code
stringlengths
101
5.91M
def simple_while(A: dace.int32[10]): i = 0 while (i < 10): A[i] += (2 * i) i += 1
class GrailEntityDisambFeature(): def __init__(self, pid, input_ids, token_type_ids, target_idx): self.pid = pid self.candidate_input_ids = input_ids self.candidate_token_type_ids = token_type_ids self.target_idx = target_idx
def save_summary(epoch: int, global_step: int, accuracies: List[utils.AverageMeter], duration: timedelta, tracking_file: str, mode: str, top=(1,)): result: Dict[(str, Any)] = OrderedDict() result['timestamp'] = datetime.now() result['mode'] = mode result['epoch'] = epoch result['global_step'] = global_step result['duration'] = duration for (k, acc) in zip(top, accuracies): result[f'top{k}_accuracy'] = acc.avg utils.save_result(result, tracking_file)
class ROCExplanation(ExplanationBase): def __init__(self): super().__init__() self.explanations = {} def add(self, fpr: Dict, tpr: Dict, auc: Dict): self.explanations = {'fpr': fpr, 'tpr': tpr, 'auc': auc} def get_explanations(self): return self.explanations def plot(self, class_names=None, linewidth=2, **kwargs): import matplotlib.pyplot as plt import matplotlib.colors as mcolors fpr = self.explanations['fpr'] tpr = self.explanations['tpr'] auc = self.explanations['auc'] colors = list(mcolors.TABLEAU_COLORS.values()) fig = plt.figure() plt.plot(fpr['micro'], tpr['micro'], label='Micro-average ROC curve (area = {:0.2f})'.format(auc['micro']), color='deeppink', linestyle=':', linewidth=linewidth) plt.plot(fpr['macro'], tpr['macro'], label='Macro-average ROC curve (area = {:0.2f})'.format(auc['macro']), color='navy', linestyle=':', linewidth=linewidth) for i in range((len(fpr) - 2)): label = (class_names[i] if (class_names is not None) else i) plt.plot(fpr[i], tpr[i], color=colors[(i % len(colors))], linewidth=linewidth, label='ROC curve of class {} (area = {:0.2f})'.format(label, auc[i])) plt.plot([0, 1], [0, 1], 'k--', linewidth=linewidth) plt.xlim([0.0, 1.0]) plt.ylim([0.0, 1.05]) plt.xlabel('False Positive Rate') plt.ylabel('True Positive Rate') plt.title('ROC Curves') plt.legend(loc='lower right') plt.grid() return fig def _plotly_figure(self, class_names=None, linewidth=2, **kwargs): import plotly.graph_objects as go fpr = self.explanations['fpr'] tpr = self.explanations['tpr'] auc = self.explanations['auc'] fig = go.Figure() fig.add_trace(go.Scatter(x=fpr['micro'], y=tpr['micro'], name='Micro-average ROC curve (area = {:0.2f})'.format(auc['micro']), line=dict(width=linewidth))) fig.add_trace(go.Scatter(x=fpr['macro'], y=tpr['macro'], name='Macro-average ROC curve (area = {:0.2f})'.format(auc['macro']), line=dict(width=linewidth))) for i in range((len(fpr) - 2)): label = (class_names[i] if (class_names is not None) else i) fig.add_trace(go.Scatter(x=fpr[i], y=tpr[i], name='ROC curve of class {} (area = {:0.2f})'.format(label, auc[i]), line=dict(width=linewidth))) fig.add_trace(go.Scatter(x=[0, 1], y=[0, 1], line=dict(color='black', dash='dash', width=linewidth), name='Baseline')) fig.update_layout(xaxis_title='False Positive Rate', yaxis_title='True Positive Rate', title={'text': 'ROC Curves'}) return fig def plotly_plot(self, class_names=None, **kwargs): return DashFigure(self._plotly_figure(class_names, **kwargs)) def ipython_plot(self, class_names=None, **kwargs): import plotly plotly.offline.iplot(self._plotly_figure(class_names, **kwargs)) def from_dict(cls, d): e = d['explanations'] for metric in ['fpr', 'tpr', 'auc']: r = {} for (key, value) in e[metric].items(): try: key = int(key) except: pass r[key] = value e[metric] = r exp = ROCExplanation() exp.explanations = e return exp
def numpy_to_hls_code(ndarray, dtype, hls_var_name, pack_innermost_dim=True, no_decl=False): hls_dtype = dtype.get_hls_datatype_str() if ((type(ndarray) != np.ndarray) or (ndarray.dtype != np.float32)): ndarray = np.asarray(ndarray, dtype=np.float32) if pack_innermost_dim: idimlen = ndarray.shape[(- 1)] idimbits = (idimlen * dtype.bitwidth()) idimbits = roundup_to_integer_multiple(idimbits, 4) ndarray = pack_innermost_dim_as_hex_string(ndarray, dtype, idimbits) hls_dtype = ('ap_uint<%d>' % idimbits) ndims = ndarray.ndim ret = ('%s %s' % (hls_dtype, hls_var_name)) for d in range(ndims): ret += ('[%d]' % ndarray.shape[d]) orig_printops = np.get_printoptions() np.set_printoptions(threshold=sys.maxsize) def elem2str(x): if ((type(x) == str) or (type(x) == np.str_) or (type(x) == np.str)): return ('%s("%s", 16)' % (hls_dtype, x)) elif (type(x) == np.float32): if dtype.is_integer(): return str(int(x)) else: return str(x) else: raise Exception('Unsupported type for numpy_to_hls_code') strarr = np.array2string(ndarray, separator=', ', formatter={'all': elem2str}) np.set_printoptions(**orig_printops) strarr = strarr.replace('[', '{').replace(']', '}') if no_decl: ret = (strarr + ';') else: ret = (((ret + ' = \n') + strarr) + ';') return ret
def plot_avg_clustering(G_times, fname): max_time = len(G_times) t = list(range(0, max_time)) avg_clustering = [] for G in G_times: avg_clustering.append(nx.average_clustering(G)) plt.rcParams.update({'figure.autolayout': True}) plt.rc('xtick', labelsize='x-small') plt.rc('ytick', labelsize='x-small') fig = plt.figure(figsize=(4, 2)) ax = fig.add_subplot(1, 1, 1) ax.plot(t, avg_clustering, marker='o', color='#78f542', ls='solid', linewidth=0.5, markersize=1) ax.set_xlabel('time', fontsize=8) outliers = find_rarity_windowed_outlier(avg_clustering, percent_ranked=0.05, window=5, initial_period=10) outliers.sort() for xc in outliers: plt.axvline(x=xc, color='k', linestyle=':', linewidth=0.5) ax.set_ylabel('average clustering coefficient', fontsize=8) plt.title('plotting temporal average clustering coefficient ', fontsize='x-small') plt.savefig((fname + 'clustering.pdf'), pad_inches=0) return outliers
class FiniteWordPath_all_iter_with_caching(WordDatatype_iter_with_caching, FiniteWordPath_all, FiniteWord_class): pass
def BetsyRoss(): E = 'abcdefghijk' CC = {2: ['acfg', 'bdgh', 'cehi', 'befj', 'adij', 'dfk', 'egk', 'ahk', 'bik', 'cjk'], 3: [E]} M = CircuitClosuresMatroid(groundset=E, circuit_closures=CC) M.rename(('BetsyRoss: ' + repr(M))) return M
def test_Detector_init(): (detector, parent, tl) = create_detector(dark_count=10) tl.init() assert (len(tl.events) == 2)
def Q_calc(TP, TN, FP, FN): try: OR = ((TP * TN) / (FP * FN)) result = ((OR - 1) / (OR + 1)) return result except (ZeroDivisionError, TypeError): return 'None'
class DateTimeField(fields.DateTimeField): def __init__(self, *args, **kwargs): if (not has_timezone): raise ImportError('DateTimeField requires Django >= 1.5') super(DateTimeField, self).__init__(*args, **kwargs) def process_formdata(self, valuelist): super(DateTimeField, self).process_formdata(valuelist) date = self.data if (settings.USE_TZ and (date is not None) and timezone.is_naive(date)): current_timezone = timezone.get_current_timezone() self.data = timezone.make_aware(date, current_timezone) def _value(self): date = self.data if (settings.USE_TZ and isinstance(date, datetime.datetime) and timezone.is_aware(date)): self.data = timezone.localtime(date) return super(DateTimeField, self)._value()
def get_params(argv='1'): print('SET: {}'.format(argv)) params = dict(quick_test=True, finetune_mode=False, pretrained_model_weights='models/1_1_foa_dev_split6_model.h5', dataset_dir='/scratch/asignal/partha/DCASE2022_SELD_dataset', feat_label_dir='/scratch/asignal/partha/DCASE2022_SELD_dataset/seld_feat_label', model_dir='models/', dcase_output_dir='results/', mode='dev', dataset='foa', fs=24000, hop_len_s=0.02, label_hop_len_s=0.1, max_audio_len_s=60, nb_mel_bins=64, use_salsalite=False, fmin_doa_salsalite=50, fmax_doa_salsalite=2000, fmax_spectra_salsalite=9000, multi_accdoa=False, thresh_unify=15, label_sequence_length=50, batch_size=128, dropout_rate=0.05, nb_cnn2d_filt=64, f_pool_size=[4, 4, 2], nb_rnn_layers=2, rnn_size=128, self_attn=False, nb_heads=4, nb_fnn_layers=1, fnn_size=128, nb_epochs=100, lr=0.001, average='macro', lad_doa_thresh=20) if (argv == '1'): print('USING DEFAULT PARAMETERS\n') elif (argv == '2'): print('FOA + ACCDOA\n') params['quick_test'] = False params['dataset'] = 'foa' params['multi_accdoa'] = False elif (argv == '3'): print('FOA + multi ACCDOA\n') params['quick_test'] = False params['dataset'] = 'foa' params['multi_accdoa'] = True elif (argv == '4'): print('MIC + GCC + ACCDOA\n') params['quick_test'] = False params['dataset'] = 'mic' params['use_salsalite'] = False params['multi_accdoa'] = False elif (argv == '5'): print('MIC + SALSA + ACCDOA\n') params['quick_test'] = False params['dataset'] = 'mic' params['use_salsalite'] = True params['multi_accdoa'] = False elif (argv == '6'): print('MIC + GCC + multi ACCDOA\n') params['quick_test'] = False params['dataset'] = 'mic' params['use_salsalite'] = False params['multi_accdoa'] = True elif (argv == '7'): print('MIC + SALSA + multi ACCDOA\n') params['quick_test'] = False params['dataset'] = 'mic' params['use_salsalite'] = True params['multi_accdoa'] = True elif (argv == '999'): print('QUICK TEST MODE\n') params['quick_test'] = True else: print('ERROR: unknown argument {}'.format(argv)) exit() feature_label_resolution = int((params['label_hop_len_s'] // params['hop_len_s'])) params['feature_sequence_length'] = (params['label_sequence_length'] * feature_label_resolution) params['t_pool_size'] = [feature_label_resolution, 1, 1] params['patience'] = int(params['nb_epochs']) if ('2020' in params['dataset_dir']): params['unique_classes'] = 14 elif ('2021' in params['dataset_dir']): params['unique_classes'] = 12 elif ('2022' in params['dataset_dir']): params['unique_classes'] = 13 for (key, value) in params.items(): print('\t{}: {}'.format(key, value)) return params
class Decoder(Network): def __init__(self, output_width, output_height, output_depth, stride=2, kernel=5, final_dim=64, scope_name='decoder', *args, **kwargs): super(Decoder, self).__init__(*args, scope_name=scope_name, **kwargs) self.output_width = output_width self.output_height = output_height self.output_depth = output_depth self.stride = stride self.kernel = kernel self.final_dim = final_dim def build(self, z, train): with tf.variable_scope(self.scope_name, reuse=tf.AUTO_REUSE): batch_size = tf.shape(z)[0] layers = [z] with tf.variable_scope('layer0'): layers.append(linear(layers[(- 1)], 1024)) layers.append(batch_norm()(layers[(- 1)], train=train)) layers.append(tf.nn.relu(layers[(- 1)])) with tf.variable_scope('layer1'): layers.append(linear(layers[(- 1)], (((self.final_dim * 4) * 4) * 4))) layers.append(batch_norm()(layers[(- 1)], train=train)) layers.append(tf.nn.relu(layers[(- 1)])) layers.append(tf.reshape(layers[(- 1)], ((- 1), 4, 4, (self.final_dim * 4)))) with tf.variable_scope('layer2'): layers.append(deconv2d(layers[(- 1)], [batch_size, 8, 8, (self.final_dim * 2)], d_h=self.stride, d_w=self.stride, k_h=self.kernel, k_w=self.kernel)) layers.append(batch_norm()(layers[(- 1)], train=train)) layers.append(tf.nn.relu(layers[(- 1)])) with tf.variable_scope('layer3'): layers.append(deconv2d(layers[(- 1)], [batch_size, 16, 16, self.final_dim], d_h=self.stride, d_w=self.stride, k_h=self.kernel, k_w=self.kernel)) layers.append(batch_norm()(layers[(- 1)], train=train)) layers.append(tf.nn.relu(layers[(- 1)])) with tf.variable_scope('layer4'): layers.append(deconv2d(layers[(- 1)], [batch_size, self.output_height, self.output_width, self.output_depth], d_h=self.stride, d_w=self.stride, k_h=self.kernel, k_w=self.kernel)) layers.append(tf.nn.tanh(layers[(- 1)])) return (layers[(- 1)], layers)
def _launch_worker(exp_key, worker_id, host, port, result_db_name): command = 'hyperopt-mongo-worker --mongo={h}:{p}/{db} --poll-interval=10 --exp-key={key} > hyperopt_worker{id}.log 2>&1' command = command.format(h=host, p=port, db=result_db_name, key=exp_key, id=worker_id) fail = os.system(command) if fail: raise RuntimeError('Problem starting hyperopt-mongo-worker.')
def arg_str2bool(v): if isinstance(v, bool): return v elif (v.lower() in ('yes', 'true', 't', 'y', '1')): return True elif (v.lower() in ('no', 'false', 'f', 'n', '0')): return False else: raise argparse.ArgumentTypeError('Boolean value expected.')
class KaHFM_model(keras.Model): def __init__(self, user_factors, item_factors, learning_rate=0.001, l_w=0, l_b=0, name='NNBPRMF', **kwargs): super().__init__(name=name, **kwargs) tf.random.set_seed(42) self._learning_rate = learning_rate self.l_w = l_w self.l_b = l_b self.initializer = tf.initializers.GlorotUniform() self.Bi = tf.Variable(tf.zeros(item_factors.shape[0]), name='Bi', dtype=tf.float32) self.Gu = tf.Variable(user_factors, name='Gu', dtype=tf.float32) self.Gi = tf.Variable(item_factors, name='Gi', dtype=tf.float32) self.optimizer = tf.optimizers.Adam(self._learning_rate) def call(self, inputs, training=None, **kwargs): (user, item) = inputs beta_i = tf.squeeze(tf.nn.embedding_lookup(self.Bi, item)) gamma_u = tf.squeeze(tf.nn.embedding_lookup(self.Gu, user)) gamma_i = tf.squeeze(tf.nn.embedding_lookup(self.Gi, item)) xui = (beta_i + tf.reduce_sum((gamma_u * gamma_i), 1)) return (xui, beta_i, gamma_u, gamma_i) def train_step(self, batch): with tf.GradientTape() as tape: (user, pos, neg) = batch (xu_pos, beta_pos, gamma_u, gamma_pos) = self.call(inputs=(user, pos), training=True) (xu_neg, beta_neg, gamma_u, gamma_neg) = self.call(inputs=(user, neg), training=True) difference = tf.clip_by_value((xu_pos - xu_neg), (- 80.0), .0) loss = tf.reduce_sum(tf.nn.softplus((- difference))) reg_loss = (((self.l_w * tf.reduce_sum([tf.nn.l2_loss(gamma_u), tf.nn.l2_loss(gamma_pos), tf.nn.l2_loss(gamma_neg)])) + (self.l_b * tf.nn.l2_loss(beta_pos))) + ((self.l_b * tf.nn.l2_loss(beta_neg)) / 10)) loss += reg_loss grads = tape.gradient(loss, [self.Bi, self.Gu, self.Gi]) self.optimizer.apply_gradients(zip(grads, [self.Bi, self.Gu, self.Gi])) return loss def predict_all(self): return (self.Bi + tf.matmul(self.Gu, self.Gi, transpose_b=True)) def predict_batch(self, start, stop): return (self.Bi + tf.matmul(self.Gu[start:stop], self.Gi, transpose_b=True)) def predict(self, inputs, training=False, **kwargs): (logits, _) = self.call(inputs=inputs, training=True) return logits def get_top_k(self, preds, train_mask, k=100): return tf.nn.top_k(tf.where(train_mask, preds, (- np.inf)), k=k, sorted=True) def get_config(self): raise NotImplementedError
def exec_command(command, execute_in='', use_shell=None, use_tee=None, _with_python=1, **env): warnings.warn('exec_command is deprecated since NumPy v1.17, use subprocess.Popen instead', DeprecationWarning, stacklevel=1) log.debug(('exec_command(%r,%s)' % (command, ','.join([('%s=%r' % kv) for kv in env.items()])))) if (use_tee is None): use_tee = (os.name == 'posix') if (use_shell is None): use_shell = (os.name == 'posix') execute_in = os.path.abspath(execute_in) oldcwd = os.path.abspath(os.getcwd()) if (__name__[(- 12):] == 'exec_command'): exec_dir = os.path.dirname(os.path.abspath(__file__)) elif os.path.isfile('exec_command.py'): exec_dir = os.path.abspath('.') else: exec_dir = os.path.abspath(sys.argv[0]) if os.path.isfile(exec_dir): exec_dir = os.path.dirname(exec_dir) if (oldcwd != execute_in): os.chdir(execute_in) log.debug(('New cwd: %s' % execute_in)) else: log.debug(('Retaining cwd: %s' % oldcwd)) oldenv = _preserve_environment(list(env.keys())) _update_environment(**env) try: st = _exec_command(command, use_shell=use_shell, use_tee=use_tee, **env) finally: if (oldcwd != execute_in): os.chdir(oldcwd) log.debug(('Restored cwd to %s' % oldcwd)) _update_environment(**oldenv) return st
class AffineGroupElement(MultiplicativeGroupElement): def __init__(self, parent, A, b=0, convert=True, check=True): try: A = A.matrix() except AttributeError: pass if (is_Matrix(A) and (A.nrows() == A.ncols() == (parent.degree() + 1))): g = A d = parent.degree() A = g.submatrix(0, 0, d, d) b = [g[(i, d)] for i in range(d)] convert = True if convert: A = parent.matrix_space()(A) b = parent.vector_space()(b) if check: if (not is_Matrix(A)): raise TypeError('A must be a matrix') if (not (A.parent() is parent.matrix_space())): raise TypeError(('A must be an element of ' + str(parent.matrix_space()))) if (not (b.parent() is parent.vector_space())): raise TypeError(('b must be an element of ' + str(parent.vector_space()))) parent._element_constructor_check(A, b) super().__init__(parent) self._A = A self._b = b def A(self): return self._A def b(self): return self._b _method def matrix(self): A = self._A b = self._b parent = self.parent() d = parent.degree() from sage.matrix.constructor import matrix, zero_matrix, block_matrix zero = zero_matrix(parent.base_ring(), 1, d) one = matrix(parent.base_ring(), [[1]]) m = block_matrix(2, 2, [A, b.column(), zero, one]) m.set_immutable() return m _matrix_ = matrix def _repr_(self): A = str(self._A) b = str(self._b.column()) deg = self.parent().degree() indices = range(deg) s = [] for (Ai, bi, i) in zip(A.splitlines(), b.splitlines(), indices): if (i == (deg // 2)): s.append(((('x |-> ' + Ai) + ' x + ') + bi)) else: s.append((((' ' + Ai) + ' ') + bi)) return '\n'.join(s) def _latex_(self): return ((('\\vec{x}\\mapsto ' + self.A()._latex_()) + '\\vec{x} + ') + self.b().column()._latex_()) def _ascii_art_(self): from sage.typeset.ascii_art import ascii_art deg = self.parent().degree() A = ascii_art(self._A, baseline=(deg // 2)) b = ascii_art(self._b.column(), baseline=(deg // 2)) return (((ascii_art('x |-> ') + A) + ascii_art(' x + ')) + b) def _unicode_art_(self): from sage.typeset.unicode_art import unicode_art deg = self.parent().degree() A = unicode_art(self._A, baseline=(deg // 2)) b = unicode_art(self._b.column(), baseline=(deg // 2)) return (((unicode_art('x ') + A) + unicode_art(' x + ')) + b) def _mul_(self, other): parent = self.parent() A = (self._A * other._A) b = (self._b + (self._A * other._b)) return parent.element_class(parent, A, b, check=False) def __call__(self, v): parent = self.parent() if (v in parent.vector_space()): return ((self._A * v) + self._b) from sage.rings.polynomial.polynomial_element import Polynomial if (isinstance(v, Polynomial) and (parent.degree() == 1)): ring = v.parent() return ring([self._A[(0, 0)], self._b[0]]) from sage.rings.polynomial.multi_polynomial import MPolynomial if (isinstance(v, MPolynomial) and (parent.degree() == v.parent().ngens())): ring = v.parent() from sage.modules.free_module_element import vector image_coords = ((self._A * vector(ring, ring.gens())) + self._b) return v(*image_coords) import sage.geometry.abc if isinstance(v, sage.geometry.abc.Polyhedron): return ((self._A * v) + self._b) v = parent.vector_space()(v) return ((self._A * v) + self._b) def _act_on_(self, x, self_on_left): if self_on_left: return self(x) def __invert__(self): parent = self.parent() A = parent.matrix_space()((~ self._A)) b = ((- A) * self.b()) return parent.element_class(parent, A, b, check=False) def _richcmp_(self, other, op): lx = self._A rx = other._A if (lx != rx): return richcmp_not_equal(lx, rx, op) return richcmp(self._b, other._b, op) def list(self): return [r.list() for r in self.matrix().rows()]
def gen_expr_simps(simps: LeanExprSimps, at_var: Optional[str]=None, indent: int=0) -> List[str]: lines = [] simp_at = (f' at {at_var}' if (at_var is not None) else '') if (0 < len(simps.const_div_rw)): lines.append(((' ' * indent) + f"try {{ simp only [{', '.join(simps.const_div_rw)}]{simp_at} }},")) if (0 < len(simps.add_comm)): lines.append(((' ' * indent) + f"try {{ simp only [{', '.join(simps.add_comm)}]{simp_at} }},")) return lines
def test_sum_add_bad_node_raise_type_error(): var1 = optplan.Parameter() var2 = optplan.Parameter() sum1 = optplan.Sum(functions=[var1, var2]) with pytest.raises(TypeError, match='add a node'): (sum1 + optplan.SimulationSpace())
def get_key(paragraphs, question, reasoningType): return (paragraphs.replace('\n', '').replace(' ', '').lower(), question.lower(), reasoningType)
class Partition5(nn.Module): LAYER_SCOPES = ['T5ForConditionalGeneration/T5Stack[decoder]/T5Block[3]', 'T5ForConditionalGeneration/T5Stack[decoder]/T5Block[4]', 'T5ForConditionalGeneration/T5Stack[decoder]/T5Block[5]', 'T5ForConditionalGeneration/T5Stack[decoder]/T5LayerNorm[final_layer_norm]', 'T5ForConditionalGeneration/T5Stack[decoder]/Dropout[dropout]', 'T5ForConditionalGeneration/Linear[lm_head]', 'T5ForConditionalGeneration/CrossEntropyLoss[lm_loss]'] TENSORS = [] def __init__(self, layers, tensors, device='cuda:5'): super().__init__() for (idx, layer_scope) in enumerate(self.LAYER_SCOPES): self.add_module(f'l_{idx}', layers[layer_scope]) b = p = 0 for tensor_scope in self.TENSORS: tensor = tensors[tensor_scope] if isinstance(tensor, nn.Parameter): self.register_parameter(f'p_{p}', tensor) p += 1 else: self.register_buffer(f'b_{b}', tensor) b += 1 self.device = torch.device(device) self.input_structure = [1, 1, 1, 1, 1, 1, 1] self.lookup = {'l_0': 'decoder.3', 'l_1': 'decoder.4', 'l_2': 'decoder.5', 'l_3': 'decoder.final_layer_norm', 'l_4': 'decoder.dropout', 'l_5': 'lm_head', 'l_6': 'lm_loss'} self.to(self.device) def forward(self, *args): (decoder_attention_mask, inverted_encoder_attention_mask, lm_labels, x0, x1, x2, x3) = unflatten(args, self.input_structure) t_0 = self.l_0(x3, attention_mask=decoder_attention_mask, position_bias=x1, encoder_hidden_states=x0, encoder_attention_mask=inverted_encoder_attention_mask, encoder_decoder_position_bias=x2) t_0 = self.l_1(t_0, attention_mask=decoder_attention_mask, position_bias=x1, encoder_hidden_states=x0, encoder_attention_mask=inverted_encoder_attention_mask, encoder_decoder_position_bias=x2) t_0 = self.l_2(t_0, attention_mask=decoder_attention_mask, position_bias=x1, encoder_hidden_states=x0, encoder_attention_mask=inverted_encoder_attention_mask, encoder_decoder_position_bias=x2) t_0 = self.l_3(t_0) t_0 = self.l_4(t_0) t_0 = (t_0 * 0.) t_0 = self.l_5(t_0) t_1 = t_0.size((- 1)) t_1 = t_0.view((- 1), t_1) t_0 = lm_labels.view((- 1)) t_0 = self.l_6(t_1, t_0) return (t_0,) def state_dict(self, *args, **kwargs): return state_dict(self, *args, **kwargs) def load_state_dict(self, state): return load_state_dict(self, state) def named_parameters(self, recurse=True): return named_parameters(self, recurse=recurse) def named_buffers(self, recurse=True): return named_buffers(self, recurse=recurse) def cpu(self): return cpu(self) def cuda(self, device=None): return cuda(self, device=device) def to(self, *args, **kwargs): return to(self, *args, **kwargs)
def dump_example(dataset_name): print('Converting {:}.h5 ...'.format(dataset_name)) file = h5py.File(os.path.join(path, 'traindata', '{:}.h5'.format(dataset_name)), 'r') for (seq_idx, seq_name) in enumerate(file): if (dataset_name == 'scenes11_train'): scale = 0.4 else: scale = 1 if (((dataset_name == 'sun3d_train_1.6m_to_infm') and (seq_idx == 7)) or ((dataset_name == 'sun3d_train_0.4m_to_0.8m') and (seq_idx == 15)) or ((dataset_name == 'scenes11_train') and ((seq_idx == 2758) or (seq_idx == 4691) or (seq_idx == 7023) or (seq_idx == 11157) or (seq_idx == 17168) or (seq_idx == 19595)))): continue print('Processing sequence {:d}/{:d}'.format(seq_idx, len(file))) dump_dir = os.path.join(path, '../train', ((dataset_name + '_') + '{:05d}'.format(seq_idx))) if (not os.path.isdir(dump_dir)): os.mkdir(dump_dir) dump_dir = Path(dump_dir) sequence = file[seq_name]['frames']['t0'] poses = [] for (f_idx, f_name) in enumerate(sequence): frame = sequence[f_name] for dt_type in frame: dataset = frame[dt_type] img = dataset[...] if (dt_type == 'camera'): if (f_idx == 0): intrinsics = np.array([[img[0], 0, img[3]], [0, img[1], img[4]], [0, 0, 1]]) pose = np.array([[img[5], img[8], img[11], (img[14] * scale)], [img[6], img[9], img[12], (img[15] * scale)], [img[7], img[10], img[13], (img[16] * scale)]]) poses.append(pose.tolist()) elif (dt_type == 'depth'): dimension = dataset.attrs['extents'] depth = np.array(np.frombuffer(decompress(img.tobytes(), ((dimension[0] * dimension[1]) * 2)), dtype=np.float16)).astype(np.float32) depth = (depth.reshape(dimension[0], dimension[1]) * scale) dump_depth_file = (dump_dir / '{:04d}.npy'.format(f_idx)) np.save(dump_depth_file, depth) elif (dt_type == 'image'): img = imageio.imread(img.tobytes()) dump_img_file = (dump_dir / '{:04d}.jpg'.format(f_idx)) scipy.misc.imsave(dump_img_file, img) dump_cam_file = (dump_dir / 'cam.txt') np.savetxt(dump_cam_file, intrinsics) poses_file = (dump_dir / 'poses.txt') np.savetxt(poses_file, np.array(poses).reshape((- 1), 12), fmt='%.6e') if (len(dump_dir.files('*.jpg')) < 2): dump_dir.rmtree()
_mapper() def modify_in_place(x: DataPoint) -> DataPoint: x.d['my_key'] = 0 return Row(num=x.num, d=x.d, d_new=x.d)
class _Encoder(nn.Module): def __init__(self, imageSize): super(_Encoder, self).__init__() n = math.log2(imageSize) assert (n == round(n)), 'imageSize must be a power of 2' assert (n >= 3), 'imageSize must be at least 8' n = int(n) self.conv1 = nn.Conv2d((ngf * (2 ** (n - 3))), nz, 4) self.conv2 = nn.Conv2d((ngf * (2 ** (n - 3))), nz, 4) self.encoder = nn.Sequential() self.encoder.add_module('input-conv', nn.Conv2d(nc, ngf, 4, 2, 1, bias=False)) self.encoder.add_module('input-relu', nn.LeakyReLU(0.2, inplace=True)) for i in range((n - 3)): self.encoder.add_module('pyramid.{0}-{1}.conv'.format((ngf * (2 ** i)), (ngf * (2 ** (i + 1)))), nn.Conv2d((ngf * (2 ** i)), (ngf * (2 ** (i + 1))), 4, 2, 1, bias=False)) self.encoder.add_module('pyramid.{0}.batchnorm'.format((ngf * (2 ** (i + 1)))), nn.BatchNorm2d((ngf * (2 ** (i + 1))))) self.encoder.add_module('pyramid.{0}.relu'.format((ngf * (2 ** (i + 1)))), nn.LeakyReLU(0.2, inplace=True)) def forward(self, input): output = self.encoder(input) return [self.conv1(output), self.conv2(output)]
class TestParameters(unittest.TestCase): def test_parameters(self): gdb.execute('set cy_colorize_code on') assert libcython.parameters.colorize_code gdb.execute('set cy_colorize_code off') assert (not libcython.parameters.colorize_code)
_ENCODERS.register_module() class DarkNet53(nn.Module): def __init__(self, freeze_layer=2, pretrained='./data/weights/darknet.weights', out_layer=(6, 8, 13)): super(DarkNet53, self).__init__() self.fp16_enabled = False assert isinstance(out_layer, tuple) self.out_layer = out_layer self.darknet = nn.ModuleList([*darknet_conv((3, 32), (32, 64), (3, 3), (1, 2)), DarknetBlock(64), *darknet_conv((64,), (128,), (3,), (2,)), DarknetBlock(128, num_block=2), *darknet_conv((128,), (256,), (3,), (2,)), DarknetBlock(256, num_block=8), *darknet_conv((256,), (512,), (3,), (2,)), DarknetBlock(512, num_block=8), *darknet_conv((512,), (1024,), (3,), (2,)), DarknetBlock(1024, num_block=4), DarknetBlock(1024, num_block=2, shortcut=False), *darknet_conv((1024, 512), (512, 1024), (1, 3), (1, 1))]) if (pretrained is not None): parse_yolo_weights(self, pretrained, len(self.darknet)) if is_main(): logger = get_root_logger() logger.info(f'load pretrained visual backbone from {pretrained}') self.do_train = False if (freeze_layer is not None): freeze_params(self.darknet[:(- freeze_layer)]) else: self.do_train = True _fp32(apply_to=('img',)) def forward(self, img, y): x = [] for (i, mod) in enumerate(self.darknet): img = mod(img) if (i in self.out_layer): x.append(img) if (len(self.out_layer) == 1): return x[0] else: return x
def batch_normalization(x, beta, gamma, mean, variance, axes=[1], decay_rate=0.9, eps=1e-05, batch_stat=True, output_stat=False, n_outputs=None): from .function_bases import batch_normalization as batch_normalization_base n_outputs = (3 if output_stat else 1) axes = _force_list(axes) axes = [(a + (len(x.shape) * (a < 0))) for a in axes] assert (batch_stat or (not output_stat)) if ((not batch_stat) and ((mean is None) or (variance is None))): raise ValueError('If batch_stat is False, mean and variable must not be None.') (_, _, mean, variance) = _create_bn_dummy_vars(x, axes, beta, gamma, mean, variance) if (batch_stat and ((mean.parent or variance.parent) is not None)): raise ValueError('if batch_stat is True, mean and variable must not have a parent function.') no_scale = (gamma is None) no_bias = (beta is None) if (len(axes) == 1): return batch_normalization_base(x, beta, gamma, mean, variance, axes=axes, decay_rate=decay_rate, eps=eps, batch_stat=batch_stat, no_scale=no_scale, no_bias=no_bias, n_outputs=n_outputs) in_adapter = BatchNormalizationInOutAdapter(x.ndim, axes) param_adapter = BatchNormalizationInOutAdapter(x.ndim, axes) inp = in_adapter(x) if (beta is not None): beta = param_adapter(beta) if (gamma is not None): gamma = param_adapter(gamma) mean = param_adapter(mean) variance = param_adapter(variance) axis = (x.ndim - len(axes)) if (n_outputs == 1): out = batch_normalization_base(inp, beta, gamma, mean, variance, axes=[axis], decay_rate=decay_rate, eps=eps, batch_stat=batch_stat, no_scale=no_scale, no_bias=no_bias, n_outputs=n_outputs) return in_adapter.inv(out) (out, mean, variance) = batch_normalization_base(inp, beta, gamma, mean, variance, axes=[axis], decay_rate=decay_rate, eps=eps, batch_stat=batch_stat, no_scale=no_scale, no_bias=no_bias, n_outputs=n_outputs) out = in_adapter.inv(out) mean = param_adapter.inv(mean) variance = param_adapter.inv(variance) return (out, mean, variance)
def supersample(clip, d, nframes): def fl(gf, t): tt = np.linspace((t - d), (t + d), nframes) avg = np.mean((1.0 * np.array([gf(t_) for t_ in tt], dtype='uint16')), axis=0) return avg.astype('uint8') return clip.fl(fl)
def get_layers(in_index, in_channels, embed_dims, channels, embed_neck_cfg, embed_cfg, fusion_cfg): embed_layers = {} for (i, in_channels, embed_dim) in zip(in_index, in_channels, embed_dims): if (i == in_index[(- 1)]): embed_layers[str(i)] = build_layer(in_channels, embed_dim, **embed_neck_cfg) else: embed_layers[str(i)] = build_layer(in_channels, embed_dim, **embed_cfg) embed_layers = nn.ModuleDict(embed_layers) fuse_layer = build_layer(sum(embed_dims), channels, **fusion_cfg) return (embed_layers, fuse_layer)
def cleaner_mimic(text, spacy=True): text = re.sub('\\s+', ' ', text.strip()) if spacy: text = [t.text.lower() for t in nlp(text)] else: text = [t.lower() for t in text.split()] text = ' '.join(text) text = re.sub('\\[\\s*\\*\\s*\\*(.*?)\\*\\s*\\*\\s*\\]', ' <DE> ', text) text = re.sub('([^a-zA-Z0-9])(\\s*\\1\\s*)+', '\\1 ', text) text = re.sub('\\s+', ' ', text.strip()) text = [('qqq' if any((char.isdigit() for char in word)) else word) for word in text.split(' ')] return ' '.join(text)
class _ReBenchDB(_ConcretePersistence): def __init__(self, configurator, data_store, ui): super(_ReBenchDB, self).__init__(data_store, ui) self._configurator = configurator self._rebench_db = configurator.get_rebench_db_connector() self._lock = Lock() self._cache_for_seconds = 30 self._cache = {} self._last_send = time() def set_start_time(self, start_time): assert (self._start_time is None) self._start_time = start_time def load_data(self, runs, discard_run_data): raise RuntimeError('Does not yet support data loading from ReBenchDB') def persist_data_point(self, data_point): with self._lock: if (data_point.run_id not in self._cache): self._cache[data_point.run_id] = [] self._cache[data_point.run_id].append(data_point) def send_data(self): current_time = time() time_past = (current_time - self._last_send) self.ui.debug_output_info('ReBenchDB: data last send {seconds}s ago\n', seconds=round(time_past, 2)) if (time_past >= self._cache_for_seconds): self._send_data_and_empty_cache() self._last_send = time() def _send_data_and_empty_cache(self): if self._cache: if self._send_data(self._cache): self._cache = {} def _send_data(self, cache): self.ui.debug_output_info('ReBenchDB: Prepare data for sending\n') num_measurements = 0 all_data = [] criteria = {} for (run_id, data_points) in cache.items(): dp_data = [] for dp in data_points: measurements = dp.measurements_as_dict(criteria) num_measurements += len(measurements['m']) dp_data.append(measurements) all_data.append({'runId': run_id.as_dict(), 'd': dp_data}) criteria_index = [] for (c, idx) in criteria.items(): criteria_index.append({'c': c[0], 'u': c[1], 'i': idx}) self.ui.debug_output_info('ReBenchDB: Sending {num_m} measures. startTime: {st}\n', num_m=num_measurements, st=self._start_time) return self._rebench_db.send_results({'data': all_data, 'criteria': criteria_index, 'env': determine_environment(), 'startTime': self._start_time, 'source': determine_source_details(self._configurator)}, num_measurements) def close(self): with self._lock: self._send_data_and_empty_cache()
class Experiment(): def __init__(self, experiment_id, params): self._experiment_id = experiment_id self._cluster_spec = params['cluster_spec'] self._policy = params['policy'] self._seed = int(params['seed']) self._lam = (- 1) self._num_total_jobs = (- 1) self._profiling_percentage = (- 1) self._num_reference_models = (- 1) self._average_jct = None self._utilization = None self._makespan = None self._total_cost = None self._num_SLO_violations = None if ('lam' in params): self._lam = float(params['lam']) if ('num_total_jobs' in params): self._num_total_jobs = int(params['num_total_jobs']) if ('profiling_percentage' in params): self._profiling_percentage = float(params['profiling_percentage']) if ('num_reference_models' in params): self._num_reference_models = int(params['num_reference_models']) def update_results(self, results): self._average_jct = float(results['average JCT']) self._utilization = float(results['utilization']) if ('makespan' in results): self._makespan = float(results['makespan']) if ('total_cost' in results): self._total_cost = float(results['total_cost'][1:]) if ('num_SLO_violations' in results): self._num_SLO_violations = int(results['num_SLO_violations'])
('data.dmlab', 'class') class DmlabData(base.ImageTfdsData): def __init__(self, data_dir=None): dataset_builder = tfds.builder('dmlab:2.0.1', data_dir=data_dir) tfds_splits = {'train': 'train', 'val': 'validation', 'trainval': 'train+validation', 'test': 'test', 'train800': 'train[:800]', 'val200': 'validation[:200]', 'train800val200': 'train[:800]+validation[:200]'} train_count = dataset_builder.info.splits['train'].num_examples val_count = dataset_builder.info.splits['validation'].num_examples test_count = dataset_builder.info.splits['test'].num_examples num_samples_splits = {'train': train_count, 'val': val_count, 'trainval': (train_count + val_count), 'test': test_count, 'train800': 800, 'val200': 200, 'train800val200': 1000} super(DmlabData, self).__init__(dataset_builder=dataset_builder, tfds_splits=tfds_splits, num_samples_splits=num_samples_splits, num_preprocessing_threads=400, shuffle_buffer_size=10000, base_preprocess_fn=base.make_get_and_cast_tensors_fn({'image': ('image', None), 'label': ('label', None)}), num_classes=dataset_builder.info.features['label'].num_classes, image_key='image')
def _load_split_txt(path): with open(path, 'r') as f: return list(map((lambda s: str(s.split()[0])), f.readlines()))
class BlackBodySimpleSourceRelativistic(BlackBodySimpleSource): def from_model(cls, model, *args, **kwargs): return cls(model.time_explosion, model.r_inner[0], model.t_inner.value, *args, **kwargs) def __init__(self, time_explosion=None, **kwargs): self.time_explosion = time_explosion super().__init__(**kwargs) def set_state_from_model(self, model): self.time_explosion = model.time_explosion super().set_state_from_model(model) def create_packets(self, no_of_packets): if ((self.radius is None) or (self.time_explosion is None)): raise ValueError("Black body Radius or Time of Explosion isn't set") self.beta = ((self.radius / self.time_explosion) / const.c).to('') return super().create_packets(no_of_packets) def create_packet_mus(self, no_of_packets): z = self.rng.random(no_of_packets) beta = self.beta return ((- beta) + np.sqrt((((beta ** 2) + ((2 * beta) * z)) + z))) def create_packet_energies(self, no_of_packets): beta = self.beta gamma = (1.0 / np.sqrt((1 - (beta ** 2)))) static_inner_boundary2cmf_factor = (((2 * beta) + 1) / (1 - (beta ** 2))) energies = (np.ones(no_of_packets) / no_of_packets) return ((energies * static_inner_boundary2cmf_factor) / gamma)
def repo_list(recipe_folder='tests/recipes', field='HF_repo'): HF_repos = [] for recipe_csvfile in os.listdir(recipe_folder): if (recipe_csvfile in __skip_list): continue with open(os.path.join(recipe_folder, recipe_csvfile), newline='') as csvf: reader = csv.DictReader(csvf, delimiter=',', skipinitialspace=True) for row in reader: if (len(row[field]) > 0): repos = row[field].split(' ') for repo in repos: if (len(repo) > 0): HF_repos.append(repo) HF_repos = set(HF_repos) return HF_repos
def test_case57(): url = (brokerIp + '/ngsi-ld/v1/entityOperations/upsert') headers = {'Content-Type': 'application/json', 'Link': '<{{link}}>; rel=" type="application/ld+json"'} r = requests.post(url, data=json.dumps(ld_data.subdata48), headers=headers) print(r.content) assert (r.status_code == 404)
def gaussian_measure_full(mean, cov, f): if (not is_pos_def(cov)): logger.warn(f'cov={cov} not positive definite') L = cholesky(cov) def integrand(x): y = ((L x) + mean) return (norm_pdf(x) * f(y)) K = mean.shape[0] lim = ([[(- 10), 10]] * K) integral = nquad(integrand, lim)[0] return integral
def all_reduce_losses(losses): (names, values) = ([], []) for (k, v) in losses.items(): names.append(k) values.append(v) values = torch.cat([v.view(1) for v in values], dim=0) dist.all_reduce(values, dist.ReduceOp.SUM) values.div_(dist.get_world_size()) values = torch.chunk(values, values.size(0), dim=0) return OrderedDict(((k, v.view(())) for (k, v) in zip(names, values)))
def test_ticket_701(): arr = numpy.arange(4).reshape((2, 2)) def func(x): return numpy.min(x) res = ndimage.generic_filter(arr, func, size=(1, 1)) res2 = ndimage.generic_filter(arr, func, size=1) assert_equal(res, res2)
def test_KMaxPooling(): with CustomObjectScope({'KMaxPooling': sequence.KMaxPooling}): layer_test(sequence.KMaxPooling, kwargs={'k': 3, 'axis': 1}, input_shape=(BATCH_SIZE, SEQ_LENGTH, EMBEDDING_SIZE, 2))
class BSNSNmat(SpectralMatrix): def assemble(self, method): (test, trial) = (self.testfunction, self.trialfunction) assert isinstance(test[0], SN) assert isinstance(trial[0], SN) N = test[0].N k = np.arange((N - 2), dtype=float) alpha = (((k * (k + 1)) / (k + 2)) / (k + 3)) d0 = get_norm_sq(test[0], trial[0], method) d = {0: (d0[:(- 2)] + ((alpha ** 2) * d0[2:])), 2: ((- alpha[:(- 2)]) * d0[2:(- 2)])} d[(- 2)] = d[2].copy() return d
def seed_worker(worker_id): clear_logging() worker_seed = (torch.initial_seed() % (2 ** 32)) seed_everything(worker_seed)
def load_weight(sess, data, include=[]): for scope in include: for v in tf.compat.v1.global_variables(): if ((v.name in data.keys()) and (scope in v.name)): if (v.shape == data[v.name].shape): sess.run(v.assign(data[v.name])) print('load weight: ', v.name)
def get_training_roidb(imdb): if cfg.TRAIN.USE_FLIPPED: print('Appending horizontally-flipped training examples...') imdb.append_flipped_images() print('done') print('Preparing training data...') rdl_roidb.prepare_roidb(imdb) print('done') return imdb.roidb
class TFGroupViTPreTrainedModel(metaclass=DummyObject): _backends = ['tf'] def __init__(self, *args, **kwargs): requires_backends(self, ['tf'])
def _is_exception(obj): if (not inspect.isclass(obj)): return False return issubclass(obj, Exception)
class NormalizedClassifier(nn.Module): def __init__(self): super().__init__() self.weight = nn.Parameter(torch.Tensor(1501, 2048)) self.weight.data.uniform_((- 1), 1).renorm_(2, 0, 1e-05).mul_(100000.0) def forward(self, x): w = self.weight x = nn.functional.normalize(x, p=2, dim=1) w = nn.functional.normalize(w, p=2, dim=1) return nn.functional.linear(x, w)
_test() def test_matmul_np(): def matmul_np(A: dace.float64[(128, 64)], B: dace.float64[(64, 32)], C: dace.float64[(128, 32)]): C[:] = (A B) A = np.random.rand(128, 64).astype(np.float64) B = np.random.rand(64, 32).astype(np.float64) C = np.random.rand(128, 32).astype(np.float64) sdfg = matmul_np.to_sdfg() sdfg.apply_transformations([FPGATransformSDFG]) from dace.libraries.blas import Gemm Gemm.default_implementation = 'FPGA1DSystolic' sdfg.expand_library_nodes() sdfg.apply_transformations_repeated([InlineSDFG]) C_regression = (A B) sdfg(A=A, B=B, C=C) assert np.allclose(C, C_regression, atol=1e-06) return sdfg
class Optimizer(object): def __init__(self, cost, params): self.cost = cost self.params = params self.updates = self._updates() def _updates(self): raise NotImplementedError()
def read_posetrack_keypoints(output_folder): people = dict() for (idx, result_file) in enumerate(sorted(os.listdir(output_folder))): json_file = osp.join(output_folder, result_file) data = json.load(open(json_file)) for person in data['people']: person_id = person['person_id'][0] joints2d = person['pose_keypoints_2d'] if (person_id in people.keys()): people[person_id]['joints2d'].append(joints2d) people[person_id]['frames'].append(idx) else: people[person_id] = {'joints2d': [], 'frames': []} people[person_id]['joints2d'].append(joints2d) people[person_id]['frames'].append(idx) for k in people.keys(): people[k]['joints2d'] = np.array(people[k]['joints2d']).reshape((len(people[k]['joints2d']), (- 1), 3)) people[k]['frames'] = np.array(people[k]['frames']) return people
def log_likelihood(mu, var, x, muq, varq, a, mask_flat, config): if (config.out_distr == 'bernoulli'): log_lik = log_bernoulli(x, mu, eps=1e-06) elif (config.out_distr == 'gaussian'): log_lik = log_gaussian(x, mu, var) log_lik = tf.reduce_sum(log_lik, 1) log_lik = tf.multiply(mask_flat, log_lik) if (config.ll_keep_prob < 1.0): log_lik = tf.layers.dropout(log_lik, config.ll_keep_prob) num_el = tf.reduce_sum(mask_flat) log_px_given_a = tf.truediv(tf.reduce_sum(log_lik), num_el) if config.use_vae: log_qa_given_x = tf.reduce_sum(log_gaussian(a, muq, varq), 1) log_qa_given_x = tf.multiply(mask_flat, log_qa_given_x) log_qa_given_x = tf.truediv(tf.reduce_sum(log_qa_given_x), num_el) else: log_qa_given_x = tf.constant(0.0, dtype=tf.float32, shape=()) LL = (log_px_given_a - log_qa_given_x) return (LL, log_px_given_a, log_qa_given_x)
def write_list(out_filename, dataset): formatted_dataset = [line._asdict() for line in dataset] with open(out_filename, 'w') as fout: fout.write('[\n') for (idx, line) in enumerate(formatted_dataset): fout.write(' ') json.dump(line, fout, ensure_ascii=False) if (idx < (len(formatted_dataset) - 1)): fout.write(',') fout.write('\n') fout.write(']\n')
class ConcatChannel(SOFactor): n_next = 1 def __init__(self, Ns, axis=0): self.Ns = Ns self.axis = axis self.repr_init() self.n_prev = len(Ns) self.N = sum(Ns) def sample(self, *Zs): if (len(Zs) != self.n_prev): raise ValueError(f'expect {self.n_prev} arrays') for (k, Z) in enumerate(Zs): if (Z.shape[self.axis] != self.Ns[k]): raise ValueError(f'expect Z k={k} array of dimension {self.Ns[k]} along axis {self.axis} but got array of dimension {Z.shape[self.axis]}') X = np.concatenate(Zs, axis=self.axis) assert (X.shape[self.axis] == self.N) return X def math(self): return '$\\oplus$' def second_moment(self, *tau_zs): if (len(tau_zs) != self.n_prev): raise ValueError(f'expect {self.n_prev} tau_zs') tau_x = (sum(((N * tau_z) for (N, tau_z) in zip(self.Ns, tau_zs))) / self.N) return tau_x def compute_forward_posterior(self, az, bz, ax, bx): (rz, vz) = self.compute_backward_posterior(az, bz, ax, bx) rx = np.concatenate(rz, axis=self.axis) vx = (sum(((N * v) for (N, v) in zip(self.Ns, vz))) / self.N) return (rx, vx) def _compute_ak_bk(self, az, bz, ax, bx): for (N, Z) in zip(self.Ns, bz): assert (bz.shape[self.axis] == N) assert (bx.shape[self.axis] == self.N) idx = ([0] + list(np.cumsum(self.Ns))) bx_subs = [np.take(bx, range(idx_min, idx_max), axis=self.axis) for (idx_min, idx_max) in zip(idx[:(- 1)], idx[1:])] ak = [(a + ax) for a in az] bk = [(b + bx_sub) for (b, bx_sub) in zip(bz, bx_subs)] return (ak, bk) def compute_backward_posterior(self, az, bz, ax, bx): (ak, bk) = self._compute_ak_bk(az, bz, ax, bx) vz = [(1 / a) for a in ak] rz = [(b / a) for (a, b) in zip(ak, bk)] return (rz, vz) def compute_forward_error(self, az, ax, tau_z): vz = self.compute_backward_error(az, ax, tau_z) vx = (sum(((N * v) for (N, v) in zip(self.Ns, vz))) / self.N) return vx def compute_backward_error(self, az, ax, tau_z): ak = [(a + ax) for a in az] vz = [(1 / ak) for a in ak] return vz def compute_log_partition(self, az, bz, ax, bx): (ak, bk) = self._compute_ak_bk(az, bz, ax, bx) logZ = sum([(0.5 * np.sum((((b ** 2) / a) + np.log(((2 * np.pi) / a))))) for (a, b) in zip(ak, bk)]) return logZ def compute_free_energy(self, az, ax, tau_z): raise NotImplementedError
def write_label_file(labels_to_class_names, dataset_dir, filename=LABELS_FILENAME): labels_filename = os.path.join(dataset_dir, filename) with tf.gfile.Open(labels_filename, 'w') as f: for label in labels_to_class_names: class_name = labels_to_class_names[label] f.write(('%d:%s\n' % (label, class_name)))
def main(): args = create_argparser().parse_args() logger.log(f'args: {args}') dist_util.setup_dist() logger.configure() logger.log('creating 2d model and diffusion...') (model, diffusion) = create_model_and_diffusion_2d(**args_to_dict(args, model_and_diffusion_defaults_2d().keys())) model.to(dist_util.dev()) schedule_sampler = create_named_schedule_sampler(args.schedule_sampler, diffusion) logger.log('creating 2d data loader...') data = load_2d_data(batch_size=args.batch_size, shape=list(Synthetic2DType)[args.task]) logger.log('training 2d model...') TrainLoop(model=model, diffusion=diffusion, data=data, batch_size=args.batch_size, microbatch=args.microbatch, lr=args.lr, ema_rate=args.ema_rate, log_interval=args.log_interval, save_interval=args.save_interval, resume_checkpoint=args.resume_checkpoint, use_fp16=args.use_fp16, fp16_scale_growth=args.fp16_scale_growth, schedule_sampler=schedule_sampler, weight_decay=args.weight_decay, lr_anneal_steps=args.lr_anneal_steps).run_loop()
class Partition0(nn.Module): LAYER_SCOPES = ['BertForQuestionAnswering/BertModel[bert]/BertEmbeddings[embeddings]/Embedding[word_embeddings]', 'BertForQuestionAnswering/BertModel[bert]/BertEmbeddings[embeddings]/Embedding[position_embeddings]', 'BertForQuestionAnswering/BertModel[bert]/BertEmbeddings[embeddings]/Embedding[token_type_embeddings]', 'BertForQuestionAnswering/BertModel[bert]/BertEmbeddings[embeddings]/LayerNorm[LayerNorm]', 'BertForQuestionAnswering/BertModel[bert]/BertEmbeddings[embeddings]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[0]/BertAttention[attention]/BertSelfAttention[self]/Linear[query]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[0]/BertAttention[attention]/BertSelfAttention[self]/Linear[key]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[0]/BertAttention[attention]/BertSelfAttention[self]/Linear[value]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[0]/BertAttention[attention]/BertSelfAttention[self]/Softmax[softmax]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[0]/BertAttention[attention]/BertSelfAttention[self]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[0]/BertAttention[attention]/BertSelfOutput[output]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[0]/BertAttention[attention]/BertSelfOutput[output]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[0]/BertAttention[attention]/BertSelfOutput[output]/LayerNorm[LayerNorm]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[0]/BertIntermediate[intermediate]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[0]/BertOutput[output]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[0]/BertOutput[output]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[0]/BertOutput[output]/LayerNorm[LayerNorm]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[1]/BertAttention[attention]/BertSelfAttention[self]/Linear[query]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[1]/BertAttention[attention]/BertSelfAttention[self]/Linear[key]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[1]/BertAttention[attention]/BertSelfAttention[self]/Linear[value]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[1]/BertAttention[attention]/BertSelfAttention[self]/Softmax[softmax]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[1]/BertAttention[attention]/BertSelfAttention[self]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[1]/BertAttention[attention]/BertSelfOutput[output]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[1]/BertAttention[attention]/BertSelfOutput[output]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[1]/BertAttention[attention]/BertSelfOutput[output]/LayerNorm[LayerNorm]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[1]/BertIntermediate[intermediate]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[1]/BertOutput[output]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[1]/BertOutput[output]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[1]/BertOutput[output]/LayerNorm[LayerNorm]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[2]/BertAttention[attention]/BertSelfAttention[self]/Linear[query]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[2]/BertAttention[attention]/BertSelfAttention[self]/Linear[key]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[2]/BertAttention[attention]/BertSelfAttention[self]/Linear[value]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[2]/BertAttention[attention]/BertSelfAttention[self]/Softmax[softmax]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[2]/BertAttention[attention]/BertSelfAttention[self]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[2]/BertAttention[attention]/BertSelfOutput[output]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[2]/BertAttention[attention]/BertSelfOutput[output]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[2]/BertAttention[attention]/BertSelfOutput[output]/LayerNorm[LayerNorm]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[2]/BertIntermediate[intermediate]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[2]/BertOutput[output]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[2]/BertOutput[output]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[2]/BertOutput[output]/LayerNorm[LayerNorm]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[3]/BertAttention[attention]/BertSelfAttention[self]/Linear[query]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[3]/BertAttention[attention]/BertSelfAttention[self]/Linear[key]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[3]/BertAttention[attention]/BertSelfAttention[self]/Linear[value]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[3]/BertAttention[attention]/BertSelfAttention[self]/Softmax[softmax]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[3]/BertAttention[attention]/BertSelfAttention[self]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[3]/BertAttention[attention]/BertSelfOutput[output]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[3]/BertAttention[attention]/BertSelfOutput[output]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[3]/BertAttention[attention]/BertSelfOutput[output]/LayerNorm[LayerNorm]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[3]/BertIntermediate[intermediate]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[3]/BertOutput[output]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[3]/BertOutput[output]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[3]/BertOutput[output]/LayerNorm[LayerNorm]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[4]/BertAttention[attention]/BertSelfAttention[self]/Linear[query]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[4]/BertAttention[attention]/BertSelfAttention[self]/Linear[key]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[4]/BertAttention[attention]/BertSelfAttention[self]/Linear[value]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[4]/BertAttention[attention]/BertSelfAttention[self]/Softmax[softmax]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[4]/BertAttention[attention]/BertSelfAttention[self]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[4]/BertAttention[attention]/BertSelfOutput[output]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[4]/BertAttention[attention]/BertSelfOutput[output]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[4]/BertAttention[attention]/BertSelfOutput[output]/LayerNorm[LayerNorm]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[4]/BertIntermediate[intermediate]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[4]/BertOutput[output]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[4]/BertOutput[output]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[4]/BertOutput[output]/LayerNorm[LayerNorm]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[5]/BertAttention[attention]/BertSelfAttention[self]/Linear[query]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[5]/BertAttention[attention]/BertSelfAttention[self]/Linear[key]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[5]/BertAttention[attention]/BertSelfAttention[self]/Linear[value]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[5]/BertAttention[attention]/BertSelfAttention[self]/Softmax[softmax]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[5]/BertAttention[attention]/BertSelfAttention[self]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[5]/BertAttention[attention]/BertSelfOutput[output]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[5]/BertAttention[attention]/BertSelfOutput[output]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[5]/BertAttention[attention]/BertSelfOutput[output]/LayerNorm[LayerNorm]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[5]/BertIntermediate[intermediate]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[5]/BertOutput[output]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[5]/BertOutput[output]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[5]/BertOutput[output]/LayerNorm[LayerNorm]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[6]/BertAttention[attention]/BertSelfAttention[self]/Linear[query]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[6]/BertAttention[attention]/BertSelfAttention[self]/Linear[key]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[6]/BertAttention[attention]/BertSelfAttention[self]/Linear[value]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[6]/BertAttention[attention]/BertSelfAttention[self]/Softmax[softmax]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[6]/BertAttention[attention]/BertSelfAttention[self]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[6]/BertAttention[attention]/BertSelfOutput[output]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[6]/BertAttention[attention]/BertSelfOutput[output]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[6]/BertAttention[attention]/BertSelfOutput[output]/LayerNorm[LayerNorm]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[6]/BertIntermediate[intermediate]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[6]/BertOutput[output]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[6]/BertOutput[output]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[6]/BertOutput[output]/LayerNorm[LayerNorm]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[7]/BertAttention[attention]/BertSelfAttention[self]/Linear[query]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[7]/BertAttention[attention]/BertSelfAttention[self]/Linear[key]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[7]/BertAttention[attention]/BertSelfAttention[self]/Linear[value]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[7]/BertAttention[attention]/BertSelfAttention[self]/Softmax[softmax]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[7]/BertAttention[attention]/BertSelfAttention[self]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[7]/BertAttention[attention]/BertSelfOutput[output]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[7]/BertAttention[attention]/BertSelfOutput[output]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[7]/BertAttention[attention]/BertSelfOutput[output]/LayerNorm[LayerNorm]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[7]/BertIntermediate[intermediate]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[7]/BertOutput[output]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[7]/BertOutput[output]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[7]/BertOutput[output]/LayerNorm[LayerNorm]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[8]/BertAttention[attention]/BertSelfAttention[self]/Linear[query]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[8]/BertAttention[attention]/BertSelfAttention[self]/Linear[key]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[8]/BertAttention[attention]/BertSelfAttention[self]/Linear[value]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[8]/BertAttention[attention]/BertSelfAttention[self]/Softmax[softmax]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[8]/BertAttention[attention]/BertSelfAttention[self]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[8]/BertAttention[attention]/BertSelfOutput[output]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[8]/BertAttention[attention]/BertSelfOutput[output]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[8]/BertAttention[attention]/BertSelfOutput[output]/LayerNorm[LayerNorm]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[8]/BertIntermediate[intermediate]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[8]/BertOutput[output]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[8]/BertOutput[output]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[8]/BertOutput[output]/LayerNorm[LayerNorm]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[9]/BertAttention[attention]/BertSelfAttention[self]/Linear[query]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[9]/BertAttention[attention]/BertSelfAttention[self]/Linear[key]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[9]/BertAttention[attention]/BertSelfAttention[self]/Linear[value]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[9]/BertAttention[attention]/BertSelfAttention[self]/Softmax[softmax]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[9]/BertAttention[attention]/BertSelfAttention[self]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[9]/BertAttention[attention]/BertSelfOutput[output]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[9]/BertAttention[attention]/BertSelfOutput[output]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[9]/BertAttention[attention]/BertSelfOutput[output]/LayerNorm[LayerNorm]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[9]/BertIntermediate[intermediate]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[9]/BertOutput[output]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[9]/BertOutput[output]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[9]/BertOutput[output]/LayerNorm[LayerNorm]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[10]/BertAttention[attention]/BertSelfAttention[self]/Linear[query]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[10]/BertAttention[attention]/BertSelfAttention[self]/Linear[key]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[10]/BertAttention[attention]/BertSelfAttention[self]/Linear[value]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[10]/BertAttention[attention]/BertSelfAttention[self]/Softmax[softmax]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[10]/BertAttention[attention]/BertSelfAttention[self]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[10]/BertAttention[attention]/BertSelfOutput[output]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[10]/BertAttention[attention]/BertSelfOutput[output]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[10]/BertAttention[attention]/BertSelfOutput[output]/LayerNorm[LayerNorm]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[10]/BertIntermediate[intermediate]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[10]/BertOutput[output]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[10]/BertOutput[output]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[10]/BertOutput[output]/LayerNorm[LayerNorm]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[11]/BertAttention[attention]/BertSelfAttention[self]/Linear[query]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[11]/BertAttention[attention]/BertSelfAttention[self]/Linear[key]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[11]/BertAttention[attention]/BertSelfAttention[self]/Linear[value]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[11]/BertAttention[attention]/BertSelfAttention[self]/Softmax[softmax]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[11]/BertAttention[attention]/BertSelfAttention[self]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[11]/BertAttention[attention]/BertSelfOutput[output]/Linear[dense]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[11]/BertAttention[attention]/BertSelfOutput[output]/Dropout[dropout]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[11]/BertAttention[attention]/BertSelfOutput[output]/LayerNorm[LayerNorm]', 'BertForQuestionAnswering/BertModel[bert]/BertEncoder[encoder]/BertLayer[11]/BertIntermediate[intermediate]/Linear[dense]'] TENSORS = [] def __init__(self, layers, tensors, device='cuda:0'): super().__init__() for (idx, layer_scope) in enumerate(self.LAYER_SCOPES): self.add_module(f'l_{idx}', layers[layer_scope]) b = p = 0 for tensor_scope in self.TENSORS: tensor = tensors[tensor_scope] if isinstance(tensor, nn.Parameter): self.register_parameter(f'p_{p}', tensor) p += 1 else: self.register_buffer(f'b_{b}', tensor) b += 1 self.device = torch.device(device) self.input_structure = [1, 1, 1] self.lookup = {'l_0': 'bert.embeddings.word_embeddings', 'l_1': 'bert.embeddings.position_embeddings', 'l_2': 'bert.embeddings.token_type_embeddings', 'l_3': 'bert.embeddings.LayerNorm', 'l_4': 'bert.embeddings.dropout', 'l_5': 'bert.encoder.0.attention.self.query', 'l_6': 'bert.encoder.0.attention.self.key', 'l_7': 'bert.encoder.0.attention.self.value', 'l_8': 'bert.encoder.0.attention.self.softmax', 'l_9': 'bert.encoder.0.attention.self.dropout', 'l_10': 'bert.encoder.0.attention.output.dense', 'l_11': 'bert.encoder.0.attention.output.dropout', 'l_12': 'bert.encoder.0.attention.output.LayerNorm', 'l_13': 'bert.encoder.0.intermediate.dense', 'l_14': 'bert.encoder.0.output.dense', 'l_15': 'bert.encoder.0.output.dropout', 'l_16': 'bert.encoder.0.output.LayerNorm', 'l_17': 'bert.encoder.1.attention.self.query', 'l_18': 'bert.encoder.1.attention.self.key', 'l_19': 'bert.encoder.1.attention.self.value', 'l_20': 'bert.encoder.1.attention.self.softmax', 'l_21': 'bert.encoder.1.attention.self.dropout', 'l_22': 'bert.encoder.1.attention.output.dense', 'l_23': 'bert.encoder.1.attention.output.dropout', 'l_24': 'bert.encoder.1.attention.output.LayerNorm', 'l_25': 'bert.encoder.1.intermediate.dense', 'l_26': 'bert.encoder.1.output.dense', 'l_27': 'bert.encoder.1.output.dropout', 'l_28': 'bert.encoder.1.output.LayerNorm', 'l_29': 'bert.encoder.2.attention.self.query', 'l_30': 'bert.encoder.2.attention.self.key', 'l_31': 'bert.encoder.2.attention.self.value', 'l_32': 'bert.encoder.2.attention.self.softmax', 'l_33': 'bert.encoder.2.attention.self.dropout', 'l_34': 'bert.encoder.2.attention.output.dense', 'l_35': 'bert.encoder.2.attention.output.dropout', 'l_36': 'bert.encoder.2.attention.output.LayerNorm', 'l_37': 'bert.encoder.2.intermediate.dense', 'l_38': 'bert.encoder.2.output.dense', 'l_39': 'bert.encoder.2.output.dropout', 'l_40': 'bert.encoder.2.output.LayerNorm', 'l_41': 'bert.encoder.3.attention.self.query', 'l_42': 'bert.encoder.3.attention.self.key', 'l_43': 'bert.encoder.3.attention.self.value', 'l_44': 'bert.encoder.3.attention.self.softmax', 'l_45': 'bert.encoder.3.attention.self.dropout', 'l_46': 'bert.encoder.3.attention.output.dense', 'l_47': 'bert.encoder.3.attention.output.dropout', 'l_48': 'bert.encoder.3.attention.output.LayerNorm', 'l_49': 'bert.encoder.3.intermediate.dense', 'l_50': 'bert.encoder.3.output.dense', 'l_51': 'bert.encoder.3.output.dropout', 'l_52': 'bert.encoder.3.output.LayerNorm', 'l_53': 'bert.encoder.4.attention.self.query', 'l_54': 'bert.encoder.4.attention.self.key', 'l_55': 'bert.encoder.4.attention.self.value', 'l_56': 'bert.encoder.4.attention.self.softmax', 'l_57': 'bert.encoder.4.attention.self.dropout', 'l_58': 'bert.encoder.4.attention.output.dense', 'l_59': 'bert.encoder.4.attention.output.dropout', 'l_60': 'bert.encoder.4.attention.output.LayerNorm', 'l_61': 'bert.encoder.4.intermediate.dense', 'l_62': 'bert.encoder.4.output.dense', 'l_63': 'bert.encoder.4.output.dropout', 'l_64': 'bert.encoder.4.output.LayerNorm', 'l_65': 'bert.encoder.5.attention.self.query', 'l_66': 'bert.encoder.5.attention.self.key', 'l_67': 'bert.encoder.5.attention.self.value', 'l_68': 'bert.encoder.5.attention.self.softmax', 'l_69': 'bert.encoder.5.attention.self.dropout', 'l_70': 'bert.encoder.5.attention.output.dense', 'l_71': 'bert.encoder.5.attention.output.dropout', 'l_72': 'bert.encoder.5.attention.output.LayerNorm', 'l_73': 'bert.encoder.5.intermediate.dense', 'l_74': 'bert.encoder.5.output.dense', 'l_75': 'bert.encoder.5.output.dropout', 'l_76': 'bert.encoder.5.output.LayerNorm', 'l_77': 'bert.encoder.6.attention.self.query', 'l_78': 'bert.encoder.6.attention.self.key', 'l_79': 'bert.encoder.6.attention.self.value', 'l_80': 'bert.encoder.6.attention.self.softmax', 'l_81': 'bert.encoder.6.attention.self.dropout', 'l_82': 'bert.encoder.6.attention.output.dense', 'l_83': 'bert.encoder.6.attention.output.dropout', 'l_84': 'bert.encoder.6.attention.output.LayerNorm', 'l_85': 'bert.encoder.6.intermediate.dense', 'l_86': 'bert.encoder.6.output.dense', 'l_87': 'bert.encoder.6.output.dropout', 'l_88': 'bert.encoder.6.output.LayerNorm', 'l_89': 'bert.encoder.7.attention.self.query', 'l_90': 'bert.encoder.7.attention.self.key', 'l_91': 'bert.encoder.7.attention.self.value', 'l_92': 'bert.encoder.7.attention.self.softmax', 'l_93': 'bert.encoder.7.attention.self.dropout', 'l_94': 'bert.encoder.7.attention.output.dense', 'l_95': 'bert.encoder.7.attention.output.dropout', 'l_96': 'bert.encoder.7.attention.output.LayerNorm', 'l_97': 'bert.encoder.7.intermediate.dense', 'l_98': 'bert.encoder.7.output.dense', 'l_99': 'bert.encoder.7.output.dropout', 'l_100': 'bert.encoder.7.output.LayerNorm', 'l_101': 'bert.encoder.8.attention.self.query', 'l_102': 'bert.encoder.8.attention.self.key', 'l_103': 'bert.encoder.8.attention.self.value', 'l_104': 'bert.encoder.8.attention.self.softmax', 'l_105': 'bert.encoder.8.attention.self.dropout', 'l_106': 'bert.encoder.8.attention.output.dense', 'l_107': 'bert.encoder.8.attention.output.dropout', 'l_108': 'bert.encoder.8.attention.output.LayerNorm', 'l_109': 'bert.encoder.8.intermediate.dense', 'l_110': 'bert.encoder.8.output.dense', 'l_111': 'bert.encoder.8.output.dropout', 'l_112': 'bert.encoder.8.output.LayerNorm', 'l_113': 'bert.encoder.9.attention.self.query', 'l_114': 'bert.encoder.9.attention.self.key', 'l_115': 'bert.encoder.9.attention.self.value', 'l_116': 'bert.encoder.9.attention.self.softmax', 'l_117': 'bert.encoder.9.attention.self.dropout', 'l_118': 'bert.encoder.9.attention.output.dense', 'l_119': 'bert.encoder.9.attention.output.dropout', 'l_120': 'bert.encoder.9.attention.output.LayerNorm', 'l_121': 'bert.encoder.9.intermediate.dense', 'l_122': 'bert.encoder.9.output.dense', 'l_123': 'bert.encoder.9.output.dropout', 'l_124': 'bert.encoder.9.output.LayerNorm', 'l_125': 'bert.encoder.10.attention.self.query', 'l_126': 'bert.encoder.10.attention.self.key', 'l_127': 'bert.encoder.10.attention.self.value', 'l_128': 'bert.encoder.10.attention.self.softmax', 'l_129': 'bert.encoder.10.attention.self.dropout', 'l_130': 'bert.encoder.10.attention.output.dense', 'l_131': 'bert.encoder.10.attention.output.dropout', 'l_132': 'bert.encoder.10.attention.output.LayerNorm', 'l_133': 'bert.encoder.10.intermediate.dense', 'l_134': 'bert.encoder.10.output.dense', 'l_135': 'bert.encoder.10.output.dropout', 'l_136': 'bert.encoder.10.output.LayerNorm', 'l_137': 'bert.encoder.11.attention.self.query', 'l_138': 'bert.encoder.11.attention.self.key', 'l_139': 'bert.encoder.11.attention.self.value', 'l_140': 'bert.encoder.11.attention.self.softmax', 'l_141': 'bert.encoder.11.attention.self.dropout', 'l_142': 'bert.encoder.11.attention.output.dense', 'l_143': 'bert.encoder.11.attention.output.dropout', 'l_144': 'bert.encoder.11.attention.output.LayerNorm', 'l_145': 'bert.encoder.11.intermediate.dense'} self.to(self.device) def forward(self, *args): (attention_mask, input_ids, token_type_ids) = unflatten(args, self.input_structure) t_0 = self.l_0(input_ids) t_1 = self.l_2(token_type_ids) t_2 = attention_mask.unsqueeze(1) t_2 = t_2.unsqueeze(2) t_2 = t_2.to(dtype=torch.float32) t_2 = (1.0 - t_2) t_2 = (t_2 * (- 10000.0)) t_3 = input_ids.size(1) t_3 = torch.arange(t_3, dtype=torch.int64, device=self.device) t_3 = t_3.unsqueeze(0) t_3 = t_3.expand_as(input_ids) t_3 = self.l_1(t_3) t_3 = (t_0 + t_3) t_1 = (t_3 + t_1) t_1 = self.l_3(t_1) t_1 = self.l_4(t_1) t_3 = self.l_5(t_1) t_0 = self.l_6(t_1) t_4 = self.l_7(t_1) t_5 = t_3.size() t_6 = t_0.size() t_7 = t_4.size() t_5 = t_5[slice(None, (- 1), None)] t_5 = (t_5 + (16, 64)) t_8 = t_5[0] t_9 = t_5[1] t_10 = t_5[2] t_5 = t_5[3] t_5 = t_3.view(t_8, t_9, t_10, t_5) t_5 = t_5.permute(0, 2, 1, 3) t_6 = t_6[slice(None, (- 1), None)] t_6 = (t_6 + (16, 64)) t_10 = t_6[0] t_9 = t_6[1] t_8 = t_6[2] t_6 = t_6[3] t_6 = t_0.view(t_10, t_9, t_8, t_6) t_6 = t_6.permute(0, 2, 1, 3) t_7 = t_7[slice(None, (- 1), None)] t_7 = (t_7 + (16, 64)) t_8 = t_7[0] t_9 = t_7[1] t_10 = t_7[2] t_7 = t_7[3] t_7 = t_4.view(t_8, t_9, t_10, t_7) t_7 = t_7.permute(0, 2, 1, 3) t_6 = t_6.transpose((- 1), (- 2)) t_6 = torch.matmul(t_5, t_6) t_5 = math.sqrt(64) t_5 = (t_6 / t_5) t_5 = (t_5 + t_2) t_5 = self.l_8(t_5) t_5 = self.l_9(t_5) t_7 = torch.matmul(t_5, t_7) t_7 = t_7.permute(0, 2, 1, 3) t_7 = t_7.contiguous() t_5 = t_7.size() t_5 = t_5[slice(None, (- 2), None)] t_5 = (t_5 + (1024,)) t_6 = t_5[0] t_10 = t_5[1] t_5 = t_5[2] t_5 = t_7.view(t_6, t_10, t_5) t_5 = self.l_10(t_5) t_5 = self.l_11(t_5) t_1 = (t_5 + t_1) t_1 = self.l_12(t_1) t_5 = self.l_13(t_1) t_5 = torch.nn.functional.gelu(t_5) t_5 = self.l_14(t_5) t_5 = self.l_15(t_5) t_1 = (t_5 + t_1) t_1 = self.l_16(t_1) t_5 = self.l_17(t_1) t_10 = self.l_18(t_1) t_6 = self.l_19(t_1) t_7 = t_5.size() t_9 = t_10.size() t_8 = t_6.size() t_7 = t_7[slice(None, (- 1), None)] t_7 = (t_7 + (16, 64)) t_4 = t_7[0] t_0 = t_7[1] t_3 = t_7[2] t_7 = t_7[3] t_7 = t_5.view(t_4, t_0, t_3, t_7) t_7 = t_7.permute(0, 2, 1, 3) t_9 = t_9[slice(None, (- 1), None)] t_9 = (t_9 + (16, 64)) t_3 = t_9[0] t_0 = t_9[1] t_4 = t_9[2] t_9 = t_9[3] t_9 = t_10.view(t_3, t_0, t_4, t_9) t_9 = t_9.permute(0, 2, 1, 3) t_8 = t_8[slice(None, (- 1), None)] t_8 = (t_8 + (16, 64)) t_4 = t_8[0] t_0 = t_8[1] t_3 = t_8[2] t_8 = t_8[3] t_8 = t_6.view(t_4, t_0, t_3, t_8) t_8 = t_8.permute(0, 2, 1, 3) t_9 = t_9.transpose((- 1), (- 2)) t_9 = torch.matmul(t_7, t_9) t_7 = math.sqrt(64) t_7 = (t_9 / t_7) t_7 = (t_7 + t_2) t_7 = self.l_20(t_7) t_7 = self.l_21(t_7) t_8 = torch.matmul(t_7, t_8) t_8 = t_8.permute(0, 2, 1, 3) t_8 = t_8.contiguous() t_7 = t_8.size() t_7 = t_7[slice(None, (- 2), None)] t_7 = (t_7 + (1024,)) t_9 = t_7[0] t_3 = t_7[1] t_7 = t_7[2] t_7 = t_8.view(t_9, t_3, t_7) t_7 = self.l_22(t_7) t_7 = self.l_23(t_7) t_1 = (t_7 + t_1) t_1 = self.l_24(t_1) t_7 = self.l_25(t_1) t_7 = torch.nn.functional.gelu(t_7) t_7 = self.l_26(t_7) t_7 = self.l_27(t_7) t_1 = (t_7 + t_1) t_1 = self.l_28(t_1) t_7 = self.l_29(t_1) t_3 = self.l_30(t_1) t_9 = self.l_31(t_1) t_8 = t_7.size() t_0 = t_3.size() t_4 = t_9.size() t_8 = t_8[slice(None, (- 1), None)] t_8 = (t_8 + (16, 64)) t_6 = t_8[0] t_10 = t_8[1] t_5 = t_8[2] t_8 = t_8[3] t_8 = t_7.view(t_6, t_10, t_5, t_8) t_8 = t_8.permute(0, 2, 1, 3) t_0 = t_0[slice(None, (- 1), None)] t_0 = (t_0 + (16, 64)) t_5 = t_0[0] t_10 = t_0[1] t_6 = t_0[2] t_0 = t_0[3] t_0 = t_3.view(t_5, t_10, t_6, t_0) t_0 = t_0.permute(0, 2, 1, 3) t_4 = t_4[slice(None, (- 1), None)] t_4 = (t_4 + (16, 64)) t_6 = t_4[0] t_10 = t_4[1] t_5 = t_4[2] t_4 = t_4[3] t_4 = t_9.view(t_6, t_10, t_5, t_4) t_4 = t_4.permute(0, 2, 1, 3) t_0 = t_0.transpose((- 1), (- 2)) t_0 = torch.matmul(t_8, t_0) t_8 = math.sqrt(64) t_8 = (t_0 / t_8) t_8 = (t_8 + t_2) t_8 = self.l_32(t_8) t_8 = self.l_33(t_8) t_4 = torch.matmul(t_8, t_4) t_4 = t_4.permute(0, 2, 1, 3) t_4 = t_4.contiguous() t_8 = t_4.size() t_8 = t_8[slice(None, (- 2), None)] t_8 = (t_8 + (1024,)) t_0 = t_8[0] t_5 = t_8[1] t_8 = t_8[2] t_8 = t_4.view(t_0, t_5, t_8) t_8 = self.l_34(t_8) t_8 = self.l_35(t_8) t_1 = (t_8 + t_1) t_1 = self.l_36(t_1) t_8 = self.l_37(t_1) t_8 = torch.nn.functional.gelu(t_8) t_8 = self.l_38(t_8) t_8 = self.l_39(t_8) t_1 = (t_8 + t_1) t_1 = self.l_40(t_1) t_8 = self.l_41(t_1) t_5 = self.l_42(t_1) t_0 = self.l_43(t_1) t_4 = t_8.size() t_10 = t_5.size() t_6 = t_0.size() t_4 = t_4[slice(None, (- 1), None)] t_4 = (t_4 + (16, 64)) t_9 = t_4[0] t_3 = t_4[1] t_7 = t_4[2] t_4 = t_4[3] t_4 = t_8.view(t_9, t_3, t_7, t_4) t_4 = t_4.permute(0, 2, 1, 3) t_10 = t_10[slice(None, (- 1), None)] t_10 = (t_10 + (16, 64)) t_7 = t_10[0] t_3 = t_10[1] t_9 = t_10[2] t_10 = t_10[3] t_10 = t_5.view(t_7, t_3, t_9, t_10) t_10 = t_10.permute(0, 2, 1, 3) t_6 = t_6[slice(None, (- 1), None)] t_6 = (t_6 + (16, 64)) t_9 = t_6[0] t_3 = t_6[1] t_7 = t_6[2] t_6 = t_6[3] t_6 = t_0.view(t_9, t_3, t_7, t_6) t_6 = t_6.permute(0, 2, 1, 3) t_10 = t_10.transpose((- 1), (- 2)) t_10 = torch.matmul(t_4, t_10) t_4 = math.sqrt(64) t_4 = (t_10 / t_4) t_4 = (t_4 + t_2) t_4 = self.l_44(t_4) t_4 = self.l_45(t_4) t_6 = torch.matmul(t_4, t_6) t_6 = t_6.permute(0, 2, 1, 3) t_6 = t_6.contiguous() t_4 = t_6.size() t_4 = t_4[slice(None, (- 2), None)] t_4 = (t_4 + (1024,)) t_10 = t_4[0] t_7 = t_4[1] t_4 = t_4[2] t_4 = t_6.view(t_10, t_7, t_4) t_4 = self.l_46(t_4) t_4 = self.l_47(t_4) t_1 = (t_4 + t_1) t_1 = self.l_48(t_1) t_4 = self.l_49(t_1) t_4 = torch.nn.functional.gelu(t_4) t_4 = self.l_50(t_4) t_4 = self.l_51(t_4) t_1 = (t_4 + t_1) t_1 = self.l_52(t_1) t_4 = self.l_53(t_1) t_7 = self.l_54(t_1) t_10 = self.l_55(t_1) t_6 = t_4.size() t_3 = t_7.size() t_9 = t_10.size() t_6 = t_6[slice(None, (- 1), None)] t_6 = (t_6 + (16, 64)) t_0 = t_6[0] t_5 = t_6[1] t_8 = t_6[2] t_6 = t_6[3] t_6 = t_4.view(t_0, t_5, t_8, t_6) t_6 = t_6.permute(0, 2, 1, 3) t_3 = t_3[slice(None, (- 1), None)] t_3 = (t_3 + (16, 64)) t_8 = t_3[0] t_5 = t_3[1] t_0 = t_3[2] t_3 = t_3[3] t_3 = t_7.view(t_8, t_5, t_0, t_3) t_3 = t_3.permute(0, 2, 1, 3) t_9 = t_9[slice(None, (- 1), None)] t_9 = (t_9 + (16, 64)) t_0 = t_9[0] t_5 = t_9[1] t_8 = t_9[2] t_9 = t_9[3] t_9 = t_10.view(t_0, t_5, t_8, t_9) t_9 = t_9.permute(0, 2, 1, 3) t_3 = t_3.transpose((- 1), (- 2)) t_3 = torch.matmul(t_6, t_3) t_6 = math.sqrt(64) t_6 = (t_3 / t_6) t_6 = (t_6 + t_2) t_6 = self.l_56(t_6) t_6 = self.l_57(t_6) t_9 = torch.matmul(t_6, t_9) t_9 = t_9.permute(0, 2, 1, 3) t_9 = t_9.contiguous() t_6 = t_9.size() t_6 = t_6[slice(None, (- 2), None)] t_6 = (t_6 + (1024,)) t_3 = t_6[0] t_8 = t_6[1] t_6 = t_6[2] t_6 = t_9.view(t_3, t_8, t_6) t_6 = self.l_58(t_6) t_6 = self.l_59(t_6) t_1 = (t_6 + t_1) t_1 = self.l_60(t_1) t_6 = self.l_61(t_1) t_6 = torch.nn.functional.gelu(t_6) t_6 = self.l_62(t_6) t_6 = self.l_63(t_6) t_1 = (t_6 + t_1) t_1 = self.l_64(t_1) t_6 = self.l_65(t_1) t_8 = self.l_66(t_1) t_3 = self.l_67(t_1) t_9 = t_6.size() t_5 = t_8.size() t_0 = t_3.size() t_9 = t_9[slice(None, (- 1), None)] t_9 = (t_9 + (16, 64)) t_10 = t_9[0] t_7 = t_9[1] t_4 = t_9[2] t_9 = t_9[3] t_9 = t_6.view(t_10, t_7, t_4, t_9) t_9 = t_9.permute(0, 2, 1, 3) t_5 = t_5[slice(None, (- 1), None)] t_5 = (t_5 + (16, 64)) t_4 = t_5[0] t_7 = t_5[1] t_10 = t_5[2] t_5 = t_5[3] t_5 = t_8.view(t_4, t_7, t_10, t_5) t_5 = t_5.permute(0, 2, 1, 3) t_0 = t_0[slice(None, (- 1), None)] t_0 = (t_0 + (16, 64)) t_10 = t_0[0] t_7 = t_0[1] t_4 = t_0[2] t_0 = t_0[3] t_0 = t_3.view(t_10, t_7, t_4, t_0) t_0 = t_0.permute(0, 2, 1, 3) t_5 = t_5.transpose((- 1), (- 2)) t_5 = torch.matmul(t_9, t_5) t_9 = math.sqrt(64) t_9 = (t_5 / t_9) t_9 = (t_9 + t_2) t_9 = self.l_68(t_9) t_9 = self.l_69(t_9) t_0 = torch.matmul(t_9, t_0) t_0 = t_0.permute(0, 2, 1, 3) t_0 = t_0.contiguous() t_9 = t_0.size() t_9 = t_9[slice(None, (- 2), None)] t_9 = (t_9 + (1024,)) t_5 = t_9[0] t_4 = t_9[1] t_9 = t_9[2] t_9 = t_0.view(t_5, t_4, t_9) t_9 = self.l_70(t_9) t_9 = self.l_71(t_9) t_1 = (t_9 + t_1) t_1 = self.l_72(t_1) t_9 = self.l_73(t_1) t_9 = torch.nn.functional.gelu(t_9) t_9 = self.l_74(t_9) t_9 = self.l_75(t_9) t_1 = (t_9 + t_1) t_1 = self.l_76(t_1) t_9 = self.l_77(t_1) t_4 = self.l_78(t_1) t_5 = self.l_79(t_1) t_0 = t_9.size() t_7 = t_4.size() t_10 = t_5.size() t_0 = t_0[slice(None, (- 1), None)] t_0 = (t_0 + (16, 64)) t_3 = t_0[0] t_8 = t_0[1] t_6 = t_0[2] t_0 = t_0[3] t_0 = t_9.view(t_3, t_8, t_6, t_0) t_0 = t_0.permute(0, 2, 1, 3) t_7 = t_7[slice(None, (- 1), None)] t_7 = (t_7 + (16, 64)) t_6 = t_7[0] t_8 = t_7[1] t_3 = t_7[2] t_7 = t_7[3] t_7 = t_4.view(t_6, t_8, t_3, t_7) t_7 = t_7.permute(0, 2, 1, 3) t_10 = t_10[slice(None, (- 1), None)] t_10 = (t_10 + (16, 64)) t_3 = t_10[0] t_8 = t_10[1] t_6 = t_10[2] t_10 = t_10[3] t_10 = t_5.view(t_3, t_8, t_6, t_10) t_10 = t_10.permute(0, 2, 1, 3) t_7 = t_7.transpose((- 1), (- 2)) t_7 = torch.matmul(t_0, t_7) t_0 = math.sqrt(64) t_0 = (t_7 / t_0) t_0 = (t_0 + t_2) t_0 = self.l_80(t_0) t_0 = self.l_81(t_0) t_10 = torch.matmul(t_0, t_10) t_10 = t_10.permute(0, 2, 1, 3) t_10 = t_10.contiguous() t_0 = t_10.size() t_0 = t_0[slice(None, (- 2), None)] t_0 = (t_0 + (1024,)) t_7 = t_0[0] t_6 = t_0[1] t_0 = t_0[2] t_0 = t_10.view(t_7, t_6, t_0) t_0 = self.l_82(t_0) t_0 = self.l_83(t_0) t_1 = (t_0 + t_1) t_1 = self.l_84(t_1) t_0 = self.l_85(t_1) t_0 = torch.nn.functional.gelu(t_0) t_0 = self.l_86(t_0) t_0 = self.l_87(t_0) t_1 = (t_0 + t_1) t_1 = self.l_88(t_1) t_0 = self.l_89(t_1) t_6 = self.l_90(t_1) t_7 = self.l_91(t_1) t_10 = t_0.size() t_8 = t_6.size() t_3 = t_7.size() t_10 = t_10[slice(None, (- 1), None)] t_10 = (t_10 + (16, 64)) t_5 = t_10[0] t_4 = t_10[1] t_9 = t_10[2] t_10 = t_10[3] t_10 = t_0.view(t_5, t_4, t_9, t_10) t_10 = t_10.permute(0, 2, 1, 3) t_8 = t_8[slice(None, (- 1), None)] t_8 = (t_8 + (16, 64)) t_9 = t_8[0] t_4 = t_8[1] t_5 = t_8[2] t_8 = t_8[3] t_8 = t_6.view(t_9, t_4, t_5, t_8) t_8 = t_8.permute(0, 2, 1, 3) t_3 = t_3[slice(None, (- 1), None)] t_3 = (t_3 + (16, 64)) t_5 = t_3[0] t_4 = t_3[1] t_9 = t_3[2] t_3 = t_3[3] t_3 = t_7.view(t_5, t_4, t_9, t_3) t_3 = t_3.permute(0, 2, 1, 3) t_8 = t_8.transpose((- 1), (- 2)) t_8 = torch.matmul(t_10, t_8) t_10 = math.sqrt(64) t_10 = (t_8 / t_10) t_10 = (t_10 + t_2) t_10 = self.l_92(t_10) t_10 = self.l_93(t_10) t_3 = torch.matmul(t_10, t_3) t_3 = t_3.permute(0, 2, 1, 3) t_3 = t_3.contiguous() t_10 = t_3.size() t_10 = t_10[slice(None, (- 2), None)] t_10 = (t_10 + (1024,)) t_8 = t_10[0] t_9 = t_10[1] t_10 = t_10[2] t_10 = t_3.view(t_8, t_9, t_10) t_10 = self.l_94(t_10) t_10 = self.l_95(t_10) t_1 = (t_10 + t_1) t_1 = self.l_96(t_1) t_10 = self.l_97(t_1) t_10 = torch.nn.functional.gelu(t_10) t_10 = self.l_98(t_10) t_10 = self.l_99(t_10) t_1 = (t_10 + t_1) t_1 = self.l_100(t_1) t_10 = self.l_101(t_1) t_9 = self.l_102(t_1) t_8 = self.l_103(t_1) t_3 = t_10.size() t_4 = t_9.size() t_5 = t_8.size() t_3 = t_3[slice(None, (- 1), None)] t_3 = (t_3 + (16, 64)) t_7 = t_3[0] t_6 = t_3[1] t_0 = t_3[2] t_3 = t_3[3] t_3 = t_10.view(t_7, t_6, t_0, t_3) t_3 = t_3.permute(0, 2, 1, 3) t_4 = t_4[slice(None, (- 1), None)] t_4 = (t_4 + (16, 64)) t_0 = t_4[0] t_6 = t_4[1] t_7 = t_4[2] t_4 = t_4[3] t_4 = t_9.view(t_0, t_6, t_7, t_4) t_4 = t_4.permute(0, 2, 1, 3) t_5 = t_5[slice(None, (- 1), None)] t_5 = (t_5 + (16, 64)) t_7 = t_5[0] t_6 = t_5[1] t_0 = t_5[2] t_5 = t_5[3] t_5 = t_8.view(t_7, t_6, t_0, t_5) t_5 = t_5.permute(0, 2, 1, 3) t_4 = t_4.transpose((- 1), (- 2)) t_4 = torch.matmul(t_3, t_4) t_3 = math.sqrt(64) t_3 = (t_4 / t_3) t_3 = (t_3 + t_2) t_3 = self.l_104(t_3) t_3 = self.l_105(t_3) t_5 = torch.matmul(t_3, t_5) t_5 = t_5.permute(0, 2, 1, 3) t_5 = t_5.contiguous() t_3 = t_5.size() t_3 = t_3[slice(None, (- 2), None)] t_3 = (t_3 + (1024,)) t_4 = t_3[0] t_0 = t_3[1] t_3 = t_3[2] t_3 = t_5.view(t_4, t_0, t_3) t_3 = self.l_106(t_3) t_3 = self.l_107(t_3) t_1 = (t_3 + t_1) t_1 = self.l_108(t_1) t_3 = self.l_109(t_1) t_3 = torch.nn.functional.gelu(t_3) t_3 = self.l_110(t_3) t_3 = self.l_111(t_3) t_1 = (t_3 + t_1) t_1 = self.l_112(t_1) t_3 = self.l_113(t_1) t_0 = self.l_114(t_1) t_4 = self.l_115(t_1) t_5 = t_3.size() t_6 = t_0.size() t_7 = t_4.size() t_5 = t_5[slice(None, (- 1), None)] t_5 = (t_5 + (16, 64)) t_8 = t_5[0] t_9 = t_5[1] t_10 = t_5[2] t_5 = t_5[3] t_5 = t_3.view(t_8, t_9, t_10, t_5) t_5 = t_5.permute(0, 2, 1, 3) t_6 = t_6[slice(None, (- 1), None)] t_6 = (t_6 + (16, 64)) t_10 = t_6[0] t_9 = t_6[1] t_8 = t_6[2] t_6 = t_6[3] t_6 = t_0.view(t_10, t_9, t_8, t_6) t_6 = t_6.permute(0, 2, 1, 3) t_7 = t_7[slice(None, (- 1), None)] t_7 = (t_7 + (16, 64)) t_8 = t_7[0] t_9 = t_7[1] t_10 = t_7[2] t_7 = t_7[3] t_7 = t_4.view(t_8, t_9, t_10, t_7) t_7 = t_7.permute(0, 2, 1, 3) t_6 = t_6.transpose((- 1), (- 2)) t_6 = torch.matmul(t_5, t_6) t_5 = math.sqrt(64) t_5 = (t_6 / t_5) t_5 = (t_5 + t_2) t_5 = self.l_116(t_5) t_5 = self.l_117(t_5) t_7 = torch.matmul(t_5, t_7) t_7 = t_7.permute(0, 2, 1, 3) t_7 = t_7.contiguous() t_5 = t_7.size() t_5 = t_5[slice(None, (- 2), None)] t_5 = (t_5 + (1024,)) t_6 = t_5[0] t_10 = t_5[1] t_5 = t_5[2] t_5 = t_7.view(t_6, t_10, t_5) t_5 = self.l_118(t_5) t_5 = self.l_119(t_5) t_1 = (t_5 + t_1) t_1 = self.l_120(t_1) t_5 = self.l_121(t_1) t_5 = torch.nn.functional.gelu(t_5) t_5 = self.l_122(t_5) t_5 = self.l_123(t_5) t_1 = (t_5 + t_1) t_1 = self.l_124(t_1) t_5 = self.l_125(t_1) t_10 = self.l_126(t_1) t_6 = self.l_127(t_1) t_7 = t_5.size() t_9 = t_10.size() t_8 = t_6.size() t_7 = t_7[slice(None, (- 1), None)] t_7 = (t_7 + (16, 64)) t_4 = t_7[0] t_0 = t_7[1] t_3 = t_7[2] t_7 = t_7[3] t_7 = t_5.view(t_4, t_0, t_3, t_7) t_7 = t_7.permute(0, 2, 1, 3) t_9 = t_9[slice(None, (- 1), None)] t_9 = (t_9 + (16, 64)) t_3 = t_9[0] t_0 = t_9[1] t_4 = t_9[2] t_9 = t_9[3] t_9 = t_10.view(t_3, t_0, t_4, t_9) t_9 = t_9.permute(0, 2, 1, 3) t_8 = t_8[slice(None, (- 1), None)] t_8 = (t_8 + (16, 64)) t_4 = t_8[0] t_0 = t_8[1] t_3 = t_8[2] t_8 = t_8[3] t_8 = t_6.view(t_4, t_0, t_3, t_8) t_8 = t_8.permute(0, 2, 1, 3) t_9 = t_9.transpose((- 1), (- 2)) t_9 = torch.matmul(t_7, t_9) t_7 = math.sqrt(64) t_7 = (t_9 / t_7) t_7 = (t_7 + t_2) t_7 = self.l_128(t_7) t_7 = self.l_129(t_7) t_8 = torch.matmul(t_7, t_8) t_8 = t_8.permute(0, 2, 1, 3) t_8 = t_8.contiguous() t_7 = t_8.size() t_7 = t_7[slice(None, (- 2), None)] t_7 = (t_7 + (1024,)) t_9 = t_7[0] t_3 = t_7[1] t_7 = t_7[2] t_7 = t_8.view(t_9, t_3, t_7) t_7 = self.l_130(t_7) t_7 = self.l_131(t_7) t_1 = (t_7 + t_1) t_1 = self.l_132(t_1) t_7 = self.l_133(t_1) t_7 = torch.nn.functional.gelu(t_7) t_7 = self.l_134(t_7) t_7 = self.l_135(t_7) t_1 = (t_7 + t_1) t_1 = self.l_136(t_1) t_7 = self.l_137(t_1) t_3 = self.l_138(t_1) t_9 = self.l_139(t_1) t_8 = t_7.size() t_0 = t_3.size() t_4 = t_9.size() t_8 = t_8[slice(None, (- 1), None)] t_8 = (t_8 + (16, 64)) t_6 = t_8[0] t_10 = t_8[1] t_5 = t_8[2] t_8 = t_8[3] t_8 = t_7.view(t_6, t_10, t_5, t_8) t_8 = t_8.permute(0, 2, 1, 3) t_0 = t_0[slice(None, (- 1), None)] t_0 = (t_0 + (16, 64)) t_5 = t_0[0] t_10 = t_0[1] t_6 = t_0[2] t_0 = t_0[3] t_0 = t_3.view(t_5, t_10, t_6, t_0) t_0 = t_0.permute(0, 2, 1, 3) t_4 = t_4[slice(None, (- 1), None)] t_4 = (t_4 + (16, 64)) t_6 = t_4[0] t_10 = t_4[1] t_5 = t_4[2] t_4 = t_4[3] t_4 = t_9.view(t_6, t_10, t_5, t_4) t_4 = t_4.permute(0, 2, 1, 3) t_0 = t_0.transpose((- 1), (- 2)) t_0 = torch.matmul(t_8, t_0) t_8 = math.sqrt(64) t_8 = (t_0 / t_8) t_8 = (t_8 + t_2) t_8 = self.l_140(t_8) t_8 = self.l_141(t_8) t_4 = torch.matmul(t_8, t_4) t_4 = t_4.permute(0, 2, 1, 3) t_4 = t_4.contiguous() t_8 = t_4.size() t_8 = t_8[slice(None, (- 2), None)] t_8 = (t_8 + (1024,)) t_0 = t_8[0] t_5 = t_8[1] t_8 = t_8[2] t_8 = t_4.view(t_0, t_5, t_8) t_8 = self.l_142(t_8) t_8 = self.l_143(t_8) t_1 = (t_8 + t_1) t_1 = self.l_144(t_1) t_8 = self.l_145(t_1) t_8 = torch.nn.functional.gelu(t_8) return list(flatten((t_2, t_1, t_8))) def state_dict(self, *args, **kwargs): return state_dict(self, *args, **kwargs) def load_state_dict(self, *args, **kwargs): return load_state_dict(self, *args, **kwargs) def named_parameters(self, *args, **kwargs): return named_parameters(self, *args, **kwargs) def named_buffers(self, *args, **kwargs): return named_buffers(self, *args, **kwargs) def cpu(self): return cpu(self) def cuda(self, device=None): return cuda(self, device=device) def to(self, *args, **kwargs): return to(self, *args, **kwargs)
class UCBVI(abc.ABC): def __init__(self, mdp, n_episodes=1, init_state=None, reg_factor=1.0, confidence_scaling_factor=(- 1.0), delta=0.05, train_every=1, throttle=int(100.0)): self.mdp = mdp self.n_episodes = n_episodes self.init_state = init_state self.reg_factor = reg_factor if (confidence_scaling_factor == (- 1.0)): confidence_scaling_factor = mdp.noise_std self.confidence_scaling_factor = confidence_scaling_factor self.delta = delta self.train_every = train_every self.throttle = throttle self.reset() def reset_upper_confidence_bounds(self): self.Q_hat = np.concatenate((np.empty((self.mdp.H, self.mdp.n_states, self.mdp.n_actions)), np.zeros((1, self.mdp.n_states, self.mdp.n_actions)))) self.exploration_bonus = np.empty((self.mdp.H, self.mdp.n_states, self.mdp.n_actions)) self.upper_confidence_bounds = np.ones((self.mdp.H, self.mdp.n_states, self.mdp.n_actions)) def reset_regrets(self): self.regrets = np.empty(self.n_episodes) (self.V_star, self.pi_star) = self.mdp.optimal_policy() def reset_policy(self): self.policy = np.empty((self.mdp.H, self.mdp.n_states)).astype('int') def reset_state_action_reward_buffer(self): self.buffer_states = np.empty((self.mdp.H + 1)).astype('int') self.buffer_actions = np.empty(self.mdp.H).astype('int') self.buffer_rewards = np.empty(self.mdp.H) if (self.init_state is not None): self.state = self.init_state else: self.state = np.random.choice(self.mdp.states) def reset_A_inv(self): self.A_inv = np.array([(np.eye(self.approximator_dim) / self.reg_factor) for _ in range(self.mdp.H)]).reshape(self.mdp.H, self.approximator_dim, self.approximator_dim) def reset_grad_approx(self): self.grad_approx = np.zeros((self.mdp.n_states, self.mdp.n_actions, self.approximator_dim)) def take_action(self): self.policy[self.mdp.iteration] = np.argmax(self.upper_confidence_bounds[self.mdp.iteration], axis=1).astype('int') self.action = self.policy[(self.mdp.iteration, self.state)] def reset(self): pass def approximator_dim(self): pass def confidence_multiplier(self): pass def update_output_gradient(self): pass def train(self): pass def predict(self): pass def update_confidence_bounds(self): self.update_output_gradient() self.exploration_bonus[self.mdp.iteration] = (np.array([np.sqrt(((self.grad_approx[(s, a)].T self.A_inv[self.mdp.iteration]) self.grad_approx[(s, a)])) for (s, a) in itertools.product(self.mdp.states, self.mdp.actions)]).reshape(self.mdp.n_states, self.mdp.n_actions) * self.confidence_multiplier) self.predict() self.upper_confidence_bounds[self.mdp.iteration] = np.clip((self.Q_hat[self.mdp.iteration] + self.exploration_bonus[self.mdp.iteration]), None, self.mdp.H) def update_A_inv(self): self.A_inv[self.mdp.iteration] = inv_sherman_morrison(self.grad_approx[(self.state, self.action)], self.A_inv[self.mdp.iteration]) def run(self): postfix = {'total regret': 0.0} with tqdm(total=self.n_episodes, postfix=postfix) as pbar: for k in range(self.n_episodes): self.mdp.reset_iteration('backward') for h in reversed(range(self.mdp.H)): if (k > 0): self.action = self.buffer_actions[h] self.reward = self.buffer_rewards[h] self.state = self.buffer_states[h] self.update_A_inv() if ((k % self.train_every) == 0): self.train() self.update_confidence_bounds() self.mdp.iteration -= 1 self.mdp.reset_iteration('forward') self.state = self.init_state self.buffer_states[0] = self.state for h in range(self.mdp.H): self.take_action() self.reward = self.mdp.rewards[(h, self.state, self.action)] self.state = self.mdp.new_state(self.state, self.action) self.buffer_actions[h] = self.action self.buffer_rewards[h] = self.reward self.buffer_states[(h + 1)] = self.state self.mdp.iteration += 1 V = self.mdp.evaluate_policy(self.policy) self.regrets[k] = (self.V_star[(0, self.init_state)] - V[(0, self.init_state)]) postfix['total regret'] += self.regrets[k] if ((k % self.throttle) == 0): pbar.set_postfix(postfix) pbar.update(self.throttle)
def get_kitchen_benchmark_goals(): object_goal_vals = {'bottom_burner': [(- 0.88), (- 0.01)], 'light_switch': [(- 0.69), (- 0.05)], 'slide_cabinet': [0.37], 'hinge_cabinet': [0.0, 0.5], 'microwave': [(- 0.5)], 'kettle': [(- 0.23), 0.75, 1.62]} object_goal_idxs = {'bottom_burner': [9, 10], 'light_switch': [17, 18], 'slide_cabinet': [19], 'hinge_cabinet': [20, 21], 'microwave': [22], 'kettle': [23, 24, 25]} base_task_names = ['bottom_burner', 'light_switch', 'slide_cabinet', 'hinge_cabinet', 'microwave', 'kettle'] goal_configs = [] for i in range(6): goal_configs.append([base_task_names[i]]) for (i, j) in combinations([1, 2, 3, 5], 2): goal_configs.append([base_task_names[i], base_task_names[j]]) obs_element_goals = [] obs_element_indices = [] for objects in goal_configs: _goal = np.concatenate([object_goal_vals[obj] for obj in objects]) _goal_idxs = np.concatenate([object_goal_idxs[obj] for obj in objects]) obs_element_goals.append(_goal) obs_element_indices.append(_goal_idxs) return (obs_element_goals, obs_element_indices, goal_configs)
def PreActResNet18(num_channels=3): return PreActResNet(PreActBlock, [2, 2, 2, 2], num_channels=num_channels)
def line_search(f, x0, dx, g0, alpha, condition, max_steps=10, c1=0.1): assert (0 < alpha < 1) f0 = f(x0) for _ in range(max_steps): x = (x0 + dx) if ((f(x) > (f0 + ((c1 * g0.T) dx))) and condition(x)): return x dx *= alpha print('Line search failed, returning x0') return x0
class R1_mAP(Metric): def __init__(self, num_query, max_rank=50, feat_norm='yes'): super(R1_mAP, self).__init__() self.num_query = num_query self.max_rank = max_rank self.feat_norm = feat_norm def reset(self): self.feats = [] self.pids = [] self.camids = [] def update(self, output): (feat, pid, camid) = output self.feats.append(feat) self.pids.extend(np.asarray(pid)) self.camids.extend(np.asarray(camid)) def compute(self): feats = torch.cat(self.feats, dim=0) if (self.feat_norm == 'yes'): print('The test feature is normalized') feats = torch.nn.functional.normalize(feats, dim=1, p=2) qf = feats[:self.num_query] q_pids = np.asarray(self.pids[:self.num_query]) q_camids = np.asarray(self.camids[:self.num_query]) gf = feats[self.num_query:] g_pids = np.asarray(self.pids[self.num_query:]) g_camids = np.asarray(self.camids[self.num_query:]) (m, n) = (qf.shape[0], gf.shape[0]) distmat = (torch.pow(qf, 2).sum(dim=1, keepdim=True).expand(m, n) + torch.pow(gf, 2).sum(dim=1, keepdim=True).expand(n, m).t()) distmat.addmm_(qf, gf.t(), beta=1, alpha=(- 2)) distmat = distmat.cpu().numpy() (cmc, mAP) = eval_func(distmat, q_pids, g_pids, q_camids, g_camids) return (cmc, mAP)
class NovelViewSynthesizeModel(object): def __init__(self, output_dir): self.output_dir = mkdir(output_dir) self.si_out_dir = self.output_dir self.num_preds_si = 0 def imitate(self, src_infos: Dict[(str, Any)], ref_infos: Dict[(str, Any)]) -> List[str]: raise NotImplementedError def build_model(self): raise NotImplementedError def terminate(self): raise NotImplementedError def personalization(self, src_infos): processed_src_infos = src_infos return processed_src_infos
class BinaryMorphology2D(): param_names = ['shape', 'footprint', 'radius', 'decomposition'] params = [((512, 512),), ('square', 'diamond', 'octagon', 'disk', 'ellipse', 'star'), (1, 3, 5, 15, 25, 40), (None, 'sequence', 'separable', 'crosses')] def setup(self, shape, footprint, radius, decomposition): rng = np.random.default_rng(123) self.image = (rng.standard_normal(shape) < 3.5) fp_func = getattr(morphology, footprint) allow_sequence = ('rectangle', 'square', 'diamond', 'octagon', 'disk') allow_separable = ('rectangle', 'square') allow_crosses = ('disk', 'ellipse') allow_decomp = tuple(((set(allow_sequence) | set(allow_separable)) | set(allow_crosses))) footprint_kwargs = {} if ((decomposition == 'sequence') and (footprint not in allow_sequence)): raise NotImplementedError('decomposition unimplemented') elif ((decomposition == 'separable') and (footprint not in allow_separable)): raise NotImplementedError('separable decomposition unavailable') elif ((decomposition == 'crosses') and (footprint not in allow_crosses)): raise NotImplementedError('separable decomposition unavailable') if (footprint in allow_decomp): footprint_kwargs['decomposition'] = decomposition if (footprint in ['rectangle', 'square']): size = ((2 * radius) + 1) self.footprint = fp_func(size, **footprint_kwargs) elif (footprint in ['diamond', 'disk']): self.footprint = fp_func(radius, **footprint_kwargs) elif (footprint == 'star'): a = max(((2 * radius) // 3), 1) self.footprint = fp_func(a, **footprint_kwargs) elif (footprint == 'octagon'): m = n = max(((2 * radius) // 3), 1) self.footprint = fp_func(m, n, **footprint_kwargs) elif (footprint == 'ellipse'): if (radius > 1): self.footprint = fp_func((radius - 1), (radius + 1), **footprint_kwargs) else: self.footprint = fp_func(radius, radius, **footprint_kwargs) def time_erosion(self, shape, footprint, radius, *args): morphology.binary_erosion(self.image, self.footprint)
def get_args(): parser = argparse.ArgumentParser() train_inten.add_inten_train_args(parser) nn_utils.add_hyperopt_args(parser) return parser.parse_args()
class Cipher(Element): def __init__(self, parent, key): Element.__init__(self, parent) self._key = key def __eq__(self, right): return ((type(self) is type(right)) and (self.parent() == right.parent()) and (self._key == right._key)) def _repr_(self): return ('Cipher on %s' % self.parent().cipher_domain()) def key(self): return self._key def domain(self): return self.parent().cipher_domain() def codomain(self): return self.parent().cipher_codomain()
def test_data_dependency_5(): module_block = BasicBlock([Instr('LOAD_BUILD_CLASS'), Instr('LOAD_CONST', arg=dummy_code_object), Instr('LOAD_CONST', arg='Foo'), Instr('MAKE_FUNCTION', arg=0), Instr('LOAD_CONST', arg='Foo'), Instr('CALL_FUNCTION', arg=2), Instr('STORE_NAME', arg='Foo'), Instr('LOAD_GLOBAL', arg='Foo'), Instr('CALL_FUNCTION', arg=0), Instr('STORE_FAST', arg='ob'), Instr('LOAD_CONST', arg=1), Instr('LOAD_FAST', arg='ob'), Instr('STORE_ATTR', arg='attr1'), Instr('LOAD_FAST', arg='ob'), Instr('STORE_FAST', arg='result'), Instr('LOAD_FAST', arg='result'), Instr('RETURN_VALUE')]) class_attr_block = BasicBlock([Instr('LOAD_CONST', arg=None), Instr('RETURN_VALUE')]) expected_instructions = [] expected_instructions.extend(module_block) expected_instructions.extend(class_attr_block) module = 'tests.fixtures.slicer.partial_cover_dependency' sliced_instructions = slice_module_at_return(module) assert (len(sliced_instructions) == len(expected_instructions)) assert compare(sliced_instructions, expected_instructions)
class InputFeatures_eval(object): def __init__(self, input_ids, input_mask, segment_ids, label_id, label_disf_id, label_sing_id): self.input_ids = input_ids self.input_mask = input_mask self.segment_ids = segment_ids self.label_id = label_id self.label_disf_id = label_disf_id self.label_sing_id = label_sing_id
def decode_pose(root_score, root_id, root_image_coord, scores, offsets, output_stride, displacements_fwd, displacements_bwd): num_parts = scores.shape[2] num_edges = len(PARENT_CHILD_TUPLES) instance_keypoint_scores = np.zeros(num_parts) instance_keypoint_coords = np.zeros((num_parts, 2)) instance_keypoint_scores[root_id] = root_score instance_keypoint_coords[root_id] = root_image_coord for edge in reversed(range(num_edges)): (target_keypoint_id, source_keypoint_id) = PARENT_CHILD_TUPLES[edge] if ((instance_keypoint_scores[source_keypoint_id] > 0.0) and (instance_keypoint_scores[target_keypoint_id] == 0.0)): (score, coords) = traverse_to_targ_keypoint(edge, instance_keypoint_coords[source_keypoint_id], target_keypoint_id, scores, offsets, output_stride, displacements_bwd) instance_keypoint_scores[target_keypoint_id] = score instance_keypoint_coords[target_keypoint_id] = coords for edge in range(num_edges): (source_keypoint_id, target_keypoint_id) = PARENT_CHILD_TUPLES[edge] if ((instance_keypoint_scores[source_keypoint_id] > 0.0) and (instance_keypoint_scores[target_keypoint_id] == 0.0)): (score, coords) = traverse_to_targ_keypoint(edge, instance_keypoint_coords[source_keypoint_id], target_keypoint_id, scores, offsets, output_stride, displacements_fwd) instance_keypoint_scores[target_keypoint_id] = score instance_keypoint_coords[target_keypoint_id] = coords return (instance_keypoint_scores, instance_keypoint_coords)
def xla_available() -> bool: try: return (find_spec('torch_xla') is not None) except ModuleNotFoundError: return False
class Extractor(ModelBase): def __init__(self, config: ExtractorConfig): super().__init__(config) def _get_full_embedded(self, batch: Batch): embedded = [] if hasattr(self, 'ohots'): ohots_embedded = [self.ohots[f](batch.ohots[f]) for f in self.ohots] embedded.extend(ohots_embedded) if hasattr(self, 'mhots'): mhots_embedded = [self.mhots[f](batch.mhots[f]) for f in self.mhots] embedded.extend(mhots_embedded) if hasattr(self, 'nested_ohots'): nested_ohots_embedded = [self.nested_ohots[f](**batch.nested_ohots[f], seq_lens=batch.seq_lens) for f in self.nested_ohots] embedded.extend(nested_ohots_embedded) return torch.cat(embedded, dim=(- 1)) def _get_full_hidden(self, batch: Batch): full_hidden = [] if any([hasattr(self, name) for name in ExtractorConfig._embedder_names]): embedded = self._get_full_embedded(batch) if hasattr(self, 'intermediate1'): full_hidden.append(self.intermediate1(embedded, batch.mask)) else: full_hidden.append(embedded) for name in ExtractorConfig._pretrained_names: if hasattr(self, name): full_hidden.append(getattr(self, name)(**getattr(batch, name))) full_hidden = torch.cat(full_hidden, dim=(- 1)) if hasattr(self, 'intermediate2'): return self.intermediate2(full_hidden, batch.mask) else: return full_hidden def pretrained_parameters(self): params = [] if hasattr(self, 'elmo'): params.extend(self.elmo.elmo._elmo_lstm.parameters()) if hasattr(self, 'bert_like'): params.extend(self.bert_like.bert_like.parameters()) if hasattr(self, 'flair_fw'): params.extend(self.flair_fw.flair_lm.parameters()) if hasattr(self, 'flair_bw'): params.extend(self.flair_bw.flair_lm.parameters()) return params def forward2states(self, batch: Batch): return {'full_hidden': self._get_full_hidden(batch)}
class MBv3LatencyTable(LatencyTable): def query(self, l_type: str, input_shape, output_shape, mid=None, ks=None, stride=None, id_skip=None, se=None, h_swish=None): infos = [l_type, ('input:%s' % self.repr_shape(input_shape)), ('output:%s' % self.repr_shape(output_shape))] if (l_type in ('expanded_conv',)): assert (None not in (mid, ks, stride, id_skip, se, h_swish)) infos += [('expand:%d' % mid), ('kernel:%d' % ks), ('stride:%d' % stride), ('idskip:%d' % id_skip), ('se:%d' % se), ('hs:%d' % h_swish)] key = '-'.join(infos) return self.lut[key]['mean'] def predict_network_latency(self, net, image_size=224): predicted_latency = 0 predicted_latency += self.query('Conv', [image_size, image_size, 3], [((image_size + 1) // 2), ((image_size + 1) // 2), net.first_conv.out_channels]) fsize = ((image_size + 1) // 2) for block in net.blocks: mb_conv = block.conv shortcut = block.shortcut if (mb_conv is None): continue if (shortcut is None): idskip = 0 else: idskip = 1 out_fz = int((((fsize - 1) / mb_conv.stride) + 1)) block_latency = self.query('expanded_conv', [fsize, fsize, mb_conv.in_channels], [out_fz, out_fz, mb_conv.out_channels], mid=mb_conv.depth_conv.conv.in_channels, ks=mb_conv.kernel_size, stride=mb_conv.stride, id_skip=idskip, se=(1 if mb_conv.use_se else 0), h_swish=(1 if (mb_conv.act_func == 'h_swish') else 0)) predicted_latency += block_latency fsize = out_fz predicted_latency += self.query('Conv_1', [fsize, fsize, net.final_expand_layer.in_channels], [fsize, fsize, net.final_expand_layer.out_channels]) predicted_latency += self.query('AvgPool2D', [fsize, fsize, net.final_expand_layer.out_channels], [1, 1, net.final_expand_layer.out_channels]) predicted_latency += self.query('Conv_2', [1, 1, net.feature_mix_layer.in_channels], [1, 1, net.feature_mix_layer.out_channels]) predicted_latency += self.query('Logits', [1, 1, net.classifier.in_features], [net.classifier.out_features]) return predicted_latency def predict_network_latency_given_config(self, net_config, image_size=224): predicted_latency = 0 predicted_latency += self.query('Conv', [image_size, image_size, 3], [((image_size + 1) // 2), ((image_size + 1) // 2), net_config['first_conv']['out_channels']]) fsize = ((image_size + 1) // 2) for block in net_config['blocks']: mb_conv = (block['mobile_inverted_conv'] if ('mobile_inverted_conv' in block) else block['conv']) shortcut = block['shortcut'] if (mb_conv is None): continue if (shortcut is None): idskip = 0 else: idskip = 1 out_fz = int((((fsize - 1) / mb_conv['stride']) + 1)) if (mb_conv['mid_channels'] is None): mb_conv['mid_channels'] = round((mb_conv['in_channels'] * mb_conv['expand_ratio'])) block_latency = self.query('expanded_conv', [fsize, fsize, mb_conv['in_channels']], [out_fz, out_fz, mb_conv['out_channels']], mid=mb_conv['mid_channels'], ks=mb_conv['kernel_size'], stride=mb_conv['stride'], id_skip=idskip, se=(1 if mb_conv['use_se'] else 0), h_swish=(1 if (mb_conv['act_func'] == 'h_swish') else 0)) predicted_latency += block_latency fsize = out_fz predicted_latency += self.query('Conv_1', [fsize, fsize, net_config['final_expand_layer']['in_channels']], [fsize, fsize, net_config['final_expand_layer']['out_channels']]) predicted_latency += self.query('AvgPool2D', [fsize, fsize, net_config['final_expand_layer']['out_channels']], [1, 1, net_config['final_expand_layer']['out_channels']]) predicted_latency += self.query('Conv_2', [1, 1, net_config['feature_mix_layer']['in_channels']], [1, 1, net_config['feature_mix_layer']['out_channels']]) predicted_latency += self.query('Logits', [1, 1, net_config['classifier']['in_features']], [net_config['classifier']['out_features']]) return predicted_latency def count_flops_given_config(net_config, image_size=224): flops = 0 flops += count_conv_flop(((image_size + 1) // 2), 3, net_config['first_conv']['out_channels'], 3, 1) fsize = ((image_size + 1) // 2) for block in net_config['blocks']: mb_conv = (block['mobile_inverted_conv'] if ('mobile_inverted_conv' in block) else block['conv']) if (mb_conv is None): continue out_fz = int((((fsize - 1) / mb_conv['stride']) + 1)) if (mb_conv['mid_channels'] is None): mb_conv['mid_channels'] = round((mb_conv['in_channels'] * mb_conv['expand_ratio'])) if (mb_conv['expand_ratio'] != 1): flops += count_conv_flop(fsize, mb_conv['in_channels'], mb_conv['mid_channels'], 1, 1) flops += count_conv_flop(out_fz, mb_conv['mid_channels'], mb_conv['mid_channels'], mb_conv['kernel_size'], mb_conv['mid_channels']) if mb_conv['use_se']: se_mid = make_divisible((mb_conv['mid_channels'] // 4), divisor=MyNetwork.CHANNEL_DIVISIBLE) flops += count_conv_flop(1, mb_conv['mid_channels'], se_mid, 1, 1) flops += count_conv_flop(1, se_mid, mb_conv['mid_channels'], 1, 1) flops += count_conv_flop(out_fz, mb_conv['mid_channels'], mb_conv['out_channels'], 1, 1) fsize = out_fz flops += count_conv_flop(fsize, net_config['final_expand_layer']['in_channels'], net_config['final_expand_layer']['out_channels'], 1, 1) flops += count_conv_flop(1, net_config['feature_mix_layer']['in_channels'], net_config['feature_mix_layer']['out_channels'], 1, 1) flops += count_conv_flop(1, net_config['classifier']['in_features'], net_config['classifier']['out_features'], 1, 1) return (flops / 1000000.0)
class AutoUpliftTX(BaseAutoUplift): __MAP_META_TO_STAGES__: Dict[(str, List[MetaLearnerStage])] = {'TLearner': [MetaLearnerStage(name='outcome_control'), MetaLearnerStage(name='outcome_treatment')], 'XLearner': [MetaLearnerStage(name='outcome_control'), MetaLearnerStage(name='outcome_treatment'), MetaLearnerStage(name='propensity', params={'task': Task('binary')}), MetaLearnerStage(name='effect_control', params={'task': Task('reg')}, prev_stage=MetaLearnerStage(name='outcome_treatment')), MetaLearnerStage(name='effect_treatment', params={'task': Task('reg')}, prev_stage=MetaLearnerStage(name='outcome_control'))]} def __init__(self, base_task: Task, baselearners: Optional[Union[(List[BaseLearnerWrapper], Dict[(MLStageFullName, List[BaseLearnerWrapper])])]]=None, metalearners: List[str]=[], metric: Union[(str, TUpliftMetric, Callable)]='adj_qini', increasing_metric: bool=True, test_size: float=0.2, timeout: Optional[int]=None, timeout_single_learner: Optional[int]=None, cpu_limit: int=4, gpu_ids: Optional[str]='all', random_state: int=42): assert all(((ml in self.__MAP_META_TO_STAGES__) for ml in metalearners)), 'Currently available for {}.'.format(self.__MAP_META_TO_STAGES__) super().__init__(base_task, metric, True, increasing_metric, test_size, timeout, None, timeout_single_learner, cpu_limit, gpu_ids, random_state) self.baselearners = baselearners self.metalearners = (metalearners if (len(metalearners) > 0) else list(self.__MAP_META_TO_STAGES__)) self._best_metalearner: MetaLearner self._best_metalearner_wrap: MetaLearnerWrapper self._trained_stage_baselearners: Dict[(MLStageFullName, List[TrainedStageBaseLearner])] = defaultdict(list) self._metalearner_metrics: Dict[(TrainedMetaLearnerFullName, float)] = {} self._n_run_l2 = 3 def fit(self, data: DataFrame, roles: dict, verbose: int=0): (train_data, test_data, test_treatment, test_target) = self._prepare_data(data, roles) self._timer.start() for stage_info in self._generate_stage_baselearner_candidates(): self._evaluate(stage_info, train_data, test_data, roles, verbose) if self._timer.time_limit_exceeded(): logger.warning("Time of training exceeds 'timeout': {} > {}.".format(self._timer.time_spent, self.timeout)) break self._calculate_metalearners_metrics(test_treatment, test_target) self._set_best_metalearner() def predict(self, data: DataFrame) -> Tuple[(np.ndarray, ...)]: assert (self._best_metalearner is not None), "First call 'self.fit(...)', to choose best metalearner." return self._best_metalearner.predict(data) def create_best_metalearner(self, need_report: bool=True, update_metalearner_params: Dict[(str, Any)]={}, update_baselearner_params: Dict[(str, Any)]={}) -> Union[(MetaLearner, ReportDecoUplift)]: assert (len(self._trained_stage_baselearners) > 0), "First call 'self.fit(...), to choose best metalearner." ml_wrap = deepcopy(self._best_metalearner_wrap) if (len(update_metalearner_params) > 0): ml_wrap.update_params(update_metalearner_params) if (len(update_baselearner_params) > 0): ml_wrap.update_baselearner_params(update_baselearner_params) best_metalearner_raw = ml_wrap() if need_report: if isinstance(self.metric, str): rdu = ReportDecoUplift() best_metalearner_raw = rdu(best_metalearner_raw) else: logger.warning("Report doesn't work with custom metric, return just best_metalearner.") return best_metalearner_raw def get_metalearners_ranting(self) -> DataFrame: (metalearner_names, params, metrics) = ([], [], []) for (ml_name, metric) in self._metalearner_metrics.items(): metalearner_names.append(ml_name[0]) params.append(ml_name[1]) metrics.append(metric) rating_table = DataFrame({'MetaLearner': metalearner_names, 'Parameters': params, 'Metrics': metrics}) rating_table['Rank'] = rating_table['Metrics'].rank(method='first', ascending=(not self.increasing_metric)) rating_table.sort_values('Rank', inplace=True) rating_table.reset_index(drop=True, inplace=True) return rating_table def _generate_stage_baselearner_candidates(self) -> Generator[(Tuple[(Tuple[(MetaLearnerStage, BaseLearnerWrapper)], ...)], None, None)]: stage_baselearners = self._set_stage_baselearners() stage_by_levels = defaultdict(list) for stage in stage_baselearners: stage_by_levels[len(stage)].append(stage) pool_iter_levels = defaultdict(list) bls_level_1 = zip_longest(*(deepcopy(bls) for (stage, bls) in stage_baselearners.items() if (stage in stage_by_levels[1]))) first_run = True n_runs = 0 for bls in bls_level_1: for (stage_name_l1, bl_l1) in zip(stage_by_levels[1], bls): stage_l1 = self._extract_stage(stage_name_l1) for stage_name_l2 in stage_by_levels[2]: if (stage_name_l2[0:1] == stage_name_l1): pool_iter_levels[2].append((stage_name_l2, bl_l1, iter(deepcopy(stage_baselearners[stage_name_l2])))) n_runs += 1 (yield ((stage_l1, bl_l1),)) if (not first_run): n_stage_iters = len(pool_iter_levels[2]) if (n_stage_iters == 0): continue for _ in range(min(n_stage_iters, self._n_run_l2)): n_stage_iters = len(pool_iter_levels[2]) idx = np.random.randint(0, n_stage_iters, 1)[0] try: (stage_name_l2, bl_l1, bls_iter) = pool_iter_levels[2][idx] bl_l2 = next(bls_iter) stage_l1 = self._extract_stage(stage_name_l2[0:1]) stage_l2 = self._extract_stage(stage_name_l2) n_runs += 1 (yield ((stage_l1, bl_l1), (stage_l2, bl_l2))) except StopIteration: pool_iter_levels[2].pop(idx) if (n_runs >= len(stage_by_levels[1])): first_run = False while (len(pool_iter_levels[2]) > 0): n_stage_iters = len(pool_iter_levels[2]) try: idx = np.random.randint(0, n_stage_iters, 1)[0] (stage_name_l2, bl_l1, bls_iter) = pool_iter_levels[2][idx] bl_l2 = next(bls_iter) stage_l1 = self._extract_stage(stage_name_l2[0:1]) stage_l2 = self._extract_stage(stage_name_l2) (yield ((stage_l1, bl_l1), (stage_l2, bl_l2))) except StopIteration: pool_iter_levels[2].pop(idx) def _extract_stages(self) -> Generator[(Tuple[(str, MetaLearnerStage)], None, None)]: for (ml_name, ml_stages) in self.__MAP_META_TO_STAGES__.items(): for stage in ml_stages: (yield (ml_name, stage)) def _extract_stage(self, full_name: MLStageFullName) -> MetaLearnerStage: for (ml_name, stage) in self._extract_stages(): if (stage.full_name() == full_name): return stage raise Exception("Can't find stage {}".format(full_name)) def _set_stage_baselearners(self) -> Dict[(MLStageFullName, List[BaseLearnerWrapper])]: stage_baselearners = {} if isinstance(self.baselearners, dict): stage_baselearners = {k: deepcopy(self._validate_baselearners(v)) for (k, v) in self.baselearners.items()} all_stages_full_names = set((stage.full_name() for (ml, ml_stages) in self.__MAP_META_TO_STAGES__.items() for stage in ml_stages)) if (len(stage_baselearners) != len(all_stages_full_names)): timeout = self._calculate_single_bl_timeout((len(stage_baselearners) > 0)) if (isinstance(self.baselearners, list) and (len(self.baselearners) > 0)): baselearners = deepcopy(self._validate_baselearners(self.baselearners)) elif ((self.baselearners is None) or isinstance(self.baselearners, dict)): baselearners = deepcopy(self.__default_learners(tab_params={'timeout': timeout})) remain_stages_full_names = (all_stages_full_names - set(stage_baselearners)) bin_baselearners: List[BaseLearnerWrapper] = [] reg_baselearners: List[BaseLearnerWrapper] = [] raw_baselearners: List[BaseLearnerWrapper] = [] for bl in baselearners: if ('task' in bl.params): if (bl.params['task'].name == 'binary'): bin_baselearners.append(bl) if (bl.params['task'].name == 'reg'): reg_baselearners.append(bl) else: raw_baselearners.append(bl) for full_name in remain_stages_full_names: stage = self._extract_stage(full_name) stage_task = (stage.params['task'] if ('task' in stage.params) else self.base_task) filled_baselearners = deepcopy(raw_baselearners) for idx in range(len(filled_baselearners)): filled_baselearners[idx].params['task'] = stage_task baselearners_on_stage = filled_baselearners if (stage_task.name == 'binary'): baselearners_on_stage.extend(bin_baselearners) elif (stage_task.name == 'reg'): baselearners_on_stage.extend(reg_baselearners) stage_baselearners[full_name] = baselearners_on_stage return stage_baselearners def _calculate_single_bl_timeout(self, specify_stages: bool) -> Optional[int]: timeout: Optional[int] = None if (not specify_stages): if (self.timeout is not None): timeout = int((self.timeout / ((2 * ('TLearner' in self.metalearners)) + (5 * ('XLearner' in self.metalearners))))) elif (self.timeout_single_learner is not None): timeout = self.timeout_single_learner else: timeout = self.timeout_single_learner return timeout def _validate_baselearners(self, baselearners: List[BaseLearnerWrapper]): k2n: Dict[(str, List[int])] = {} for (idx, bl) in enumerate(baselearners): bl_name = bl.name k2n.setdefault(bl_name, []) k2n[bl_name].append(idx) is_name_duplicates = any(((len(idxs) > 1) for (bl_name, idxs) in k2n.items())) if is_name_duplicates: logger.warning('Naming of baselearner should be unique.') logger.warning("Name of baselearner would be updated with postfix '__order_idx_bl__'.") renaming_by_idxs: Dict[(int, str)] = {} for (bl_name, idxs) in k2n.items(): if (len(idxs) > 1): for idx in idxs: renaming_by_idxs[idx] = '{}__#{}__'.format(bl_name, idx) baselearners_t = [] for (idx, bl) in enumerate(baselearners): if (idx in renaming_by_idxs): bl.name = renaming_by_idxs[idx] baselearners_t.append(bl) return baselearners_t else: return baselearners def _evaluate(self, stage_info: Tuple[(Tuple[(MetaLearnerStage, BaseLearnerWrapper)], ...)], train: DataFrame, test: DataFrame, roles: dict, verbose: int=0): (train_data, train_roles) = self._prepare_data_for_stage(stage_info, train, roles) (ml_stage, bl_wrap) = stage_info[(- 1)] (prev_stage, prev_bl_wrap) = (None, None) if (len(stage_info) == 2): (prev_stage, prev_bl_wrap) = stage_info[0] bl = bl_wrap() bl.fit_predict(train_data, train_roles, verbose) test_pred = bl.predict(test).data.ravel() tsbl = TrainedStageBaseLearner(stage_bl=bl_wrap, prev_stage_bl=prev_bl_wrap, trained_model=bl, pred=test_pred) self._trained_stage_baselearners[ml_stage.full_name()].append(tsbl) def _prepare_data_for_stage(self, stage_info: Tuple[(Tuple[(MetaLearnerStage, BaseLearnerWrapper)], ...)], train: DataFrame, roles: dict) -> Tuple[(DataFrame, Dict)]: (treatment_role, treatment_col) = uplift_utils._get_treatment_role(roles) (target_role, target_col) = uplift_utils._get_target_role(roles) stage_name = stage_info[(- 1)][0].name if (len(stage_info) == 1): if (stage_name == 'propensity'): train_roles = deepcopy(roles) train_roles.pop(treatment_role) train_roles.pop(target_role) train_roles['target'] = treatment_col train_data = train.drop(target_col, axis=1) elif (stage_name == 'outcome_control'): train_roles = deepcopy(roles) train_roles.pop(treatment_role) train_data = train[(train[treatment_col] == 0)].drop(treatment_col, axis=1) elif (stage_name == 'outcome_treatment'): train_roles = deepcopy(roles) train_roles.pop(treatment_role) train_data = train[(train[treatment_col] == 1)].drop(treatment_col, axis=1) else: raise Exception('Wrong l1 stage name') elif (len(stage_info) == 2): train_roles = deepcopy(roles) train_roles.pop(treatment_role) (prev_ml_stage, prev_bl_wrap) = stage_info[0] prev_stage_bl = [bl.trained_model for bl in self._trained_stage_baselearners[prev_ml_stage.full_name()] if (bl.stage_bl.name == prev_bl_wrap.name)][0] if (stage_name == 'effect_control'): train_data = train[(train[treatment_col] == 0)].drop(treatment_col, axis=1) opposite_gr_pred = prev_stage_bl.predict(train_data).data.ravel() train_data[target_col] = (opposite_gr_pred - train_data[target_col]) elif (stage_name == 'effect_treatment'): train_data = train[(train[treatment_col] == 1)].drop(treatment_col, axis=1) opposite_gr_pred = prev_stage_bl.predict(train_data).data.ravel() train_data[target_col] = (train_data[target_col] - opposite_gr_pred) else: raise Exception('Wrong l2 stage name') return (train_data, train_roles) def _calculate_metalearners_metrics(self, test_target: np.ndarray, test_treatment: np.ndarray): for set_ml_stage_bls in self._bl_for_ml(): for (ml_name, stage_bls) in set_ml_stage_bls.items(): sbls = tuple(sorted([(stage_name, bl.stage_bl.name) for (stage_name, bl) in stage_bls.items()])) trained_ml_full_name = (ml_name, sbls) uplift_pred = self._metalearner_predict(ml_name, stage_bls) metric_value = self.calculate_metric(test_target, uplift_pred, test_treatment) self._metalearner_metrics[trained_ml_full_name] = metric_value def _bl_for_ml(self) -> Generator[(Dict[(str, Dict[(MLStageFullName, TrainedStageBaseLearner)])], None, None)]: ready_metalearners = [] for (ml_name, ml_stages) in self.__MAP_META_TO_STAGES__.items(): ready_metalearners.append(all(((s.full_name() in self._trained_stage_baselearners) for s in ml_stages))) if (not any(ready_metalearners)): raise Exception('No one metalearner can predict.') stage_baselearners = {} for (stage_fullname, bls) in self._trained_stage_baselearners.items(): stage_baselearners[stage_fullname] = bls stage_names = list(stage_baselearners.keys()) bls_prd = product(*(bls for (_, bls) in stage_baselearners.items())) for bl in bls_prd: set_bls = dict(zip(stage_names, bl)) set_ml_with_sbls = {} for (ml_name, ml_stages) in self.__MAP_META_TO_STAGES__.items(): ml_bls = {} for ml_stage in ml_stages: ml_stage_full_name = ml_stage.full_name() if (ml_stage_full_name in set_bls): trained_sbl = set_bls[ml_stage_full_name] if (trained_sbl.prev_stage_bl is None): ml_bls[ml_stage_full_name] = trained_sbl else: if (not (ml_stage_full_name[0:1] in set_bls)): continue if (trained_sbl.prev_stage_bl.name == set_bls[ml_stage_full_name[0:1]].stage_bl.name): ml_bls[ml_stage_full_name] = set_bls[ml_stage_full_name] if (len(ml_bls) == len(ml_stages)): set_ml_with_sbls[ml_name] = ml_bls (yield set_ml_with_sbls) def _metalearner_predict(self, metalearner: str, baselearners: Dict[(MLStageFullName, TrainedStageBaseLearner)]) -> np.ndarray: if (metalearner == 'TLearner'): control_pred = baselearners[('outcome_control',)].pred treatment_pred = baselearners[('outcome_treatment',)].pred uplift_pred = (treatment_pred - control_pred) elif (metalearner == 'XLearner'): control_pred = baselearners[('outcome_treatment', 'effect_control')].pred treatment_pred = baselearners[('outcome_control', 'effect_treatment')].pred propensity_pred = baselearners[('propensity',)].pred uplift_pred = ((propensity_pred * treatment_pred) + ((1 - propensity_pred) * control_pred)) else: raise Exception() return uplift_pred.ravel() def _set_best_metalearner(self): best_metric_value = None best_candidate = None for (k, v) in self._metalearner_metrics.items(): if (best_metric_value is None): best_candidate = k best_metric_value = v elif ((self.increasing_metric and (best_metric_value < v)) or ((not self.increasing_metric) and (best_metric_value > v))): best_candidate = k best_metric_value = v (ml_name, stages_params) = best_candidate stages_params = dict(stages_params) self._best_metalearner = self._init_metalearner(ml_name, stages_params) self._best_metalearner_wrap = self._create_metalearner_wrap(ml_name, stages_params) def _init_metalearner(self, metalearner_name: str, bls: Dict[(MLStageFullName, str)]) -> MetaLearner: ml: Optional[MetaLearner] = None if (metalearner_name == 'TLearner'): ocl = self._get_trained_bl(('outcome_control',), bls[('outcome_control',)]).trained_model otl = self._get_trained_bl(('outcome_treatment',), bls[('outcome_treatment',)]).trained_model ml = TLearner(control_learner=ocl, treatment_learner=otl) elif (metalearner_name == 'XLearner'): ocl = self._get_trained_bl(('outcome_control',), bls[('outcome_control',)]).trained_model otl = self._get_trained_bl(('outcome_treatment',), bls[('outcome_treatment',)]).trained_model pl = self._get_trained_bl(('propensity',), bls[('propensity',)]).trained_model ecl = self._get_trained_bl(('outcome_treatment', 'effect_control'), bls[('outcome_treatment', 'effect_control')]).trained_model etl = self._get_trained_bl(('outcome_control', 'effect_treatment'), bls[('outcome_control', 'effect_treatment')]).trained_model ml = XLearner(outcome_learners=[ocl, otl], effect_learners=[ecl, etl], propensity_learner=pl) else: raise Exception() return ml def _get_trained_bl(self, metalearner_stage: MLStageFullName, baselearner_name: str): for bl in self._trained_stage_baselearners[metalearner_stage]: if (bl.stage_bl.name == baselearner_name): return bl raise Exception("There isn't baselearner {}".format(baselearner_name)) def _create_metalearner_wrap(self, metalearner_name: str, bls: Dict[(MLStageFullName, str)]) -> MetaLearnerWrapper: ml_wrap_name = '__ML__{ML}'.format(ML=metalearner_name) ml_wrap: Optional[MetaLearnerWrapper] = None if (metalearner_name == 'TLearner'): ocl = self._get_trained_bl('outcome_control', bls[('outcome_control',)]).stage_bl otl = self._get_trained_bl(('outcome_treatment',), bls[('outcome_treatment',)]).stage_bl ml_wrap = MetaLearnerWrapper(name=ml_wrap_name, klass=TLearner, params={'control_learner': ocl, 'treatment_learner': otl}) elif (metalearner_name == 'XLearner'): ocl = self._get_trained_bl(('outcome_control',), bls[('outcome_control',)]).stage_bl otl = self._get_trained_bl(('outcome_treatment',), bls[('outcome_treatment',)]).stage_bl pl = self._get_trained_bl(('propensity',), bls[('propensity',)]).stage_bl ecl = self._get_trained_bl(('outcome_treatment', 'effect_control'), bls[('outcome_treatment', 'effect_control')]).stage_bl etl = self._get_trained_bl(('outcome_control', 'effect_treatment'), bls[('outcome_control', 'effect_treatment')]).stage_bl ml_wrap = MetaLearnerWrapper(name=ml_wrap_name, klass=XLearner, params={'outcome_learners': [ocl, otl], 'effect_learners': [ecl, etl], 'propensity_learner': pl}) else: raise Exception() return ml_wrap def __default_learners(self, lin_params: Optional[Dict[(str, Any)]]=None, tab_params: Optional[Dict[(str, Any)]]=None) -> List[BaseLearnerWrapper]: default_lin_params = {'cpu_limit': self.cpu_limit} default_tab_params = {'timeout': None, 'cpu_limit': self.cpu_limit, 'gpu_ids': self.gpu_ids} return [BaseLearnerWrapper(name='__Linear__', klass=uplift_utils.create_linear_automl, params=(default_lin_params if (lin_params is None) else lin_params)), BaseLearnerWrapper(name='__Tabular__', klass=TabularAutoML, params=(default_tab_params if (tab_params is None) else tab_params))]
class AutoContrast(DauphinTransform): def __init__(self, name=None, prob=1.0, level=0): super().__init__(name, prob, level) def transform(self, pil_img, label, **kwargs): return (ImageOps.autocontrast(pil_img), label)
class CAtlas(): def __init__(self, cdbg_directory, catlas_directory, load_domfile=True, load_sizefile=False, min_abund=0.0): self.cdbg_dir = cdbg_directory self.name = catlas_directory self.parent = {} self.children = defaultdict(set) self.levels = {} self._cdbg_to_catlas = {} catlas_file = os.path.join(catlas_directory, 'catlas.csv') self.__load_catlas(catlas_file) if load_domfile: domfile = os.path.join(catlas_directory, 'first_doms.txt') self.__load_first_level(domfile) if (load_sizefile is not None): sizefile = os.path.join(cdbg_directory, 'contigs.info.csv') self.__load_size_info(sizefile, min_abund) def __load_catlas(self, catlas_file): self.max_level = (- 1) self.root = (- 1) fp = open(catlas_file, 'rt') for line in fp: (node_id, cdbg_id, level, children) = line.strip().split(',') node_id = int(node_id) children = children.strip() if children: children = children.split(' ') children = set(map(int, children)) self.children[node_id] = children for child in children: self.parent[child] = node_id level = int(level) self.levels[node_id] = level if (level > self.max_level): self.max_level = level self.root = node_id if (level == 1): self._cdbg_to_catlas[int(cdbg_id)] = node_id fp.close() def __load_first_level(self, domfile): self.layer1_to_cdbg = {} self.cdbg_to_layer1 = {} fp = open(domfile, 'rt') for line in fp: (dom_node, *beneath) = line.strip().split(' ') dom_node = int(dom_node) beneath = set(map(int, beneath)) equiv_cdbg_to_catlas = self._cdbg_to_catlas[dom_node] self.layer1_to_cdbg[equiv_cdbg_to_catlas] = beneath for cdbg_id in beneath: self.cdbg_to_layer1[cdbg_id] = equiv_cdbg_to_catlas fp.close() def __load_size_info(self, sizefile, min_abund): cdbg_sizes = defaultdict(int) weighted_cdbg_sizes = defaultdict(float) with open(sizefile, 'rt') as fp: reader = csv.DictReader(fp) for row in reader: contig_id = int(row['contig_id']) n_kmers = int(row['n_kmers']) mean_abund = float(row['mean_abund']) if ((not min_abund) or (mean_abund >= min_abund)): cdbg_sizes[contig_id] = n_kmers weighted_cdbg_sizes[contig_id] = (mean_abund * n_kmers) self.cdbg_sizes = cdbg_sizes self.weighted_cdbg_sizes = weighted_cdbg_sizes self.kmer_sizes = {} self.weighted_kmer_sizes = {} for node_id in self: level = self.levels[node_id] if (level == 1): total_kmers = 0 total_weighted_kmers = 0 for cdbg_node in self.layer1_to_cdbg.get(node_id): total_kmers += cdbg_sizes[cdbg_node] total_weighted_kmers += weighted_cdbg_sizes[cdbg_node] self.kmer_sizes[node_id] = total_kmers self.weighted_kmer_sizes[node_id] = total_weighted_kmers else: sub_size = 0 sub_weighted_size = 0 for child_id in self.children[node_id]: sub_size += self.kmer_sizes[child_id] sub_weighted_size += self.weighted_kmer_sizes[child_id] self.kmer_sizes[node_id] = sub_size self.weighted_kmer_sizes[node_id] = sub_weighted_size def __iter__(self): Q = [self.root] pos = 0 while (pos < len(self.levels)): v = Q[pos] Q.extend(self.children[v]) pos += 1 for v in reversed(Q): (yield v) def __len__(self): return len(self.parent) def decorate_with_shadow_sizes(self): self.shadow_sizes = {} for node_id in self: level = self.levels[node_id] if (level == 1): self.shadow_sizes[node_id] = len(self.layer1_to_cdbg[node_id]) else: sub_size = 0 for child_id in self.children[node_id]: sub_size += self.shadow_sizes[child_id] self.shadow_sizes[node_id] = sub_size def decorate_with_index_sizes(self, index): self.index_sizes = index.build_catlas_node_sizes(self) def leaves(self, nodes: List[int]=None) -> Set[int]: if (nodes is None): nodes = [self.root] leaves = set() seen_nodes = set() def add_to_shadow(node_id: int): if (node_id in seen_nodes): return seen_nodes.add(node_id) children_ids = self.children[node_id] if (len(children_ids) == 0): leaves.add(node_id) else: for child in children_ids: add_to_shadow(child) for node in nodes: add_to_shadow(node) return leaves def shadow(self, nodes: List[int]) -> Set[int]: leaves = self.leaves(nodes) shadow = set() for leaf in leaves: shadow.update(self.layer1_to_cdbg[leaf]) return shadow
class MultiEnvWrapper(gym.Wrapper): def __init__(self, envs, sample_strategy=uniform_random_strategy): self._sample_strategy = sample_strategy self._num_tasks = len(envs) self._active_task_index = None self._observation_space = None super().__init__(envs[0]) self._task_envs = [] for env in envs: if (env.observation_space.shape != self.env.observation_space.shape): raise ValueError('Observation space of all envs should be same.') if (env.action_space.shape != self.env.action_space.shape): raise ValueError('Action space of all envs should be same.') self._task_envs.append(env) self.env.spec.observation_space = self.observation_space def num_tasks(self): return len(self._task_envs) def task_space(self): one_hot_ub = np.ones(self.num_tasks) one_hot_lb = np.zeros(self.num_tasks) return akro.Box(one_hot_lb, one_hot_ub) def active_task_index(self): return self._active_task_index def observation_space(self): (task_lb, task_ub) = self.task_space.bounds (env_lb, env_ub) = self._observation_space.bounds return akro.Box(np.concatenate([task_lb, env_lb]), np.concatenate([task_ub, env_ub])) _space.setter def observation_space(self, observation_space): self._observation_space = observation_space def active_task_one_hot(self): one_hot = np.zeros(self.task_space.shape) index = (self.active_task_index or 0) one_hot[index] = self.task_space.high[index] return one_hot def reset(self, **kwargs): self._active_task_index = self._sample_strategy(self._num_tasks, self._active_task_index) self.env = self._task_envs[self._active_task_index] obs = self.env.reset(**kwargs) oh_obs = self._obs_with_one_hot(obs) return oh_obs def step(self, action): (obs, reward, done, info) = self.env.step(action) oh_obs = self._obs_with_one_hot(obs) info['task_id'] = self._active_task_index return (oh_obs, reward, done, info) def close(self): for env in self._task_envs: env.close() def _obs_with_one_hot(self, obs): oh_obs = np.concatenate([self.active_task_one_hot, obs]) return oh_obs
class foldnorm_gen(rv_continuous): def _argcheck(self, c): return (c >= 0) def _shape_info(self): return [_ShapeInfo('c', False, (0, np.inf), (True, False))] def _rvs(self, c, size=None, random_state=None): return abs((random_state.standard_normal(size) + c)) def _pdf(self, x, c): return (_norm_pdf((x + c)) + _norm_pdf((x - c))) def _cdf(self, x, c): sqrt_two = np.sqrt(2) return (0.5 * (sc.erf(((x - c) / sqrt_two)) + sc.erf(((x + c) / sqrt_two)))) def _sf(self, x, c): return (_norm_sf((x - c)) + _norm_sf((x + c))) def _stats(self, c): c2 = (c * c) expfac = (np.exp(((- 0.5) * c2)) / np.sqrt((2.0 * np.pi))) mu = ((2.0 * expfac) + (c * sc.erf((c / np.sqrt(2))))) mu2 = ((c2 + 1) - (mu * mu)) g1 = (2.0 * ((((mu * mu) * mu) - (c2 * mu)) - expfac)) g1 /= np.power(mu2, 1.5) g2 = (((c2 * (c2 + 6.0)) + 3) + ((8.0 * expfac) * mu)) g2 += (((2.0 * (c2 - 3.0)) - (3.0 * (mu ** 2))) * (mu ** 2)) g2 = ((g2 / (mu2 ** 2.0)) - 3.0) return (mu, mu2, g1, g2)
_REGISTRY.register() class PartialiLIDS(ImageDataset): dataset_name = 'partialilids' def __init__(self, root='datasets'): self.root = root self.query_dir = osp.join(self.root, 'PartialiLIDS/query') self.gallery_dir = osp.join(self.root, 'PartialiLIDS/gallery') (query, gallery) = process_test(self.query_dir, self.gallery_dir) ImageDataset.__init__(self, [], query, gallery)
class EvaluatorConfig(metaclass=AutodocABCMeta): _timedelta_keys = ['train_window', 'retrain_freq', 'cadence'] def __init__(self, train_window: float=None, retrain_freq: float=None, cadence: float=None): self.train_window = train_window self.retrain_freq = retrain_freq self.cadence = cadence def train_window(self) -> Union[(pd.Timedelta, pd.DateOffset, None)]: return self._train_window _window.setter def train_window(self, train_window): self._train_window = to_offset(train_window) def retrain_freq(self) -> Union[(pd.Timedelta, pd.DateOffset, None)]: return self._retrain_freq _freq.setter def retrain_freq(self, retrain_freq): self._retrain_freq = to_offset(retrain_freq) def cadence(self) -> Union[(pd.Timedelta, pd.DateOffset)]: if (self._cadence is None): return self.retrain_freq return self._cadence def cadence(self, cadence): self._cadence = to_offset(cadence) def horizon(self) -> pd.DateOffset: return self.cadence def to_dict(self): config_dict = {} for (key, value) in self.__dict__.items(): k_strip = key.lstrip('_') if ((k_strip in self._timedelta_keys) and (value is not None)): config_dict[k_strip] = ((value.microseconds / 1000000.0) if isinstance(value, pd.Timedelta) else value.freqstr) else: config_dict[k_strip] = value return config_dict
def get_joint_slot_correctness(preds, class_types, label_maps, key_class_label_id='class_label_id', key_class_prediction='class_prediction', key_start_pos='start_pos', key_start_prediction='start_prediction', key_end_pos='end_pos', key_end_prediction='end_prediction', key_refer_id='refer_id', key_refer_prediction='refer_prediction', key_slot_groundtruth='slot_groundtruth', key_slot_prediction='slot_prediction'): for pred in preds: guid = pred['guid'] turn_gt_class = pred[key_class_label_id] turn_pd_class = pred[key_class_prediction] gt_start_pos = pred[key_start_pos] pd_start_pos = pred[key_start_prediction] gt_end_pos = pred[key_end_pos] pd_end_pos = pred[key_end_prediction] gt_refer = pred[key_refer_id] pd_refer = pred[key_refer_prediction] gt_slot = pred[key_slot_groundtruth] pd_slot = pred[key_slot_prediction] gt_slot = tokenize(gt_slot) pd_slot = tokenize(pd_slot) joint_gt_slot = gt_slot if (guid[(- 1)] == '0'): joint_pd_slot = 'none' if (turn_pd_class == class_types.index('none')): pass elif (turn_pd_class == class_types.index('dontcare')): joint_pd_slot = 'dontcare' elif (turn_pd_class == class_types.index('copy_value')): joint_pd_slot = pd_slot elif (('true' in class_types) and (turn_pd_class == class_types.index('true'))): joint_pd_slot = 'true' elif (('false' in class_types) and (turn_pd_class == class_types.index('false'))): joint_pd_slot = 'false' elif (('refer' in class_types) and (turn_pd_class == class_types.index('refer'))): if (pd_slot[0:3] == ' '): if (pd_slot[3:] != 'none'): joint_pd_slot = check_slot_inform(joint_gt_slot, pd_slot[3:], label_maps) elif (pd_slot[0:2] == ''): if (pd_slot[2:] != 'none'): joint_pd_slot = check_slot_inform(joint_gt_slot, pd_slot[2:], label_maps) elif (pd_slot != 'none'): joint_pd_slot = pd_slot elif (('inform' in class_types) and (turn_pd_class == class_types.index('inform'))): if (pd_slot[0:3] == ' '): if (pd_slot[3:] != 'none'): joint_pd_slot = check_slot_inform(joint_gt_slot, pd_slot[3:], label_maps) elif (pd_slot[0:2] == ''): if (pd_slot[2:] != 'none'): joint_pd_slot = check_slot_inform(joint_gt_slot, pd_slot[2:], label_maps) else: print('ERROR: Unexpected slot value format. Aborting.') exit() else: print('ERROR: Unexpected class_type. Aborting.') exit() if (joint_gt_slot == joint_pd_slot): pred_flag = 1.0 elif ((joint_gt_slot != 'none') and (joint_gt_slot != 'dontcare') and (joint_gt_slot != 'true') and (joint_gt_slot != 'false') and (joint_gt_slot in label_maps)): no_match = True for variant in label_maps[joint_gt_slot]: if (variant == joint_pd_slot): no_match = False pred_flag = 1.0 break if no_match: pred_flag = 0.0 else: pred_flag = 0.0 return (pred_flag, joint_pd_slot, joint_gt_slot)
class RawData(TypedDict): label: ThreeLabels supporting_sentences: list[list[int]] claim: str evidence: list[str] meta: dict
class local_mem(Structure): _fields_ = [('raw_ptr', POINTER(ctypes.c_char)), ('mem_arr', POINTER(POINTER(ctypes.c_uint32))), ('count', ctypes.c_int32), ('size_per_mem', ctypes.c_int32), ('align_num', ctypes.c_int32), ('need_free', ctypes.c_int32)]
class GeneralEdgeAttConvv1(nn.Module): def __init__(self, dim_in, dim_out, bias=False, **kwargs): super(GeneralEdgeAttConvv1, self).__init__() self.model = GeneralEdgeAttConvv1Layer(dim_in, dim_out, bias=bias) def forward(self, batch): batch.node_feature = self.model(batch.node_feature, batch.edge_index, edge_feature=batch.edge_feature) return batch
def main(unused_argv): def _is_valid_num_shards(num_shards): return ((num_shards < FLAGS.num_threads) or (not (num_shards % FLAGS.num_threads))) assert _is_valid_num_shards(FLAGS.train_shards), 'Please make the FLAGS.num_threads commensurate with FLAGS.train_shards' assert _is_valid_num_shards(FLAGS.val_shards), 'Please make the FLAGS.num_threads commensurate with FLAGS.val_shards' assert _is_valid_num_shards(FLAGS.test_shards), 'Please make the FLAGS.num_threads commensurate with FLAGS.test_shards' if (not tf.gfile.IsDirectory(FLAGS.output_dir)): tf.gfile.MakeDirs(FLAGS.output_dir) mscoco_train_dataset = _load_and_process_metadata(FLAGS.train_captions_file, FLAGS.train_image_dir) mscoco_val_dataset = _load_and_process_metadata(FLAGS.val_captions_file, FLAGS.val_image_dir) train_cutoff = int((0.85 * len(mscoco_val_dataset))) val_cutoff = int((0.9 * len(mscoco_val_dataset))) train_dataset = (mscoco_train_dataset + mscoco_val_dataset[0:train_cutoff]) val_dataset = mscoco_val_dataset[train_cutoff:val_cutoff] test_dataset = mscoco_val_dataset[val_cutoff:] train_captions = [c for image in train_dataset for c in image.captions] vocab = _create_vocab(train_captions) _process_dataset('train', train_dataset, vocab, FLAGS.train_shards) _process_dataset('val', val_dataset, vocab, FLAGS.val_shards) _process_dataset('test', test_dataset, vocab, FLAGS.test_shards)
def dconv_flops_counter_hook(dconv_module, input, output): input = input[0] batch_size = input.shape[0] output_dims = list(output.shape[2:]) (m_channels, in_channels, kernel_dim1, _) = dconv_module.weight.shape (out_channels, _, kernel_dim2, _) = dconv_module.projection.shape conv_per_position_flops1 = (((kernel_dim1 ** 2) * in_channels) * m_channels) conv_per_position_flops2 = (((kernel_dim2 ** 2) * out_channels) * m_channels) active_elements_count = (batch_size * np.prod(output_dims)) overall_conv_flops = ((conv_per_position_flops1 + conv_per_position_flops2) * active_elements_count) overall_flops = overall_conv_flops dconv_module.__flops__ += int(overall_flops)
class TypecastNode(ExprNode): subexprs = ['operand'] base_type = declarator = type = None def type_dependencies(self, env): return () def infer_type(self, env): if (self.type is None): base_type = self.base_type.analyse(env) (_, self.type) = self.declarator.analyse(base_type, env) return self.type def analyse_types(self, env): if (self.type is None): base_type = self.base_type.analyse(env) (_, self.type) = self.declarator.analyse(base_type, env) if self.operand.has_constant_result(): self.calculate_constant_result() if self.type.is_cfunction: error(self.pos, 'Cannot cast to a function type') self.type = PyrexTypes.error_type self.operand = self.operand.analyse_types(env) if (self.type is PyrexTypes.c_bint_type): return self.operand.coerce_to_boolean(env) to_py = self.type.is_pyobject from_py = self.operand.type.is_pyobject if (from_py and (not to_py) and self.operand.is_ephemeral()): if ((not self.type.is_numeric) and (not self.type.is_cpp_class)): error(self.pos, 'Casting temporary Python object to non-numeric non-Python type') if (to_py and (not from_py)): if ((self.type is bytes_type) and self.operand.type.is_int): return CoerceIntToBytesNode(self.operand, env) elif self.operand.type.can_coerce_to_pyobject(env): self.result_ctype = py_object_type self.operand = self.operand.coerce_to(self.type, env) else: if self.operand.type.is_ptr: if (not (self.operand.type.base_type.is_void or self.operand.type.base_type.is_struct)): error(self.pos, 'Python objects cannot be cast from pointers of primitive types') else: warning(self.pos, ('No conversion from %s to %s, python object pointer used.' % (self.operand.type, self.type))) self.operand = self.operand.coerce_to_simple(env) elif (from_py and (not to_py)): if self.type.create_from_py_utility_code(env): self.operand = self.operand.coerce_to(self.type, env) elif self.type.is_ptr: if (not (self.type.base_type.is_void or self.type.base_type.is_struct)): error(self.pos, 'Python objects cannot be cast to pointers of primitive types') else: warning(self.pos, ('No conversion from %s to %s, python object pointer used.' % (self.type, self.operand.type))) elif (from_py and to_py): if self.typecheck: self.operand = PyTypeTestNode(self.operand, self.type, env, notnone=True) elif isinstance(self.operand, SliceIndexNode): self.operand = self.operand.coerce_to(self.type, env) elif (self.type.is_complex and self.operand.type.is_complex): self.operand = self.operand.coerce_to_simple(env) elif self.operand.type.is_fused: self.operand = self.operand.coerce_to(self.type, env) if (self.type.is_ptr and self.type.base_type.is_cfunction and self.type.base_type.nogil): op_type = self.operand.type if op_type.is_ptr: op_type = op_type.base_type if (op_type.is_cfunction and (not op_type.nogil)): warning(self.pos, 'Casting a GIL-requiring function into a nogil function circumvents GIL validation', 1) return self def is_simple(self): return self.operand.is_simple() def is_ephemeral(self): return self.operand.is_ephemeral() def nonlocally_immutable(self): return (self.is_temp or self.operand.nonlocally_immutable()) def nogil_check(self, env): if (self.type and self.type.is_pyobject and self.is_temp): self.gil_error() def check_const(self): return self.operand.check_const() def calculate_constant_result(self): self.constant_result = self.calculate_result_code(self.operand.constant_result) def calculate_result_code(self, operand_result=None): if (operand_result is None): operand_result = self.operand.result() if self.type.is_complex: operand_result = self.operand.result() if self.operand.type.is_complex: real_part = self.type.real_type.cast_code(('__Pyx_CREAL(%s)' % operand_result)) imag_part = self.type.real_type.cast_code(('__Pyx_CIMAG(%s)' % operand_result)) else: real_part = self.type.real_type.cast_code(operand_result) imag_part = '0' return ('%s(%s, %s)' % (self.type.from_parts, real_part, imag_part)) else: return self.type.cast_code(operand_result) def get_constant_c_result_code(self): operand_result = self.operand.get_constant_c_result_code() if operand_result: return self.type.cast_code(operand_result) def result_as(self, type): if (self.type.is_pyobject and (not self.is_temp)): return self.operand.result_as(type) else: return ExprNode.result_as(self, type) def generate_result_code(self, code): if self.is_temp: code.putln(('%s = (PyObject *)%s;' % (self.result(), self.operand.result()))) code.put_incref(self.result(), self.ctype())
def set_working_device(device_name: str): device_manager = DeviceManager() device_manager.set_device(device_name)
def formatannotation(annotation, base_module=None): if isinstance(annotation, type): if (annotation.__module__ in ('builtins', '__builtin__', base_module)): return annotation.__name__ return ((annotation.__module__ + '.') + annotation.__name__) return repr(annotation)
def build_model(): g = tf.Graph() with g.as_default(), tf.device(tf.train.replica_device_setter(FLAGS.ps_tasks)): (inputs, labels) = imagenet_input(is_training=True) with slim.arg_scope(mobilenet_v1.mobilenet_v1_arg_scope(is_training=True)): (logits, _) = mobilenet_v1.mobilenet_v1(inputs, is_training=True, depth_multiplier=FLAGS.depth_multiplier, num_classes=FLAGS.num_classes) tf.losses.softmax_cross_entropy(labels, logits) if FLAGS.quantize: tf.contrib.quantize.create_training_graph(quant_delay=get_quant_delay()) total_loss = tf.losses.get_total_loss(name='total_loss') num_epochs_per_decay = 2.5 imagenet_size = 1271167 decay_steps = int(((imagenet_size / FLAGS.batch_size) * num_epochs_per_decay)) learning_rate = tf.train.exponential_decay(get_learning_rate(), tf.train.get_or_create_global_step(), decay_steps, _LEARNING_RATE_DECAY_FACTOR, staircase=True) opt = tf.train.GradientDescentOptimizer(learning_rate) train_tensor = slim.learning.create_train_op(total_loss, optimizer=opt) slim.summaries.add_scalar_summary(total_loss, 'total_loss', 'losses') slim.summaries.add_scalar_summary(learning_rate, 'learning_rate', 'training') return (g, train_tensor)
def dstn(x, type=2, s=None, axes=None, norm=None, overwrite_x=False, workers=None, orthogonalize=None): return _execute(_pocketfft.dstn, x, type, s, axes, norm, overwrite_x, workers, orthogonalize)
def load_conv2d(state_dict, name_pth, name_tf): h5f = h5py.File((('dump/InceptionV4/' + name_tf) + '.h5'), 'r') state_dict[(name_pth + '.conv.weight')] = torch.from_numpy(h5f['weights'][()]).permute(3, 2, 0, 1) out_planes = state_dict[(name_pth + '.conv.weight')].size(0) state_dict[(name_pth + '.bn.weight')] = torch.ones(out_planes) state_dict[(name_pth + '.bn.bias')] = torch.from_numpy(h5f['beta'][()]) state_dict[(name_pth + '.bn.running_mean')] = torch.from_numpy(h5f['mean'][()]) state_dict[(name_pth + '.bn.running_var')] = torch.from_numpy(h5f['var'][()]) h5f.close()
def capture_time(): start = time.perf_counter() done = False def fn(): if done: return (end - start) else: return (time.perf_counter() - start) (yield fn) end = time.time()
_module() class SRREDSMultipleGTDataset(BaseSRDataset): def __init__(self, lq_folder, gt_folder, num_input_frames, pipeline, scale, val_partition='official', repeat=1, test_mode=False): self.repeat = repeat if (not isinstance(repeat, int)): raise TypeError(f'"repeat" must be an integer, but got {type(repeat)}.') super().__init__(pipeline, scale, test_mode) self.lq_folder = str(lq_folder) self.gt_folder = str(gt_folder) self.num_input_frames = num_input_frames self.val_partition = val_partition self.data_infos = self.load_annotations() def load_annotations(self): keys = [f'{i:03d}' for i in range(0, 270)] if (self.val_partition == 'REDS4'): val_partition = ['000', '011', '015', '020'] elif (self.val_partition == 'official'): val_partition = [f'{i:03d}' for i in range(240, 270)] else: raise ValueError(f'Wrong validation partition {self.val_partition}.Supported ones are ["official", "REDS4"]') if self.test_mode: keys = [v for v in keys if (v in val_partition)] keys *= self.repeat else: keys = [v for v in keys if (v not in val_partition)] data_infos = [] for key in keys: data_infos.append(dict(lq_path=self.lq_folder, gt_path=self.gt_folder, key=key, sequence_length=100, num_input_frames=self.num_input_frames)) return data_infos
class IndicatorMin(OptimizationFunction): def __init__(self, objective: OptimizationFunction, beta: float=0, power: float=2): super().__init__(objective) self.obj = objective self.beta = beta self.power = power def eval(self, input_vals: List[np.ndarray]) -> np.ndarray: return ((input_vals[0] < self.beta) * (abs((self.beta - input_vals[0])) ** self.power)) def grad(self, input_vals: List[np.ndarray], grad_val: np.ndarray) -> List[np.ndarray]: beta_diff = (self.beta - input_vals[0]) graph_beta = (((((input_vals[0] < self.beta) * self.power) * (abs(beta_diff) ** (self.power - 1))) * np.sign(beta_diff)) * (- 1)) return [(grad_val * graph_beta)] def __str__(self): return 'I_min({0}-{1})**{2}'.format(self.obj, self.beta, self.power)
class Inferencer(ABC): def load_model(self, path: (str | Path)) -> Any: raise NotImplementedError def pre_process(self, image: np.ndarray) -> (np.ndarray | Tensor): raise NotImplementedError def forward(self, image: (np.ndarray | Tensor)) -> (np.ndarray | Tensor): raise NotImplementedError def post_process(self, predictions: (np.ndarray | Tensor), metadata: (dict[(str, Any)] | None)) -> dict[(str, Any)]: raise NotImplementedError def predict(self, image: ((str | Path) | np.ndarray), metadata: (dict[(str, Any)] | None)=None) -> ImageResult: if (metadata is None): if hasattr(self, 'metadata'): metadata = getattr(self, 'metadata') else: metadata = {} if isinstance(image, (str, Path)): image_arr: np.ndarray = read_image(image) else: image_arr = image metadata['image_shape'] = image_arr.shape[:2] processed_image = self.pre_process(image_arr) predictions = self.forward(processed_image) output = self.post_process(predictions, metadata=metadata) return ImageResult(image=image_arr, pred_score=output['pred_score'], pred_label=output['pred_label'], anomaly_map=output['anomaly_map'], pred_mask=output['pred_mask'], pred_boxes=output['pred_boxes'], box_labels=output['box_labels']) def _superimpose_segmentation_mask(metadata: dict, anomaly_map: np.ndarray, image: np.ndarray) -> np.ndarray: pred_mask = compute_mask(anomaly_map, 0.5) image_height = metadata['image_shape'][0] image_width = metadata['image_shape'][1] pred_mask = cv2.resize(pred_mask, (image_width, image_height)) boundaries = find_boundaries(pred_mask) outlines = dilation(boundaries, np.ones((7, 7))) image[outlines] = [255, 0, 0] return image def __call__(self, image: np.ndarray) -> ImageResult: return self.predict(image) def _normalize(pred_scores: (Tensor | np.float32), metadata: (dict | DictConfig), anomaly_maps: ((Tensor | np.ndarray) | None)=None) -> tuple[(((np.ndarray | Tensor) | None), float)]: if (('min' in metadata) and ('max' in metadata)): if (anomaly_maps is not None): anomaly_maps = normalize_min_max(anomaly_maps, metadata['pixel_threshold'], metadata['min'], metadata['max']) pred_scores = normalize_min_max(pred_scores, metadata['image_threshold'], metadata['min'], metadata['max']) if (('pixel_mean' in metadata.keys()) and ('pixel_std' in metadata.keys())): if (anomaly_maps is not None): anomaly_maps = standardize(anomaly_maps, metadata['pixel_mean'], metadata['pixel_std'], center_at=metadata['image_mean']) anomaly_maps = normalize_cdf(anomaly_maps, metadata['pixel_threshold']) if (('image_mean' in metadata.keys()) and ('image_std' in metadata.keys())): pred_scores = standardize(pred_scores, metadata['image_mean'], metadata['image_std']) pred_scores = normalize_cdf(pred_scores, metadata['image_threshold']) return (anomaly_maps, float(pred_scores)) def _load_metadata(self, path: ((str | Path) | None)=None) -> (dict | DictConfig): metadata: (dict[(str, ((float | np.ndarray) | Tensor))] | DictConfig) = {} if (path is not None): config = OmegaConf.load(path) metadata = cast(DictConfig, config) return metadata
class AuxiliaryHeadCIFAR(nn.Module): def __init__(self, C, num_classes): super(AuxiliaryHeadCIFAR, self).__init__() self.features = nn.Sequential(nn.ReLU(inplace=True), nn.AvgPool2d(5, stride=3, padding=0, count_include_pad=False), nn.Conv2d(C, 128, 1, bias=False), nn.BatchNorm2d(128), nn.ReLU(inplace=True), nn.Conv2d(128, 768, 2, bias=False), nn.BatchNorm2d(768), nn.ReLU(inplace=True)) self.classifier = nn.Linear(768, num_classes) def forward(self, x): x = self.features(x) x = self.classifier(x.contiguous().view(x.size(0), (- 1))) return x
class BipedalWalkerExperiment(QDExperiment): def reinit(self): super().reinit() self.env_name = self.config['game']['env_name'] self.init_model() self.update_dimension() def init_model(self): self.model = Model(self.config['game']) def update_dimension(self): self.algo.dimension = self.model.param_count def eval_fn(self, ind, render_mode=False): env = make_env(self.env_name) self.model.set_model_params(ind) scores = simulate(self.model, env, render_mode=render_mode, num_episode=self.config['indv_eps']) ind.fitness.values = (scores[self.fitness_type],) ind.features.values = [scores[x] for x in self.features_list] return ind
def get_model(model_type: str, **kwargs: Union[(int, float)]) -> torch.nn.Module: if (model_type == 'deeptime'): model = deeptime(datetime_feats=kwargs['datetime_feats']) else: raise ValueError(f'Unknown model type {model_type}') return model
def add_model_args(parser): group = parser.add_argument_group('Model configuration') from fairseq.models import ARCH_MODEL_REGISTRY group.add_argument('--arch', '-a', default='fconv', metavar='ARCH', required=True, choices=ARCH_MODEL_REGISTRY.keys(), help='Model Architecture') return group