JorgeVB/code_unlearning_dataset · Datasets at Hugging Face

text stringlengths 81 112k
Binds the symbols to construct executors. This is necessary before one can perform computation with the module. Parameters ---------- data_shapes : list of (str, tuple) or DataDesc objects Typically is ``data_iter.provide_data``. Can also be a list of (data name,...
Find MXNet dynamic library files. Returns ------- lib_path : list(string) List of all found path to the libraries. def find_lib_path(): """Find MXNet dynamic library files. Returns ------- lib_path : list(string) List of all found path to the libraries. """ lib_fro...
Find MXNet included header files. Returns ------- incl_path : string Path to the header files. def find_include_path(): """Find MXNet included header files. Returns ------- incl_path : string Path to the header files. """ incl_from_env = os.environ.get('MXNET_INCLU...
Generate a greyscale captcha image representing number string Parameters ---------- captcha_str: str string a characters for captcha image Returns ------- numpy.ndarray Generated greyscale image in np.ndarray float type with values normalized to ...
Generates a character string of digits. Number of digits are between self.num_digit_min and self.num_digit_max Returns ------- str def get_rand(num_digit_min, num_digit_max): """Generates a character string of digits. Number of digits are between self.num_digit_min and s...
Generate a random captcha image sample Returns ------- (numpy.ndarray, str) Tuple of image (numpy ndarray) and character string of digits used to generate the image def _gen_sample(self): """Generate a random captcha image sample Returns ------- (nump...
Registers a new optimizer. Once an optimizer is registered, we can create an instance of this optimizer with `create_optimizer` later. Examples -------- >>> @mx.optimizer.Optimizer.register ... class MyOptimizer(mx.optimizer.Optimizer): ... pass >>>...
Instantiates an optimizer with a given name and kwargs. .. note:: We can use the alias `create` for ``Optimizer.create_optimizer``. Parameters ---------- name: str Name of the optimizer. Should be the name of a subclass of Optimizer. Case insensitive. k...
Creates auxiliary state for a given weight, including FP32 high precision copy if original weight is FP16. This method is provided to perform automatic mixed precision training for optimizers that do not support it themselves. Parameters ---------- index : int ...
Updates the given parameter using the corresponding gradient and state. Mixed precision version. Parameters ---------- index : int The unique index of the parameter into the individual learning rates and weight decays. Learning rates and weight decay ...
Sets an individual learning rate multiplier for each parameter. If you specify a learning rate multiplier for a parameter, then the learning rate for the parameter will be set as the product of the global learning rate `self.lr` and its multiplier. .. note:: The default learning rate m...
Sets an individual weight decay multiplier for each parameter. By default, if `param_idx2name` was provided in the constructor, the weight decay multipler is set as 0 for all parameters whose name don't end with ``_weight`` or ``_gamma``. .. note:: The default weight decay mult...
Sets the number of the currently handled device. Parameters ---------- device_id : int The number of current device. def _set_current_context(self, device_id): """Sets the number of the currently handled device. Parameters ---------- device_id : int...
Updates num_update. Parameters ---------- index : int or list of int The index to be updated. def _update_count(self, index): """Updates num_update. Parameters ---------- index : int or list of int The index to be updated. """ ...
Gets the learning rates given the indices of the weights. Parameters ---------- indices : list of int Indices corresponding to weights. Returns ------- lrs : list of float Learning rates for those indices. def _get_lrs(self, indices): ""...
Gets weight decays for indices. Returns 0 for non-weights if the name of weights are provided for `__init__`. Parameters ---------- indices : list of int Indices of weights. Returns ------- wds : list of float Weight decays for those indi...
sync state context. def sync_state_context(self, state, context): """sync state context.""" if isinstance(state, NDArray): return state.as_in_context(context) elif isinstance(state, (tuple, list)): synced_state = (self.sync_state_context(i, context) for i in state) ...
Sets updater states. def set_states(self, states): """Sets updater states.""" states = pickle.loads(states) if isinstance(states, tuple) and len(states) == 2: self.states, self.optimizer = states else: self.states = states self.states_synced = dict.fromke...
Gets updater states. Parameters ---------- dump_optimizer : bool, default False Whether to also save the optimizer itself. This would also save optimizer information such as learning rate and weight decay schedules. def get_states(self, dump_optimizer=False): ""...
Preprocess: Convert a video into the mouth images def preprocess(from_idx, to_idx, _params): """ Preprocess: Convert a video into the mouth images """ source_exts = '*.mpg' src_path = _params['src_path'] tgt_path = _params['tgt_path'] face_predictor_path = './shape_predictor_68_face_landmar...
Read from frames def from_frames(self, path): """ Read from frames """ frames_path = sorted([os.path.join(path, x) for x in os.listdir(path)]) frames = [ndimage.imread(frame_path) for frame_path in frames_path] self.handle_type(frames) return self
Read from videos def from_video(self, path): """ Read from videos """ frames = self.get_video_frames(path) self.handle_type(frames) return self
Config video types def handle_type(self, frames): """ Config video types """ if self.vtype == 'mouth': self.process_frames_mouth(frames) elif self.vtype == 'face': self.process_frames_face(frames) else: raise Exception('Video type not ...
Preprocess from frames using face detector def process_frames_face(self, frames): """ Preprocess from frames using face detector """ detector = dlib.get_frontal_face_detector() predictor = dlib.shape_predictor(self.face_predictor_path) mouth_frames = self.get_frames_mout...
Preprocess from frames using mouth detector def process_frames_mouth(self, frames): """ Preprocess from frames using mouth detector """ self.face = np.array(frames) self.mouth = np.array(frames) self.set_data(frames)
Get frames using mouth crop def get_frames_mouth(self, detector, predictor, frames): """ Get frames using mouth crop """ mouth_width = 100 mouth_height = 50 horizontal_pad = 0.19 normalize_ratio = None mouth_frames = [] for frame in frames: ...
Get video frames def get_video_frames(self, path): """ Get video frames """ videogen = skvideo.io.vreader(path) frames = np.array([frame for frame in videogen]) return frames
Prepare the input of model def set_data(self, frames): """ Prepare the input of model """ data_frames = [] for frame in frames: #frame H x W x C frame = frame.swapaxes(0, 1) # swap width and height to form format W x H x C if len(frame.shape) ...
Resets the iterator to the beginning of the data. def reset(self): """Resets the iterator to the beginning of the data.""" self.curr_idx = 0 random.shuffle(self.idx) for buck in self.data: np.random.shuffle(buck)
Returns the next batch of data. def next(self): """Returns the next batch of data.""" if self.curr_idx == len(self.idx): raise StopIteration i, j = self.idx[self.curr_idx] self.curr_idx += 1 audio_paths = [] texts = [] for duration, audio_path, text ...
Subtract ImageNet mean pixel-wise from a BGR image. def subtract_imagenet_mean_preprocess_batch(batch): """Subtract ImageNet mean pixel-wise from a BGR image.""" batch = F.swapaxes(batch,0, 1) (r, g, b) = F.split(batch, num_outputs=3, axis=0) r = r - 123.680 g = g - 116.779 b = b - 103.939 ...
Not necessary in practice def imagenet_clamp_batch(batch, low, high): """ Not necessary in practice """ F.clip(batch[:,0,:,:],low-123.680, high-123.680) F.clip(batch[:,1,:,:],low-116.779, high-116.779) F.clip(batch[:,2,:,:],low-103.939, high-103.939)
Create a linear regression network for performing SVRG optimization. :return: an instance of mx.io.NDArrayIter :return: an instance of mx.mod.svrgmodule for performing SVRG optimization def create_network(batch_size, update_freq): """Create a linear regression network for performing SVRG optimization. ...
Function to evaluate accuracy of any data iterator passed to it as an argument def evaluate_accuracy(data_iterator, net): """Function to evaluate accuracy of any data iterator passed to it as an argument""" acc = mx.metric.Accuracy() for data, label in data_iterator: output = net(data) pred...
Function responsible for running the training the model. def train(train_dir=None, train_csv=None, epochs=30, batch_size=32): """Function responsible for running the training the model.""" if not train_dir or not os.path.exists(train_dir) or not train_csv: warnings.warn("No train directory could be fo...
Set size limit on bulk execution. Bulk execution bundles many operators to run together. This can improve performance when running a lot of small operators sequentially. Parameters ---------- size : int Maximum number of operators that can be bundled in a bulk. Returns -------...
calculate LM score of child beam by taking score from parent beam and bigram probability of last two chars def applyLM(parentBeam, childBeam, classes, lm): """ calculate LM score of child beam by taking score from parent beam and bigram probability of last two chars """ if lm and not childBeam.lmApplie...
add beam if it does not yet exist def addBeam(beamState, labeling): """ add beam if it does not yet exist """ if labeling not in beamState.entries: beamState.entries[labeling] = BeamEntry()
beam search as described by the paper of Hwang et al. and the paper of Graves et al. def ctcBeamSearch(mat, classes, lm, k, beamWidth): """ beam search as described by the paper of Hwang et al. and the paper of Graves et al. """ blankIdx = len(classes) maxT, maxC = mat.shape # initialise beam...
length-normalise LM score def norm(self): """ length-normalise LM score """ for (k, _) in self.entries.items(): labelingLen = len(self.entries[k].labeling) self.entries[k].prText = self.entries[k].prText ** (1.0 / (labelingLen if labelingLen else 1.0))
return beam-labelings, sorted by probability def sort(self): """ return beam-labelings, sorted by probability """ beams = [v for (_, v) in self.entries.items()] sortedBeams = sorted(beams, reverse=True, key=lambda x: x.prTotal*x.prText) return [x.labeling for x in sorted...
the localisation network in lenet-stn, it will increase acc about more than 1%, when num-epoch >=15 def get_loc(data, attr={'lr_mult':'0.01'}): """ the localisation network in lenet-stn, it will increase acc about more than 1%, when num-epoch >=15 """ loc = mx.symbol.Convolution(data=data, num_...
wrapper for initialize a detector Parameters: ---------- net : str test network name prefix : str load model prefix epoch : int load model epoch data_shape : int resize image shape mean_pixels : tuple (float, float, float) mean pixel values (R, G, B) ...
parse # classes and class_names if applicable def parse_class_names(class_names): """ parse # classes and class_names if applicable """ if len(class_names) > 0: if os.path.isfile(class_names): # try to open it to read class names with open(class_names, 'r') as f: ...
Parse string to tuple or int def parse_data_shape(data_shape_str): """Parse string to tuple or int""" ds = data_shape_str.strip().split(',') if len(ds) == 1: data_shape = (int(ds[0]), int(ds[0])) elif len(ds) == 2: data_shape = (int(ds[0]), int(ds[1])) else: raise ValueError...
A lenet style net, takes difference of each frame as input. def get_lenet(): """ A lenet style net, takes difference of each frame as input. """ source = mx.sym.Variable("data") source = (source - 128) * (1.0/128) frames = mx.sym.SliceChannel(source, num_outputs=30) diffs = [frames[i+1] - frame...
Custom evaluation metric on CRPS. def CRPS(label, pred): """ Custom evaluation metric on CRPS. """ for i in range(pred.shape[0]): for j in range(pred.shape[1] - 1): if pred[i, j] > pred[i, j + 1]: pred[i, j + 1] = pred[i, j] return np.sum(np.square(label - pred)) / l...
Run encoding to encode the label into the CDF target. def encode_label(label_data): """Run encoding to encode the label into the CDF target. """ systole = label_data[:, 1] diastole = label_data[:, 2] systole_encode = np.array([ (x < np.arange(600)) for x in systole ], dtype=np.u...
coco ann: [u'segmentation', u'area', u'iscrowd', u'image_id', u'bbox', u'category_id', u'id'] iscrowd: crowd instances are handled by marking their overlaps with all categories to -1 and later excluded in training bbox: [x1, y1, w, h] :param index: coco image ...
example results [{"image_id": 42, "category_id": 18, "bbox": [258.15,41.29,348.26,243.78], "score": 0.236}, ...] def _write_coco_results(self, _coco, detections): """ example results [{"image_id": 42, "category_id": 18, "bbox": [258.15,41.29,348...
Draw random samples from an approximately log-uniform or Zipfian distribution. This operation randomly samples num_sampled candidates the range of integers [0, range_max). The elements of sampled_candidates are drawn with replacement from the base distribution. The base distribution for this operator is...
Run a for loop with user-defined computation over NDArrays on dimension 0. This operator simulates a for loop and body has the computation for an iteration of the for loop. It runs the computation in body on each slice from the input NDArrays. body takes two arguments as input and outputs a tuple of t...
Run a while loop with user-defined computation and loop condition. This operator simulates a while loop which iterately does customized computation as long as the condition is satisfied. `loop_vars` is a list of NDArrays on which the computation uses. `cond` is a user-defined function, used as the lo...
Run an if-then-else using user-defined condition and computation This operator simulates a if-like branch which chooses to do one of the two customized computations according to the specified condition. `pred` is a scalar MXNet NDArray, indicating which branch of computation should be used. `then...
Performs an element-wise check to determine if the NDArray contains an infinite element or not. Parameters ---------- input : NDArray An N-D NDArray. Returns ------- output: NDArray The output NDarray, with same shape as input, where 1 indicates the array element is ...
LSTM Cell symbol def vanilla_lstm(num_hidden, indata, prev_state, param, seqidx, layeridx, is_batchnorm=False, gamma=None, beta=None, name=None): """LSTM Cell symbol""" i2h = mx.sym.FullyConnected(data=indata, weight=param.i2h_weight, bias=param.i...
LSTM Cell symbol def lstm(num_hidden, indata, prev_state, param, seqidx, layeridx, dropout=0., num_hidden_proj=0, is_batchnorm=False, gamma=None, beta=None, name=None): """LSTM Cell symbol""" # dropout input if dropout > 0.: indata = mx.sym.Dropout(data=indata, p=dropout) i2h = mx.sym...
read, resize, transform image, return im_tensor, im_info, gt_boxes roi_rec should have keys: ["image", "boxes", "gt_classes", "flipped"] 0 --- x (width, second dim of im) \| y (height, first dim of im) def get_image(roi_rec, short, max_size, mean, std): """ read, resize, transform image, return ...
Return BGR image read by opencv def imdecode(image_path): """Return BGR image read by opencv""" import os assert os.path.exists(image_path), image_path + ' not found' im = cv2.imread(image_path) return im
only resize input image to target size and return scale :param im: BGR image input by opencv :param short: one dimensional size (the short side) :param max_size: one dimensional max size (the long side) :return: resized image (NDArray) and scale (float) def resize(im, short, max_size): """ only...
transform into mxnet tensor, subtract pixel size and transform to correct format :param im: [height, width, channel] in BGR :param mean: [RGB pixel mean] :param std: [RGB pixel std var] :return: [batch, channel, height, width] def transform(im, mean, std): """ transform into mxnet tensor, ...
transform from mxnet im_tensor to ordinary RGB image im_tensor is limited to one image :param im_tensor: [batch, channel, height, width] :param mean: [RGB pixel mean] :param std: [RGB pixel std var] :return: im [height, width, channel(RGB)] def transform_inverse(im_tensor, mean, std): """ t...
vertically stack tensors by adding a new axis expand dims if only 1 tensor :param tensor_list: list of tensor to be stacked vertically :param pad: label to pad with :return: tensor with max shape def tensor_vstack(tensor_list, pad=0): """ vertically stack tensors by adding a new axis expand...
Get distance matrix given a matrix. Used in testing. def get_distance_matrix(x): """Get distance matrix given a matrix. Used in testing.""" square = nd.sum(x ** 2.0, axis=1, keepdims=True) distance_square = square + square.transpose() - (2.0 * nd.dot(x, x.transpose())) return nd.sqrt(distance_square)
Evaluate embeddings based on Recall@k. def evaluate_emb(emb, labels): """Evaluate embeddings based on Recall@k.""" d_mat = get_distance_matrix(emb) d_mat = d_mat.asnumpy() labels = labels.asnumpy() names = [] accs = [] for k in [1, 2, 4, 8, 16]: names.append('Recall@%d' % k) ...
Get learning rate based on schedule. def get_lr(lr, epoch, steps, factor): """Get learning rate based on schedule.""" for s in steps: if epoch >= s: lr *= factor return lr
Training function. def train(epochs, ctx): """Training function.""" if isinstance(ctx, mx.Context): ctx = [ctx] net.initialize(mx.init.Xavier(magnitude=2), ctx=ctx) opt_options = {'learning_rate': opt.lr, 'wd': opt.wd} if opt.optimizer == 'sgd': opt_options['momentum'] = 0.9 if...
Returns symbol for LSTM model up to loss/softmax def _lstm_unroll_base(num_lstm_layer, seq_len, num_hidden): """ Returns symbol for LSTM model up to loss/softmax""" param_cells = [] last_states = [] for i in range(num_lstm_layer): param_cells.append(LSTMParam(i2h_weight=mx.sym.Variable("l%d_i2h...
Adds Symbol.contrib.ctc_loss on top of pred symbol and returns the resulting symbol def _add_warp_ctc_loss(pred, seq_len, num_label, label): """ Adds Symbol.contrib.ctc_loss on top of pred symbol and returns the resulting symbol """ label = mx.sym.Reshape(data=label, shape=(-1,)) label = mx.sym.Cast(data=l...
Adds Symbol.WapCTC on top of pred symbol and returns the resulting symbol def _add_mxnet_ctc_loss(pred, seq_len, label): """ Adds Symbol.WapCTC on top of pred symbol and returns the resulting symbol """ pred_ctc = mx.sym.Reshape(data=pred, shape=(-4, seq_len, -1, 0)) loss = mx.sym.contrib.ctc_loss(data=pr...
Adds CTC loss on top of pred symbol and returns the resulting symbol def _add_ctc_loss(pred, seq_len, num_label, loss_type): """ Adds CTC loss on top of pred symbol and returns the resulting symbol """ label = mx.sym.Variable('label') if loss_type == 'warpctc': print("Using WarpCTC Loss") s...
Creates an unrolled LSTM symbol for inference if loss_type is not specified, and for training if loss_type is specified. loss_type must be one of 'ctc' or 'warpctc' Parameters ---------- num_lstm_layer: int seq_len: int num_hidden: int num_label: int loss_type: str 'ctc' or 'war...
Returns name and shape of init states of LSTM network Parameters ---------- batch_size: list of tuple of str and tuple of int and int num_lstm_layer: int num_hidden: int Returns ------- list of tuple of str and tuple of int and int def init_states(batch_size, num_lstm_layer, num_hidde...
ctypes implementation of imperative invoke wrapper def _imperative_invoke(handle, ndargs, keys, vals, out): """ctypes implementation of imperative invoke wrapper""" if out is not None: original_output = out if isinstance(out, NDArrayBase): out = (out,) num_output = ctypes.c_...
Set status to training/not training. When training, graph will be constructed for gradient computation. Operators will also run with ctx.is_train=True. For example, Dropout will drop inputs randomly when is_train=True while simply passing through if is_train=False. Parameters ---------- is_trai...
Compute the gradients of outputs w.r.t variables. Parameters ---------- outputs: list of NDArray out_grads: list of NDArray or None def backward(outputs, out_grads=None, retain_graph=False): """Compute the gradients of outputs w.r.t variables. Parameters ---------- outputs: list of ND...
Return function that computes both gradient of arguments and loss value. Parameters ---------- func: a python function The forward (loss) function. argnum: an int or a list of int The index of argument to calculate gradient for. Returns ------- grad_and_loss_func: a python ...
Return function that computes gradient of arguments. Parameters ---------- func: a python function The forward (loss) function. argnum: an int or a list of int The index of argument to calculate gradient for. Returns ------- grad_func: a python function A function t...
Splits an NDArray into `num_slice` slices along `batch_axis`. Usually used for data parallelism where each slices is sent to one device (i.e. GPU). Parameters ---------- data : NDArray A batch of data. num_slice : int Number of desired slices. batch_axis : int, default 0 ...
Splits an NDArray into `len(ctx_list)` slices along `batch_axis` and loads each slice to one context in `ctx_list`. Parameters ---------- data : NDArray A batch of data. ctx_list : list of Context A list of Contexts. batch_axis : int, default 0 The axis along which to sl...
Rescales NDArrays so that the sum of their 2-norm is smaller than `max_norm`. Parameters ---------- arrays : list of NDArray max_norm : float check_isfinite : bool, default True If True, check that the total_norm is finite (not nan or inf). This requires a blocking .asscalar() cal...
Indent string def _indent(s_, numSpaces): """Indent string """ s = s_.split('\n') if len(s) == 1: return s_ first = s.pop(0) s = [first] + [(numSpaces * ' ') + line for line in s] s = '\n'.join(s) return s
Check whether the sha1 hash of the file content matches the expected hash. Parameters ---------- filename : str Path to the file. sha1_hash : str Expected sha1 hash in hexadecimal digits. Returns ------- bool Whether the file content matches the expected hash. def ...
Download an given URL Parameters ---------- url : str URL to download path : str, optional Destination path to store downloaded file. By default stores to the current directory with same name as in url. overwrite : bool, optional Whether to overwrite destination file...
Return the base URL for Gluon dataset and model repository. def _get_repo_url(): """Return the base URL for Gluon dataset and model repository.""" default_repo = 'https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/' repo_url = os.environ.get('MXNET_GLUON_REPO', default_repo) if repo_url[-1] != ...
Return the URL for hosted file in Gluon repository. Parameters ---------- namespace : str Namespace of the file. filename : str Name of the file def _get_repo_file_url(namespace, filename): """Return the URL for hosted file in Gluon repository. Parameters ---------- na...
Print at most `limit` elements of list. def _brief_print_list(lst, limit=7): """Print at most `limit` elements of list.""" lst = list(lst) if len(lst) > limit: return _brief_print_list(lst[:limit//2], limit) + ', ..., ' + \ _brief_print_list(lst[-limit//2:], limit) return ', '.join(...
Create a symbol function by handle and function name. def _make_symbol_function(handle, name, func_name): """Create a symbol function by handle and function name.""" code, doc_str = _generate_symbol_function_code(handle, name, func_name) local = {} exec(code, None, local) # pylint: disable=exec-used ...
Generate row ids based on the current mini-batch def batch_row_ids(data_batch): """ Generate row ids based on the current mini-batch """ item = data_batch.data[0] user = data_batch.data[1] return {'user_weight': user.astype(np.int64), 'item_weight': item.astype(np.int64)}
Generate row ids for all rows def all_row_ids(data_batch): """ Generate row ids for all rows """ all_users = mx.nd.arange(0, MOVIELENS['max_user'], dtype='int64') all_movies = mx.nd.arange(0, MOVIELENS['max_movie'], dtype='int64') return {'user_weight': all_users, 'item_weight': all_movies}
Convert caffe model Parameters ---------- prototxt_fname : str Filename of the prototxt model definition caffemodel_fname : str Filename of the binary caffe model output_prefix : str, optinoal If given, then save the converted MXNet into output_prefx+'.json' and ...
Parse Caffe prototxt into symbol string def _parse_proto(prototxt_fname): """Parse Caffe prototxt into symbol string """ proto = caffe_parser.read_prototxt(prototxt_fname) # process data layer input_name, input_dim, layers = _get_input(proto) # only support single input, so always use `data` a...
Convert caffe model definition into Symbol Parameters ---------- prototxt_fname : str Filename of the prototxt file Returns ------- Symbol Converted Symbol tuple Input shape def convert_symbol(prototxt_fname): """Convert caffe model definition into Symbol ...
r"""VGG model from the `"Very Deep Convolutional Networks for Large-Scale Image Recognition" <https://arxiv.org/abs/1409.1556>`_ paper. Parameters ---------- num_layers : int Number of layers for the variant of densenet. Options are 11, 13, 16, 19. pretrained : bool, default False W...
check function consistency with uniform random numbers def check_with_uniform(uf, arg_shapes, dim=None, npuf=None, rmin=-10, type_list=[np.float32]): """check function consistency with uniform random numbers""" if isinstance(arg_shapes, int): assert dim shape = tuple(np.random.randint(1, int(10...
Remove images without usable rois def filter_roidb(self): """Remove images without usable rois""" num_roidb = len(self._roidb) self._roidb = [roi_rec for roi_rec in self._roidb if len(roi_rec['gt_classes'])] num_after = len(self._roidb) logger.info('filter roidb: {} -> {}'.forma...
Only flip boxes coordinates, images will be flipped when loading into network def append_flipped_images(self): """Only flip boxes coordinates, images will be flipped when loading into network""" logger.info('%s append flipped images to roidb' % self._name) roidb_flipped = [] for roi_rec...
r"""Return location for the pretrained on local file system. This function will download from online model zoo when model cannot be found or has mismatch. The root directory will be created if it doesn't exist. Parameters ---------- name : str Name of the model. root : str, default $MX...
r"""Purge all pretrained model files in local file store. Parameters ---------- root : str, default '$MXNET_HOME/models' Location for keeping the model parameters. def purge(root=os.path.join(base.data_dir(), 'models')): r"""Purge all pretrained model files in local file store. Parameters...
given image index, find out full path Parameters: ---------- index: int index of a specific image Returns: ---------- full path of this image def image_path_from_index(self, index): """ given image index, find out full path Parameter...

Binds the symbols to construct executors. This is necessary before one can perform computation with the module. Parameters ---------- data_shapes : list of (str, tuple) or DataDesc objects Typically is ``data_iter.provide_data``. Can also be a list of (data name,...

Find MXNet dynamic library files. Returns ------- lib_path : list(string) List of all found path to the libraries. def find_lib_path(): """Find MXNet dynamic library files. Returns ------- lib_path : list(string) List of all found path to the libraries. """ lib_fro...

Find MXNet included header files. Returns ------- incl_path : string Path to the header files. def find_include_path(): """Find MXNet included header files. Returns ------- incl_path : string Path to the header files. """ incl_from_env = os.environ.get('MXNET_INCLU...

Generate a greyscale captcha image representing number string Parameters ---------- captcha_str: str string a characters for captcha image Returns ------- numpy.ndarray Generated greyscale image in np.ndarray float type with values normalized to ...

Generates a character string of digits. Number of digits are between self.num_digit_min and self.num_digit_max Returns ------- str def get_rand(num_digit_min, num_digit_max): """Generates a character string of digits. Number of digits are between self.num_digit_min and s...

Generate a random captcha image sample Returns ------- (numpy.ndarray, str) Tuple of image (numpy ndarray) and character string of digits used to generate the image def _gen_sample(self): """Generate a random captcha image sample Returns ------- (nump...

Registers a new optimizer. Once an optimizer is registered, we can create an instance of this optimizer with `create_optimizer` later. Examples -------- >>> @mx.optimizer.Optimizer.register ... class MyOptimizer(mx.optimizer.Optimizer): ... pass >>>...

Instantiates an optimizer with a given name and kwargs. .. note:: We can use the alias `create` for ``Optimizer.create_optimizer``. Parameters ---------- name: str Name of the optimizer. Should be the name of a subclass of Optimizer. Case insensitive. k...

Creates auxiliary state for a given weight, including FP32 high precision copy if original weight is FP16. This method is provided to perform automatic mixed precision training for optimizers that do not support it themselves. Parameters ---------- index : int ...

Updates the given parameter using the corresponding gradient and state. Mixed precision version. Parameters ---------- index : int The unique index of the parameter into the individual learning rates and weight decays. Learning rates and weight decay ...

Sets an individual learning rate multiplier for each parameter. If you specify a learning rate multiplier for a parameter, then the learning rate for the parameter will be set as the product of the global learning rate `self.lr` and its multiplier. .. note:: The default learning rate m...

Sets an individual weight decay multiplier for each parameter. By default, if `param_idx2name` was provided in the constructor, the weight decay multipler is set as 0 for all parameters whose name don't end with ``_weight`` or ``_gamma``. .. note:: The default weight decay mult...

Sets the number of the currently handled device. Parameters ---------- device_id : int The number of current device. def _set_current_context(self, device_id): """Sets the number of the currently handled device. Parameters ---------- device_id : int...

Updates num_update. Parameters ---------- index : int or list of int The index to be updated. def _update_count(self, index): """Updates num_update. Parameters ---------- index : int or list of int The index to be updated. """ ...

Gets the learning rates given the indices of the weights. Parameters ---------- indices : list of int Indices corresponding to weights. Returns ------- lrs : list of float Learning rates for those indices. def _get_lrs(self, indices): ""...

Gets weight decays for indices. Returns 0 for non-weights if the name of weights are provided for `__init__`. Parameters ---------- indices : list of int Indices of weights. Returns ------- wds : list of float Weight decays for those indi...

sync state context. def sync_state_context(self, state, context): """sync state context.""" if isinstance(state, NDArray): return state.as_in_context(context) elif isinstance(state, (tuple, list)): synced_state = (self.sync_state_context(i, context) for i in state) ...

Sets updater states. def set_states(self, states): """Sets updater states.""" states = pickle.loads(states) if isinstance(states, tuple) and len(states) == 2: self.states, self.optimizer = states else: self.states = states self.states_synced = dict.fromke...

Gets updater states. Parameters ---------- dump_optimizer : bool, default False Whether to also save the optimizer itself. This would also save optimizer information such as learning rate and weight decay schedules. def get_states(self, dump_optimizer=False): ""...

Preprocess: Convert a video into the mouth images def preprocess(from_idx, to_idx, _params): """ Preprocess: Convert a video into the mouth images """ source_exts = '*.mpg' src_path = _params['src_path'] tgt_path = _params['tgt_path'] face_predictor_path = './shape_predictor_68_face_landmar...

Read from frames def from_frames(self, path): """ Read from frames """ frames_path = sorted([os.path.join(path, x) for x in os.listdir(path)]) frames = [ndimage.imread(frame_path) for frame_path in frames_path] self.handle_type(frames) return self

Read from videos def from_video(self, path): """ Read from videos """ frames = self.get_video_frames(path) self.handle_type(frames) return self

Config video types def handle_type(self, frames): """ Config video types """ if self.vtype == 'mouth': self.process_frames_mouth(frames) elif self.vtype == 'face': self.process_frames_face(frames) else: raise Exception('Video type not ...

Preprocess from frames using face detector def process_frames_face(self, frames): """ Preprocess from frames using face detector """ detector = dlib.get_frontal_face_detector() predictor = dlib.shape_predictor(self.face_predictor_path) mouth_frames = self.get_frames_mout...

Preprocess from frames using mouth detector def process_frames_mouth(self, frames): """ Preprocess from frames using mouth detector """ self.face = np.array(frames) self.mouth = np.array(frames) self.set_data(frames)

Get frames using mouth crop def get_frames_mouth(self, detector, predictor, frames): """ Get frames using mouth crop """ mouth_width = 100 mouth_height = 50 horizontal_pad = 0.19 normalize_ratio = None mouth_frames = [] for frame in frames: ...

Get video frames def get_video_frames(self, path): """ Get video frames """ videogen = skvideo.io.vreader(path) frames = np.array([frame for frame in videogen]) return frames

Prepare the input of model def set_data(self, frames): """ Prepare the input of model """ data_frames = [] for frame in frames: #frame H x W x C frame = frame.swapaxes(0, 1) # swap width and height to form format W x H x C if len(frame.shape) ...

Resets the iterator to the beginning of the data. def reset(self): """Resets the iterator to the beginning of the data.""" self.curr_idx = 0 random.shuffle(self.idx) for buck in self.data: np.random.shuffle(buck)

Returns the next batch of data. def next(self): """Returns the next batch of data.""" if self.curr_idx == len(self.idx): raise StopIteration i, j = self.idx[self.curr_idx] self.curr_idx += 1 audio_paths = [] texts = [] for duration, audio_path, text ...

Subtract ImageNet mean pixel-wise from a BGR image. def subtract_imagenet_mean_preprocess_batch(batch): """Subtract ImageNet mean pixel-wise from a BGR image.""" batch = F.swapaxes(batch,0, 1) (r, g, b) = F.split(batch, num_outputs=3, axis=0) r = r - 123.680 g = g - 116.779 b = b - 103.939 ...

Not necessary in practice def imagenet_clamp_batch(batch, low, high): """ Not necessary in practice """ F.clip(batch[:,0,:,:],low-123.680, high-123.680) F.clip(batch[:,1,:,:],low-116.779, high-116.779) F.clip(batch[:,2,:,:],low-103.939, high-103.939)

Create a linear regression network for performing SVRG optimization. :return: an instance of mx.io.NDArrayIter :return: an instance of mx.mod.svrgmodule for performing SVRG optimization def create_network(batch_size, update_freq): """Create a linear regression network for performing SVRG optimization. ...

Function to evaluate accuracy of any data iterator passed to it as an argument def evaluate_accuracy(data_iterator, net): """Function to evaluate accuracy of any data iterator passed to it as an argument""" acc = mx.metric.Accuracy() for data, label in data_iterator: output = net(data) pred...

Function responsible for running the training the model. def train(train_dir=None, train_csv=None, epochs=30, batch_size=32): """Function responsible for running the training the model.""" if not train_dir or not os.path.exists(train_dir) or not train_csv: warnings.warn("No train directory could be fo...

Set size limit on bulk execution. Bulk execution bundles many operators to run together. This can improve performance when running a lot of small operators sequentially. Parameters ---------- size : int Maximum number of operators that can be bundled in a bulk. Returns -------...

calculate LM score of child beam by taking score from parent beam and bigram probability of last two chars def applyLM(parentBeam, childBeam, classes, lm): """ calculate LM score of child beam by taking score from parent beam and bigram probability of last two chars """ if lm and not childBeam.lmApplie...

add beam if it does not yet exist def addBeam(beamState, labeling): """ add beam if it does not yet exist """ if labeling not in beamState.entries: beamState.entries[labeling] = BeamEntry()

beam search as described by the paper of Hwang et al. and the paper of Graves et al. def ctcBeamSearch(mat, classes, lm, k, beamWidth): """ beam search as described by the paper of Hwang et al. and the paper of Graves et al. """ blankIdx = len(classes) maxT, maxC = mat.shape # initialise beam...

length-normalise LM score def norm(self): """ length-normalise LM score """ for (k, _) in self.entries.items(): labelingLen = len(self.entries[k].labeling) self.entries[k].prText = self.entries[k].prText ** (1.0 / (labelingLen if labelingLen else 1.0))

return beam-labelings, sorted by probability def sort(self): """ return beam-labelings, sorted by probability """ beams = [v for (_, v) in self.entries.items()] sortedBeams = sorted(beams, reverse=True, key=lambda x: x.prTotal*x.prText) return [x.labeling for x in sorted...

the localisation network in lenet-stn, it will increase acc about more than 1%, when num-epoch >=15 def get_loc(data, attr={'lr_mult':'0.01'}): """ the localisation network in lenet-stn, it will increase acc about more than 1%, when num-epoch >=15 """ loc = mx.symbol.Convolution(data=data, num_...

wrapper for initialize a detector Parameters: ---------- net : str test network name prefix : str load model prefix epoch : int load model epoch data_shape : int resize image shape mean_pixels : tuple (float, float, float) mean pixel values (R, G, B) ...

parse # classes and class_names if applicable def parse_class_names(class_names): """ parse # classes and class_names if applicable """ if len(class_names) > 0: if os.path.isfile(class_names): # try to open it to read class names with open(class_names, 'r') as f: ...

Parse string to tuple or int def parse_data_shape(data_shape_str): """Parse string to tuple or int""" ds = data_shape_str.strip().split(',') if len(ds) == 1: data_shape = (int(ds[0]), int(ds[0])) elif len(ds) == 2: data_shape = (int(ds[0]), int(ds[1])) else: raise ValueError...

A lenet style net, takes difference of each frame as input. def get_lenet(): """ A lenet style net, takes difference of each frame as input. """ source = mx.sym.Variable("data") source = (source - 128) * (1.0/128) frames = mx.sym.SliceChannel(source, num_outputs=30) diffs = [frames[i+1] - frame...

Custom evaluation metric on CRPS. def CRPS(label, pred): """ Custom evaluation metric on CRPS. """ for i in range(pred.shape[0]): for j in range(pred.shape[1] - 1): if pred[i, j] > pred[i, j + 1]: pred[i, j + 1] = pred[i, j] return np.sum(np.square(label - pred)) / l...

Run encoding to encode the label into the CDF target. def encode_label(label_data): """Run encoding to encode the label into the CDF target. """ systole = label_data[:, 1] diastole = label_data[:, 2] systole_encode = np.array([ (x < np.arange(600)) for x in systole ], dtype=np.u...

coco ann: [u'segmentation', u'area', u'iscrowd', u'image_id', u'bbox', u'category_id', u'id'] iscrowd: crowd instances are handled by marking their overlaps with all categories to -1 and later excluded in training bbox: [x1, y1, w, h] :param index: coco image ...

example results [{"image_id": 42, "category_id": 18, "bbox": [258.15,41.29,348.26,243.78], "score": 0.236}, ...] def _write_coco_results(self, _coco, detections): """ example results [{"image_id": 42, "category_id": 18, "bbox": [258.15,41.29,348...

Draw random samples from an approximately log-uniform or Zipfian distribution. This operation randomly samples *num_sampled* candidates the range of integers [0, range_max). The elements of sampled_candidates are drawn with replacement from the base distribution. The base distribution for this operator is...

Run a for loop with user-defined computation over NDArrays on dimension 0. This operator simulates a for loop and body has the computation for an iteration of the for loop. It runs the computation in body on each slice from the input NDArrays. body takes two arguments as input and outputs a tuple of t...

Run a while loop with user-defined computation and loop condition. This operator simulates a while loop which iterately does customized computation as long as the condition is satisfied. `loop_vars` is a list of NDArrays on which the computation uses. `cond` is a user-defined function, used as the lo...

Run an if-then-else using user-defined condition and computation This operator simulates a if-like branch which chooses to do one of the two customized computations according to the specified condition. `pred` is a scalar MXNet NDArray, indicating which branch of computation should be used. `then...

Performs an element-wise check to determine if the NDArray contains an infinite element or not. Parameters ---------- input : NDArray An N-D NDArray. Returns ------- output: NDArray The output NDarray, with same shape as input, where 1 indicates the array element is ...

LSTM Cell symbol def vanilla_lstm(num_hidden, indata, prev_state, param, seqidx, layeridx, is_batchnorm=False, gamma=None, beta=None, name=None): """LSTM Cell symbol""" i2h = mx.sym.FullyConnected(data=indata, weight=param.i2h_weight, bias=param.i...

LSTM Cell symbol def lstm(num_hidden, indata, prev_state, param, seqidx, layeridx, dropout=0., num_hidden_proj=0, is_batchnorm=False, gamma=None, beta=None, name=None): """LSTM Cell symbol""" # dropout input if dropout > 0.: indata = mx.sym.Dropout(data=indata, p=dropout) i2h = mx.sym...

read, resize, transform image, return im_tensor, im_info, gt_boxes roi_rec should have keys: ["image", "boxes", "gt_classes", "flipped"] 0 --- x (width, second dim of im) | y (height, first dim of im) def get_image(roi_rec, short, max_size, mean, std): """ read, resize, transform image, return ...

Return BGR image read by opencv def imdecode(image_path): """Return BGR image read by opencv""" import os assert os.path.exists(image_path), image_path + ' not found' im = cv2.imread(image_path) return im

only resize input image to target size and return scale :param im: BGR image input by opencv :param short: one dimensional size (the short side) :param max_size: one dimensional max size (the long side) :return: resized image (NDArray) and scale (float) def resize(im, short, max_size): """ only...

transform into mxnet tensor, subtract pixel size and transform to correct format :param im: [height, width, channel] in BGR :param mean: [RGB pixel mean] :param std: [RGB pixel std var] :return: [batch, channel, height, width] def transform(im, mean, std): """ transform into mxnet tensor, ...

transform from mxnet im_tensor to ordinary RGB image im_tensor is limited to one image :param im_tensor: [batch, channel, height, width] :param mean: [RGB pixel mean] :param std: [RGB pixel std var] :return: im [height, width, channel(RGB)] def transform_inverse(im_tensor, mean, std): """ t...

vertically stack tensors by adding a new axis expand dims if only 1 tensor :param tensor_list: list of tensor to be stacked vertically :param pad: label to pad with :return: tensor with max shape def tensor_vstack(tensor_list, pad=0): """ vertically stack tensors by adding a new axis expand...

Get distance matrix given a matrix. Used in testing. def get_distance_matrix(x): """Get distance matrix given a matrix. Used in testing.""" square = nd.sum(x ** 2.0, axis=1, keepdims=True) distance_square = square + square.transpose() - (2.0 * nd.dot(x, x.transpose())) return nd.sqrt(distance_square)

Evaluate embeddings based on Recall@k. def evaluate_emb(emb, labels): """Evaluate embeddings based on Recall@k.""" d_mat = get_distance_matrix(emb) d_mat = d_mat.asnumpy() labels = labels.asnumpy() names = [] accs = [] for k in [1, 2, 4, 8, 16]: names.append('Recall@%d' % k) ...

Get learning rate based on schedule. def get_lr(lr, epoch, steps, factor): """Get learning rate based on schedule.""" for s in steps: if epoch >= s: lr *= factor return lr

Training function. def train(epochs, ctx): """Training function.""" if isinstance(ctx, mx.Context): ctx = [ctx] net.initialize(mx.init.Xavier(magnitude=2), ctx=ctx) opt_options = {'learning_rate': opt.lr, 'wd': opt.wd} if opt.optimizer == 'sgd': opt_options['momentum'] = 0.9 if...

Returns symbol for LSTM model up to loss/softmax def _lstm_unroll_base(num_lstm_layer, seq_len, num_hidden): """ Returns symbol for LSTM model up to loss/softmax""" param_cells = [] last_states = [] for i in range(num_lstm_layer): param_cells.append(LSTMParam(i2h_weight=mx.sym.Variable("l%d_i2h...

Adds Symbol.contrib.ctc_loss on top of pred symbol and returns the resulting symbol def _add_warp_ctc_loss(pred, seq_len, num_label, label): """ Adds Symbol.contrib.ctc_loss on top of pred symbol and returns the resulting symbol """ label = mx.sym.Reshape(data=label, shape=(-1,)) label = mx.sym.Cast(data=l...

Adds Symbol.WapCTC on top of pred symbol and returns the resulting symbol def _add_mxnet_ctc_loss(pred, seq_len, label): """ Adds Symbol.WapCTC on top of pred symbol and returns the resulting symbol """ pred_ctc = mx.sym.Reshape(data=pred, shape=(-4, seq_len, -1, 0)) loss = mx.sym.contrib.ctc_loss(data=pr...

Adds CTC loss on top of pred symbol and returns the resulting symbol def _add_ctc_loss(pred, seq_len, num_label, loss_type): """ Adds CTC loss on top of pred symbol and returns the resulting symbol """ label = mx.sym.Variable('label') if loss_type == 'warpctc': print("Using WarpCTC Loss") s...

Creates an unrolled LSTM symbol for inference if loss_type is not specified, and for training if loss_type is specified. loss_type must be one of 'ctc' or 'warpctc' Parameters ---------- num_lstm_layer: int seq_len: int num_hidden: int num_label: int loss_type: str 'ctc' or 'war...

Returns name and shape of init states of LSTM network Parameters ---------- batch_size: list of tuple of str and tuple of int and int num_lstm_layer: int num_hidden: int Returns ------- list of tuple of str and tuple of int and int def init_states(batch_size, num_lstm_layer, num_hidde...

ctypes implementation of imperative invoke wrapper def _imperative_invoke(handle, ndargs, keys, vals, out): """ctypes implementation of imperative invoke wrapper""" if out is not None: original_output = out if isinstance(out, NDArrayBase): out = (out,) num_output = ctypes.c_...

Set status to training/not training. When training, graph will be constructed for gradient computation. Operators will also run with ctx.is_train=True. For example, Dropout will drop inputs randomly when is_train=True while simply passing through if is_train=False. Parameters ---------- is_trai...

Compute the gradients of outputs w.r.t variables. Parameters ---------- outputs: list of NDArray out_grads: list of NDArray or None def backward(outputs, out_grads=None, retain_graph=False): """Compute the gradients of outputs w.r.t variables. Parameters ---------- outputs: list of ND...

Return function that computes both gradient of arguments and loss value. Parameters ---------- func: a python function The forward (loss) function. argnum: an int or a list of int The index of argument to calculate gradient for. Returns ------- grad_and_loss_func: a python ...

Return function that computes gradient of arguments. Parameters ---------- func: a python function The forward (loss) function. argnum: an int or a list of int The index of argument to calculate gradient for. Returns ------- grad_func: a python function A function t...

Splits an NDArray into `num_slice` slices along `batch_axis`. Usually used for data parallelism where each slices is sent to one device (i.e. GPU). Parameters ---------- data : NDArray A batch of data. num_slice : int Number of desired slices. batch_axis : int, default 0 ...

Splits an NDArray into `len(ctx_list)` slices along `batch_axis` and loads each slice to one context in `ctx_list`. Parameters ---------- data : NDArray A batch of data. ctx_list : list of Context A list of Contexts. batch_axis : int, default 0 The axis along which to sl...

Rescales NDArrays so that the sum of their 2-norm is smaller than `max_norm`. Parameters ---------- arrays : list of NDArray max_norm : float check_isfinite : bool, default True If True, check that the total_norm is finite (not nan or inf). This requires a blocking .asscalar() cal...

Indent string def _indent(s_, numSpaces): """Indent string """ s = s_.split('\n') if len(s) == 1: return s_ first = s.pop(0) s = [first] + [(numSpaces * ' ') + line for line in s] s = '\n'.join(s) return s

Check whether the sha1 hash of the file content matches the expected hash. Parameters ---------- filename : str Path to the file. sha1_hash : str Expected sha1 hash in hexadecimal digits. Returns ------- bool Whether the file content matches the expected hash. def ...

Download an given URL Parameters ---------- url : str URL to download path : str, optional Destination path to store downloaded file. By default stores to the current directory with same name as in url. overwrite : bool, optional Whether to overwrite destination file...

Return the base URL for Gluon dataset and model repository. def _get_repo_url(): """Return the base URL for Gluon dataset and model repository.""" default_repo = 'https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/' repo_url = os.environ.get('MXNET_GLUON_REPO', default_repo) if repo_url[-1] != ...

Return the URL for hosted file in Gluon repository. Parameters ---------- namespace : str Namespace of the file. filename : str Name of the file def _get_repo_file_url(namespace, filename): """Return the URL for hosted file in Gluon repository. Parameters ---------- na...

Print at most `limit` elements of list. def _brief_print_list(lst, limit=7): """Print at most `limit` elements of list.""" lst = list(lst) if len(lst) > limit: return _brief_print_list(lst[:limit//2], limit) + ', ..., ' + \ _brief_print_list(lst[-limit//2:], limit) return ', '.join(...

Create a symbol function by handle and function name. def _make_symbol_function(handle, name, func_name): """Create a symbol function by handle and function name.""" code, doc_str = _generate_symbol_function_code(handle, name, func_name) local = {} exec(code, None, local) # pylint: disable=exec-used ...

Generate row ids based on the current mini-batch def batch_row_ids(data_batch): """ Generate row ids based on the current mini-batch """ item = data_batch.data[0] user = data_batch.data[1] return {'user_weight': user.astype(np.int64), 'item_weight': item.astype(np.int64)}

Generate row ids for all rows def all_row_ids(data_batch): """ Generate row ids for all rows """ all_users = mx.nd.arange(0, MOVIELENS['max_user'], dtype='int64') all_movies = mx.nd.arange(0, MOVIELENS['max_movie'], dtype='int64') return {'user_weight': all_users, 'item_weight': all_movies}

Convert caffe model Parameters ---------- prototxt_fname : str Filename of the prototxt model definition caffemodel_fname : str Filename of the binary caffe model output_prefix : str, optinoal If given, then save the converted MXNet into output_prefx+'.json' and ...

Parse Caffe prototxt into symbol string def _parse_proto(prototxt_fname): """Parse Caffe prototxt into symbol string """ proto = caffe_parser.read_prototxt(prototxt_fname) # process data layer input_name, input_dim, layers = _get_input(proto) # only support single input, so always use `data` a...

Convert caffe model definition into Symbol Parameters ---------- prototxt_fname : str Filename of the prototxt file Returns ------- Symbol Converted Symbol tuple Input shape def convert_symbol(prototxt_fname): """Convert caffe model definition into Symbol ...

r"""VGG model from the `"Very Deep Convolutional Networks for Large-Scale Image Recognition" <https://arxiv.org/abs/1409.1556>`_ paper. Parameters ---------- num_layers : int Number of layers for the variant of densenet. Options are 11, 13, 16, 19. pretrained : bool, default False W...

check function consistency with uniform random numbers def check_with_uniform(uf, arg_shapes, dim=None, npuf=None, rmin=-10, type_list=[np.float32]): """check function consistency with uniform random numbers""" if isinstance(arg_shapes, int): assert dim shape = tuple(np.random.randint(1, int(10...

Remove images without usable rois def filter_roidb(self): """Remove images without usable rois""" num_roidb = len(self._roidb) self._roidb = [roi_rec for roi_rec in self._roidb if len(roi_rec['gt_classes'])] num_after = len(self._roidb) logger.info('filter roidb: {} -> {}'.forma...

Only flip boxes coordinates, images will be flipped when loading into network def append_flipped_images(self): """Only flip boxes coordinates, images will be flipped when loading into network""" logger.info('%s append flipped images to roidb' % self._name) roidb_flipped = [] for roi_rec...

r"""Return location for the pretrained on local file system. This function will download from online model zoo when model cannot be found or has mismatch. The root directory will be created if it doesn't exist. Parameters ---------- name : str Name of the model. root : str, default $MX...

r"""Purge all pretrained model files in local file store. Parameters ---------- root : str, default '$MXNET_HOME/models' Location for keeping the model parameters. def purge(root=os.path.join(base.data_dir(), 'models')): r"""Purge all pretrained model files in local file store. Parameters...

given image index, find out full path Parameters: ---------- index: int index of a specific image Returns: ---------- full path of this image def image_path_from_index(self, index): """ given image index, find out full path Parameter...