|
| | hdf5_data_reader (bool shuffle=true) |
| |
| | hdf5_data_reader (const hdf5_data_reader &) |
| |
| hdf5_data_reader & | operator= (const hdf5_data_reader &) |
| |
| hdf5_data_reader * | copy () const override |
| |
| void | copy_members (const hdf5_data_reader &rhs) |
| |
| | ~hdf5_data_reader () override |
| |
| bool | has_conduit_output () override |
| |
| std::string | get_type () const override |
| |
| void | print_metadata (std::ostream &os=std::cout) |
| | Prints metadata and data-types for all field-names. More...
|
| |
| void | load () override |
| |
| bool | fetch_conduit_node (conduit::Node &sample, int data_id) override |
| |
| void | set_experiment_schema_filename (std::string fn) |
| | Sets the name of the yaml experiment file. More...
|
| |
| const std::string & | get_experiment_schema_filename () |
| | Returns the name of the yaml experiment file. More...
|
| |
| void | set_data_schema_filename (std::string fn) |
| | Sets the name of the yaml data file. More...
|
| |
| const std::string & | get_data_schema_filename () |
| | Returns the name of the yaml data file. More...
|
| |
| const std::vector< El::Int > | get_data_dims () const override |
| | Get the dimensions of the data. More...
|
| |
| int | get_linearized_data_size () const override |
| | Get the linearized size (i.e. number of elements) in a sample. More...
|
| |
| int | get_linearized_response_size () const override |
| | Get the linearized size (i.e. number of elements) in a response. More...
|
| |
| int | get_linearized_label_size () const override |
| | Get the linearized size (i.e. number of elements) in a label. More...
|
| |
| int | get_num_labels () const override |
| | Return the number of labels (classes) in this dataset. More...
|
| |
| int | get_num_responses () const override |
| | Return the number of responses in this dataset. More...
|
| |
| conduit::Node | get_experiment_schema () const |
| | this method is made public for testing More...
|
| |
| conduit::Node | get_data_schema () const |
| | this method is made public for testing More...
|
| |
| void | set_experiment_schema (const conduit::Node &s) |
| | this method is made public for testing More...
|
| |
| void | set_data_schema (const conduit::Node &s) |
| | this method is made public for testing More...
|
| |
| std::unordered_map< std::string, conduit::Node > | get_node_map () const |
| | this method is made public for testing More...
|
| |
| void | adjust_metadata (conduit::Node *root) |
| | this method is made public for testing More...
|
| |
| | data_reader_sample_list (bool shuffle=true) |
| |
| | data_reader_sample_list (const data_reader_sample_list &) |
| |
| data_reader_sample_list & | operator= (const data_reader_sample_list &) |
| |
| | ~data_reader_sample_list () override |
| |
| data_reader_sample_list * | copy () const override |
| |
| void | copy_members (const data_reader_sample_list &rhs) |
| |
| std::string | get_type () const override |
| |
| std::pair< file_handle_type, sample_name_type > | open_file (size_t index) |
| | Open the file and get the sample name for the given index. More...
|
| |
| void | close_file (size_t index_in) |
| |
| void | shuffle_indices (rng_gen &gen) override |
| |
| void | load () override |
| |
| sample_list_hdf5< std::string > & | get_sample_list () |
| |
| sample_type | get_sample (size_t index) |
| |
| | generic_data_reader (bool shuffle=true) |
| |
| | generic_data_reader (const generic_data_reader &)=default |
| |
| generic_data_reader & | operator= (const generic_data_reader &)=default |
| |
| virtual | ~generic_data_reader () |
| |
| template<class Archive > |
| void | serialize (Archive &ar) |
| |
| void | set_comm (lbann_comm *comm) |
| | set the comm object More...
|
| |
| lbann_comm * | get_comm () const |
| | returns a (possibly nullptr) to comm More...
|
| |
| void | set_file_dir (std::string s) |
| |
| void | set_local_file_dir (std::string s) |
| |
| void | set_max_files_to_load (size_t n) |
| |
| std::string | get_file_dir () const |
| |
| std::string | get_local_file_dir () const |
| |
| void | set_data_sample_list (std::string s) |
| |
| std::string | get_data_sample_list () const |
| |
| void | keep_sample_order (bool same_order=false) |
| |
| void | set_data_filename (std::string s) |
| |
| std::string | get_data_filename () const |
| |
| void | set_label_filename (std::string s) |
| |
| std::string | get_label_filename () const |
| |
| void | set_shuffle (bool b) |
| |
| bool | is_shuffled () const |
| |
| void | set_shuffled_indices (const std::vector< int > &indices) |
| |
| const std::vector< int > & | get_shuffled_indices () const |
| |
| void | set_first_n (int n) |
| |
| void | set_absolute_sample_count (size_t s) |
| |
| void | set_use_fraction (double s) |
| |
| virtual void | set_execution_mode_split_fraction (execution_mode m, double s) |
| |
| virtual void | set_role (std::string role) |
| |
| std::string | get_role () const |
| |
| virtual void | setup (int num_io_threads, observer_ptr< thread_pool > io_thread_pool) |
| |
| int | fetch (std::map< data_field_type, CPUMat *> &input_buffers, El::Matrix< El::Int > &indices_fetched, size_t mb_size) |
| | Fetch a mini-batch worth of data, including samples, labels, responses (as appropriate) More...
|
| |
| int | fetch (std::vector< conduit::Node > &samples, El::Matrix< El::Int > &indices_fetched, size_t mb_size) |
| |
| virtual bool | has_data_field (data_field_type data_field) const |
| | Check to see if the data reader supports this specific data field. More...
|
| |
| virtual bool | has_labels () const |
| |
| virtual bool | has_responses () const |
| |
| void | set_has_data_field (data_field_type const data_field, const bool b) |
| | Whether or not a data reader has a data field. More...
|
| |
| virtual void | set_has_labels (const bool b) |
| | Whether or not a data reader has labels. More...
|
| |
| virtual void | set_has_responses (const bool b) |
| | Whether or not a data reader has a response field. More...
|
| |
| void | start_data_store_mini_batch_exchange () |
| |
| void | finish_data_store_mini_batch_exchange () |
| |
| virtual bool | update (bool is_active_reader) |
| |
| virtual std::vector< El::Int > | get_slice_points (const slice_points_mode var_category, bool &is_supported) |
| |
| virtual bool | position_valid () const |
| | True if the data reader's current position is valid. More...
|
| |
| virtual bool | position_is_overrun () const |
| |
| bool | at_new_epoch () const |
| | True if the data reader is at the start of an epoch. More...
|
| |
| void | set_mini_batch_size (const int s) |
| | Set the mini batch size. More...
|
| |
| int | get_mini_batch_size () const |
| | Get the mini batch size. More...
|
| |
| int | get_loaded_mini_batch_size () const |
| | Get the loaded mini-batch size. More...
|
| |
| int | get_current_mini_batch_size () const |
| | Get the current mini-batch size. More...
|
| |
| int | get_mini_batch_max () const |
| | Return the full mini_batch_size. More...
|
| |
| void | set_stride_to_next_mini_batch (const int s) |
| | Set the mini batch stride. More...
|
| |
| int | get_stride_to_next_mini_batch () const |
| | Return the mini batch stride. More...
|
| |
| void | set_sample_stride (const int s) |
| | Set the sample stride. More...
|
| |
| int | get_sample_stride () const |
| | Return the sample stride. More...
|
| |
| void | set_iteration_stride (const int s) |
| | Set the iteration stride. More...
|
| |
| int | get_iteration_stride () const |
| | Return the iteration stride. More...
|
| |
| virtual void | set_base_offset (const int s) |
| | Return the base offset. More...
|
| |
| int | get_base_offset () const |
| | Return the base offset. More...
|
| |
| void | set_last_mini_batch_size (const int s) |
| | Set the last mini batch size. More...
|
| |
| int | get_last_mini_batch_size () const |
| | Return the last mini batch size. More...
|
| |
| void | set_stride_to_last_mini_batch (const int s) |
| | Set the last mini batch stride. More...
|
| |
| int | get_stride_to_last_mini_batch () const |
| | Return the last mini batch stride. More...
|
| |
| void | set_num_parallel_readers (const int s) |
| | Set the number of parallel readers per model. More...
|
| |
| int | get_num_parallel_readers () const |
| | Return the number of parallel readers per model. More...
|
| |
| virtual void | set_reset_mini_batch_index (const int s) |
| | Set the starting mini-batch index for the epoch. More...
|
| |
| int | get_reset_mini_batch_index () const |
| | Return the starting mini-batch index for the epoch. More...
|
| |
| int | get_loaded_mini_batch_index () const |
| | Return the current mini-batch index for the epoch. More...
|
| |
| int | get_current_mini_batch_index () const |
| | Return the current mini-batch index for the epoch. More...
|
| |
| void | set_initial_position () |
| | Set the current position based on the base and model offsets. More...
|
| |
| int | get_position () const |
| | Get the current position in the data reader. More...
|
| |
| int | get_next_position () const |
| | Get the next position in the data reader. More...
|
| |
| int * | get_indices () |
| | Get a pointer to the start of the shuffled indices. More...
|
| |
| virtual int | get_num_data () const |
| | Get the number of samples in this dataset. More...
|
| |
| int | get_num_unused_data (execution_mode m) const |
| | Get the number of unused samples in this dataset. More...
|
| |
| int * | get_unused_data (execution_mode m) |
| | Get a pointer to the start of the unused sample indices. More...
|
| |
| const std::vector< int > & | get_unused_indices (execution_mode m) |
| |
| void | set_num_iterations_per_epoch (int num_iterations_per_epoch) |
| | Set the number of iterations in each epoch. More...
|
| |
| int | get_num_iterations_per_epoch () const |
| | Get the number of iterations in each epoch. More...
|
| |
| int | get_current_step_in_epoch () const |
| |
| void | resize_shuffled_indices () |
| |
| void | select_subset_of_data () |
| |
| virtual void | use_unused_index_set (execution_mode m) |
| |
| virtual bool | has_list_per_model () const |
| | Does the data reader have a unique sample list per model. More...
|
| |
| virtual bool | has_list_per_trainer () const |
| | Does the data reader have a unique sample list per trainer. More...
|
| |
| bool | save_to_checkpoint_shared (persist &p, execution_mode mode) |
| | Given directory to store checkpoint files, write state to file and add to number of bytes written. More...
|
| |
| bool | load_from_checkpoint_shared (persist &p, execution_mode mode) |
| | Given directory to store checkpoint files, read state from file and add to number of bytes read. More...
|
| |
| bool | save_to_checkpoint_distributed (persist &p, execution_mode mode) |
| |
| bool | load_from_checkpoint_distributed (persist &p, execution_mode mode) |
| | Given directory to store checkpoint files, read state from file and add to number of bytes read. More...
|
| |
| const data_store_conduit & | get_data_store () const |
| | returns a const ref to the data store More...
|
| |
| data_store_conduit & | get_data_store () |
| | returns a non-const ref to the data store More...
|
| |
| data_store_conduit * | get_data_store_ptr () const |
| |
| void | setup_data_store (int mini_batch_size) |
| |
| void | instantiate_data_store () |
| |
| virtual void | preload_data_store () |
| |
| void | set_gan_labelling (bool has_gan_labelling) |
| |
| void | set_gan_label_value (int gan_label_value) |
| |
| void | set_data_store (data_store_conduit *g) |
| | support of data store functionality More...
|
| |
| virtual bool | data_store_active () const |
| |
| virtual bool | priming_data_store () const |
| |
| virtual void | post_update () |
| |
| void | set_transform_pipeline (transform::transform_pipeline &&tp) |
| |
| void | print_get_methods (const std::string filename) |
| | Print the return values from various get_X methods to file. More...
|
| |
| size_t | get_num_indices_to_use () const |
| |
| void | set_use_data_store (bool s) |
| |
|
| const std::vector< El::Int > | get_data_dims (std::string name="") const |
| |
| int | get_linearized_size (data_field_type const &data_field) const override |
| |
| void | load_sample_schema (conduit::Schema &s) |
| |
| void | parse_schemas () |
| |
| void | get_schema_ptrs (conduit::Node *starting_node, std::unordered_map< std::string, conduit::Node *> &schema_name_map) |
| |
| void | get_leaves (conduit::Node *node_in, std::unordered_map< std::string, conduit::Node *> &leaves_out) |
| |
| void | get_leaves_multi (conduit::Node *node_in, std::unordered_map< std::string, conduit::Node *> &leaves_out) |
| |
| void | do_preload_data_store () override |
| |
| void | load_sample (conduit::Node &node, hid_t file_handle, const std::string &sample_name, bool ignore_failure=false) |
| |
| void | load_sample_from_sample_list (conduit::Node &node, size_t index, bool ignore_failure=false) |
| |
| void | pack_data (conduit::Node &node_in_out) |
| |
| void | load_schema (std::string filename, conduit::Node &schema) |
| |
| void | pack (conduit::Node &node, size_t index) |
| |
| conduit::Node | merge_metadata_nodes (const conduit::Node *node_A, const conduit::Node *node_B) |
| |
| void | build_packing_map (conduit::Node &node) |
| |
| void | repack_image (conduit::Node &node, const std::string &path, const conduit::Node &metadata) |
| |
| void | coerce (const conduit::Node &metadata, hid_t file_handle, const std::string &original_path, const std::string &new_pathname, conduit::Node &node) |
| |
| void | normalize (conduit::Node &node, const std::string &path, const conduit::Node &metadata) |
| |
| void | construct_linearized_size_lookup_tables () |
| |
| void | construct_linearized_size_lookup_tables (conduit::Node &node) |
| |
| void | test_that_all_nodes_contain_metadata (conduit::Node &node) |
| |
| bool | get_delete_packed_fields () |
| |
| void | set_delete_packed_fields (bool flag) |
| |
| template<typename T > |
| void | pack (std::string const &group_name, conduit::Node &node, size_t index) |
| |
| bool | is_composite_node (const conduit::Node &node) const |
| |
|
| using | sample_name_type = typename sample_list_hdf5< std::string > ::name_t |
| |
| using | sample_file_id_type = typename sample_list_hdf5< std::string > ::sample_file_id_t |
| |
| using | sample_type = std::pair< sample_file_id_type, sample_name_type > |
| |
| using | file_handle_type = typename sample_list_hdf5< std::string > ::file_handle_t |
| |
| using | unused_index_map_t = std::map< execution_mode, std::vector< int > > |
| |
| int | m_mini_batch_size |
| |
| int | m_current_pos |
| |
| int | m_stride_to_next_mini_batch |
| |
| int | m_base_offset |
| |
| int | m_sample_stride |
| |
| int | m_iteration_stride |
| | Stride used by parallel data readers within the model. More...
|
| |
| std::vector< int > | m_shuffled_indices |
| |
| unused_index_map_t | m_unused_indices |
| | Record of the indicies that are not being used for training. More...
|
| |
| int | m_last_mini_batch_size |
| |
| int | m_stride_to_last_mini_batch |
| |
| int | m_reset_mini_batch_index |
| | The index at which this data reader starts its epoch. More...
|
| |
| int | m_loaded_mini_batch_idx |
| | The index of the current mini-batch that has been loaded. More...
|
| |
| int | m_current_mini_batch_idx |
| |
| int | m_num_iterations_per_epoch |
| |
| int | m_num_parallel_readers |
| | How many iterations all readers will execute. More...
|
| |
| size_t | m_max_files_to_load |
| | How many parallel readers are being used. More...
|
| |
| std::string | m_file_dir |
| |
| std::string | m_local_file_dir |
| |
| std::string | m_data_sample_list |
| |
| std::string | m_data_fn |
| |
| std::string | m_label_fn |
| |
| bool | m_shuffle |
| |
| size_t | m_absolute_sample_count |
| |
| std::map< execution_mode, double > | m_execution_mode_split_fraction |
| |
| double | m_use_fraction |
| |
| int | m_first_n |
| |
| std::string | m_role |
| |
| void | load_list_of_samples (const std::string sample_list_file) |
| |
| void | load_list_of_samples_from_archive (const std::string &sample_list_archive) |
| |
| size_t | get_absolute_sample_count () const |
| |
| double | get_use_fraction () const |
| |
| double | get_execution_mode_split_fraction (execution_mode m) const |
| |
| virtual bool | fetch_data_block (std::map< data_field_type, CPUMat *> &input_buffers, El::Int block_offset, El::Int block_stride, El::Int mb_size, El::Matrix< El::Int > &indices_fetched) |
| |
| bool | fetch_data_block_conduit (std::vector< conduit::Node > &samples, El::Int block_offset, El::Int block_stride, El::Int mb_size, El::Matrix< El::Int > &indices_fetched) |
| |
| virtual bool | fetch_data_field (data_field_type data_field, CPUMat &Y, int data_id, int mb_idx) |
| | Called by fetch_data, fetch_label, fetch_response. More...
|
| |
| virtual bool | fetch_datum (CPUMat &X, int data_id, int mb_idx) |
| |
| virtual bool | fetch_label (CPUMat &Y, int data_id, int mb_idx) |
| |
| virtual bool | fetch_response (CPUMat &Y, int data_id, int mb_idx) |
| |
| CPUMat | create_datum_view (CPUMat &X, const int mb_idx) |
| |
| virtual void | preprocess_data_source (int tid) |
| |
| virtual void | postprocess_data_source (int tid) |
| |
| virtual void | shuffle_indices () |
| | Shuffle indices (uses the data_seq_generator) More...
|
| |
| void | error_check_counts () const |
| |
| sample_list_hdf5< std::string > | m_sample_list |
| |
| bool | m_verbose = false |
| |
| std::unordered_set< int > | m_using_random_node |
| |
| data_store_conduit * | m_data_store |
| |
| lbann_comm * | m_comm |
| |
| bool | m_use_data_store = false |
| |
| std::map< data_field_type, bool > | m_supported_input_types |
| | Holds a true value for each input data type that is supported. Use an ordered map so that checkpoints are stable. More...
|
| |
| bool | m_gan_labelling |
| |
| int | m_gan_label_value |
| |
| observer_ptr< thread_pool > | m_io_thread_pool |
| |
| bool | m_keep_sample_order |
| |
| transform::transform_pipeline | m_transform_pipeline |
| |
| bool | m_issue_warning |
| |