colossalai.context.process_group_initializer.initializer_2p5d

class colossalai.context.process_group_initializer.initializer_2p5d.Initializer_2p5D_ROW(tesseract_dim, tesseract_dep, *args)

2.5d tensor parallel initialization among rows.

Parameters

• tesseract_dim (int) – The dimension of tesseract.

• tesseract_dep (int) – The dimension of depth.

• rank (int) – The rank of current process.

• world_size (int) – Size of whole communication world.

• config (Config) – Running configuration.

• data_parallel_size (int) – Size of data parallel.

• pipeline_parallel_size (int) – Size of pipeline parallel.

• tensor_parallel_size (int) – Size of tensor parallel.

init_dist_group()

Initialize 2.5D tensor row parallel groups, and assign local_ranks and groups to each gpu.

Returns

2.5D tensor row parallelism’s information in a tuple.

Return type

Tuple (local_rank, group_world_size, process_group, ranks_in_group, mode)

class colossalai.context.process_group_initializer.initializer_2p5d.Initializer_2p5D_Col(tesseract_dim, tesseract_dep, *args)

2.5d tensor parallel initialization among cols.

Parameters

• tesseract_dim (int) – The dimension of tesseract.

• tesseract_dep (int) – The dimension of depth.

• rank (int) – The rank of current process.

• world_size (int) – Size of whole communication world.

• config (Config) – Running configuration.

• data_parallel_size (int) – Size of data parallel.

• pipeline_parallel_size (int) – Size of pipeline parallel.

• tensor_parallel_size (int) – Size of tensor parallel.

init_dist_group()

Initialize 2.5D tensor col parallel groups, and assign local_ranks and groups to each gpu.

Returns

2.5D tensor col parallelism’s information in a tuple.

Return type

Tuple (local_rank, group_world_size, process_group, ranks_in_group, mode)

class colossalai.context.process_group_initializer.initializer_2p5d.Initializer_2p5D_Dep(tesseract_dim, tesseract_dep, *args)

2.5D tensor parallel initialization among depths.

Parameters

• tesseract_dim (int) – The dimension of tesseract.

• tesseract_dep (int) – The dimension of depth.

• rank (int) – The rank of current process.

• world_size (int) – Size of whole communication world.

• config (Config) – Running configuration.

• data_parallel_size (int) – Size of data parallel.

• pipeline_parallel_size (int) – Size of pipeline parallel.

• tensor_parallel_size (int) – Size of tensor parallel.

init_dist_group()

Initialize 2.5D tensor depth parallel groups, and assign local_ranks and groups to each gpu.

Returns

2.5D tensor depth parallelism’s information in a tuple.

Return type

Tuple (local_rank, group_world_size, process_group, ranks_in_group, mode)

class colossalai.context.process_group_initializer.initializer_2p5d.Initializer_2p5D_XZ(tesseract_dim, tesseract_dep, *args)

2.5d tensor parallel initialization among cols times dep.

Parameters

• tesseract_dim (int) – The dimension of tesseract.

• tesseract_dep (int) – The dimension of depth.

• rank (int) – The rank of current process.

• world_size (int) – Size of whole communication world.

• config (Config) – Running configuration.

• data_parallel_size (int) – Size of data parallel.

• pipeline_parallel_size (int) – Size of pipeline parallel.

• tensor_parallel_size (int) – Size of tensor parallel.

init_dist_group()

Initialize 2.5D tensor colXdepth parallel groups, and assign local_ranks and groups to each gpu.

Returns

2.5D tensor colXdepth parallelism’s information in a tuple.

Return type

Tuple (local_rank, group_world_size, process_group, ranks_in_group, mode)