INSTRUCTION
stringlengths
1
46.3k
RESPONSE
stringlengths
75
80.2k
Helper function to obtain the shape of an array
def get_input_shape(sym, proto_obj): """Helper function to obtain the shape of an array""" arg_params = proto_obj.arg_dict aux_params = proto_obj.aux_dict model_input_shape = [data[1] for data in proto_obj.model_metadata.get('input_tensor_data')] data_names = [data[0] for data in proto_obj.model_...
r"""Resize image with OpenCV. .. note:: `imresize` uses OpenCV (not the CV2 Python library). MXNet must have been built with USE_OPENCV=1 for `imresize` to work. Parameters ---------- src : NDArray source image w : int, required Width of resized image. h : int, required ...
def imresize(src, w, h, *args, **kwargs): r"""Resize image with OpenCV. .. note:: `imresize` uses OpenCV (not the CV2 Python library). MXNet must have been built with USE_OPENCV=1 for `imresize` to work. Parameters ---------- src : NDArray source image w : int, required ...
Decode an image to an NDArray. .. note:: `imdecode` uses OpenCV (not the CV2 Python library). MXNet must have been built with USE_OPENCV=1 for `imdecode` to work. Parameters ---------- buf : str/bytes/bytearray or numpy.ndarray Binary image data as string or numpy ndarray. flag : in...
def imdecode(buf, *args, **kwargs): """Decode an image to an NDArray. .. note:: `imdecode` uses OpenCV (not the CV2 Python library). MXNet must have been built with USE_OPENCV=1 for `imdecode` to work. Parameters ---------- buf : str/bytes/bytearray or numpy.ndarray Binary image dat...
Scales down crop size if it's larger than image size. If width/height of the crop is larger than the width/height of the image, sets the width/height to the width/height of the image. Parameters ---------- src_size : tuple of int Size of the image in (width, height) format. size : tupl...
def scale_down(src_size, size): """Scales down crop size if it's larger than image size. If width/height of the crop is larger than the width/height of the image, sets the width/height to the width/height of the image. Parameters ---------- src_size : tuple of int Size of the image in ...
Pad image border with OpenCV. Parameters ---------- src : NDArray source image top : int, required Top margin. bot : int, required Bottom margin. left : int, required Left margin. right : int, required Right margin. type : int, optional, default='...
def copyMakeBorder(src, top, bot, left, right, *args, **kwargs): """Pad image border with OpenCV. Parameters ---------- src : NDArray source image top : int, required Top margin. bot : int, required Bottom margin. left : int, required Left margin. right :...
Get the interpolation method for resize functions. The major purpose of this function is to wrap a random interp method selection and a auto-estimation method. Parameters ---------- interp : int interpolation method for all resizing operations Possible values: 0: Nearest Ne...
def _get_interp_method(interp, sizes=()): """Get the interpolation method for resize functions. The major purpose of this function is to wrap a random interp method selection and a auto-estimation method. Parameters ---------- interp : int interpolation method for all resizing operation...
Resizes shorter edge to size. .. note:: `resize_short` uses OpenCV (not the CV2 Python library). MXNet must have been built with OpenCV for `resize_short` to work. Resizes the original image by setting the shorter edge to size and setting the longer edge accordingly. Resizing function is called...
def resize_short(src, size, interp=2): """Resizes shorter edge to size. .. note:: `resize_short` uses OpenCV (not the CV2 Python library). MXNet must have been built with OpenCV for `resize_short` to work. Resizes the original image by setting the shorter edge to size and setting the longer edg...
Crop src at fixed location, and (optionally) resize it to size. Parameters ---------- src : NDArray Input image x0 : int Left boundary of the cropping area y0 : int Top boundary of the cropping area w : int Width of the cropping area h : int Height of...
def fixed_crop(src, x0, y0, w, h, size=None, interp=2): """Crop src at fixed location, and (optionally) resize it to size. Parameters ---------- src : NDArray Input image x0 : int Left boundary of the cropping area y0 : int Top boundary of the cropping area w : int ...
Crops the image `src` to the given `size` by trimming on all four sides and preserving the center of the image. Upsamples if `src` is smaller than `size`. .. note:: This requires MXNet to be compiled with USE_OPENCV. Parameters ---------- src : NDArray Binary source image data. siz...
def center_crop(src, size, interp=2): """Crops the image `src` to the given `size` by trimming on all four sides and preserving the center of the image. Upsamples if `src` is smaller than `size`. .. note:: This requires MXNet to be compiled with USE_OPENCV. Parameters ---------- src : NDAr...
Normalize src with mean and std. Parameters ---------- src : NDArray Input image mean : NDArray RGB mean to be subtracted std : NDArray RGB standard deviation to be divided Returns ------- NDArray An `NDArray` containing the normalized image.
def color_normalize(src, mean, std=None): """Normalize src with mean and std. Parameters ---------- src : NDArray Input image mean : NDArray RGB mean to be subtracted std : NDArray RGB standard deviation to be divided Returns ------- NDArray An `NDAr...
Randomly crop src with size. Randomize area and aspect ratio. Parameters ---------- src : NDArray Input image size : tuple of (int, int) Size of the crop formatted as (width, height). area : float in (0, 1] or tuple of (float, float) If tuple, minimum area and maximum area t...
def random_size_crop(src, size, area, ratio, interp=2, **kwargs): """Randomly crop src with size. Randomize area and aspect ratio. Parameters ---------- src : NDArray Input image size : tuple of (int, int) Size of the crop formatted as (width, height). area : float in (0, 1] or ...
Creates an augmenter list. Parameters ---------- data_shape : tuple of int Shape for output data resize : int Resize shorter edge if larger than 0 at the begining rand_crop : bool Whether to enable random cropping other than center crop rand_resize : bool Whether...
def CreateAugmenter(data_shape, resize=0, rand_crop=False, rand_resize=False, rand_mirror=False, mean=None, std=None, brightness=0, contrast=0, saturation=0, hue=0, pca_noise=0, rand_gray=0, inter_method=2): """Creates an augmenter list. Parameters ---------- dat...
Saves the Augmenter to string Returns ------- str JSON formatted string that describes the Augmenter.
def dumps(self): """Saves the Augmenter to string Returns ------- str JSON formatted string that describes the Augmenter. """ return json.dumps([self.__class__.__name__.lower(), self._kwargs])
Override the default to avoid duplicate dump.
def dumps(self): """Override the default to avoid duplicate dump.""" return [self.__class__.__name__.lower(), [x.dumps() for x in self.ts]]
Resets the iterator to the beginning of the data.
def reset(self): """Resets the iterator to the beginning of the data.""" if self.seq is not None and self.shuffle: random.shuffle(self.seq) if self.last_batch_handle != 'roll_over' or \ self._cache_data is None: if self.imgrec is not None: self...
Resets the iterator and ignore roll over data
def hard_reset(self): """Resets the iterator and ignore roll over data""" if self.seq is not None and self.shuffle: random.shuffle(self.seq) if self.imgrec is not None: self.imgrec.reset() self.cur = 0 self._allow_read = True self._cache_data = Non...
Helper function for reading in next sample.
def next_sample(self): """Helper function for reading in next sample.""" if self._allow_read is False: raise StopIteration if self.seq is not None: if self.cur < self.num_image: idx = self.seq[self.cur] else: if self.last_batch_...
Helper function for batchifying data
def _batchify(self, batch_data, batch_label, start=0): """Helper function for batchifying data""" i = start batch_size = self.batch_size try: while i < batch_size: label, s = self.next_sample() data = self.imdecode(s) try: ...
Decodes a string or byte string to an NDArray. See mx.img.imdecode for more details.
def imdecode(self, s): """Decodes a string or byte string to an NDArray. See mx.img.imdecode for more details.""" def locate(): """Locate the image file/index if decode fails.""" if self.seq is not None: idx = self.seq[(self.cur % self.num_image) - 1] ...
Reads an input image `fname` and returns the decoded raw bytes. Examples -------- >>> dataIter.read_image('Face.jpg') # returns decoded raw bytes.
def read_image(self, fname): """Reads an input image `fname` and returns the decoded raw bytes. Examples -------- >>> dataIter.read_image('Face.jpg') # returns decoded raw bytes. """ with open(os.path.join(self.path_root, fname), 'rb') as fin: img = fin.read()...
evaluate accuracy
def facc(label, pred): """ evaluate accuracy """ pred = pred.ravel() label = label.ravel() return ((pred > 0.5) == label).mean()
Convert character vectors to integer vectors.
def word_to_vector(word): """ Convert character vectors to integer vectors. """ vector = [] for char in list(word): vector.append(char2int(char)) return vector
Convert integer vectors to character vectors.
def vector_to_word(vector): """ Convert integer vectors to character vectors. """ word = "" for vec in vector: word = word + int2char(vec) return word
Convert integer vectors to character vectors for batch.
def char_conv(out): """ Convert integer vectors to character vectors for batch. """ out_conv = list() for i in range(out.shape[0]): tmp_str = '' for j in range(out.shape[1]): if int(out[i][j]) >= 0: tmp_char = int2char(int(out[i][j])) if in...
Add a pooling layer to the model. This is our own implementation of add_pooling since current CoreML's version (0.5.0) of builder doesn't provide support for padding types apart from valid. This support will be added in the next release of coremltools. When that happens, this can be removed. Par...
def add_pooling_with_padding_types(builder, name, height, width, stride_height, stride_width, layer_type, padding_type, input_name, output_name, padding_top = 0, padding_bottom = 0, padding_left = 0, padding_right = 0, same_padding_asymmetry_mode = 'BOTTOM_RIGHT_HEAVY', exclude_pad_a...
Get path to all the frame in view SAX and contain complete frames
def get_frames(root_path): """Get path to all the frame in view SAX and contain complete frames""" ret = [] for root, _, files in os.walk(root_path): root=root.replace('\\','/') files=[s for s in files if ".dcm" in s] if len(files) == 0 or not files[0].endswith(".dcm") or root.find("sax") ...
crop center and resize
def crop_resize(img, size): """crop center and resize""" if img.shape[0] < img.shape[1]: img = img.T # we crop image from center short_egde = min(img.shape[:2]) yy = int((img.shape[0] - short_egde) / 2) xx = int((img.shape[1] - short_egde) / 2) crop_img = img[yy : yy + short_egde, xx : xx + ...
construct and return generator
def get_generator(): """ construct and return generator """ g_net = gluon.nn.Sequential() with g_net.name_scope(): g_net.add(gluon.nn.Conv2DTranspose( channels=512, kernel_size=4, strides=1, padding=0, use_bias=False)) g_net.add(gluon.nn.BatchNorm()) g_net.add(gluon.nn.L...
construct and return descriptor
def get_descriptor(ctx): """ construct and return descriptor """ d_net = gluon.nn.Sequential() with d_net.name_scope(): d_net.add(SNConv2D(num_filter=64, kernel_size=4, strides=2, padding=1, in_channels=3, ctx=ctx)) d_net.add(gluon.nn.LeakyReLU(0.2)) d_net.add(SNConv2D(num_filter=1...
spectral normalization
def _spectral_norm(self): """ spectral normalization """ w = self.params.get('weight').data(self.ctx) w_mat = nd.reshape(w, [w.shape[0], -1]) _u = self.u.data(self.ctx) _v = None for _ in range(POWER_ITERATION): _v = nd.L2Normalization(nd.dot(_u, w_mat)) ...
Compute the length of the output sequence after 1D convolution along time. Note that this function is in line with the function used in Convolution1D class from Keras. Params: input_length (int): Length of the input sequence. filter_size (int): Width of the convolution kernel. ...
def conv_output_length(input_length, filter_size, border_mode, stride, dilation=1): """ Compute the length of the output sequence after 1D convolution along time. Note that this function is in line with the function used in Convolution1D class from Keras. Params: i...
Compute the spectrogram for a real signal. The parameters follow the naming convention of matplotlib.mlab.specgram Args: samples (1D array): input audio signal fft_length (int): number of elements in fft window sample_rate (scalar): sample rate hop_length (int): hop length (r...
def spectrogram(samples, fft_length=256, sample_rate=2, hop_length=128): """ Compute the spectrogram for a real signal. The parameters follow the naming convention of matplotlib.mlab.specgram Args: samples (1D array): input audio signal fft_length (int): number of elements in fft win...
Calculate the log of linear spectrogram from FFT energy Params: filename (str): Path to the audio file step (int): Step size in milliseconds between windows window (int): FFT window size in milliseconds max_freq (int): Only FFT bins corresponding to frequencies between [0...
def spectrogram_from_file(filename, step=10, window=20, max_freq=None, eps=1e-14, overwrite=False, save_feature_as_csvfile=False): """ Calculate the log of linear spectrogram from FFT energy Params: filename (str): Path to the audio file step (int): Step size in millise...
generate random cropping boxes according to parameters if satifactory crops generated, apply to ground-truth as well Parameters: ---------- label : numpy.array (n x 5 matrix) ground-truths Returns: ---------- list of (crop_box, label) tuples, if fail...
def sample(self, label): """ generate random cropping boxes according to parameters if satifactory crops generated, apply to ground-truth as well Parameters: ---------- label : numpy.array (n x 5 matrix) ground-truths Returns: ---------- ...
check if overlap with any gt box is larger than threshold
def _check_satisfy(self, rand_box, gt_boxes): """ check if overlap with any gt box is larger than threshold """ l, t, r, b = rand_box num_gt = gt_boxes.shape[0] ls = np.ones(num_gt) * l ts = np.ones(num_gt) * t rs = np.ones(num_gt) * r bs = np.ones...
generate random padding boxes according to parameters if satifactory padding generated, apply to ground-truth as well Parameters: ---------- label : numpy.array (n x 5 matrix) ground-truths Returns: ---------- list of (crop_box, label) tuples, if fai...
def sample(self, label): """ generate random padding boxes according to parameters if satifactory padding generated, apply to ground-truth as well Parameters: ---------- label : numpy.array (n x 5 matrix) ground-truths Returns: ---------- ...
Measure time cost of running a function
def measure_cost(repeat, scipy_trans_lhs, scipy_dns_lhs, func_name, *args, **kwargs): """Measure time cost of running a function """ mx.nd.waitall() args_list = [] for arg in args: args_list.append(arg) start = time.time() if scipy_trans_lhs: args_list[0] = np.transpose(args_...
Print information about the annotation file. :return:
def info(self): """ Print information about the annotation file. :return: """ for key, value in self.dataset['info'].items(): print('{}: {}'.format(key, value))
filtering parameters. default skips that filter. :param catNms (str array) : get cats for given cat names :param supNms (str array) : get cats for given supercategory names :param catIds (int array) : get cats for given cat ids :return: ids (int array) : integer array of cat ids
def getCatIds(self, catNms=[], supNms=[], catIds=[]): """ filtering parameters. default skips that filter. :param catNms (str array) : get cats for given cat names :param supNms (str array) : get cats for given supercategory names :param catIds (int array) : get cats for given...
Load anns with the specified ids. :param ids (int array) : integer ids specifying anns :return: anns (object array) : loaded ann objects
def loadAnns(self, ids=[]): """ Load anns with the specified ids. :param ids (int array) : integer ids specifying anns :return: anns (object array) : loaded ann objects """ if type(ids) == list: return [self.anns[id] for id in ids] elif type(ids)...
Load cats with the specified ids. :param ids (int array) : integer ids specifying cats :return: cats (object array) : loaded cat objects
def loadCats(self, ids=[]): """ Load cats with the specified ids. :param ids (int array) : integer ids specifying cats :return: cats (object array) : loaded cat objects """ if type(ids) == list: return [self.cats[id] for id in ids] elif type(ids)...
Load anns with the specified ids. :param ids (int array) : integer ids specifying img :return: imgs (object array) : loaded img objects
def loadImgs(self, ids=[]): """ Load anns with the specified ids. :param ids (int array) : integer ids specifying img :return: imgs (object array) : loaded img objects """ if type(ids) == list: return [self.imgs[id] for id in ids] elif type(ids) ...
Display the specified annotations. :param anns (array of object): annotations to display :return: None
def showAnns(self, anns): """ Display the specified annotations. :param anns (array of object): annotations to display :return: None """ if len(anns) == 0: return 0 if 'segmentation' in anns[0] or 'keypoints' in anns[0]: datasetType = 'inst...
Download COCO images from mscoco.org server. :param tarDir (str): COCO results directory name imgIds (list): images to be downloaded :return:
def download(self, tarDir = None, imgIds = [] ): ''' Download COCO images from mscoco.org server. :param tarDir (str): COCO results directory name imgIds (list): images to be downloaded :return: ''' if tarDir is None: print('Please specify targe...
Convert result data from a numpy array [Nx7] where each row contains {imageID,x1,y1,w,h,score,class} :param data (numpy.ndarray) :return: annotations (python nested list)
def loadNumpyAnnotations(self, data): """ Convert result data from a numpy array [Nx7] where each row contains {imageID,x1,y1,w,h,score,class} :param data (numpy.ndarray) :return: annotations (python nested list) """ print('Converting ndarray to lists...') assert...
Convert annotation which can be polygons, uncompressed RLE to RLE. :return: binary mask (numpy 2D array)
def annToRLE(self, ann): """ Convert annotation which can be polygons, uncompressed RLE to RLE. :return: binary mask (numpy 2D array) """ t = self.imgs[ann['image_id']] h, w = t['height'], t['width'] segm = ann['segmentation'] if type(segm) == list: ...
Save cnn model Returns ---------- callback: A callback function that can be passed as epoch_end_callback to fit
def save_model(): """Save cnn model Returns ---------- callback: A callback function that can be passed as epoch_end_callback to fit """ if not os.path.exists("checkpoint"): os.mkdir("checkpoint") return mx.callback.do_checkpoint("checkpoint/checkpoint", args.save_period)
Construct highway net Parameters ---------- data: Returns ---------- Highway Networks
def highway(data): """Construct highway net Parameters ---------- data: Returns ---------- Highway Networks """ _data = data high_weight = mx.sym.Variable('high_weight') high_bias = mx.sym.Variable('high_bias') high_fc = mx.sym.FullyConnected(data=data, weight=high_weight...
Train cnn model Parameters ---------- symbol_data: symbol train_iterator: DataIter Train DataIter valid_iterator: DataIter Valid DataIter data_column_names: list of str Defaults to ('data') for a typical model used in image classific...
def train(symbol_data, train_iterator, valid_iterator, data_column_names, target_names): """Train cnn model Parameters ---------- symbol_data: symbol train_iterator: DataIter Train DataIter valid_iterator: DataIter Valid DataIter data_column_names: lis...
Collate data into batch.
def default_batchify_fn(data): """Collate data into batch.""" if isinstance(data[0], nd.NDArray): return nd.stack(*data) elif isinstance(data[0], tuple): data = zip(*data) return [default_batchify_fn(i) for i in data] else: data = np.asarray(data) return nd.array(...
Collate data into batch. Use shared memory for stacking.
def default_mp_batchify_fn(data): """Collate data into batch. Use shared memory for stacking.""" if isinstance(data[0], nd.NDArray): out = nd.empty((len(data),) + data[0].shape, dtype=data[0].dtype, ctx=context.Context('cpu_shared', 0)) return nd.stack(*data, out=out) ...
Move data into new context.
def _as_in_context(data, ctx): """Move data into new context.""" if isinstance(data, nd.NDArray): return data.as_in_context(ctx) elif isinstance(data, (list, tuple)): return [_as_in_context(d, ctx) for d in data] return data
Worker loop for multiprocessing DataLoader.
def worker_loop_v1(dataset, key_queue, data_queue, batchify_fn): """Worker loop for multiprocessing DataLoader.""" while True: idx, samples = key_queue.get() if idx is None: break batch = batchify_fn([dataset[i] for i in samples]) data_queue.put((idx, batch))
Fetcher loop for fetching data from queue and put in reorder dict.
def fetcher_loop_v1(data_queue, data_buffer, pin_memory=False, pin_device_id=0, data_buffer_lock=None): """Fetcher loop for fetching data from queue and put in reorder dict.""" while True: idx, batch = data_queue.get() if idx is None: break if pin_memory: ...
Function for processing data in worker process.
def _worker_fn(samples, batchify_fn, dataset=None): """Function for processing data in worker process.""" # pylint: disable=unused-argument # it is required that each worker process has to fork a new MXIndexedRecordIO handle # preserving dataset as global variable can save tons of overhead and is safe i...
Send object
def send(self, obj): """Send object""" buf = io.BytesIO() ForkingPickler(buf, pickle.HIGHEST_PROTOCOL).dump(obj) self.send_bytes(buf.getvalue())
Assign next batch workload to workers.
def _push_next(self): """Assign next batch workload to workers.""" r = next(self._iter, None) if r is None: return self._key_queue.put((self._sent_idx, r)) self._sent_idx += 1
Shutdown internal workers by pushing terminate signals.
def shutdown(self): """Shutdown internal workers by pushing terminate signals.""" if not self._shutdown: # send shutdown signal to the fetcher and join data queue first # Remark: loop_fetcher need to be joined prior to the workers. # otherwise, the the fet...
Assign next batch workload to workers.
def _push_next(self): """Assign next batch workload to workers.""" r = next(self._iter, None) if r is None: return async_ret = self._worker_pool.apply_async( self._worker_fn, (r, self._batchify_fn, self._dataset)) self._data_buffer[self._sent_idx] = async_...
Returns ctype arrays for the key-value args, and the whether string keys are used. For internal use only.
def _ctype_key_value(keys, vals): """ Returns ctype arrays for the key-value args, and the whether string keys are used. For internal use only. """ if isinstance(keys, (tuple, list)): assert(len(keys) == len(vals)) c_keys = [] c_vals = [] use_str_keys = None f...
Returns ctype arrays for keys and values(converted to strings) in a dictionary
def _ctype_dict(param_dict): """ Returns ctype arrays for keys and values(converted to strings) in a dictionary """ assert(isinstance(param_dict, dict)), \ "unexpected type for param_dict: " + str(type(param_dict)) c_keys = c_array(ctypes.c_char_p, [c_str(k) for k in param_dict.keys()]) ...
A wrapper for the user-defined handle.
def _updater_wrapper(updater): """A wrapper for the user-defined handle.""" def updater_handle(key, lhs_handle, rhs_handle, _): """ ctypes function """ lhs = _ndarray_cls(NDArrayHandle(lhs_handle)) rhs = _ndarray_cls(NDArrayHandle(rhs_handle)) updater(key, lhs, rhs) return up...
Creates a new KVStore. For single machine training, there are two commonly used types: ``local``: Copies all gradients to CPU memory and updates weights there. ``device``: Aggregates gradients and updates weights on GPUs. With this setting, the KVStore also attempts to use GPU peer-to-peer communicat...
def create(name='local'): """Creates a new KVStore. For single machine training, there are two commonly used types: ``local``: Copies all gradients to CPU memory and updates weights there. ``device``: Aggregates gradients and updates weights on GPUs. With this setting, the KVStore also attempts t...
Initializes a single or a sequence of key-value pairs into the store. For each key, one must `init` it before calling `push` or `pull`. When multiple workers invoke `init` for the same key, only the value supplied by worker with rank `0` is used. This function returns after data has bee...
def init(self, key, value): """ Initializes a single or a sequence of key-value pairs into the store. For each key, one must `init` it before calling `push` or `pull`. When multiple workers invoke `init` for the same key, only the value supplied by worker with rank `0` is used. This fun...
Pushes a single or a sequence of key-value pairs into the store. This function returns immediately after adding an operator to the engine. The actual operation is executed asynchronously. If there are consecutive pushes to the same key, there is no guarantee on the serialization of pushes. ...
def push(self, key, value, priority=0): """ Pushes a single or a sequence of key-value pairs into the store. This function returns immediately after adding an operator to the engine. The actual operation is executed asynchronously. If there are consecutive pushes to the same key, there ...
Pulls a single value or a sequence of values from the store. This function returns immediately after adding an operator to the engine. Subsequent attempts to read from the `out` variable will be blocked until the pull operation completes. `pull` is executed asynchronously after all pre...
def pull(self, key, out=None, priority=0, ignore_sparse=True): """ Pulls a single value or a sequence of values from the store. This function returns immediately after adding an operator to the engine. Subsequent attempts to read from the `out` variable will be blocked until the pull op...
Pulls a single RowSparseNDArray value or a sequence of RowSparseNDArray values \ from the store with specified row_ids. When there is only one row_id, KVStoreRowSparsePull \ is invoked just once and the result is broadcast to all the rest of outputs. `row_sparse_pull` is executed asynchronously...
def row_sparse_pull(self, key, out=None, priority=0, row_ids=None): """ Pulls a single RowSparseNDArray value or a sequence of RowSparseNDArray values \ from the store with specified row_ids. When there is only one row_id, KVStoreRowSparsePull \ is invoked just once and the result is broadcast t...
Specifies type of low-bit quantization for gradient compression \ and additional arguments depending on the type of compression being used. 2bit Gradient Compression takes a positive float `threshold`. The technique works by thresholding values such that positive values in the gradient...
def set_gradient_compression(self, compression_params): """ Specifies type of low-bit quantization for gradient compression \ and additional arguments depending on the type of compression being used. 2bit Gradient Compression takes a positive float `threshold`. The technique works by t...
Registers an optimizer with the kvstore. When using a single machine, this function updates the local optimizer. If using multiple machines and this operation is invoked from a worker node, it will serialized the optimizer with pickle and send it to all servers. The function returns aft...
def set_optimizer(self, optimizer): """ Registers an optimizer with the kvstore. When using a single machine, this function updates the local optimizer. If using multiple machines and this operation is invoked from a worker node, it will serialized the optimizer with pickle and send it ...
Returns the type of this kvstore. Returns ------- type : str the string type
def type(self): """ Returns the type of this kvstore. Returns ------- type : str the string type """ kv_type = ctypes.c_char_p() check_call(_LIB.MXKVStoreGetType(self.handle, ctypes.byref(kv_type))) return py_str(kv_type.value)
Returns the rank of this worker node. Returns ------- rank : int The rank of this node, which is in range [0, num_workers())
def rank(self): """ Returns the rank of this worker node. Returns ------- rank : int The rank of this node, which is in range [0, num_workers()) """ rank = ctypes.c_int() check_call(_LIB.MXKVStoreGetRank(self.handle, ctypes.byref(rank))) retur...
Returns the number of worker nodes. Returns ------- size :int The number of worker nodes.
def num_workers(self): """Returns the number of worker nodes. Returns ------- size :int The number of worker nodes. """ size = ctypes.c_int() check_call(_LIB.MXKVStoreGetGroupSize(self.handle, ctypes.byref(size))) return size.value
Saves the optimizer (updater) state to a file. This is often used when checkpointing the model during training. Parameters ---------- fname : str Path to the output states file. dump_optimizer : bool, default False Whether to also save the optimizer itsel...
def save_optimizer_states(self, fname, dump_optimizer=False): """Saves the optimizer (updater) state to a file. This is often used when checkpointing the model during training. Parameters ---------- fname : str Path to the output states file. dump_optimizer :...
Loads the optimizer (updater) state from the file. Parameters ---------- fname : str Path to input states file.
def load_optimizer_states(self, fname): """Loads the optimizer (updater) state from the file. Parameters ---------- fname : str Path to input states file. """ assert self._updater is not None, "Cannot load states for distributed training" self._update...
Sets a push updater into the store. This function only changes the local store. When running on multiple machines one must use `set_optimizer`. Parameters ---------- updater : function The updater function. Examples -------- >>> def update(k...
def _set_updater(self, updater): """Sets a push updater into the store. This function only changes the local store. When running on multiple machines one must use `set_optimizer`. Parameters ---------- updater : function The updater function. Exampl...
Sends a command to all server nodes. Sending command to a server node will cause that server node to invoke ``KVStoreServer.controller`` to execute the command. This function returns after the command has been executed on all server nodes. Parameters ---------- ...
def _send_command_to_servers(self, head, body): """Sends a command to all server nodes. Sending command to a server node will cause that server node to invoke ``KVStoreServer.controller`` to execute the command. This function returns after the command has been executed on all server ...
Add a module to the chain. Parameters ---------- module : BaseModule The new module to add. kwargs : ``**keywords`` All the keyword arguments are saved as meta information for the added module. The currently known meta includes - `take_la...
def add(self, module, **kwargs): """Add a module to the chain. Parameters ---------- module : BaseModule The new module to add. kwargs : ``**keywords`` All the keyword arguments are saved as meta information for the added module. The currently...
Gets current parameters. Returns ------- (arg_params, aux_params) A pair of dictionaries each mapping parameter names to NDArray values. This is a merged dictionary of all the parameters in the modules.
def get_params(self): """Gets current parameters. Returns ------- (arg_params, aux_params) A pair of dictionaries each mapping parameter names to NDArray values. This is a merged dictionary of all the parameters in the modules. """ assert self.bin...
Initializes parameters. Parameters ---------- initializer : Initializer arg_params : dict Default ``None``. Existing parameters. This has higher priority than `initializer`. aux_params : dict Default ``None``. Existing auxiliary states. This h...
def init_params(self, initializer=Uniform(0.01), arg_params=None, aux_params=None, allow_missing=False, force_init=False, allow_extra=False): """Initializes parameters. Parameters ---------- initializer : Initializer arg_params : dict Default ``No...
Binds the symbols to construct executors. This is necessary before one can perform computation with the module. Parameters ---------- data_shapes : list of (str, tuple) Typically is `data_iter.provide_data`. label_shapes : list of (str, tuple) Typically i...
def bind(self, data_shapes, label_shapes=None, for_training=True, inputs_need_grad=False, force_rebind=False, shared_module=None, grad_req='write'): """Binds the symbols to construct executors. This is necessary before one can perform computation with the module. Param...
Installs and initializes optimizers. Parameters ---------- kvstore : str or KVStore Default `'local'`. optimizer : str or Optimizer Default `'sgd'` optimizer_params : dict Default ``(('learning_rate', 0.01),)``. The default value is not a dict...
def init_optimizer(self, kvstore='local', optimizer='sgd', optimizer_params=(('learning_rate', 0.01),), force_init=False): """Installs and initializes optimizers. Parameters ---------- kvstore : str or KVStore Default `'local'`. ...
Forward computation. Parameters ---------- data_batch : DataBatch is_train : bool Default is ``None``, in which case `is_train` is take as ``self.for_training``.
def forward(self, data_batch, is_train=None): """Forward computation. Parameters ---------- data_batch : DataBatch is_train : bool Default is ``None``, in which case `is_train` is take as ``self.for_training``. """ assert self.binded and self.params_i...
Backward computation.
def backward(self, out_grads=None): """Backward computation.""" assert self.binded and self.params_initialized for i_layer, module in reversed(list(zip(range(len(self._modules)), self._modules))): module.backward(out_grads=out_grads) if i_layer == 0: brea...
Updates parameters according to installed optimizer and the gradient computed in the previous forward-backward cycle.
def update(self): """Updates parameters according to installed optimizer and the gradient computed in the previous forward-backward cycle. """ assert self.binded and self.params_initialized and self.optimizer_initialized for module in self._modules: module.update()
Gets outputs from a previous forward computation. Parameters ---------- merge_multi_context : bool Default is ``True``. In the case when data-parallelism is used, the outputs will be collected from multiple devices. A ``True`` value indicate that we should me...
def get_outputs(self, merge_multi_context=True): """Gets outputs from a previous forward computation. Parameters ---------- merge_multi_context : bool Default is ``True``. In the case when data-parallelism is used, the outputs will be collected from multiple devi...
Gets the gradients with respect to the inputs of the module. Parameters ---------- merge_multi_context : bool Default is ``True``. In the case when data-parallelism is used, the outputs will be collected from multiple devices. A ``True`` value indicate that we ...
def get_input_grads(self, merge_multi_context=True): """Gets the gradients with respect to the inputs of the module. Parameters ---------- merge_multi_context : bool Default is ``True``. In the case when data-parallelism is used, the outputs will be collected fro...
Evaluates and accumulates evaluation metric on outputs of the last forward computation. Parameters ---------- eval_metric : EvalMetric labels : list of NDArray Typically ``data_batch.label``.
def update_metric(self, eval_metric, labels, pre_sliced=False): """Evaluates and accumulates evaluation metric on outputs of the last forward computation. Parameters ---------- eval_metric : EvalMetric labels : list of NDArray Typically ``data_batch.label``. ...
Installs monitor on all executors.
def install_monitor(self, mon): """Installs monitor on all executors.""" assert self.binded for module in self._modules: module.install_monitor(mon)
Generate the iterator of mnist dataset
def get_iterator(data_shape, use_caffe_data): """Generate the iterator of mnist dataset""" def get_iterator_impl_mnist(args, kv): """return train and val iterators for mnist""" # download data get_mnist_ubyte() flat = False if len(data_shape) != 1 else True train = mx.io...
The function is used to run predictions on the audio files in the directory `pred_directory`. Parameters ---------- net: The model that has been trained. prediction_dir: string, default ./Test The directory that contains the audio files on which predictions are to be made
def predict(prediction_dir='./Test'): """The function is used to run predictions on the audio files in the directory `pred_directory`. Parameters ---------- net: The model that has been trained. prediction_dir: string, default ./Test The directory that contains the audio files on wh...
Thread loop for generating data Parameters ---------- proc_id: int Process id alive: multiprocessing.Value variable for signaling whether process should continue or not queue: multiprocessing.Queue queue for passing data back fn: funct...
def _proc_loop(proc_id, alive, queue, fn): """Thread loop for generating data Parameters ---------- proc_id: int Process id alive: multiprocessing.Value variable for signaling whether process should continue or not queue: multiprocessing.Queue ...
Start processes if not already started
def _init_proc(self): """Start processes if not already started""" if not self.proc: self.proc = [ mp.Process(target=self._proc_loop, args=(i, self.alive, self.queue, self.fn)) for i in range(self.num_proc) ] self.alive.value = True ...
Resets the generator by stopping all processes
def reset(self): """Resets the generator by stopping all processes""" self.alive.value = False qsize = 0 try: while True: self.queue.get(timeout=0.1) qsize += 1 except QEmptyExcept: pass print("Queue size on reset: {...
Create a base class with a metaclass.
def with_metaclass(meta, *bases): """Create a base class with a metaclass.""" # This requires a bit of explanation: the basic idea is to make a dummy # metaclass for one level of class instantiation that replaces itself with # the actual metaclass. class metaclass(type): def __new__(cls, na...
Load library by searching possible path.
def _load_lib(): """Load library by searching possible path.""" lib_path = libinfo.find_lib_path() lib = ctypes.CDLL(lib_path[0], ctypes.RTLD_LOCAL) # DMatrix functions lib.MXGetLastError.restype = ctypes.c_char_p return lib
Create ctypes array from a Python array. Parameters ---------- ctype : ctypes data type Data type of the array we want to convert to, such as mx_float. values : tuple or list Data content. Returns ------- out : ctypes array Created ctypes array. Examples -...
def c_array(ctype, values): """Create ctypes array from a Python array. Parameters ---------- ctype : ctypes data type Data type of the array we want to convert to, such as mx_float. values : tuple or list Data content. Returns ------- out : ctypes array Create...
Create ctypes const void ** from a list of MXNet objects with handles. Parameters ---------- objs : list of NDArray/Symbol. MXNet objects. Returns ------- (ctypes.c_void_p * len(objs)) A void ** pointer that can be passed to C API.
def c_handle_array(objs): """Create ctypes const void ** from a list of MXNet objects with handles. Parameters ---------- objs : list of NDArray/Symbol. MXNet objects. Returns ------- (ctypes.c_void_p * len(objs)) A void ** pointer that can be passed to C API. """ a...
Convert a ctypes pointer to a numpy array. The resulting NumPy array shares the memory with the pointer. Parameters ---------- cptr : ctypes.POINTER(mx_float) pointer to the memory region shape : tuple Shape of target `NDArray`. Returns ------- out : numpy_array ...
def ctypes2numpy_shared(cptr, shape): """Convert a ctypes pointer to a numpy array. The resulting NumPy array shares the memory with the pointer. Parameters ---------- cptr : ctypes.POINTER(mx_float) pointer to the memory region shape : tuple Shape of target `NDArray`. Re...
Build argument docs in python style. arg_names : list of str Argument names. arg_types : list of str Argument type information. arg_descs : list of str Argument description information. remove_dup : boolean, optional Whether remove duplication or not. Returns ...
def build_param_doc(arg_names, arg_types, arg_descs, remove_dup=True): """Build argument docs in python style. arg_names : list of str Argument names. arg_types : list of str Argument type information. arg_descs : list of str Argument description information. remove_dup :...
Append the definition position to each function contained in module. Examples -------- # Put the following codes at the end of a file add_fileline_to_docstring(__name__)
def add_fileline_to_docstring(module, incursive=True): """Append the definition position to each function contained in module. Examples -------- # Put the following codes at the end of a file add_fileline_to_docstring(__name__) """ def _add_fileline(obj): """Add fileinto to a objec...