MXNet Plugin API reference

class nvidia.dali.plugin.mxnet.DALIClassificationIterator(pipelines, size=-1, reader_name=None, data_name='data', label_name='softmax_label', data_layout='NCHW', fill_last_batch=None, auto_reset=False, squeeze_labels=True, dynamic_shape=False, last_batch_padded=False, last_batch_policy=LastBatchPolicy.FILL, prepare_first_batch=True)

DALI iterator for classification tasks for MXNet. It returns 2 outputs (data and label) in the form of MXNet’s DataBatch of NDArrays.

Calling

DALIClassificationIterator(pipelines, reader_name, data_name, label_name, data_layout)

is equivalent to calling

DALIGenericIterator(pipelines,
                    [(data_name, DALIClassificationIterator.DATA_TAG),
                     (label_name, DALIClassificationIterator.LABEL_TAG)],
                    reader_name,
                    data_layout)
Parameters:
  • pipelines (list of Pipeline) – List of pipelines to use

  • size (int, default = -1) – Number of samples in the shard for the wrapped pipeline (if there is more than one it is a sum) Providing -1 means that the iterator will work until StopIteration is raised from the inside of iter_setup(). The options last_batch_policy and last_batch_padded don’t work in such case. It works with only one pipeline inside the iterator. Mutually exclusive with reader_name argument

  • reader_name (str, default = None) – Name of the reader which will be queried for the shard size, number of shards and all other properties necessary to count properly the number of relevant and padded samples that iterator needs to deal with. It automatically sets last_batch_padded accordingly to match the reader’s configuration.

  • data_name (str, optional, default = 'data') – Data name for provided symbols.

  • label_name (str, optional, default = 'softmax_label') – Label name for provided symbols.

  • data_layout (str, optional, default = 'NCHW') – Either ‘NHWC’ or ‘NCHW’ - layout of the pipeline outputs.

  • auto_reset (string or bool, optional, default = False) –

    Whether the iterator resets itself for the next epoch or it requires reset() to be called explicitly.

    It can be one of the following values:

    • "no", False or None - at the end of epoch StopIteration is raised and reset() needs to be called

    • "yes" or "True"- at the end of epoch StopIteration is raised but reset() is called internally automatically

  • squeeze_labels ((DEPRECATED) bool, optional, default = True) – Whether the iterator should squeeze the labels before copying them to the ndarray. This argument is deprecated and will be removed from future releases.

  • dynamic_shape (any, optional,) – Parameter used only for backward compatibility.

  • fill_last_batch (bool, optional, default = None) –

    Deprecated Please use last_batch_policy instead

    Whether to fill the last batch with data up to ‘self.batch_size’. The iterator would return the first integer multiple of self._num_gpus * self.batch_size entries which exceeds ‘size’. Setting this flag to False will cause the iterator to return exactly ‘size’ entries.

  • last_batch_policy (optional, default = LastBatchPolicy.FILL) – What to do with the last batch when there are not enough samples in the epoch to fully fill it. See nvidia.dali.plugin.base_iterator.LastBatchPolicy(). Both FILL and PARTIAL would return a full batch but the pad property value of the returned array would differ.

  • last_batch_padded (bool, optional, default = False) – Whether the last batch provided by DALI is padded with the last sample or it just wraps up. In the conjunction with last_batch_policy it tells if the iterator returning last batch with data only partially filled with data from the current epoch is dropping padding samples or samples from the next epoch (it doesn’t literally drop but sets pad field of ndarray so the following code could use it to drop the data). If set to False next epoch will end sooner as data from it was consumed but dropped. If set to True next epoch would be the same length as the first one. For this to happen, the option pad_last_batch in the reader needs to be set to True as well. It is overwritten when reader_name argument is provided

  • prepare_first_batch (bool, optional, default = True) – Whether DALI should buffer the first batch right after the creation of the iterator, so one batch is already prepared when the iterator is prompted for the data

Example

With the data set [1,2,3,4,5,6,7] and the batch size 2:

last_batch_policy = LastBatchPolicy.PARTIAL, last_batch_padded = True -> last batch = [7, 7] and MXNet array property .pad=1, next iteration will return [1, 2]

last_batch_policy = LastBatchPolicy.PARTIAL, last_batch_padded = False -> last batch = [7, 1] and MXNet array property .pad=1, next iteration will return [2, 3]

last_batch_policy = LastBatchPolicy.FILL, last_batch_padded = True -> last batch = [7, 7] and MXNet array property .pad=0, next iteration will return [1, 2]

last_batch_policy = LastBatchPolicy.FILL, last_batch_padded = False -> last batch = [7, 1] and MXNet array property .pad=0, next iteration will return [2, 3]

last_batch_policy = LastBatchPolicy.DROP, last_batch_padded = True -> last batch = [5, 6], next iteration will return [1, 2]

last_batch_policy = LastBatchPolicy.DROP, last_batch_padded = False -> last batch = [5, 6], next iteration will return [2, 3]

checkpoints()

Returns the current checkpoints of the pipelines.

getdata()

Get data of current batch.

Returns:

The data of the current batch.

Return type:

list of NDArray

getindex()

Get index of the current batch.

Returns:

index – The indices of examples in the current batch.

Return type:

numpy.array

getlabel()

Get label of the current batch.

Returns:

The label of the current batch.

Return type:

list of NDArray

getpad()

Get the number of padding examples in the current batch.

Returns:

Number of padding examples in the current batch.

Return type:

int

iter_next()

Move to the next batch.

Returns:

Whether the move is successful.

Return type:

boolean

next()

Returns the next batch of data.

reset()

Resets the iterator after the full epoch. DALI iterators do not support resetting before the end of the epoch and will ignore such request.

property size
class nvidia.dali.plugin.mxnet.DALIGenericIterator(pipelines, output_map, size=-1, reader_name=None, data_layout='NCHW', fill_last_batch=None, auto_reset=False, squeeze_labels=True, dynamic_shape=False, last_batch_padded=False, last_batch_policy=LastBatchPolicy.FILL, prepare_first_batch=True)

General DALI iterator for MXNet. It can return any number of outputs from the DALI pipeline in the form of MXNet’s DataBatch of NDArrays.

Parameters:
  • pipelines (list of Pipeline) – List of pipelines to use

  • output_map (list of (str, str)) – List of pairs (output_name, tag) which maps consecutive outputs of DALI pipelines to proper field in MXNet’s DataBatch. tag is one of DALIGenericIterator.DATA_TAG and DALIGenericIterator.LABEL_TAG mapping given output for data or label correspondingly. output_names should be distinct.

  • size (int, default = -1) – Number of samples in the shard for the wrapped pipeline (if there is more than one it is a sum) Providing -1 means that the iterator will work until StopIteration is raised from the inside of iter_setup(). The options last_batch_policy and last_batch_padded don’t work in such case. It works with only one pipeline inside the iterator. Mutually exclusive with reader_name argument

  • reader_name (str, default = None) – Name of the reader which will be queried for the shard size, number of shards and all other properties necessary to count properly the number of relevant and padded samples that iterator needs to deal with. It automatically sets last_batch_padded accordingly to match the reader’s configuration.

  • data_layout (str, optional, default = 'NCHW') – Either ‘NHWC’ or ‘NCHW’ - layout of the pipeline outputs.

  • auto_reset (string or bool, optional, default = False) –

    Whether the iterator resets itself for the next epoch or it requires reset() to be called explicitly.

    It can be one of the following values:

    • "no", False or None - at the end of epoch StopIteration is raised

    and reset() needs to be called * "yes" or "True"- at the end of epoch StopIteration is raised but reset() is called internally automatically

  • squeeze_labels ((DEPRECATED) bool, optional, default = False) – Whether the iterator should squeeze the labels before copying them to the ndarray. This argument is deprecated and will be removed from future releases.

  • dynamic_shape (any, optional,) – Parameter used only for backward compatibility.

  • fill_last_batch (bool, optional, default = None) –

    Deprecated Please use last_batch_policy instead

    Whether to fill the last batch with data up to ‘self.batch_size’. The iterator would return the first integer multiple of self._num_gpus * self.batch_size entries which exceeds ‘size’. Setting this flag to False will cause the iterator to return exactly ‘size’ entries.

  • last_batch_policy (optional, default = LastBatchPolicy.FILL) – What to do with the last batch when there are not enough samples in the epoch to fully fill it. See nvidia.dali.plugin.base_iterator.LastBatchPolicy(). Both FILL and PARTIAL would return a full batch but the pad property value of the returned array would differ.

  • last_batch_padded (bool, optional, default = False) – Whether the last batch provided by DALI is padded with the last sample or it just wraps up. In the conjunction with last_batch_policy it tells if the iterator returning last batch with data only partially filled with data from the current epoch is dropping padding samples or samples from the next epoch (it doesn’t literally drop but sets pad field of ndarray so the following code could use it to drop the data). If set to False next epoch will end sooner as data from it was consumed but dropped. If set to True next epoch would be the same length as the first one. For this to happen, the option pad_last_batch in the reader needs to be set to True as well. It is overwritten when reader_name argument is provided

  • prepare_first_batch (bool, optional, default = True) – Whether DALI should buffer the first batch right after the creation of the iterator, so one batch is already prepared when the iterator is prompted for the data

Example

With the data set [1,2,3,4,5,6,7] and the batch size 2:

last_batch_policy = LastBatchPolicy.PARTIAL, last_batch_padded = True -> last batch = [7, 7] and MXNet array property .pad=1, next iteration will return [1, 2]

last_batch_policy = LastBatchPolicy.PARTIAL, last_batch_padded = False -> last batch = [7, 1] and MXNet array property .pad=1, next iteration will return [2, 3]

last_batch_policy = LastBatchPolicy.FILL, last_batch_padded = True -> last batch = [7, 7] and MXNet array property .pad=0, next iteration will return [1, 2]

last_batch_policy = LastBatchPolicy.FILL, last_batch_padded = False -> last batch = [7, 1] and MXNet array property .pad=0, next iteration will return [2, 3]

last_batch_policy = LastBatchPolicy.DROP, last_batch_padded = True -> last batch = [5, 6], next iteration will return [1, 2]

last_batch_policy = LastBatchPolicy.DROP, last_batch_padded = False -> last batch = [5, 6], next iteration will return [2, 3]

checkpoints()

Returns the current checkpoints of the pipelines.

getdata()

Get data of current batch.

Returns:

The data of the current batch.

Return type:

list of NDArray

getindex()

Get index of the current batch.

Returns:

index – The indices of examples in the current batch.

Return type:

numpy.array

getlabel()

Get label of the current batch.

Returns:

The label of the current batch.

Return type:

list of NDArray

getpad()

Get the number of padding examples in the current batch.

Returns:

Number of padding examples in the current batch.

Return type:

int

iter_next()

Move to the next batch.

Returns:

Whether the move is successful.

Return type:

boolean

next()

Returns the next batch of data.

reset()

Resets the iterator after the full epoch. DALI iterators do not support resetting before the end of the epoch and will ignore such request.

property size
class nvidia.dali.plugin.mxnet.DALIGluonIterator(pipelines, size=-1, reader_name=None, output_types=None, auto_reset=False, fill_last_batch=None, last_batch_padded=False, last_batch_policy=LastBatchPolicy.FILL, prepare_first_batch=True)

General DALI iterator for MXNet with Gluon API. It can return any number of outputs from the DALI pipeline in the form of per GPU tuples. These tuples consisting of NDArrays (for outputs marked as DALIGluonIterator.DENSE_TAG) and list of NDArrays (for output marked as DALIGluonIterator.SPARSE_TAG).

Parameters:
  • pipelines (list of Pipeline) – List of pipelines to use

  • size (int, default = -1) – Number of samples in the shard for the wrapped pipeline (if there is more than one it is a sum) Providing -1 means that the iterator will work until StopIteration is raised from the inside of iter_setup(). The options last_batch_policy and last_batch_padded don’t work in such case. It works with only one pipeline inside the iterator. Mutually exclusive with reader_name argument

  • reader_name (str, default = None) – Name of the reader which will be queried for the shard size, number of shards and all other properties necessary to count properly the number of relevant and padded samples that iterator needs to deal with. It automatically sets last_batch_padded accordingly to match the reader’s configuration.

  • output_types (list of str, optional, default = None) – List of tags indicating whether the pipeline(s) output batch is uniform (all the samples have the same size) or not. Batch output marked as the former will be returned as a single NDArray, the latter will be returned as a list of NDArray. Must be either DALIGluonIterator.DENSE_TAG or DALIGluonIterator.SPARSE_TAG. Length of output_types must match the number of output of the pipeline(s). If not set, all outputs are considered to be marked with DALIGluonIterator.DENSE_TAG.

  • auto_reset (string or bool, optional, default = False) –

    Whether the iterator resets itself for the next epoch or it requires reset() to be called explicitly.

    It can be one of the following values:

    • "no", False or None - at the end of epoch StopIteration is raised and reset() needs to be called

    • "yes" or "True"- at the end of epoch StopIteration is raised but reset() is called internally automatically

  • fill_last_batch (bool, optional, default = None) –

    Deprecated Please use last_batch_policy instead

    Whether to fill the last batch with data up to ‘self.batch_size’. The iterator would return the first integer multiple of self._num_gpus * self.batch_size entries which exceeds ‘size’. Setting this flag to False will cause the iterator to return exactly ‘size’ entries.

  • last_batch_policy (optional, default = LastBatchPolicy.FILL) – What to do with the last batch when there are not enough samples in the epoch to fully fill it. See nvidia.dali.plugin.base_iterator.LastBatchPolicy()

  • last_batch_padded (bool, optional, default = False) – Whether the last batch provided by DALI is padded with the last sample or it just wraps up. In the conjunction with last_batch_policy it tells if the iterator returning last batch with data only partially filled with data from the current epoch is dropping padding samples or samples from the next epoch (it doesn’t literally drop but sets pad field of ndarray so the following code could use it to drop the data). If set to False next epoch will end sooner as data from it was consumed but dropped. If set to True next epoch would be the same length as the first one. For this to happen, the option pad_last_batch in the reader needs to be set to True as well. It is overwritten when reader_name argument is provided

  • prepare_first_batch (bool, optional, default = True) – Whether DALI should buffer the first batch right after the creation of the iterator, so one batch is already prepared when the iterator is prompted for the data

Example

With the data set [1,2,3,4,5,6,7] and the batch size 2:

last_batch_policy = LastBatchPolicy.PARTIAL, last_batch_padded = True -> last batch = [7], next iteration will return [1, 2]

last_batch_policy = LastBatchPolicy.PARTIAL, last_batch_padded = False -> last batch = [7], next iteration will return [2, 3]

last_batch_policy = LastBatchPolicy.FILL, last_batch_padded = True -> last batch = [7, 7], next iteration will return [1, 2]

last_batch_policy = LastBatchPolicy.FILL, last_batch_padded = False -> last batch = [7, 1], next iteration will return [2, 3]

last_batch_policy = LastBatchPolicy.DROP, last_batch_padded = True -> last batch = [5, 6], next iteration will return [1, 2]

last_batch_policy = LastBatchPolicy.DROP, last_batch_padded = False -> last batch = [5, 6], next iteration will return [2, 3]

checkpoints()

Returns the current checkpoints of the pipelines.

getdata()

Get data of current batch.

Returns:

The data of the current batch.

Return type:

list of NDArray

getindex()

Get index of the current batch.

Returns:

index – The indices of examples in the current batch.

Return type:

numpy.array

getlabel()

Get label of the current batch.

Returns:

The label of the current batch.

Return type:

list of NDArray

getpad()

Get the number of padding examples in the current batch.

Returns:

Number of padding examples in the current batch.

Return type:

int

iter_next()

Move to the next batch.

Returns:

Whether the move is successful.

Return type:

boolean

next()

Returns the next batch of data.

reset()

Resets the iterator after the full epoch. DALI iterators do not support resetting before the end of the epoch and will ignore such request.

property size
nvidia.dali.plugin.mxnet.feed_ndarray(dali_tensor, arr, cuda_stream=None)

Copy contents of DALI tensor to MXNet’s NDArray.

Parameters:
  • dali_tensor (nvidia.dali.backend.TensorCPU or nvidia.dali.backend.TensorGPU) – Tensor from which to copy

  • arr (mxnet.nd.NDArray) – Destination of the copy

  • cuda_stream (cudaStream_t handle or any value that can be cast to cudaStream_t.) – CUDA stream to be used for the copy (if not provided, an internal user stream will be selected) In most cases, using the default internal user stream or stream 0 is expected.

nvidia.dali.plugin.mxnet.get_mx_array(shape, ctx=None, dtype=None)