Intel® Machine Learning Scaling Library
2018
A library providing an efficient implementation of communication patterns used in deep learning.
|
The namespace containing all Intel MLSL classes and types. More...
Classes | |
struct | QuantParams |
A struct to hold quantization parameters. More... | |
class | CommBlockInfo |
A class to hold block information for activations packing/unpacking. More... | |
class | Activation |
A wrapper class for operation input and output activations. More... | |
class | ParameterSet |
A wrapper class for operation parameters. More... | |
class | Distribution |
A class to hold information about the parallelism scheme being used. More... | |
class | OperationRegInfo |
A class to hold Operation registration information. More... | |
class | Operation |
A class to hold information about learnable parameters (parameter sets) and activations corresponding to a certain operation of the computational graph. More... | |
class | Statistics |
A class to measure and store performance statistics of communication among processes that perform computation in the computational graph. More... | |
class | Session |
A class to represent a collection of Operation objects with the same global mini-batch size. More... | |
class | Environment |
A singleton object that holds global Intel MLSL functions. More... | |
Typedefs | |
typedef int | CommReq |
Enumerations | |
enum | DataType { DT_FLOAT = 0, DT_DOUBLE = 1, DT_BYTE = 2 } |
enum | PhaseType { PT_TRAIN = 0, PT_TEST = 1 } |
enum | GroupType { GT_DATA = 0, GT_MODEL = 1, GT_GLOBAL = 2 } |
Groups of processes supported by Intel MLSL. More... | |
enum | ReductionType { RT_SUM = 0, RT_MIN = 1, RT_MAX = 2 } |
enum | OpType { OT_CC = 0, OT_BIAS = 1, OT_ACT = 2, OT_POOL = 3, OT_SPLIT = 4, OT_CONCAT = 5, OT_BCAST = 6, OT_REDUCE = 7, OT_DATA = 8, OT_EVAL = 9 } |
Compute operation types. More... | |
enum | CompressionType { CT_NONE = 0, CT_QUANTIZATION = 1 } |
The namespace containing all Intel MLSL classes and types.
typedef int MLSL::CommReq |
Type definition for a communication request type.
enum MLSL::DataType |
enum MLSL::GroupType |
Groups of processes supported by Intel MLSL.
A group represents a subset of processes handling the model.
The data group contains processes working on the same part of parameters for different input batches (i.e. doing data parallelism) and communicating to exchange gradients with respect to parameters.
The model group contains processes working on different parts of parameters for the same input batch (i.e. doing model parallelism) and communicating to exchange activations and gradients with respect to activations.
The global group contains all Intel MLSL processes.
Enumerator | |
---|---|
GT_DATA |
data group |
GT_MODEL |
model group |
GT_GLOBAL |
global group |
enum MLSL::OpType |
Compute operation types.
Each operation type defines specific relationship between the input and output activations and associated parameters (weights or biases).
(abbreviations: IA – input activation, OA – output activation)
enum MLSL::PhaseType |
enum MLSL::ReductionType |
Reduction operations for Distribution::Reduce(), Distribution::AllReduce() and Distribution::ReduceScatter().
Enumerator | |
---|---|
RT_SUM | |
RT_MIN | |
RT_MAX |