_checkpoints is not None) or (args.num_update_checkpoints is not None))), '--checkpoint-upper-bound requires --num-epoch-checkpoints or --num-update-checkpoints' assert ((args.num_epoch_checkpoints is None) or (args.num_update_checkpoints is None)), 'Cannot combine --num-epoch-checkpoints and --num-update-checkpoints' if (num is not None): args.inputs = last_n_checkpoints(args.inputs, num, is_update_based, upper_bound=args.checkpoint_upper_bound) print('averaging checkpoints: ', args.inputs) new_state = average_checkpoints(args.inputs) with PathManager.open(args.output, 'wb') as f: torch.save(new_state, f) print('Finished writing averaged checkpoint to {}'.format(args.output))
def plot_dimension_interval(ax, cmap, dim_: int, x_box: List[float], y_box: List[float], x_color='#FFB570', y_color='#67AB9F', label=False): if (not label): ax.hlines(dim_, x_box[0], x_box[1], x_color, lw=10) ax.hlines(dim_, y_box[0], y_box[1], y_color, lw=7) else: ax.hlines(dim_, x_box[0], x_box[1], x_color, lw=10, label='X Box Intervals', cmap=cmap) ax.hlines(dim_, y_box[0], y_box[1], y_color, lw=7, label='Y Box Intervals', cmap=cmap)
def register_all_cityscapes(root='datasets'): for (key, (image_dir, gt_dir)) in _RAW_CITYSCAPES_SPLITS.items(): meta = _get_builtin_metadata('cityscapes') image_dir = os.path.join(root, image_dir) gt_dir = os.path.join(root, gt_dir) inst_key = key.format(task='instance_seg') DatasetCatalog.register(inst_key, (lambda x=image_dir, y=gt_dir: load_cityscapes_instances(x, y, from_json=True, to_polygons=True))) MetadataCatalog.get(inst_key).set(image_dir=image_dir, gt_dir=gt_dir, evaluator_type='cityscapes', **meta) sem_key = key.format(task='sem_seg') DatasetCatalog.register(sem_key, (lambda x=image_dir, y=gt_dir: load_cityscapes_semantic(x, y))) MetadataCatalog.get(sem_key).set(image_dir=image_dir, gt_dir=gt_dir, evaluator_type='sem_seg', **meta)
class Boco(): def __init__(self, name): self.name = name def validate(self): assert self.computeLoss, 'You need to specify a function to compute the loss'
def get_video_loader(use_petrel_backend: bool=True, enable_mc: bool=True, conf_path: str=None): if (petrel_backend_imported and use_petrel_backend): _client = Client(conf_path=conf_path, enable_mc=enable_mc) else: _client = None def _loader(video_path): if ((_client is not None) and ('s3:' in video_path)): video_path = io.BytesIO(_client.get(video_path)) vr = VideoReader(video_path, num_threads=1, ctx=cpu(0)) return vr return _loader
def create_split_mesh(cat_desc, edge_length_threshold, overwrite=False, start_threshold=None): from shapenet.core import cat_desc_to_id from shapenet.core.meshes.config import get_mesh_config from template_ffd.templates.ids import get_template_ids cat_id = cat_desc_to_id(cat_desc) example_ids = get_template_ids(cat_id) config = get_mesh_config(edge_length_threshold) init = (None if (start_threshold is None) else get_mesh_config(start_threshold)) config.create_cat_data(cat_id, example_ids, overwrite, init)
class LevelsFilter(logging.Filter): def __init__(self, levels): self.levels = [getattr(logging, level) for level in levels] def filter(self, record): return (record.levelno in self.levels)
def reorder_image(img, input_order='HWC'): if (input_order not in ['HWC', 'CHW']): raise ValueError(f"Wrong input_order {input_order}. Supported input_orders are 'HWC' and 'CHW'") if (len(img.shape) == 2): img = img[(..., None)] return img if (input_order == 'CHW'): img = img.transpose(1, 2, 0) return img
def update_config_from_file(filename, base_cfg=None): exp_config = None with open(filename) as f: exp_config = edict(yaml.safe_load(f)) if (base_cfg is not None): _update_config(base_cfg, exp_config) else: _update_config(cfg, exp_config)
class ADFF(nn.Module): def __init__(self, block_channel, adff_num_features=1280, rpd_num_features=2048): super(ADFF, self).__init__() rpd_num_features = (rpd_num_features // 2) print('block_channel:', block_channel) self.upsample_scale1to5 = _UpProjection(num_input_features=block_channel[0], num_output_features=(adff_num_features // 5)) self.upsample_scale2to5 = _UpProjection(num_input_features=block_channel[1], num_output_features=(adff_num_features // 5)) self.upsample_scale3to5 = _UpProjection(num_input_features=block_channel[2], num_output_features=(adff_num_features // 5)) self.upsample_scale4to5 = _UpProjection(num_input_features=block_channel[3], num_output_features=(adff_num_features // 5)) self.upsample_scale5to5 = _UpProjection(num_input_features=block_channel[4], num_output_features=(adff_num_features // 5)) self.conv_scale5 = nn.Conv2d(adff_num_features, rpd_num_features, kernel_size=3, stride=1, padding=1, bias=False) self.bn_scale5 = nn.BatchNorm2d(rpd_num_features) adff_num_features = (adff_num_features // 2) rpd_num_features = (rpd_num_features // 2) self.upsample_scale1to4 = _UpProjection(num_input_features=block_channel[0], num_output_features=(adff_num_features // 5)) self.upsample_scale2to4 = _UpProjection(num_input_features=block_channel[1], num_output_features=(adff_num_features // 5)) self.upsample_scale3to4 = _UpProjection(num_input_features=block_channel[2], num_output_features=(adff_num_features // 5)) self.upsample_scale4to4 = _UpProjection(num_input_features=block_channel[3], num_output_features=(adff_num_features // 5)) self.upsample_scale5to4 = _UpProjection(num_input_features=block_channel[4], num_output_features=(adff_num_features // 5)) self.conv_scale4 = nn.Conv2d(adff_num_features, rpd_num_features, kernel_size=3, stride=1, padding=1, bias=False) self.bn_scale4 = nn.BatchNorm2d(rpd_num_features) adff_num_features = (adff_num_features // 2) rpd_num_features = (rpd_num_features // 2) self.upsample_scale1to3 = _UpProjection(num_input_features=block_channel[0], num_output_features=(adff_num_features // 5)) self.upsample_scale2to3 = _UpProjection(num_input_features=block_channel[1], num_output_features=(adff_num_features // 5)) self.upsample_scale3to3 = _UpProjection(num_input_features=block_channel[2], num_output_features=(adff_num_features // 5)) self.upsample_scale4to3 = _UpProjection(num_input_features=block_channel[3], num_output_features=(adff_num_features // 5)) self.upsample_scale5to3 = _UpProjection(num_input_features=block_channel[4], num_output_features=(adff_num_features // 5)) self.conv_scale3 = nn.Conv2d(adff_num_features, rpd_num_features, kernel_size=3, stride=1, padding=1, bias=False) self.bn_scale3 = nn.BatchNorm2d(rpd_num_features) adff_num_features = (adff_num_features // 2) rpd_num_features = (rpd_num_features // 2) self.upsample_scale1to2 = _UpProjection(num_input_features=block_channel[0], num_output_features=(adff_num_features // 5)) self.upsample_scale2to2 = _UpProjection(num_input_features=block_channel[1], num_output_features=(adff_num_features // 5)) self.upsample_scale3to2 = _UpProjection(num_input_features=block_channel[2], num_output_features=(adff_num_features // 5)) self.upsample_scale4to2 = _UpProjection(num_input_features=block_channel[3], num_output_features=(adff_num_features // 5)) self.upsample_scale5to2 = _UpProjection(num_input_features=block_channel[4], num_output_features=(adff_num_features // 5)) self.conv_scale2 = nn.Conv2d(adff_num_features, rpd_num_features, kernel_size=3, stride=1, padding=1, bias=False) self.bn_scale2 = nn.BatchNorm2d(rpd_num_features) adff_num_features = (adff_num_features // 2) rpd_num_features = (rpd_num_features // 2) self.upsample_scale1to1 = _UpProjection(num_input_features=block_channel[0], num_output_features=(adff_num_features // 5)) self.upsample_scale2to1 = _UpProjection(num_input_features=block_channel[1], num_output_features=(adff_num_features // 5)) self.upsample_scale3to1 = _UpProjection(num_input_features=block_channel[2], num_output_features=(adff_num_features // 5)) self.upsample_scale4to1 = _UpProjection(num_input_features=block_channel[3], num_output_features=(adff_num_features // 5)) self.upsample_scale5to1 = _UpProjection(num_input_features=block_channel[4], num_output_features=(adff_num_features // 5)) self.conv_scale1 = nn.Conv2d(adff_num_features, rpd_num_features, kernel_size=3, stride=1, padding=1, bias=False) self.bn_scale1 = nn.BatchNorm2d(rpd_num_features) def forward(self, feature_pyramid): scale1_size = [feature_pyramid[0].size(2), feature_pyramid[0].size(3)] scale2_size = [feature_pyramid[1].size(2), feature_pyramid[1].size(3)] scale3_size = [feature_pyramid[2].size(2), feature_pyramid[2].size(3)] scale4_size = [feature_pyramid[3].size(2), feature_pyramid[3].size(3)] scale5_size = [feature_pyramid[4].size(2), feature_pyramid[4].size(3)] scale_1to5 = self.upsample_scale1to5(feature_pyramid[0], scale5_size) scale_2to5 = self.upsample_scale2to5(feature_pyramid[1], scale5_size) scale_3to5 = self.upsample_scale3to5(feature_pyramid[2], scale5_size) scale_4to5 = self.upsample_scale4to5(feature_pyramid[3], scale5_size) scale_5to5 = self.upsample_scale5to5(feature_pyramid[4], scale5_size) scale5_mff = torch.cat((scale_1to5, scale_2to5, scale_3to5, scale_4to5, scale_5to5), 1) scale5_mff = F.relu(self.bn_scale5(self.conv_scale5(scale5_mff))) scale_1to4 = self.upsample_scale1to4(feature_pyramid[0], scale4_size) scale_2to4 = self.upsample_scale2to4(feature_pyramid[1], scale4_size) scale_3to4 = self.upsample_scale3to4(feature_pyramid[2], scale4_size) scale_4to4 = self.upsample_scale4to4(feature_pyramid[3], scale4_size) scale_5to4 = self.upsample_scale5to4(feature_pyramid[4], scale4_size) scale4_mff = torch.cat((scale_1to4, scale_2to4, scale_3to4, scale_4to4, scale_5to4), 1) scale4_mff = F.relu(self.bn_scale4(self.conv_scale4(scale4_mff))) scale_1to3 = self.upsample_scale1to3(feature_pyramid[0], scale3_size) scale_2to3 = self.upsample_scale2to3(feature_pyramid[1], scale3_size) scale_3to3 = self.upsample_scale3to3(feature_pyramid[2], scale3_size) scale_4to3 = self.upsample_scale4to3(feature_pyramid[3], scale3_size) scale_5to3 = self.upsample_scale5to3(feature_pyramid[4], scale3_size) scale3_mff = torch.cat((scale_1to3, scale_2to3, scale_3to3, scale_4to3, scale_5to3), 1) scale3_mff = F.relu(self.bn_scale3(self.conv_scale3(scale3_mff))) scale_1to2 = self.upsample_scale1to2(feature_pyramid[0], scale2_size) scale_2to2 = self.upsample_scale2to2(feature_pyramid[1], scale2_size) scale_3to2 = self.upsample_scale3to2(feature_pyramid[2], scale2_size) scale_4to2 = self.upsample_scale4to2(feature_pyramid[3], scale2_size) scale_5to2 = self.upsample_scale5to2(feature_pyramid[4], scale2_size) scale2_mff = torch.cat((scale_1to2, scale_2to2, scale_3to2, scale_4to2, scale_5to2), 1) scale2_mff = F.relu(self.bn_scale2(self.conv_scale2(scale2_mff))) scale_1to1 = self.upsample_scale1to1(feature_pyramid[0], scale1_size) scale_2to1 = self.upsample_scale2to1(feature_pyramid[1], scale1_size) scale_3to1 = self.upsample_scale3to1(feature_pyramid[2], scale1_size) scale_4to1 = self.upsample_scale4to1(feature_pyramid[3], scale1_size) scale_5to1 = self.upsample_scale5to1(feature_pyramid[4], scale1_size) scale1_mff = torch.cat((scale_1to1, scale_2to1, scale_3to1, scale_4to1, scale_5to1), 1) scale1_mff = F.relu(self.bn_scale1(self.conv_scale1(scale1_mff))) fused_feature_pyramid = [scale1_mff, scale2_mff, scale3_mff, scale4_mff, scale5_mff] return fused_feature_pyramid
def __median_wilcoxon_to_latex(indicator_name: str, wilcoxon_data: pd.DataFrame, caption: str, label): indicator_data = wilcoxon_data[(wilcoxon_data['Indicator'] == indicator_name)] problems = pd.unique(indicator_data['Problem']) algorithms = pd.unique(indicator_data['Algorithm']) num_columns = len(algorithms) columns = algorithms alignment = 'c' col_format = '{}|{}'.format(alignment, (alignment * num_columns)) column_labels = ['\\textbf{{{0}}}'.format(label.replace('_', '\\_')) for label in columns] output = io.StringIO() output.write('\\documentclass{article}\n') output.write('\\usepackage[utf8]{inputenc}\n') output.write('\\usepackage{tabularx}\n') output.write('\\usepackage{colortbl}\n') output.write('\\usepackage[table*]{xcolor}\n') output.write('\\xdefinecolor{gray95}{gray}{0.65}\n') output.write('\\xdefinecolor{gray25}{gray}{0.8}\n') output.write('\\title{Median and Wilcoxon}\n') output.write('\\author{}\n') output.write('\\begin{document}\n') output.write('\\maketitle\n') output.write('\\section{Table}\n') output.write('\\begin{table}[!htp]\n') output.write(' \\caption{{{}}}\n'.format(caption)) output.write(' \\label{{{}}}\n'.format(label)) output.write(' \\centering\n') output.write(' \\begin{tiny}\n') output.write((' \\begin{tabular}{%s}\n' % col_format)) output.write(' & {} \\\\\\hline\n'.format(' & '.join(column_labels))) counters = {} for algorithm in algorithms: counters[algorithm] = [0, 0, 0] for problem in problems: values = [] for algorithm in algorithms: row = indicator_data[((indicator_data['Problem'] == problem) & (indicator_data['Algorithm'] == algorithm))] value = '{:.2e}({:.2e})'.format(row['Median'].tolist()[0], row['IQR'].tolist()[0]) if (algorithm != algorithms[(- 1)]): if (row['TestResult'].tolist()[0] == '-'): value = '{{{}-}}'.format(value) counters[algorithm][2] = (counters[algorithm][2] + 1) elif (row['TestResult'].tolist()[0] == '+'): value = '{{{}+}}'.format(value) counters[algorithm][0] = (counters[algorithm][0] + 1) else: value = '{{{}\\approx}}'.format(value) counters[algorithm][1] = (counters[algorithm][1] + 1) values.append(value) medians = indicator_data[(indicator_data['Problem'] == problem)]['Median'] iqrs = indicator_data[(indicator_data['Problem'] == problem)]['IQR'] pairs = list(zip(medians, iqrs)) indexes = sorted(range(len(pairs)), key=(lambda x: pairs[x])) if check_minimization(indicator_name): best = indexes[0] second_best = indexes[1] else: best = indexes[(- 1)] second_best = indexes[(- 2)] values[best] = ('\\cellcolor{gray95} ' + values[best]) values[second_best] = ('\\cellcolor{gray25} ' + values[second_best]) output.write('\\textbf{{{0}}} & ${1}$ \\\\\n'.format(problem, ' $ & $ '.join([str(val).replace('e-', 'e\\makebox[0.1cm]{-}').replace('e+', 'e\\makebox[0.1cm]{+}') for val in values]))) counter_summary = [] for algorithm in algorithms[:(- 1)]: counter_summary.append(counters[algorithm]) output.write(' \\hline\n') output.write('\\textbf{{{0}}} & ${1}$ \\\\\n'.format('$+/\\approx/-$', ' $ & $ '.join([((((str(val[0]) + '/') + str(val[1])) + '/') + str(val[2])) for val in counter_summary]))) output.write(' \\end{tabular}\n') output.write(' \\end{tiny}\n') output.write('\\end{table}\n') output.write('\\end{document}') return output.getvalue()
class TFFlaubertForTokenClassification(): def __init__(self, *args, **kwargs): requires_tf(self) def from_pretrained(self, *args, **kwargs): requires_tf(self)
def ResNet50(include_top=False, weights='imagenet', input_tensor=None, input_shape=None, pooling=None, classes=1000, **kwargs): def stack_fn(x): x = stack1(x, 64, 3, stride1=1, name='conv2') x = stack1(x, 128, 4, name='conv3') x = stack1(x, 256, 6, name='conv4') x = stack1(x, 512, 3, name='conv5', stride1=1, dilation=2) return x return ResNet(stack_fn, False, True, 'resnet50', include_top, weights, input_tensor, input_shape, pooling, classes, **kwargs)
def maybe_dict_from_checkpoint(ckpt_path=None, ckpt_dict=None): assert ((ckpt_path is not None) or (ckpt_dict is not None)) if (ckpt_dict is None): ckpt_dict = load_dict_from_checkpoint(ckpt_path) return ckpt_dict
class IntLinear(nn.Module): def __init__(self, in_features, out_features, bias=True, p=0, update_step=3000, bits=8, method='histogram'): super(IntLinear, self).__init__() self.in_features = int(in_features) self.out_features = int(out_features) self.weight = torch.nn.Parameter(torch.Tensor(out_features, in_features)) self.chosen_bias = bias if self.chosen_bias: self.bias = torch.nn.Parameter(torch.Tensor(out_features)) else: self.register_parameter('bias', None) self.reset_parameters() self.p = p self.bits = bits self.method = method self.update_step = update_step self.counter = 0 def reset_parameters(self): nn.init.xavier_uniform_(self.weight) if self.chosen_bias: nn.init.constant_(self.bias, 0.0) return def forward(self, input): p = (self.p if self.training else 1) if ((self.counter % self.update_step) == 0): self.scale = None self.zero_point = None self.counter += 1 (weight_quantized, self.scale, self.zero_point) = emulate_int(self.weight.detach(), bits=self.bits, method=self.method, scale=self.scale, zero_point=self.zero_point) mask = torch.zeros_like(self.weight) mask.bernoulli_((1 - p)) noise = (weight_quantized - self.weight).masked_fill(mask.bool(), 0) clamp_low = ((- self.scale) * self.zero_point) clamp_high = (self.scale * (((2 ** self.bits) - 1) - self.zero_point)) weight = (torch.clamp(self.weight, clamp_low.item(), clamp_high.item()) + noise.detach()) output = F.linear(input, weight, self.bias) return output def extra_repr(self): return 'in_features={}, out_features={}, bias={}, quant_noise={}, bits={}, method={}'.format(self.in_features, self.out_features, (self.bias is not None), self.p, self.bits, self.method)
def hook_adapavgpool2d(m, x, y): x = x[0] out_size = _pair(m.output_size) k = (torch.Tensor(list(x.size()[2:])) / torch.Tensor(out_size)) k = torch.prod(torch.ceil(k)).item() flops_per_ele = k flops = (flops_per_ele * y.numel()) return int(flops)
def set_seed(seed): import random import tensorflow as tf seed %= random.seed(seed) np.random.seed(seed) tf.compat.v1.set_random_seed(seed) print(('using seed %s' % str(seed)))
class ContinuousMctsPolicies(AbstractMctsPolicies): def __init__(self, sample=None, initialize=None, score_c=1.0, pw_C=1.0, pw_alpha=0.25, *args, **kwargs): super(ContinuousMctsPolicies, self).__init__(*args, score=ContinuousRaveScore(score_c), widen=ProgressiveWiden(pw_C, pw_alpha), extract=MostVisitedExtract(), sample=sample, initialize=initialize, **kwargs) raise NotImplementedError('this does not yet exist')
class UP(TDAgent): def __init__(self, eval_points=10000, leverage=1.0, W=None): super(UP, self).__init__() self.eval_points = eval_points self.leverage = leverage self.W = W def init_portfolio(self, X): m = X.shape[1] self.W = np.matrix(mc_simplex((m - 1), self.eval_points)) self.S = np.matrix(np.ones(self.W.shape[0])).T leverage = max(self.leverage, (1.0 / m)) stretch = ((leverage - (1.0 / m)) / (1.0 - (1.0 / m))) self.W = (((self.W - (1.0 / m)) * stretch) + (1.0 / m)) def decide_by_history(self, x, last_b): x = self.get_last_rpv(x) x = np.reshape(x, (1, x.size)) if (self.W is None): self.init_portfolio(x) self.S = np.multiply(self.S, (self.W * np.matrix(x).T)) b = (self.W.T * self.S) pv = (b / np.sum(b)) pvn = np.ravel(pv) return pvn
def makeGraph(root): g = nx.DiGraph() nid = 0 (name, nid) = get_name(root, nid) g.add_node(name) good = [] bad = [] mid = [name] for child in root.children: nid = add_to_graph(g, name, child, good, bad, mid, nid) return (g, good, bad, mid)
def get_model(): model = Sequential() model.add(Conv2D(64, (3, 3), activation='relu', input_shape=(80, 80, 3))) model.add(MaxPooling2D(pool_size=(2, 2))) model.add(Conv2D(64, (3, 3), activation='relu')) model.add(MaxPooling2D(pool_size=(2, 2))) model.add(Flatten()) model.add(Dense(64, activation='relu')) model.add(Dense(1, activation='sigmoid')) return model
class SemiNet(torch.nn.Module): def __init__(self, in_channels=3, taskcla=None): super(SemiNet, self).__init__() self.conv1 = torch.nn.Conv2d(in_channels, 32, kernel_size=3, stride=1, padding=1) self.GN1 = torch.nn.GroupNorm(32, 32) self.BN1 = torch.nn.BatchNorm2d(32) self.conv2 = torch.nn.Conv2d(32, 64, kernel_size=3, stride=1, padding=1) self.conv3 = torch.nn.Conv2d(64, 128, kernel_size=3, stride=1, padding=1) self.GN2 = torch.nn.GroupNorm(32, 128) self.BN2 = torch.nn.BatchNorm2d(128) self.conv4 = torch.nn.Conv2d(128, 128, kernel_size=3, stride=1, padding=1) self.conv5 = torch.nn.Conv2d(128, 256, kernel_size=3, stride=1, padding=1) self.GN3 = torch.nn.GroupNorm(32, 256) self.BN3 = torch.nn.BatchNorm2d(256) self.conv6 = torch.nn.Conv2d(256, 256, kernel_size=3, stride=1, padding=1) self.maxpool = torch.nn.MaxPool2d(2) self.relu = torch.nn.ReLU() self.drop1 = torch.nn.Dropout(0.05) self.drop2 = torch.nn.Dropout(0.1) self.drop3 = torch.nn.Dropout(0.1) self.fc1 = torch.nn.Linear(4096, 1024) self.fc2 = torch.nn.Linear(1024, 512) self.fc3 = torch.nn.Linear(512, 10) return def forward(self, x): h = self.relu(self.BN1(self.conv1(x))) h = self.maxpool(self.relu(self.conv2(h))) h = self.relu(self.BN2(self.conv3(h))) h = self.drop1(self.maxpool(self.relu(self.conv4(h)))) h = self.relu(self.BN3(self.conv5(h))) h = self.maxpool(self.relu(self.conv6(h))) h = h.view(x.size(0), (- 1)) h = self.drop2(self.relu(self.fc1(h))) h = self.drop3(self.relu(self.fc2(h))) y = self.fc3(h) return y
def unset_hook(f: Callable[([Any], Any)]) -> Callable[([Any], Any)]: (f) def unset_hook_wrapper(self, **kwargs): f(self, **kwargs) self.attribution_model.is_hooked = False return unset_hook_wrapper
def gen_convs(inchannel, outchannel, bn=False): (yield nn.Conv2d(inchannel, outchannel, 3, padding=1)) if bn: (yield nn.BatchNorm2d(outchannel)) (yield nn.ReLU(inplace=True))
def lazily_load_dataset(corpus_type, opt): assert (corpus_type in ['train', 'valid']) def _lazy_dataset_loader(pt_file, corpus_type): dataset = torch.load(pt_file) logger.info(('Loading %s dataset from %s, number of examples: %d' % (corpus_type, pt_file, len(dataset)))) return dataset pt = (((opt.data + '.') + corpus_type) + '.pt') (yield _lazy_dataset_loader(pt, corpus_type))
class LlamaTokenizer(PreTrainedTokenizer): vocab_files_names = VOCAB_FILES_NAMES pretrained_vocab_files_map = PRETRAINED_VOCAB_FILES_MAP model_input_names = ['input_ids', 'attention_mask'] def __init__(self, vocab_file, unk_token='<unk>', bos_token='<s>', eos_token='</s>', sp_model_kwargs: Optional[Dict[(str, Any)]]=None, add_bos_token=True, add_eos_token=False, decode_with_prefix_space=False, clean_up_tokenization_spaces=False, **kwargs): self.sp_model_kwargs = ({} if (sp_model_kwargs is None) else sp_model_kwargs) super().__init__(bos_token=bos_token, eos_token=eos_token, unk_token=unk_token, clean_up_tokenization_spaces=clean_up_tokenization_spaces, **kwargs) self.vocab_file = vocab_file self.add_bos_token = add_bos_token self.add_eos_token = add_eos_token self.decode_with_prefix_space = decode_with_prefix_space self.sp_model = spm.SentencePieceProcessor(**self.sp_model_kwargs) self.sp_model.Load(vocab_file) self._no_prefix_space_tokens = None def no_prefix_space_tokens(self): if (self._no_prefix_space_tokens is None): vocab = self.convert_ids_to_tokens(list(range(self.vocab_size))) self._no_prefix_space_tokens = {i for (i, tok) in enumerate(vocab) if (not tok.startswith(''))} return self._no_prefix_space_tokens def vocab_size(self): return self.sp_model.get_piece_size() def bos_token_id(self) -> Optional[int]: return self.sp_model.bos_id() def eos_token_id(self) -> Optional[int]: return self.sp_model.eos_id() def get_vocab(self): vocab = {self.convert_ids_to_tokens(i): i for i in range(self.vocab_size)} vocab.update(self.added_tokens_encoder) return vocab def _tokenize(self, text): return self.sp_model.encode(text, out_type=str) def _convert_token_to_id(self, token): return self.sp_model.piece_to_id(token) def _convert_id_to_token(self, index): token = self.sp_model.IdToPiece(index) return token def _maybe_add_prefix_space(self, tokens, decoded): if (tokens and (tokens[0] not in self.no_prefix_space_tokens)): return (' ' + decoded) else: return decoded def convert_tokens_to_string(self, tokens): current_sub_tokens = [] out_string = '' prev_is_special = False for token in tokens: if (token in self.all_special_tokens): if (not prev_is_special): out_string += ' ' out_string += (self.sp_model.decode(current_sub_tokens) + token) prev_is_special = True current_sub_tokens = [] else: current_sub_tokens.append(token) prev_is_special = False out_string += self.sp_model.decode(current_sub_tokens) out_string = self._maybe_add_prefix_space(tokens=tokens, decoded=out_string) return out_string def save_vocabulary(self, save_directory, filename_prefix: Optional[str]=None) -> Tuple[str]: if (not os.path.isdir(save_directory)): logger.error(f'Vocabulary path ({save_directory}) should be a directory') return out_vocab_file = os.path.join(save_directory, (((filename_prefix + '-') if filename_prefix else '') + VOCAB_FILES_NAMES['vocab_file'])) if ((os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file)) and os.path.isfile(self.vocab_file)): copyfile(self.vocab_file, out_vocab_file) elif (not os.path.isfile(self.vocab_file)): with open(out_vocab_file, 'wb') as fi: content_spiece_model = self.sp_model.serialized_model_proto() fi.write(content_spiece_model) return (out_vocab_file,) def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None): if self.add_bos_token: bos_token_ids = [self.bos_token_id] else: bos_token_ids = [] output = (bos_token_ids + token_ids_0) if (token_ids_1 is not None): output = (output + token_ids_1) if self.add_eos_token: output = (output + [self.eos_token_id]) return output def get_special_tokens_mask(self, token_ids_0: List[int], token_ids_1: Optional[List[int]]=None, already_has_special_tokens: bool=False) -> List[int]: if already_has_special_tokens: return super().get_special_tokens_mask(token_ids_0=token_ids_0, token_ids_1=token_ids_1, already_has_special_tokens=True) if (token_ids_1 is None): return (([1] + ([0] * len(token_ids_0))) + [1]) return (((([1] + ([0] * len(token_ids_0))) + [1, 1]) + ([0] * len(token_ids_1))) + [1]) def create_token_type_ids_from_sequences(self, token_ids_0: List[int], token_ids_1: Optional[List[int]]=None) -> List[int]: eos = [self.eos_token_id] if (token_ids_1 is None): return (len((token_ids_0 + eos)) * [0]) return (len((((token_ids_0 + eos) + token_ids_1) + eos)) * [0])
class PPM(nn.ModuleList): def __init__(self, pool_scales, in_channels, channels, conv_cfg, norm_cfg, act_cfg, align_corners, **kwargs): super(PPM, self).__init__() self.pool_scales = pool_scales self.align_corners = align_corners self.in_channels = in_channels self.channels = channels self.conv_cfg = conv_cfg self.norm_cfg = norm_cfg self.act_cfg = act_cfg for pool_scale in pool_scales: self.append(nn.Sequential(nn.AdaptiveAvgPool2d(pool_scale), ConvModule(self.in_channels, self.channels, 1, conv_cfg=self.conv_cfg, norm_cfg=self.norm_cfg, act_cfg=self.act_cfg, **kwargs))) def forward(self, x): ppm_outs = [] for ppm in self: ppm_out = ppm(x) upsampled_ppm_out = resize(ppm_out, size=x.size()[2:], mode='bilinear', align_corners=self.align_corners) ppm_outs.append(upsampled_ppm_out) return ppm_outs
def get_connected_molecules(molecules): connected = [] for mol in molecules: if is_connected(mol): connected.append(mol) return connected
def test_parse_dimensions(): valid_examples = [('3x3', (3, 3)), ('4x2', (4, 2))] for (inp, expect) in valid_examples: out = parse_dimensions(inp) assert (out == expect), (out, '!=', expect)
_materialize('tensorflow') class NHWCConv2dValidPad(NHWCConv2d): def __init__(self, out_channels: Union[(int, z3.ExprRef)], stride: Union[(int, z3.ExprRef)], dilation_h: Union[(int, z3.ExprRef)], dilation_w: Union[(int, z3.ExprRef)]): super().__init__(out_channels, stride, dilation_h, dilation_w, 'VALID')
def set_log_dir(root_dir, exp_name): path_dict = {} os.makedirs(root_dir, exist_ok=True) exp_path = os.path.join(root_dir, exp_name) now = datetime.now(dateutil.tz.tzlocal()) timestamp = now.strftime('%Y_%m_%d_%H_%M_%S') prefix = ((exp_path + '_') + timestamp) os.makedirs(prefix) path_dict['prefix'] = prefix ckpt_path = os.path.join(prefix, 'Model') os.makedirs(ckpt_path) path_dict['ckpt_path'] = ckpt_path log_path = os.path.join(prefix, 'Log') os.makedirs(log_path) path_dict['log_path'] = log_path sample_path = os.path.join(prefix, 'Samples') os.makedirs(sample_path) path_dict['sample_path'] = sample_path return path_dict
def load_config(path='configs/default.yaml') -> dict: with open(path, 'r') as ymlfile: cfg = yaml.safe_load(ymlfile) return cfg
def handle_signal(signum, frame) -> None: log.info(f'Received interrupt signal {signum}') if waiting: sys.exit(1) global abort abort = True
def parse_args(): parser = argparse.ArgumentParser(description='Train a detector') parser.add_argument('config', help='train config file path') parser.add_argument('--work_dir', help='the dir to save logs and models') parser.add_argument('--resume_from', help='the checkpoint file to resume from') parser.add_argument('--validate', action='store_true', help='whether to evaluate the checkpoint during training') group_gpus = parser.add_mutually_exclusive_group() group_gpus.add_argument('--gpus', type=int, help='number of gpus to use (only applicable to non-distributed training)') group_gpus.add_argument('--gpu-ids', type=int, nargs='+', help='ids of gpus to use (only applicable to non-distributed training)') parser.add_argument('--seed', type=int, default=None, help='random seed') parser.add_argument('--deterministic', action='store_true', help='whether to set deterministic options for CUDNN backend.') parser.add_argument('--launcher', choices=['none', 'pytorch', 'slurm', 'mpi'], default='none', help='job launcher') parser.add_argument('--local_rank', type=int, default=0) parser.add_argument('--autoscale-lr', action='store_true', help='automatically scale lr with the number of gpus') args = parser.parse_args() if ('LOCAL_RANK' not in os.environ): os.environ['LOCAL_RANK'] = str(args.local_rank) return args
def grad(y, x, create_graph=True, keepdim=False): N = (y.size(0) if (len(y.size()) == 2) else 1) Ny = y.size((- 1)) Nx = x.size((- 1)) z = torch.ones_like(y[(..., 0)]) dy = [] for i in range(Ny): dy.append(torch.autograd.grad(y[(..., i)], x, grad_outputs=z, create_graph=create_graph)[0]) shape = np.array([N, Ny])[(2 - len(y.size())):] shape = (list(shape) if keepdim else list(shape[(shape > 1)])) return torch.cat(dy, dim=(- 1)).view((shape + [Nx]))
def is_time(token): without_time = re.sub('(\\d)*(\\d):(\\d\\d)([aA][mM]|[pP][Mm])', '', token).strip() return (not without_time)
class TaskDataCollatorForSeq2Seq(DataCollatorForSeq2Seq): def check_uniqueness(self, samples): assert (len(np.unique(samples)) == 1) def __call__(self, features): print('#### COLLATOR DEBUG #$$$$') print(features[0].keys()) print('') tasks = [d.pop('task') for d in features] labels_list_exist = False if ('labels_list' in features[0]): labels_list_exist = True labels_list = [d.pop('labels_list') for d in features] self.check_uniqueness(tasks) output = super().__call__(features) output['task'] = tasks[0] if labels_list_exist: output['labels_list'] = labels_list return output
def get_permutations(num_images): permutations = [] for i in range(num_images): for j in range(num_images): if (i != j): permutations.append((i, j)) return np.array(permutations)
class BN(PlainNetBasicBlockClass): def __init__(self, out_channels=None, copy_from=None, no_create=False, **kwargs): super(BN, self).__init__(**kwargs) self.no_create = no_create if (copy_from is not None): assert isinstance(copy_from, nn.BatchNorm2d) self.in_channels = copy_from.weight.shape[0] self.out_channels = copy_from.weight.shape[0] assert ((out_channels is None) or (out_channels == self.out_channels)) self.netblock = copy_from else: self.in_channels = out_channels self.out_channels = out_channels if no_create: return else: self.netblock = nn.BatchNorm2d(num_features=self.out_channels) def forward(self, x): return self.netblock(x) def __str__(self): return 'BN({})'.format(self.out_channels) def __repr__(self): return 'BN({}|{})'.format(self.block_name, self.out_channels) def get_output_resolution(self, input_resolution): return input_resolution def get_FLOPs(self, input_resolution): return ((input_resolution ** 2) * self.out_channels) def get_model_size(self): return self.out_channels def set_in_channels(self, c): self.in_channels = c self.out_channels = c if (not self.no_create): self.netblock = nn.BatchNorm2d(num_features=self.out_channels) self.netblock.train() self.netblock.requires_grad_(True) def create_from_str(cls, s, no_create=False, **kwargs): assert BN.is_instance_from_str(s) idx = _get_right_parentheses_index_(s) assert (idx is not None) param_str = s[len('BN('):idx] tmp_idx = param_str.find('|') if (tmp_idx < 0): tmp_block_name = 'uuid{}'.format(uuid.uuid4().hex) else: tmp_block_name = param_str[0:tmp_idx] param_str = param_str[(tmp_idx + 1):] out_channels = int(param_str) return (BN(out_channels=out_channels, block_name=tmp_block_name, no_create=no_create), s[(idx + 1):])
class SeparableConv2d_same(nn.Module): def __init__(self, inplanes, planes, kernel_size=3, stride=1, dilation=1, bias=False): super(SeparableConv2d_same, self).__init__() self.conv1 = nn.Conv2d(inplanes, inplanes, kernel_size, stride, 0, dilation, groups=inplanes, bias=bias) self.pointwise = nn.Conv2d(inplanes, planes, 1, 1, 0, 1, 1, bias=bias) def forward(self, x): x = fixed_padding(x, self.conv1.kernel_size[0], dilation=self.conv1.dilation[0]) x = self.conv1(x) x = self.pointwise(x) return x
class MeanShift(nn.Conv2d): def __init__(self, rgb_range, rgb_mean, rgb_std, sign=(- 1)): super(MeanShift, self).__init__(3, 3, kernel_size=1) std = torch.Tensor(rgb_std) self.weight.data = torch.eye(3).view(3, 3, 1, 1) self.weight.data.div_(std.view(3, 1, 1, 1)) self.bias.data = ((sign * rgb_range) * torch.Tensor(rgb_mean)) self.bias.data.div_(std) self.requires_grad = False
class ResNetConvBackbone(nn.Module): def __init__(self, num_layers: int, pretrained: bool) -> None: super(ResNetConvBackbone, self).__init__() self.resnet = get_vanilla_resnet_model(num_layers, pretrained) def forward(self, x: torch.Tensor) -> torch.Tensor: x = self.resnet.conv1(x) x = self.resnet.bn1(x) x = self.resnet.relu(x) x = self.resnet.maxpool(x) x = self.resnet.layer1(x) x = self.resnet.layer2(x) x = self.resnet.layer3(x) x = self.resnet.layer4(x) x = self.resnet.avgpool(x) x = torch.flatten(x, 1) return x
def get_bt_sentences(train_path): pkl_path = Path(train_path).parent.joinpath(f'train_aug_bt_data.pkl') sentence_to_aug_sentence = get_backtrans_data_dict(pkl_path, train_path) return sentence_to_aug_sentence
def create_data_loader(root_dir, batch_size, nproc): dir_path = os.path.realpath(root_dir) data_transform = transforms.Compose([transforms.Resize(256), transforms.ColorJitter(), transforms.RandomCrop(224), transforms.RandomHorizontalFlip(), transforms.Resize(128), transforms.ToTensor()]) catdogs = ImageFolder(dir_path, data_transform) dataset_size = len(catdogs) if (('SANITY_CHECK' in os.environ) and (os.environ['SANITY_CHECK'] == '1')): train_size = ((2 * nproc) * batch_size) val_size = ((2 * nproc) * batch_size) test_size = ((dataset_size - train_size) - val_size) (train_set, val_set, _) = torch.utils.data.random_split(catdogs, [train_size, val_size, test_size]) else: train_split = 0.8 train_size = int((dataset_size * train_split)) val_size = (dataset_size - train_size) (train_set, val_set) = torch.utils.data.random_split(catdogs, [train_size, val_size]) train_loader = DataLoader(train_set, batch_size=batch_size, shuffle=True, num_workers=0) val_loader = DataLoader(val_set, batch_size=batch_size, shuffle=False, num_workers=0) size = (train_size, val_size) return (train_loader, val_loader, size)
def make_module(sym_model, ctx, input_desc): assert (isinstance(sym_model, tuple) and isinstance(sym_model[0], mx.symbol.Symbol)) (symnet, args, auxs) = sym_model mod = mx.module.module.Module(symbol=symnet, data_names=[d.name for d in input_desc], label_names=None, context=ctx) mod.bind(input_desc, for_training=False) mod.set_params(args, auxs, allow_missing=True) return mod
class ANY(): def __init__(self, _type): self._type = _type def __eq__(self, other): return isinstance(other, self._type) def __repr__(self): return f'ANY({self._type.__name__})'
def main(_): RANDOM_SEED = 66 np.random.seed(RANDOM_SEED) output_dir = os.path.join(FLAGS.output_dir, FLAGS.category) sample_dir = os.path.join(output_dir, 'synthesis') log_dir = os.path.join(output_dir, 'log') model_dir = os.path.join(output_dir, 'checkpoints') if tf.gfile.Exists(log_dir): tf.gfile.DeleteRecursively(log_dir) tf.gfile.MakeDirs(log_dir) if tf.gfile.Exists(sample_dir): tf.gfile.DeleteRecursively(sample_dir) tf.gfile.MakeDirs(sample_dir) if tf.gfile.Exists(model_dir): tf.gfile.DeleteRecursively(model_dir) tf.gfile.MakeDirs(model_dir) train_data = data_io.getObj(FLAGS.data_path, FLAGS.category, cube_len=FLAGS.cube_len, num_voxels=FLAGS.train_size, low_bound=0, up_bound=1) data_io.saveVoxelsToMat(train_data, ('%s/observed_data.mat' % output_dir), cmin=0, cmax=1) voxel_mean = train_data.mean() train_data = (train_data - voxel_mean) train_data = train_data[(..., np.newaxis)] FLAGS.num_batches = int(math.ceil((len(train_data) / FLAGS.batch_size))) print('Reading voxel data {}, shape: {}'.format(FLAGS.category, train_data.shape)) print(('min: %.4f\tmax: %.4f\tmean: %.4f' % (train_data.min(), train_data.max(), voxel_mean))) net = DescriptorNet3D(FLAGS) net.build_model() with tf.Session() as sess: sess.run(tf.global_variables_initializer()) sample_size = (FLAGS.sample_batch * FLAGS.num_batches) sample_voxels = np.random.randn(sample_size, FLAGS.cube_len, FLAGS.cube_len, FLAGS.cube_len, 1) saver = tf.train.Saver(max_to_keep=50) des_loss_epoch = [] rec_err_epoch = [] plt.ion() for epoch in range(FLAGS.num_epochs): d_grad_vec = [] des_loss_vec = [] rec_err_vec = [] start_time = time.time() sess.run(net.reset_grads) for i in range(FLAGS.num_batches): obs_data = train_data[(i * FLAGS.batch_size):min(len(train_data), ((i + 1) * FLAGS.batch_size))] syn_data = sample_voxels[(i * FLAGS.sample_batch):((i + 1) * FLAGS.sample_batch)] if (epoch < 100): syn = sess.run(net.langevin_descriptor_noise, feed_dict={net.syn: syn_data}) else: syn = sess.run(net.langevin_descriptor, feed_dict={net.syn: syn_data}) (des_grads, des_loss) = sess.run([net.des_grads, net.des_loss, net.update_d_grads], feed_dict={net.obs: obs_data, net.syn: syn})[:2] d_grad_vec.append(des_grads) des_loss_vec.append(des_loss) rec_err = sess.run(net.recon_err, feed_dict={net.obs: obs_data, net.syn: syn}) rec_err_vec.append(rec_err) sample_voxels[(i * FLAGS.sample_batch):((i + 1) * FLAGS.sample_batch)] = syn sess.run(net.apply_d_grads) (d_grad_mean, des_loss_mean, rec_err_mean) = (float(np.mean(d_grad_vec)), float(np.mean(des_loss_vec)), float(np.mean(rec_err_vec))) des_loss_epoch.append(des_loss_mean) rec_err_epoch.append(rec_err_mean) end_time = time.time() print(('Epoch #%d, descriptor loss: %.4f, descriptor SSD weight: %.4f, Avg MSE: %4.4f, time: %.2fs' % (epoch, des_loss_mean, d_grad_mean, rec_err_mean, (end_time - start_time)))) if ((epoch % FLAGS.log_step) == 0): if (not os.path.exists(sample_dir)): os.makedirs(sample_dir) data_io.saveVoxelsToMat((sample_voxels + voxel_mean), ('%s/sample%04d.mat' % (sample_dir, epoch)), cmin=0, cmax=1) if (not os.path.exists(model_dir)): os.makedirs(model_dir) saver.save(sess, ('%s/%s' % (model_dir, 'net.ckpt')), global_step=epoch) if (not os.path.exists(log_dir)): os.makedirs(log_dir) plt.figure(1) data_io.draw_graph(plt, des_loss_epoch, 'des_loss', log_dir, 'r') plt.figure(2) data_io.draw_graph(plt, rec_err_epoch, 'recon_error', log_dir, 'b')
def find_similar_token(token, tokens): token = re.sub('-\\d\\d$', '', token) for (i, t) in enumerate(tokens): if (token == t): return tokens[i] return None
class ConvertCommand(BaseTransformersCLICommand): def register_subcommand(parser: ArgumentParser): train_parser = parser.add_parser('convert', help='CLI tool to run convert model from original author checkpoints to Transformers PyTorch checkpoints.') train_parser.add_argument('--model_type', type=str, required=True, help="Model's type.") train_parser.add_argument('--tf_checkpoint', type=str, required=True, help='TensorFlow checkpoint path or folder.') train_parser.add_argument('--pytorch_dump_output', type=str, required=True, help='Path to the PyTorch saved model output.') train_parser.add_argument('--config', type=str, default='', help='Configuration file path or folder.') train_parser.add_argument('--finetuning_task_name', type=str, default=None, help='Optional fine-tuning task name if the TF model was a finetuned model.') train_parser.set_defaults(func=convert_command_factory) def __init__(self, model_type: str, tf_checkpoint: str, pytorch_dump_output: str, config: str, finetuning_task_name: str, *args): self._logger = logging.get_logger('transformers-cli/converting') self._logger.info(f'Loading model {model_type}') self._model_type = model_type self._tf_checkpoint = tf_checkpoint self._pytorch_dump_output = pytorch_dump_output self._config = config self._finetuning_task_name = finetuning_task_name def run(self): if (self._model_type == 'albert'): try: from ..models.albert.convert_albert_original_tf_checkpoint_to_pytorch import convert_tf_checkpoint_to_pytorch except ImportError: raise ImportError(IMPORT_ERROR_MESSAGE) convert_tf_checkpoint_to_pytorch(self._tf_checkpoint, self._config, self._pytorch_dump_output) elif (self._model_type == 'bert'): try: from ..models.bert.convert_bert_original_tf_checkpoint_to_pytorch import convert_tf_checkpoint_to_pytorch except ImportError: raise ImportError(IMPORT_ERROR_MESSAGE) convert_tf_checkpoint_to_pytorch(self._tf_checkpoint, self._config, self._pytorch_dump_output) elif (self._model_type == 'funnel'): try: from ..models.funnel.convert_funnel_original_tf_checkpoint_to_pytorch import convert_tf_checkpoint_to_pytorch except ImportError: raise ImportError(IMPORT_ERROR_MESSAGE) convert_tf_checkpoint_to_pytorch(self._tf_checkpoint, self._config, self._pytorch_dump_output) elif (self._model_type == 't5'): try: from ..models.t5.convert_t5_original_tf_checkpoint_to_pytorch import convert_tf_checkpoint_to_pytorch except ImportError: raise ImportError(IMPORT_ERROR_MESSAGE) convert_tf_checkpoint_to_pytorch(self._tf_checkpoint, self._config, self._pytorch_dump_output) elif (self._model_type == 'gpt'): from ..models.openai.convert_openai_original_tf_checkpoint_to_pytorch import convert_openai_checkpoint_to_pytorch convert_openai_checkpoint_to_pytorch(self._tf_checkpoint, self._config, self._pytorch_dump_output) elif (self._model_type == 'transfo_xl'): try: from ..models.transfo_xl.convert_transfo_xl_original_tf_checkpoint_to_pytorch import convert_transfo_xl_checkpoint_to_pytorch except ImportError: raise ImportError(IMPORT_ERROR_MESSAGE) if ('ckpt' in self._tf_checkpoint.lower()): TF_CHECKPOINT = self._tf_checkpoint TF_DATASET_FILE = '' else: TF_DATASET_FILE = self._tf_checkpoint TF_CHECKPOINT = '' convert_transfo_xl_checkpoint_to_pytorch(TF_CHECKPOINT, self._config, self._pytorch_dump_output, TF_DATASET_FILE) elif (self._model_type == 'gpt2'): try: from ..models.gpt2.convert_gpt2_original_tf_checkpoint_to_pytorch import convert_gpt2_checkpoint_to_pytorch except ImportError: raise ImportError(IMPORT_ERROR_MESSAGE) convert_gpt2_checkpoint_to_pytorch(self._tf_checkpoint, self._config, self._pytorch_dump_output) elif (self._model_type == 'xlnet'): try: from ..models.xlnet.convert_xlnet_original_tf_checkpoint_to_pytorch import convert_xlnet_checkpoint_to_pytorch except ImportError: raise ImportError(IMPORT_ERROR_MESSAGE) convert_xlnet_checkpoint_to_pytorch(self._tf_checkpoint, self._config, self._pytorch_dump_output, self._finetuning_task_name) elif (self._model_type == 'xlm'): from ..models.xlm.convert_xlm_original_pytorch_checkpoint_to_pytorch import convert_xlm_checkpoint_to_pytorch convert_xlm_checkpoint_to_pytorch(self._tf_checkpoint, self._pytorch_dump_output) elif (self._model_type == 'lxmert'): from ..models.lxmert.convert_lxmert_original_tf_checkpoint_to_pytorch import convert_lxmert_checkpoint_to_pytorch convert_lxmert_checkpoint_to_pytorch(self._tf_checkpoint, self._pytorch_dump_output) elif (self._model_type == 'rembert'): from ..models.rembert.convert_rembert_tf_checkpoint_to_pytorch import convert_rembert_tf_checkpoint_to_pytorch convert_rembert_tf_checkpoint_to_pytorch(self._tf_checkpoint, self._config, self._pytorch_dump_output) else: raise ValueError('--model_type should be selected in the list [bert, gpt, gpt2, t5, transfo_xl, xlnet, xlm, lxmert]')
class SimilarityClient(): def __init__(self, stub): self.stub = stub def search(self, id, k): request = recall_pb2.Query(userID=id, k=k) try: candidates = self.stub.searchCandidates(request) return candidates.candidate except Exception as e: logging.warning('RPC failed:{}'.format(e)) return def getMetrics(self): try: msg = self.stub.getMetrics(recall_pb2.ServerMessage()) except Exception as e: logging.warning('RPC failed:{}'.format(e)) logging.info(('Got metrics: ' + msg.str)) def resetMetrics(self): try: self.stub.resetMetrics(recall_pb2.ServerMessage()) except Exception as e: logging.warning('RPC failed:{}'.format(e))
def load_model_from_config(config, sd): model = instantiate_from_config(config) model.load_state_dict(sd, strict=False) model.cuda() model.eval() return model
class LowRankAdapter(nn.Module): def __init__(self, config): super().__init__() self.config = config self.input_dim = config.input_dim self.down_sample_size = (self.input_dim // config.reduction_factor) self.activation = Activations(config.non_linearity.lower()) self.down_sampler = LowRankLinear(self.input_dim, self.down_sample_size, w_init=config.low_rank_w_init, rank=config.low_rank_rank) self.up_sampler = LowRankLinear(self.down_sample_size, self.input_dim, w_init=config.low_rank_w_init, rank=config.low_rank_rank) def forward(self, x): z = self.down_sampler(x) z = self.activation(z) output = self.up_sampler(z) return output
def relevant_clauses(clauses, object_pair_list, converter): def _relevant(c): for l in c: triple = converter.num2triple(abs(l)) for obj_pair in object_pair_list: if ((obj_pair == [triple[1], triple[2]]) or (obj_pair == [triple[2], triple[1]])): return True return False return [c for c in clauses if _relevant(c)]
class Option(NetOption): def __init__(self, conf_path, args): super(Option, self).__init__() self.conf = ConfigFactory.parse_file(conf_path) self.save_path = self.conf['save_path'] self.dataPath = self.conf['dataPath'] self.dataset = self.conf['dataset'] self.nGPU = self.conf['nGPU'] self.GPU = self.conf['GPU'] self.visible_devices = args.gpu self.network = self.conf['network'] self.nThreads = self.conf['nThreads'] self.nEpochs = self.conf['nEpochs'] self.batchSize = self.conf['batchSize'] self.momentum = self.conf['momentum'] if (args.wd == None): self.weightDecay = float(self.conf['weightDecay']) else: self.weightDecay = float(args.wd) self.opt_type = self.conf['opt_type'] self.warmup_epochs = self.conf['warmup_epochs'] if (args.lrs == None): self.lr_S = self.conf['lr_S'] else: self.lr_S = args.lrs self.lrPolicy_S = self.conf['lrPolicy_S'] self.step_S = self.conf['step_S'] self.decayRate_S = self.conf['decayRate_S'] if (args.qw == None): self.qw = self.conf['qw'] else: self.qw = args.qw if (args.qa == None): self.qa = self.conf['qa'] else: self.qa = args.qa self.experimentID = f"autorecon_{self.network}_qwqa_{self.qw}_{self.qa}_threshold_{args.passing_threshold}_kd_scale_{args.kd_scale}_ce_scale_{args.ce_scale}_lrS_{self.lr_S}_WD_{self.weightDecay}_id_{self.conf['experimentID']}" self.nClasses = self.conf['nClasses'] self.temperature = self.conf['temperature'] self.alpha = self.conf['alpha'] self.ce_scale = args.ce_scale self.kd_scale = args.kd_scale self.latent_dim = self.conf['latent_dim'] self.img_size = self.conf['img_size'] self.channels = self.conf['channels'] self.lr_G = self.conf['lr_G'] self.lrPolicy_G = self.conf['lrPolicy_G'] self.step_G = self.conf['step_G'] self.decayRate_G = self.conf['decayRate_G'] self.b1 = self.conf['b1'] self.b2 = self.conf['b2'] self.save = args.save self.passing_threshold = args.passing_threshold self.passing_threshold_first = args.passing_threshold self.threshold_decay_rate = args.threshold_decay_rate self.threshold_decay_ep = args.threshold_decay_ep self.alpha_iter = args.alpha_iter self.adalr = args.adalr def set_save_path(self): self.save_path = (self.save_path + 'log_{}_bs{:d}_lr{:.4f}_TELCNN_baseline_opt{}_qw{:d}_qa{:d}_epoch{}_{}/'.format(self.dataset, self.batchSize, self.lr, self.opt_type, self.qw, self.qa, self.nEpochs, self.experimentID)) if os.path.exists(self.save_path): print('{} file exist!'.format(self.save_path)) if (not os.path.exists(self.save_path)): os.makedirs(self.save_path) def paramscheck(self, logger): logger.info('|===>The used PyTorch version is {}'.format(self.torch_version)) if (self.dataset in ['cifar10', 'mnist']): self.nClasses = 10 elif (self.dataset == 'cifar100'): self.nClasses = 100 elif ((self.dataset == 'imagenet') or 'thi_imgnet'): self.nClasses = 1000 elif (self.dataset == 'imagenet100'): self.nClasses = 100
def Align_LC(mjd, mjd2, data, data2, error, error2): if (len(data2) > len(data)): new_data2 = [] new_error2 = [] new_mjd2 = [] new_mjd = np.copy(mjd) new_error = np.copy(error) new_data = np.copy(data) count = 0 for index in xrange(len(data)): where = np.where((mjd2 == mjd[index])) if (np.array_equal(where[0], []) is False): new_data2.append(data2[where]) new_error2.append(error2[where]) new_mjd2.append(mjd2[where]) else: new_mjd = np.delete(new_mjd, (index - count)) new_error = np.delete(new_error, (index - count)) new_data = np.delete(new_data, (index - count)) count = (count + 1) new_data2 = np.asarray(new_data2).flatten() new_error2 = np.asarray(new_error2).flatten() else: new_data = [] new_error = [] new_mjd = [] new_mjd2 = np.copy(mjd2) new_error2 = np.copy(error2) new_data2 = np.copy(data2) count = 0 for index in xrange(len(data2)): where = np.where((mjd == mjd2[index])) if (np.array_equal(where[0], []) is False): new_data.append(data[where]) new_error.append(error[where]) new_mjd.append(mjd[where]) else: new_mjd2 = np.delete(new_mjd2, (index - count)) new_error2 = np.delete(new_error2, (index - count)) new_data2 = np.delete(new_data2, (index - count)) count = (count + 1) new_data = np.asarray(new_data).flatten() new_mjd = np.asarray(new_mjd).flatten() new_error = np.asarray(new_error).flatten() return (new_data, new_data2, new_mjd, new_error, new_error2)
def test_classifier(P, model, loader, criterion, steps, logger=None): metric_logger = MetricLogger(delimiter=' ') if (logger is None): log_ = print else: log_ = logger.log mode = model.training model.eval() acc = 0.0 acc_ema = 0.0 if hasattr(P, 'moving_inner_lr'): inner_step_ema = P.moving_inner_lr else: inner_step_ema = P.inner_lr for (n, batch) in enumerate(loader): if ((n * P.test_batch_size) > P.max_test_task): break (train_inputs, train_targets) = batch['train'] train_inputs = train_inputs.to(device, non_blocking=True) train_targets = train_targets.to(device, non_blocking=True) (test_inputs, test_targets) = batch['test'] test_inputs = test_inputs.to(device, non_blocking=True) test_targets = test_targets.to(device, non_blocking=True) for (task_idx, (train_input, train_target, test_input, test_target)) in enumerate(zip(train_inputs, train_targets, test_inputs, test_targets)): (params, loss_train) = maml_inner_adapt(model, criterion, train_input, train_target, P.inner_lr, P.inner_steps_test, first_order=True) (params_ema, loss_train_ema) = maml_inner_adapt(model, criterion, train_input, train_target, inner_step_ema, P.inner_steps_test, first_order=True, params=P.moving_average) with torch.no_grad(): outputs_test = model(test_input, params=params) outputs_test_ema = model(test_input, params=params_ema) loss = criterion(outputs_test, test_target) loss_ema = criterion(outputs_test_ema, test_target) if (not P.regression): acc = accuracy(outputs_test, test_target, topk=(1,))[0].item() acc_ema = accuracy(outputs_test_ema, test_target, topk=(1,))[0].item() elif (P.dataset == 'shapenet'): acc = (- degree_loss(outputs_test, test_target).item()) acc_ema = (- degree_loss(outputs_test_ema, test_target).item()) elif (P.dataset == 'pose'): acc = (- loss.item()) acc_ema = (- loss_ema.item()) else: raise NotImplementedError() metric_logger.meters['loss_train_ori'].update(loss_train.item()) metric_logger.meters['loss_ori'].update(loss.item()) metric_logger.meters['acc_ori'].update(acc) metric_logger.meters['loss_train'].update(loss_train_ema.item()) metric_logger.meters['loss'].update(loss_ema.item()) metric_logger.meters['acc'].update(acc_ema) metric_logger.synchronize_between_processes() log_((' * [ %.3f] [LossOutEMA %.3f] [LossInEMA %.3f]' % (metric_logger.acc.global_avg, metric_logger.loss.global_avg, metric_logger.loss_train.global_avg))) log_((' * [ %.3f] [LossOut %.3f] [LossIn %.3f]' % (metric_logger.acc_ori.global_avg, metric_logger.loss_ori.global_avg, metric_logger.loss_train_ori.global_avg))) if (logger is not None): logger.scalar_summary('eval/acc', metric_logger.acc.global_avg, steps) logger.scalar_summary('eval/loss_test', metric_logger.loss.global_avg, steps) logger.scalar_summary('eval/loss_train', metric_logger.loss_train.global_avg, steps) logger.scalar_summary('eval/acc_ori', metric_logger.acc_ori.global_avg, steps) logger.scalar_summary('eval/loss_test_ori', metric_logger.loss_ori.global_avg, steps) logger.scalar_summary('eval/loss_train_ori', metric_logger.loss_train_ori.global_avg, steps) model.train(mode) return metric_logger.acc.global_avg
def train(dataset='mnist', model_name='sl', batch_size=128, epochs=50, noise_ratio=0, asym=False, alpha=1.0, beta=1.0): print(('Dataset: %s, model: %s, batch: %s, epochs: %s, noise ratio: %s%%, asymmetric: %s, alpha: %s, beta: %s' % (dataset, model_name, batch_size, epochs, noise_ratio, asym, alpha, beta))) (X_train, y_train, y_train_clean, X_test, y_test) = get_data(dataset, noise_ratio, asym=asym, random_shuffle=False) n_images = X_train.shape[0] image_shape = X_train.shape[1:] num_classes = y_train.shape[1] print('n_images', n_images, 'num_classes', num_classes, 'image_shape:', image_shape) P = np.eye(num_classes) model = get_model(dataset, input_tensor=None, input_shape=image_shape, num_classes=num_classes) if (dataset == 'cifar-100'): optimizer = SGD(lr=0.1, decay=0.005, momentum=0.9) else: optimizer = SGD(lr=0.1, decay=0.0001, momentum=0.9) if (model_name == 'ce'): loss = cross_entropy elif (model_name == 'sl'): loss = symmetric_cross_entropy(alpha, beta) elif (model_name == 'lsr'): loss = lsr elif (model_name == 'joint'): loss = joint_optimization_loss elif (model_name == 'gce'): loss = generalized_cross_entropy elif (model_name == 'boot_hard'): loss = boot_hard elif (model_name == 'boot_soft'): loss = boot_soft elif (model_name == 'forward'): loss = forward(P) elif (model_name == 'backward'): loss = backward(P) else: print(('Model %s is unimplemented!' % model_name)) exit(0) model.compile(loss=loss, optimizer=optimizer, metrics=['accuracy']) if asym: model_save_file = ('model/asym_%s_%s_%s.{epoch:02d}.hdf5' % (model_name, dataset, noise_ratio)) else: model_save_file = ('model/%s_%s_%s.{epoch:02d}.hdf5' % (model_name, dataset, noise_ratio)) callbacks = [] if (model_name == 'sl'): cp_callback = ModelCheckpoint(model_save_file, monitor='val_loss', verbose=0, save_best_only=False, save_weights_only=True, period=1) callbacks.append(cp_callback) else: cp_callback = ModelCheckpoint(model_save_file, monitor='val_loss', verbose=0, save_best_only=False, save_weights_only=True, period=1) callbacks.append(cp_callback) lr_scheduler = get_lr_scheduler(dataset) callbacks.append(lr_scheduler) callbacks.append(SGDLearningRateTracker(model)) log_callback = LoggerCallback(model, X_train, y_train, y_train_clean, X_test, y_test, dataset, model_name, noise_ratio, asym, epochs, alpha, beta) callbacks.append(log_callback) if (dataset in ['mnist', 'svhn']): datagen = ImageDataGenerator() elif (dataset in ['cifar-10']): datagen = ImageDataGenerator(width_shift_range=0.2, height_shift_range=0.2, horizontal_flip=True) else: datagen = ImageDataGenerator(rotation_range=20, width_shift_range=0.2, height_shift_range=0.2, horizontal_flip=True) datagen.fit(X_train) model.fit_generator(datagen.flow(X_train, y_train, batch_size=batch_size), steps_per_epoch=(len(X_train) / batch_size), epochs=epochs, validation_data=(X_test, y_test), verbose=1, callbacks=callbacks)
def _shufflenetv2_mpncov(arch, pretrained, progress, *args, **kwargs): model = ShuffleNetV2_MPNCOV(*args, **kwargs) if pretrained: model_url = model_urls[arch] if (model_url is None): raise NotImplementedError('pretrained {} is not supported as of now'.format(arch)) else: state_dict = load_state_dict_from_url(model_url, progress=progress) model.load_state_dict(state_dict) return model
class RandomGaussianNoise(object): def __init__(self, p=0.5, noise_variance=(0, 0.5)): super().__init__() self.p = p self.noise_variance = noise_variance def __call__(self, img_and_mask: Tuple[(np.ndarray, np.ndarray, np.ndarray)]) -> Tuple[(np.ndarray, np.ndarray, np.ndarray)]: (img, _, mask) = img_and_mask noised_image = img mask = np.stack(([mask] * 4)) if (torch.rand(1) < self.p): if (self.noise_variance[0] == self.noise_variance[1]): variance = self.noise_variance[0] else: variance = random.uniform(self.noise_variance[0], self.noise_variance[1]) noised_image = (img + np.random.normal(0.0, variance, size=img.shape)) noised_image[(mask == 0)] = img[(mask == 0)] return (noised_image, img_and_mask[1], mask)
def validate(data_path, device, model, word2idx, entity2idx, model_name, return_hits_at_k): model.eval() data = process_text_file(data_path) answers = [] data_gen = data_generator(data=data, word2ix=word2idx, entity2idx=entity2idx) total_correct = 0 error_count = 0 hit_at_1 = 0 hit_at_5 = 0 hit_at_10 = 0 candidates_with_scores = [] writeCandidatesToFile = False for i in tqdm(range(len(data))): try: d = next(data_gen) head = d[0].to(device) question = d[1].to(device) ans = d[2] ques_len = d[3].unsqueeze(0) tail_test = torch.tensor(ans, dtype=torch.long).to(device) scores = model.get_score_ranked(head=head, sentence=question, sent_len=ques_len)[0] mask = torch.zeros(len(entity2idx)).to(device) mask[head] = 1 new_scores = (scores - (mask * 99999)) pred_ans = torch.argmax(new_scores).item() if (pred_ans == head.item()): print('Head and answer same') print(torch.max(new_scores)) print(torch.min(new_scores)) if writeCandidatesToFile: entry = {} entry['question'] = d[(- 1)] head_text = idx2entity[head.item()] entry['head'] = head_text (s, c) = torch.topk(new_scores, 200) s = s.cpu().detach().numpy() c = c.cpu().detach().numpy() cands = [] for cand in c: cands.append(idx2entity[cand]) entry['scores'] = s entry['candidates'] = cands correct_ans = [] for a in ans: correct_ans.append(idx2entity[a]) entry['answers'] = correct_ans candidates_with_scores.append(entry) if inTopk(new_scores, ans, 1): hit_at_1 += 1 if inTopk(new_scores, ans, 5): hit_at_5 += 1 if inTopk(new_scores, ans, 10): hit_at_10 += 1 if (type(ans) is int): ans = [ans] is_correct = 0 if (pred_ans in ans): total_correct += 1 is_correct = 1 else: num_incorrect += 1 q_text = d[(- 1)] answers.append(((((q_text + '\t') + str(pred_ans)) + '\t') + str(is_correct))) except: error_count += 1 accuracy = (total_correct / len(data)) if return_hits_at_k: return (answers, accuracy, (hit_at_1 / len(data)), (hit_at_5 / len(data)), (hit_at_10 / len(data))) else: return (answers, accuracy)
def test_global_context_block_1d(): N = 10 C = 128 reduction = 16 data = torch.randn(N, C, 7) model = GlobalContextBlock1D(in_channels=C, reduction=reduction) print(model) outputs = model(data) print(outputs.shape) assert (outputs.shape == (N, C, 7))
def clean_html_one_sample(sample): roles = ['human', 'gpt'] if (len(sample['conversations']) <= 1): return (sample, 1) if (sample['conversations'][0]['from'] != 'human'): sample['conversations'] = sample['conversations'][1:] if (len(sample['conversations']) <= 1): return (sample, 1) if (sample['conversations'][(- 1)]['from'] == 'human'): sample['conversations'] = sample['conversations'][:(- 1)] if (len(sample['conversations']) <= 1): return (sample, 1) for (i, c) in enumerate(sample['conversations']): if (c['from'] != roles[(i % 2)]): return (sample, 2) if contain_blocked_words(c['value']): return (sample, 3) try: new_val = html_to_markdown(c['value']) except (bs4.builder.ParserRejectedMarkup, AssertionError): return (sample, 4) c['value'] = new_val return (sample, 0)
class SetupArgs(): def __init__(self, sampler_cls, sampler_args, seed): self.sampler_cls = sampler_cls self.sampler_args = sampler_args self.seed = seed
class Bow(BaseBow): def __init__(self): super().__init__('bow', weight=30, damage=D.Dice.from_str('d2'), material=M.Wood, hit=0)
def common_config(parser): parser.add_argument('--percentage', '-pc', type=int, default=100, help='% of samples to use for % experiment, defaults to 100 for no experiment') parser.add_argument('--shuffle', '-sh', type=str_to_bool, default=True, help='shuffle test set (after splitting)') parser.add_argument('--speed_benchmark', '-speed', type=str_to_bool, default=False, help='Excecute speed benchmark') parser.add_argument('--space_benchmark', '-space', type=str_to_bool, default=False, help='Excecute space benchmark') parser.add_argument('--gaussian-blur', '-gb', type=str_to_bool, default=False, help='Apply gaussian filtering on anomaly map') parser.add_argument('--sigma', '-s', type=float, default=4, help='sigma of gaussian filter') parser.add_argument('--seed', type=int, default=10, help='Random seed') parser.add_argument('--load_pretrained', '-lp', type=str_to_bool, default=False, help='Load encoder pretrained with CCD') parser.add_argument('--disable_wandb', '-dw', type=str_to_bool, default=False, help='disable wandb logging') parser.add_argument('--eval', '-ev', type=str_to_bool, default=False, help='Evaluation mode') parser.add_argument('--no_dice', type=str_to_bool, default=False, help='do not calculate dice (used to save inference time)') parser.add_argument('--restoration', '-res', type=str_to_bool, default=False, help='VAE restoration') parser.add_argument('--image_size', type=int, default=128, help='Image size') parser.add_argument('--img_channels', type=int, default=1, help='Image channels') parser.add_argument('--center', '-cnt', type=str_to_bool, default=False, help='Whether to center the samples to [-1,1] range.') parser.add_argument('--stadardize', '-stad', type=str_to_bool, default=False, help='Whether to standardize the samples to N(0,1) dataset-wise.') parser.add_argument('--modality', '-mod', type=str, default='MRI', help='MRI sequence') parser.add_argument('--normal_split', '-ns', type=float, default=0.95, help='normal set split') parser.add_argument('--anomal_split', '-as', type=float, default=0.9, help='anomaly set split') parser.add_argument('--num_workers', type=int, default=4, help='Number of workers') parser.add_argument('--sequence', '-seq', type=str, default='t2', help='MRI sequence', choices=['t1', 't2']) parser.add_argument('--brats_t1', type=str_to_bool, default=True, help='True for BraTS T1, false for ATLAS') parser.add_argument('--slice_range', type=int, nargs='+', default=(0, 155), help='Lower and Upper slice index') parser.add_argument('--normalize', type=str_to_bool, default=False, help='Normalize images to 98th percentile and scale to [0,1]') parser.add_argument('--equalize_histogram', type=str_to_bool, default=False, help='Equalize histogram') parser.add_argument('--sup_devices', type=str_to_bool, default=False, help='Whether to include CXRs with support devices') parser.add_argument('--AP_only', type=str_to_bool, default=True, help='Whether to include only AP CXRs') parser.add_argument('--pathology', type=str, default='effusion', help='Pathology of test set.', choices=['enlarged', 'effusion', 'opacity']) parser.add_argument('--sex', type=str, default='both', help='Sex of patients', choices=['male', 'female', 'both']) parser.add_argument('--name_add', '-nam', type=str, default='', help='option to add to the name string') parser.add_argument('--log_frequency', '-lf', type=int, default=200, help='logging frequency') parser.add_argument('--val_frequency', '-vf', type=int, default=1000, help='validation frequency') parser.add_argument('--anom_val_frequency', '-avf', type=int, default=1000, help='Validation frequency on anomalous samples') parser.add_argument('--val_steps', type=int, default=100, help='validation steps') parser.add_argument('--num_images_log', '-nil', type=int, default=16, help='Number of images to log on wandb') parser.add_argument('--ssim_eval', '-ssim', type=str_to_bool, default=True, help='Whether to use SSIM on residual-based methods.') parser.add_argument('--get_images', '-img', type=str_to_bool, default=False) return parser
def get_texts(texts: List[str], already_processed: Optional[bool]=False, n_cpus: Optional[int]=None) -> List[List[str]]: num_cpus = (n_cpus if (n_cpus is not None) else os.cpu_count()) if (not already_processed): processed_texts = [' '.join(simple_preprocess(t)) for t in texts] else: processed_texts = texts tok = Tokenizer(n_cpus=num_cpus).process_all(processed_texts) return tok
class SharedMemoryWriter(StorageWriter): def __init__(self, shm_handler: SharedMemoryHandler) -> None: super().__init__() self.file_name = '' self.shm_handler = shm_handler self.metadata: Dict[(str, object)] = {} def set_up_storage_writer(self, is_coordinator: bool) -> None: pass def prepare_local_plan(self, plan: SavePlan) -> SavePlan: return plan def prepare_global_plan(self, global_plan: List[SavePlan]) -> List[SavePlan]: new_plans = [dataclasses.replace(plan, storage_data=_StoragePrefix(f'__{i}_')) for (i, plan) in enumerate(global_plan)] return new_plans def write_data(self, plan: SavePlan, planner: SavePlanner) -> Future[List[WriteResult]]: storage_plan: _StoragePrefix = plan.storage_data file_count = 0 files = [] self.file_name = f'{storage_plan.prefix}{file_count}{DEFAULT_SUFFIX}' for bucket in plan.items: files.append((self.file_name, bucket)) if self.shm_handler.no_checkpint_state(): buffer_size = _get_buffer_size(files, planner) self.shm_handler.init_shared_memory(create=True, size=buffer_size) assert (self.shm_handler.shared_memory is not None) (write_results, no_shard_data) = _write_memory_from_list(shm=self.shm_handler.shared_memory, files=files, planner=planner) self.metadata['no_shard_data'] = no_shard_data fut: Future[List[WriteResult]] = Future() fut.set_result(write_results) return fut def finish(self, metadata: Metadata, results: List[List[WriteResult]]) -> None: storage_md = dict() for wr_list in results: storage_md.update({wr.index: wr.storage_data for wr in wr_list}) metadata.storage_data = storage_md self.metadata['dcp_metadata'] = metadata
def main(args, local_rank): logging.basicConfig(format='%(asctime)s - %(levelname)s - %(name)s - %(message)s', datefmt='%m/%d/%Y %H:%M:%S', level=logging.INFO) vocabs = dict() vocabs['src'] = Vocab(args.src_vocab, 0, [BOS, EOS]) vocabs['tgt'] = Vocab(args.tgt_vocab, 0, [BOS, EOS]) if ((args.world_size == 1) or (dist.get_rank() == 0)): logger.info(args) for name in vocabs: logger.info('vocab %s, size %d, coverage %.3f', name, vocabs[name].size, vocabs[name].coverage) set_seed() torch.cuda.set_device(local_rank) device = torch.device('cuda', local_rank) if (args.arch == 'vanilla'): model = Generator(vocabs, args.embed_dim, args.ff_embed_dim, args.num_heads, args.dropout, args.enc_layers, args.dec_layers, args.label_smoothing) elif (args.arch == 'mem'): model = MemGenerator(vocabs, args.embed_dim, args.ff_embed_dim, args.num_heads, args.dropout, args.mem_dropout, args.enc_layers, args.dec_layers, args.mem_enc_layers, args.label_smoothing, args.use_mem_score) elif (args.arch == 'rg'): logger.info('start building model') logger.info('building retriever') retriever = Retriever.from_pretrained(args.num_retriever_heads, vocabs, args.retriever, args.nprobe, args.topk, local_rank, use_response_encoder=(args.rebuild_every > 0)) logger.info('building retriever + generator') model = RetrieverGenerator(vocabs, retriever, args.share_encoder, args.embed_dim, args.ff_embed_dim, args.num_heads, args.dropout, args.mem_dropout, args.enc_layers, args.dec_layers, args.mem_enc_layers, args.label_smoothing) if args.resume_ckpt: model.load_state_dict(torch.load(args.resume_ckpt)['model']) else: global_step = 0 if (args.world_size > 1): set_seed(( + dist.get_rank())) model = model.to(device) retriever_params = [v for (k, v) in model.named_parameters() if k.startswith('retriever.')] other_params = [v for (k, v) in model.named_parameters() if (not k.startswith('retriever.'))] optimizer = Adam([{'params': retriever_params, 'lr': ((args.embed_dim ** (- 0.5)) * 0.1)}, {'params': other_params, 'lr': (args.embed_dim ** (- 0.5))}], betas=(0.9, 0.98), eps=1e-09) lr_schedule = get_inverse_sqrt_schedule_with_warmup(optimizer, args.warmup_steps, args.total_train_steps) train_data = DataLoader(vocabs, args.train_data, args.per_gpu_train_batch_size, for_train=True, rank=local_rank, num_replica=args.world_size) model.eval() (step, epoch) = (0, 0) tr_stat = Statistics() logger.info('start training') model.train() best_dev_bleu = 0.0 while (global_step <= args.total_train_steps): for batch in train_data: batch = move_to_device(batch, device) if (args.arch == 'rg'): (loss, acc) = model(batch, update_mem_bias=(global_step > args.update_retriever_after)) else: (loss, acc) = model(batch) tr_stat.update({'loss': (loss.item() * batch['tgt_num_tokens']), 'tokens': batch['tgt_num_tokens'], 'acc': acc}) tr_stat.step() loss.backward() step += 1 if (not ((step % args.gradient_accumulation_steps) == ((- 1) % args.gradient_accumulation_steps))): continue if (args.world_size > 1): average_gradients(model) torch.nn.utils.clip_grad_norm_(model.parameters(), 1.0) optimizer.step() lr_schedule.step() optimizer.zero_grad() global_step += 1 if ((args.world_size == 1) or (dist.get_rank() == 0)): if ((global_step % args.print_every) == ((- 1) % args.print_every)): logger.info('epoch %d, step %d, loss %.3f, acc %.3f', epoch, global_step, (tr_stat['loss'] / tr_stat['tokens']), (tr_stat['acc'] / tr_stat['tokens'])) tr_stat = Statistics() if ((global_step % args.eval_every) == ((- 1) % args.eval_every)): model.eval() max_time_step = (256 if (global_step > (2 * args.warmup_steps)) else 5) bleus = [] for cur_dev_data in args.dev_data: dev_data = DataLoader(vocabs, cur_dev_data, args.dev_batch_size, for_train=False) bleu = validate(device, model, dev_data, beam_size=5, alpha=0.6, max_time_step=max_time_step) bleus.append(bleu) bleu = (sum(bleus) / len(bleus)) logger.info('epoch %d, step %d, dev bleu %.2f', epoch, global_step, bleu) if (bleu > best_dev_bleu): testbleus = [] for cur_test_data in args.test_data: test_data = DataLoader(vocabs, cur_test_data, args.dev_batch_size, for_train=False) testbleu = validate(device, model, test_data, beam_size=5, alpha=0.6, max_time_step=max_time_step) testbleus.append(testbleu) testbleu = (sum(testbleus) / len(testbleus)) logger.info('epoch %d, step %d, test bleu %.2f', epoch, global_step, testbleu) torch.save({'args': args, 'model': model.state_dict()}, ('%s/best.pt' % (args.ckpt,))) if (not args.only_save_best): torch.save({'args': args, 'model': model.state_dict()}, ('%s/epoch%d_batch%d_devbleu%.2f_testbleu%.2f' % (args.ckpt, epoch, global_step, bleu, testbleu))) best_dev_bleu = bleu model.train() if ((args.rebuild_every > 0) and ((global_step % args.rebuild_every) == ((- 1) % args.rebuild_every))): model.retriever.drop_index() torch.cuda.empty_cache() next_index_dir = ('%s/batch%d' % (args.ckpt, global_step)) if ((args.world_size == 1) or (dist.get_rank() == 0)): model.retriever.rebuild_index(next_index_dir) dist.barrier() else: dist.barrier() model.retriever.update_index(next_index_dir, args.nprobe) if (global_step > args.total_train_steps): break epoch += 1 logger.info('rank %d, finish training after %d steps', local_rank, global_step)
class TemporalCenterCrop(object): def __init__(self, size, padding=True, pad_method='loop'): self.size = size self.padding = padding self.pad_method = pad_method def __call__(self, frame_indices): center_index = (len(frame_indices) // 2) begin_index = max(0, (center_index - (self.size // 2))) end_index = min((begin_index + self.size), len(frame_indices)) out = list(frame_indices[begin_index:end_index]) if (self.padding == True): if (self.pad_method == 'loop'): while (len(out) < self.size): for index in out: if (len(out) >= self.size): break out.append(index) else: while (len(out) < self.size): for index in out: if (len(out) >= self.size): break out.append(index) out.sort() return out
_model_architecture('hf_gpt2', 'hf_gpt2_medium') def hf_gpt2_medium(args): args.embed_dim = getattr(args, 'embed_dim', 1024) args.num_attention_heads = getattr(args, 'num_attention_heads', 16) args.num_layers = getattr(args, 'num_layers', 24) default_architecture(args)
def get_paths(path): files = os.listdir(path) if ((LOG_FILE_NAME in files) or any([('seed' in f) for f in files])): return [path] else: return sum([get_paths(os.path.join(path, f)) for f in files], start=[])
('(float32[:], float32[:])', device=True, inline=True) def rbbox_to_corners(corners, rbbox): angle = rbbox[4] a_cos = math.cos(angle) a_sin = math.sin(angle) center_x = rbbox[0] center_y = rbbox[1] x_d = rbbox[2] y_d = rbbox[3] corners_x = cuda.local.array((4,), dtype=numba.float32) corners_y = cuda.local.array((4,), dtype=numba.float32) corners_x[0] = ((- x_d) / 2) corners_x[1] = ((- x_d) / 2) corners_x[2] = (x_d / 2) corners_x[3] = (x_d / 2) corners_y[0] = ((- y_d) / 2) corners_y[1] = (y_d / 2) corners_y[2] = (y_d / 2) corners_y[3] = ((- y_d) / 2) for i in range(4): corners[(2 * i)] = (((a_cos * corners_x[i]) + (a_sin * corners_y[i])) + center_x) corners[((2 * i) + 1)] = ((((- a_sin) * corners_x[i]) + (a_cos * corners_y[i])) + center_y)
def matrix_for_bone_from_parent(bone, ao): eb1 = ao.data.bones[bone.name] E = eb1.matrix_local ebp = ao.data.bones[bone.name].parent E_p = ebp.matrix_local return (E_p.inverted() E)
class HiResCAM(BaseCAM): def __init__(self, model, target_layers, reshape_transform=None): super(HiResCAM, self).__init__(model, target_layers, reshape_transform) def get_cam_image(self, input_tensor, target_layer, target_category, activations, grads, eigen_smooth): elementwise_activations = (grads * activations) if eigen_smooth: print("Warning: HiResCAM's faithfulness guarantees do not hold if smoothing is applied") cam = get_2d_projection(elementwise_activations) else: cam = elementwise_activations.sum(axis=1) return cam
def dfscode_to_tensor(dfscode, feature_map): (max_nodes, max_edges) = (feature_map['max_nodes'], feature_map['max_edges']) (node_forward_dict, edge_forward_dict) = (feature_map['node_forward'], feature_map['edge_forward']) (num_nodes_feat, num_edges_feat) = (len(feature_map['node_forward']), len(feature_map['edge_forward'])) dfscode_tensors = {'t1': ((max_nodes + 1) * torch.ones((max_edges + 1), dtype=torch.long)), 't2': ((max_nodes + 1) * torch.ones((max_edges + 1), dtype=torch.long)), 'v1': ((num_nodes_feat + 1) * torch.ones((max_edges + 1), dtype=torch.long)), 'e': ((num_edges_feat + 1) * torch.ones((max_edges + 1), dtype=torch.long)), 'v2': ((num_nodes_feat + 1) * torch.ones((max_edges + 1), dtype=torch.long)), 'len': len(dfscode)} for (i, code) in enumerate(dfscode): dfscode_tensors['t1'][i] = int(code[0]) dfscode_tensors['t2'][i] = int(code[1]) dfscode_tensors['v1'][i] = int(node_forward_dict[code[2]]) dfscode_tensors['e'][i] = int(edge_forward_dict[code[3]]) dfscode_tensors['v2'][i] = int(node_forward_dict[code[4]]) (dfscode_tensors['t1'][len(dfscode)], dfscode_tensors['t2'][len(dfscode)]) = (max_nodes, max_nodes) (dfscode_tensors['v1'][len(dfscode)], dfscode_tensors['v2'][len(dfscode)]) = (num_nodes_feat, num_nodes_feat) dfscode_tensors['e'][len(dfscode)] = num_edges_feat return dfscode_tensors
class TensorFlowWrapper(): def __init__(self, embedding_layer_hub_name: str) -> None: g = tensorflow.Graph() with g.as_default(): embedding_layer = tensorflow_hub.Module(embedding_layer_hub_name) self._sts_input1 = tensorflow.placeholder(tensorflow.string, shape=None) self._sts_input2 = tensorflow.placeholder(tensorflow.string, shape=None) sts_encode1 = tensorflow.nn.l2_normalize(embedding_layer(self._sts_input1), axis=1) sts_encode2 = tensorflow.nn.l2_normalize(embedding_layer(self._sts_input2), axis=1) cosine_similarities = tensorflow.reduce_sum(tensorflow.multiply(sts_encode1, sts_encode2), axis=1) clip_cosine_similarities = tensorflow.clip_by_value(cosine_similarities, (- 1.0), 1.0) self._sim_scores = (1.0 - tensorflow.acos(clip_cosine_similarities)) init_op = tensorflow.group([tensorflow.global_variables_initializer(), tensorflow.tables_initializer()]) g.finalize() self._session = tensorflow.Session(graph=g) self._session.run(init_op) def append_scores(self, sentence_pairs: pandas.DataFrame) -> None: text_a = sentence_pairs['sent_1'].fillna('').tolist() text_b = sentence_pairs['sent_2'].fillna('').tolist() scores = self._session.run(self._sim_scores, feed_dict={self._sts_input1: text_a, self._sts_input2: text_b}) sentence_pairs['score'] = scores def close(self): self._session.close()
def get_default_sess_config(mem_fraction=0.99): conf = tf.ConfigProto() conf.gpu_options.per_process_gpu_memory_fraction = mem_fraction conf.gpu_options.allocator_type = 'BFC' conf.gpu_options.allow_growth = True conf.allow_soft_placement = True return conf
def early_stopping(loss, patience): if (len(loss) < patience): return False loss = loss[(patience * (- 1)):] last = sys.float_info.min for l in loss: if (l < last): return False last = l return True
class NatPreTrainedModel(metaclass=DummyObject): _backends = ['torch'] def __init__(self, *args, **kwargs): requires_backends(self, ['torch'])
def CheckForNonStandardConstructs(filename, clean_lines, linenum, nesting_state, error): line = clean_lines.lines[linenum] if Search('printf\\s*\\(.*".*%[-+ ]?\\d*q', line): error(filename, linenum, 'runtime/printf_format', 3, '%q in format strings is deprecated. Use %ll instead.') if Search('printf\\s*\\(.*".*%\\d+\\$', line): error(filename, linenum, 'runtime/printf_format', 2, '%N$ formats are unconventional. Try rewriting to avoid them.') line = line.replace('\\\\', '') if Search('("|\\\').*\\\\(%|\\[|\\(|{)', line): error(filename, linenum, 'build/printf_format', 3, '%, [, (, and { are undefined character escapes. Unescape them.') line = clean_lines.elided[linenum] if Search('\\b(const|volatile|void|char|short|int|long|float|double|signed|unsigned|schar|u?int8|u?int16|u?int32|u?int64)\\s+(register|static|extern|typedef)\\b', line): error(filename, linenum, 'build/storage_class', 5, 'Storage class (static, extern, typedef, etc) should be first.') if Match('\\s*#\\s*endif\\s*[^/\\s]+', line): error(filename, linenum, 'build/endif_comment', 5, 'Uncommented text after #endif is non-standard. Use a comment.') if Match('\\s*class\\s+(\\w+\\s*::\\s*)+\\w+\\s*;', line): error(filename, linenum, 'build/forward_decl', 5, 'Inner-style forward declarations are invalid. Remove this line.') if Search('(\\w+|[+-]?\\d+(\\.\\d*)?)\\s*(<|>)\\?=?\\s*(\\w+|[+-]?\\d+)(\\.\\d*)?', line): error(filename, linenum, 'build/deprecated', 3, '>? and <? (max and min) operators are non-standard and deprecated.') if Search('^\\s*const\\s*string\\s*&\\s*\\w+\\s*;', line): error(filename, linenum, 'runtime/member_string_references', 2, 'const string& members are dangerous. It is much better to use alternatives, such as pointers or simple constants.') classinfo = nesting_state.InnermostClass() if ((not classinfo) or (not classinfo.seen_open_brace)): return base_classname = classinfo.name.split('::')[(- 1)] args = Match(('\\s+(?:inline\\s+)?%s\\s*\\(([^,()]+)\\)' % re.escape(base_classname)), line) if (args and (args.group(1) != 'void') and (not Match(('(const\\s+)?%s(\\s+const)?\\s*(?:<\\w+>\\s*)?&' % re.escape(base_classname)), args.group(1).strip()))): error(filename, linenum, 'runtime/explicit', 5, 'Single-argument constructors should be marked explicit.')
def ReadFile(filename, print_error=True): try: fp = open(filename) try: return fp.read() finally: fp.close() except IOError: if print_error: print(('Error reading %s: %s' % (filename, sys.exc_info()[1]))) return None
def adjust_upper_plane(x0, y0, x_minus, x_plus, y_minus, y_plus, lr=0.01, max_iter=100, print_info=True): x0 = x0.detach() y0 = y0.detach() x1 = ((x0 + x_minus) / 2).data.clone() y1 = ((y0 + y_minus) / 2).data.clone() x2 = ((x0 + x_minus) / 2).data.clone() y2 = ((y0 + y_plus) / 2).data.clone() x3 = ((x0 + x_plus) / 2).data.clone() y3 = ((y0 + y_plus) / 2).data.clone() x4 = ((x0 + x_plus) / 2).data.clone() y4 = ((y0 + y_minus) / 2).data.clone() x1.requires_grad = True y1.requires_grad = True x2.requires_grad = True y2.requires_grad = True x3.requires_grad = True y3.requires_grad = True x4.requires_grad = True y4.requires_grad = True (a, b, c) = plane(x0, y0) optimizer = optim.Adam([x1, y1, x2, y2, x3, y3, x4, y4], lr=lr) x1_best = torch.zeros(x_minus.shape, device=x_minus.device) y1_best = torch.zeros(x_minus.shape, device=x_minus.device) loss1_best = (torch.ones(x_minus.shape, device=x_minus.device) * 1000) x2_best = torch.zeros(x_minus.shape, device=x_minus.device) y2_best = torch.zeros(x_minus.shape, device=x_minus.device) loss2_best = (torch.ones(x_minus.shape, device=x_minus.device) * 1000) x3_best = torch.zeros(x_minus.shape, device=x_minus.device) y3_best = torch.zeros(x_minus.shape, device=x_minus.device) loss3_best = (torch.ones(x_minus.shape, device=x_minus.device) * 1000) x4_best = torch.zeros(x_minus.shape, device=x_minus.device) y4_best = torch.zeros(x_minus.shape, device=x_minus.device) loss4_best = (torch.ones(x_minus.shape, device=x_minus.device) * 1000) for i in range(max_iter): loss1 = ((((a * x1) + (b * y1)) + c) - (torch.tanh(x1) * torch.sigmoid(y1))) loss2 = ((((a * x2) + (b * y2)) + c) - (torch.tanh(x2) * torch.sigmoid(y2))) loss3 = ((((a * x3) + (b * y3)) + c) - (torch.tanh(x3) * torch.sigmoid(y3))) loss4 = ((((a * x4) + (b * y4)) + c) - (torch.tanh(x4) * torch.sigmoid(y4))) (qloss1, valid1) = qualification_loss_upper(x1, y1, x_minus, x_plus, y_minus, y_plus) best1 = ((loss1 < loss1_best) * valid1) x1_best[best1] = x1[best1] y1_best[best1] = y1[best1] loss1_best[best1] = loss1[best1] (qloss2, valid2) = qualification_loss_upper(x2, y2, x_minus, x_plus, y_minus, y_plus) best2 = ((loss2 < loss2_best) * valid2) x2_best[best2] = x2[best2] y2_best[best2] = y2[best2] loss2_best[best2] = loss2[best2] (qloss3, valid3) = qualification_loss_upper(x3, y3, x_minus, x_plus, y_minus, y_plus) best3 = ((loss3 < loss3_best) * valid3) x3_best[best3] = x3[best3] y3_best[best3] = y3[best3] loss3_best[best3] = loss3[best3] (qloss4, valid4) = qualification_loss_upper(x4, y4, x_minus, x_plus, y_minus, y_plus) best4 = ((loss4 < loss4_best) * valid4) x4_best[best4] = x4[best4] y4_best[best4] = y4[best4] loss4_best[best4] = loss4[best4] loss = ((loss1 * (valid1.float() + 0.1)) + qloss1) loss = ((loss + (loss2 * (valid2.float() + 0.1))) + qloss2) loss = ((loss + (loss3 * (valid3.float() + 0.1))) + qloss3) loss = ((loss + (loss4 * (valid4.float() + 0.1))) + qloss4) loss = loss.mean() optimizer.zero_grad() loss.backward() optimizer.step() if print_info: print(('1 adjust upper plane loss: %.4f' % loss.item())) return (x1_best, y1_best, x2_best, y2_best, x3_best, y3_best, x4_best, y4_best, loss1_best, loss2_best, loss3_best, loss4_best)
def parameter_net_file(): with tempfile.NamedTemporaryFile(mode='w+', delete=False) as f: f.write("name: 'pythonnet' force_backward: true\n input: 'data' input_shape { dim: 10 dim: 9 dim: 8 }\n layer { type: 'Python' name: 'layer' bottom: 'data' top: 'top'\n python_param { module: 'test_python_layer' layer: 'ParameterLayer' } }\n ") return f.name
def _metric_max_over_ground_truths(metric_fn, prediction, ground_truths): scores_for_ground_truths = [] for ground_truth in ground_truths: score = metric_fn(prediction, ground_truth) scores_for_ground_truths.append(score) return max(scores_for_ground_truths)
def nchw_to_nlc(x): assert (len(x.shape) == 4) return x.flatten(2).transpose(1, 2).contiguous()
def shared_single(dim=2): shp = tuple(([1] * dim)) return theano.shared(numpy.zeros(shp, dtype='float32'))
def get_path_iterator(root, tsv, nshard, rank, audio_col_name): with open(tsv) as f: reader = csv.DictReader(f, delimiter='\t', quotechar=None, doublequote=False, lineterminator='\n', quoting=csv.QUOTE_NONE) subpaths = [op.join(root, e[audio_col_name]) for e in reader] (start, end) = get_shard_range(len(subpaths), nshard, rank) subpaths = subpaths[start:end] def iterate(): for subpath in subpaths: (yield (op.join(root, subpath), None)) return (iterate, len(subpaths))
def main(args): (model, dataset) = (args.model, args.dataset) model_name = ClipModel.get_model_name_by_index(model) dataset_name = ImageTextData.get_data_name_by_index(dataset) args.log_file = (os.getcwd() + '/log/{}_{}_{}.txt'.format(args.mode, model_name, dataset_name)) logger = get_logger(args.log_file, args.log_file) logger.info(args) clip = ClipModel(model, logger=logger) logger.info(f'Clip model {model_name} loaded') itdata = ImageTextData(dataset, root=args.root, preprocess=clip.preprocess) train_loader = torch.utils.data.DataLoader(itdata, batch_size=args.batchsize, shuffle=True) logger.info(f'Dataset {dataset_name} loaded') if (args.mode == 'zs'): (acc, res) = clip.evaluate(train_loader) logger.info('Results: {}'.format(res)) logger.info('Accuracy: {:.2f}%'.format((acc * 100))) elif (args.mode == 'fe'): res = clip.feature_extraction(train_loader) logger.info('Feature extracted!') if (not os.path.exists('feat')): os.makedirs('feat') feat_file = 'feat/{}_{}_{}.csv'.format(args.mode, model_name, dataset_name) np.savetxt(feat_file, res, fmt='%.4f') elif (args.mode == 'ft'): test_data = ImageTextData(args.test_data, root=args.root, preprocess=clip.preprocess) test_loader = torch.utils.data.DataLoader(test_data, batch_size=args.test_batchsize, shuffle=False, drop_last=False) optimizer = optim.Adam(clip.model.parameters(), lr=args.lr, betas=(args.beta1, args.beta2), eps=args.eps, weight_decay=args.weight_decay) best_acc = clip.finetune(train_loader, test_loader, optimizer, args.nepoch, save_path='/home/jindwang/mine/clipood/model/{}_{}_{}.pt'.format(args.mode, model_name, dataset_name)) logger.info('Accuracy: {:.2f}%'.format((best_acc * 100))) else: raise NotImplementedError
def make_mlp_model(latent_dim, output_dim, num_layers, activation=tf.nn.relu, l2_regularizer_weight=0.01, bias_init_stddev=0.1): layers = ([latent_dim] * (num_layers - 1)) layers.append(output_dim) return snt.Sequential([snt.nets.MLP(layers, activation=activation, initializers={'w': tf.initializers.glorot_normal(), 'b': tf.initializers.truncated_normal(stddev=bias_init_stddev)}, activate_final=False)])
class ResNet(nn.Module): def __init__(self, block, num_blocks, num_classes=10, soft=False): super(ResNet, self).__init__() self.in_planes = 64 self.soft = soft self.downsample = nn.MaxPool2d(2) self.conv1 = nn.Conv2d(3, 64, kernel_size=3, stride=1, padding=1, bias=False) self.bn1 = nn.BatchNorm2d(64) self.layer1 = self._make_layer(block, 64, num_blocks[0], stride=1) self.layer2 = self._make_layer(block, 128, num_blocks[1], stride=2) self.layer3 = self._make_layer(block, 256, num_blocks[2], stride=2) self.layer4 = self._make_layer(block, 512, num_blocks[3], stride=2) self.linear = nn.Linear((512 * block.expansion), num_classes) def _make_layer(self, block, planes, num_blocks, stride): strides = ([stride] + ([1] * (num_blocks - 1))) layers = [] for stride in strides: layers.append(block(self.in_planes, planes, stride)) self.in_planes = (planes * block.expansion) return nn.Sequential(*layers) def forward(self, x, downsample_=True): if downsample_: x = self.downsample(x) out = F.relu(self.bn1(self.conv1(x))) out = self.layer1(out) out = self.layer2(out) out = self.layer3(out) out = self.layer4(out) out = F.avg_pool2d(out, 4) out = out.view(out.size(0), (- 1)) out = self.linear(out) if self.soft: out = F.softmax(out, dim=1) return F.log_softmax(out, dim=1) else: return F.log_softmax(out, dim=1) def forward_oltl(self, x): out = F.relu(self.bn1(self.conv1(x))) out = self.layer1(out) out = self.layer2(out) out = self.layer3(out) out = self.layer4(out) out = F.avg_pool2d(out, 4) out = out.view(out.size(0), (- 1)) out = self.linear(out) return (F.log_softmax(out, dim=1), F.softmax(out, dim=1))
class VOTLTVideo(Video): def __init__(self, name, root, video_dir, init_rect, img_names, gt_rect, load_img=False): super(VOTLTVideo, self).__init__(name, root, video_dir, init_rect, img_names, gt_rect, None, load_img) self.gt_traj = [([0] if np.isnan(bbox[0]) else bbox) for bbox in self.gt_traj] if (not load_img): img_name = os.path.join(root, self.img_names[0]) img = np.array(Image.open(img_name), np.uint8) self.width = img.shape[1] self.height = img.shape[0] self.confidence = {} def load_tracker(self, path, tracker_names=None, store=True): if (not tracker_names): tracker_names = [x.split('/')[(- 1)] for x in glob(path) if os.path.isdir(x)] if isinstance(tracker_names, str): tracker_names = [tracker_names] for name in tracker_names: traj_file = os.path.join(path, name, 'longterm', self.name, (self.name + '_001.txt')) with open(traj_file, 'r') as f: traj = [list(map(float, x.strip().split(','))) for x in f.readlines()] if store: self.pred_trajs[name] = traj confidence_file = os.path.join(path, name, 'longterm', self.name, (self.name + '_001_confidence.value')) with open(confidence_file, 'r') as f: score = [float(x.strip()) for x in f.readlines()[1:]] score.insert(0, float('nan')) if store: self.confidence[name] = score return (traj, score)
def MFM(x, name): with tf.variable_scope(name): shape = x.get_shape().as_list() res = tf.reshape(x, [(- 1), shape[1], shape[2], 2, (shape[(- 1)] // 2)]) res = tf.reduce_max(res, axis=[3]) return res
def zero_last_layer(encoder): encoder.model_z[4].weight.data.fill_(0.0) encoder.model_z[4].bias.data.fill_(0.0) return encoder