Files
2026-07-13 12:40:42 +08:00

1001 lines
32 KiB
Python

# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import annotations
from typing import TYPE_CHECKING, Literal
import paddle
from paddle import _C_ops, _legacy_C_ops
from paddle.autograd import PyLayer
from paddle.base.data_feeder import check_dtype, check_variable_and_dtype
from paddle.distributed import collective
from paddle.framework import (
LayerHelper,
_create_tensor,
in_dynamic_mode,
in_dynamic_or_pir_mode,
in_pir_mode,
)
from paddle.jit.marker import unified
from paddle.nn import Layer
from paddle.nn.utils import dygraph_utils
from ....communication.reduce import ReduceOp, _get_reduce_op
if TYPE_CHECKING:
from paddle import Tensor
from paddle._typing import ParamAttrLike, Size2
class c_identity_eager(PyLayer):
@staticmethod
def forward(ctx, tensor, group, skip_c_identity_dynamic):
ctx.group = group
if skip_c_identity_dynamic:
return tensor
else:
return _C_ops.c_identity(tensor, group.id, True, True)
@staticmethod
def backward(ctx, dy):
op_type = _get_reduce_op(ReduceOp.SUM)
ctx.group.process_group.all_reduce_on_calc_stream(dy, op_type)
return dy
class c_split_eager(PyLayer):
@staticmethod
def forward(ctx, tensor, group, rank, nranks):
ctx.group = group
ctx.nranks = nranks
return _C_ops.c_split(tensor, rank, nranks, group.id, True)
@staticmethod
def backward(ctx, dy):
group = ctx.group
out_shape = dy.shape
out_shape[0] = out_shape[0] * ctx.nranks
out = paddle.empty(out_shape, dtype=dy.dtype)
group.process_group.all_gather_into_tensor_on_calc_stream(
out,
dy,
)
return out
@unified
def _c_identity(tensor, group=None, skip_c_identity_dynamic=False):
"""
Return a copy of the tensor, mainly used with model parallel.
Args:
tensor (Tensor): The input Tensor. Its data type
should be float16, float32, float64, int32 or int64.
group (int): The id of the process group to work on.
Returns:
Tensor.
"""
if group is not None and not group.is_member():
return
ring_id = 0 if group is None else group.id
if in_dynamic_mode():
return c_identity_eager.apply(tensor, group, skip_c_identity_dynamic)
elif in_pir_mode():
return _C_ops.c_identity(tensor, ring_id, True, True)
else:
op_type = 'c_identity'
helper = LayerHelper(op_type, **locals())
out = helper.create_variable_for_type_inference(dtype=tensor.dtype)
check_variable_and_dtype(
tensor,
'tensor',
['float16', 'float32', 'float64', 'int32', 'int64', 'uint16'],
'_c_identity',
)
helper.append_op(
type=op_type,
inputs={'X': tensor},
outputs={'Out': out},
attrs={
'ring_id': ring_id,
'use_calc_stream': True,
'use_model_parallel': True,
},
)
return out
@unified
def _c_concat(tensor, group=None):
"""
Return allgather of the tensor, mainly used with model parallel.
Args:
tensor (Tensor): The input Tensor. Its data type
should be float16, float32, float64, int32 or int64.
group (int): The id of the process group to work on.
Returns:
Tensor.
"""
if group is not None and not group.is_member():
return
group = collective._get_default_group() if group is None else group
ring_id = group.id
global_rank = collective._get_global_env().rank
rank = group.rank
nranks = group.nranks
if in_dynamic_or_pir_mode():
return _C_ops.c_concat(tensor, rank, nranks, ring_id, True, True)
else:
op_type = 'c_concat'
helper = LayerHelper(op_type, **locals())
out = helper.create_variable_for_type_inference(dtype=tensor.dtype)
check_variable_and_dtype(
tensor,
'tensor',
['float16', 'float32', 'float64', 'int32', 'int64', 'uint16'],
'_c_concat',
)
helper.append_op(
type=op_type,
inputs={'X': tensor},
outputs={'Out': out},
attrs={
'ring_id': ring_id,
'use_calc_stream': True,
'use_model_parallel': True,
'nranks': nranks,
'rank': rank,
},
)
return out
@unified
def _c_split(tensor, group=None):
"""
Split tensor evenly among all members, mainly used with model parallel.
Args:
tensor (Tensor): The input Tensor. Its data type
should be float16, float32, float64, int32 or int64.
rank (int): The rank of the current process.
group (int): The id of the process group to work on.
Returns:
Tensor.
"""
if group is not None and not group.is_member():
return
ring_id = 0 if group is None else group.id
global_rank = collective._get_global_env().rank
rank = global_rank if group is None else group.get_group_rank(global_rank)
nranks = (
collective._get_global_env().world_size
if group is None
else group.nranks
)
if in_dynamic_mode():
return c_split_eager.apply(tensor, group, rank, nranks)
else:
op_type = 'c_split'
helper = LayerHelper(op_type, **locals())
out = helper.create_variable_for_type_inference(dtype=tensor.dtype)
check_variable_and_dtype(
tensor,
'tensor',
['float16', 'float32', 'float64', 'int32', 'int64', 'uint16'],
'_c_split',
)
helper.append_op(
type=op_type,
inputs={'X': tensor},
outputs={'Out': out},
attrs={
'ring_id': ring_id,
'use_calc_stream': True,
'rank': rank,
'nranks': nranks,
'use_model_parallel': True,
},
)
return out
class mp_allreduce_eager(PyLayer):
@staticmethod
def forward(
ctx,
tensor,
group,
use_calc_stream,
use_model_parallel,
op,
skip_c_identity_dynamic,
):
ctx.ring_id = group.id
ctx.skip_c_identity_dynamic = skip_c_identity_dynamic
if use_calc_stream:
op_type = _get_reduce_op(op)
group.process_group.all_reduce_on_calc_stream(tensor, op_type)
return tensor
else:
return _C_ops.all_reduce_(
tensor,
group.id,
paddle.distributed.ReduceOp.SUM,
)
@staticmethod
def backward(ctx, dy):
if ctx.skip_c_identity_dynamic:
return dy
else:
return _C_ops.c_identity(dy, ctx.ring_id, True, True)
@unified
def _mp_allreduce(
tensor,
op=ReduceOp.SUM,
group=None,
use_calc_stream=True,
use_model_parallel=True,
skip_c_identity_dynamic=False,
):
"""[it is same as allreduce above, but it supports model parallel. And it support inplace strategy]"""
if group is not None and not group.is_member():
return
if in_dynamic_mode():
group = collective._get_default_group() if group is None else group
assert op == ReduceOp.SUM, f"Unknown parameter: {op}."
return mp_allreduce_eager.apply(
tensor,
group,
use_calc_stream,
use_model_parallel,
op,
skip_c_identity_dynamic,
)
elif in_pir_mode():
ring_id = 0 if group is None else group.id
return _C_ops.mp_allreduce_sum(tensor, ring_id)
else:
ring_id = 0 if group is None else group.id
op_type = 'mp_allreduce_sum'
helper = LayerHelper(op_type, **locals())
out = helper.create_variable_for_type_inference(dtype=tensor.dtype)
check_variable_and_dtype(
tensor,
'tensor',
['float16', 'float32', 'float64', 'int32', 'int64', 'uint16'],
op_type,
)
helper.append_op(
type=op_type,
inputs={'X': tensor},
outputs={'Out': out},
attrs={
'ring_id': ring_id,
'use_calc_stream': use_calc_stream,
},
)
return out
@unified
def _c_lookup_table(table, index, start_index=0, vocab_size=-1, name=None):
"""
Lookup table according to index.
Args:
table (Tensor): The input Tensor. Its data type
should be float16, float32, float64.
index (Tensor): The index to lookup table.
start_index (int): The initial index for table range.
name (string): The name of the api
Returns:
Tensor.
"""
if in_dynamic_mode():
return _C_ops.c_embedding(table, index, start_index, vocab_size)
elif in_pir_mode():
return _C_ops.c_embedding(table, index, start_index, vocab_size)
else:
op_type = 'c_embedding'
helper = LayerHelper(op_type, **locals())
dtype = helper.input_dtype(input_param_name='table')
check_variable_and_dtype(index, 'input', ['int32', 'int64'], op_type)
tmp = helper.create_variable_for_type_inference(dtype)
helper.append_op(
type='c_embedding',
inputs={'Ids': index, 'W': table},
outputs={'Out': tmp},
attrs={"start_index": start_index, "vocab_size": vocab_size},
)
return tmp
class _Linear(Layer):
"""
Linear
"""
def __init__(
self,
in_features,
out_features,
weight_attr=None,
bias_attr=None,
name=None,
):
super().__init__()
self._dtype = self._helper.get_default_dtype()
self._weight_attr = weight_attr
self._bias_attr = bias_attr
self.weight = self.create_parameter(
shape=[in_features, out_features],
attr=self._weight_attr,
dtype=self._dtype,
is_bias=False,
)
self.bias = self.create_parameter(
shape=[out_features],
attr=self._bias_attr,
dtype=self._dtype,
is_bias=True,
)
self.name = name
def forward(self, input):
out = _linear(
x=input, weight=self.weight, bias=self.bias, name=self.name
)
return out
def extra_repr(self):
name_str = f', name={self.name}' if self.name else ''
return f'in_features={self.weight.shape[0]}, out_features={self.weight.shape[1]}, dtype={self._dtype}{name_str}'
@unified
def _c_softmax_with_cross_entropy(
logits,
label,
group=None,
return_softmax=False,
ignore_index=-100,
):
if group is not None and not group.is_member():
return
ring_id = 0 if group is None else group.id
global_rank = collective._get_global_env().rank
rank = global_rank if group is None else group.get_group_rank(global_rank)
nranks = (
collective._get_global_env().world_size
if group is None
else group.nranks
)
input_shape = list(logits.shape)
label_shape = list(label.shape)
input_dims = len(input_shape)
label_dims = len(label_shape)
if input_dims - 1 != label_dims and input_dims != label_dims:
raise ValueError(
f'Expected input_dims - 1 = label_dims or input_dims == label_dims\
(got input_dims{input_dims}, label_dims{label_dims})'
)
if input_dims - 1 == label_dims:
label = paddle.unsqueeze(label, axis=-1)
label_shape = list(label.shape)
if label_shape[-1] < 1 or label_shape[-1] > input_shape[-1] * nranks:
raise ValueError(
f'Expected label_shape[-1] >= 1 and label_shape[-1] <= input_shape[-1] * nranks\
(got label_shape[-1] = {label_shape[-1]}, input_shape[-1] = {input_shape[-1]})'
)
if in_dynamic_mode():
softmax, loss = _C_ops.c_softmax_with_cross_entropy(
logits, label, ignore_index, ring_id, rank, nranks
)
if not return_softmax:
return loss
else:
return loss, softmax
else:
attrs = {
'ring_id': ring_id,
'rank': rank,
'nranks': nranks,
'ignore_index': ignore_index,
}
helper = LayerHelper('c_softmax_with_cross_entropy', **locals())
softmax = helper.create_variable_for_type_inference(dtype=logits.dtype)
loss = helper.create_variable_for_type_inference(dtype=logits.dtype)
helper.append_op(
type='c_softmax_with_cross_entropy',
inputs={'Logits': logits, 'Label': label},
outputs={'Softmax': softmax, 'Loss': loss},
attrs=attrs,
)
if return_softmax:
return loss, softmax
return loss
@unified
def _c_softmax_with_multi_label_cross_entropy(
logits,
label,
smooth_weight,
group=None,
return_softmax=False,
ignore_index=-100,
sum_multi_label_loss=True,
):
if group is not None and not group.is_member():
return
ring_id = 0 if group is None else group.id
global_rank = collective._get_global_env().rank
rank = global_rank if group is None else group.get_group_rank(global_rank)
nranks = (
collective._get_global_env().world_size
if group is None
else group.nranks
)
input_shape = list(logits.shape)
label_shape = list(label.shape)
input_dims = len(input_shape)
label_dims = len(label_shape)
if input_dims - 1 != label_dims and input_dims != label_dims:
raise ValueError(
f'Expected input_dims - 1 = label_dims or input_dims == label_dims\
(got input_dims{input_dims}, label_dims{label_dims})'
)
if input_dims - 1 == label_dims:
label = paddle.unsqueeze(label, axis=-1)
label_shape = list(label.shape)
if label_shape[-1] < 1 or label_shape[-1] > input_shape[-1] * nranks:
raise ValueError(
f'Expected label_shape[-1] >= 1 and label_shape[-1] <= input_shape[-1] * nranks\
(got label_shape[-1] = {label_shape[-1]}, input_shape[-1] = {input_shape[-1]})'
)
if in_dynamic_mode():
softmax, loss = _legacy_C_ops.c_softmax_with_multi_label_cross_entropy(
logits,
label,
smooth_weight,
'ring_id',
ring_id,
'rank',
rank,
'nranks',
nranks,
'ignore_index',
ignore_index,
'sum_multi_label_loss',
sum_multi_label_loss,
)
if not return_softmax:
return loss
else:
return loss, softmax
else:
attrs = {
'ring_id': ring_id,
'rank': rank,
'nranks': nranks,
'ignore_index': ignore_index,
'sum_multi_label_loss': sum_multi_label_loss,
}
helper = LayerHelper(
'c_softmax_with_multi_label_cross_entropy', **locals()
)
softmax = helper.create_variable_for_type_inference(dtype=logits.dtype)
loss = helper.create_variable_for_type_inference(dtype=logits.dtype)
helper.append_op(
type='c_softmax_with_multi_label_cross_entropy',
inputs={
'Logits': logits,
'Label': label,
'SmoothWeight': smooth_weight,
},
outputs={'Softmax': softmax, 'Loss': loss},
attrs=attrs,
)
if return_softmax:
return loss, softmax
return loss
@unified
def _linear(x, weight, bias=None, name=None):
"""
Function Linear
"""
if in_dynamic_mode():
pre_bias = _create_tensor(dtype=x.dtype)
_legacy_C_ops.matmul(
x,
weight,
pre_bias,
'transpose_X',
False,
'transpose_Y',
False,
"alpha",
1,
)
return dygraph_utils._append_bias_in_dygraph(
pre_bias, bias, axis=len(x.shape) - 1
)
else:
helper = LayerHelper('linear', **locals())
dtype = x.dtype
assert len(x.shape) < 4, (
"X latitude is not supported greater than 3 now."
)
check_variable_and_dtype(
x, 'x', ['float16', 'float32', 'float64'], 'linear'
)
check_dtype(dtype, 'dtype', ['float16', 'float32', 'float64'], 'linear')
inputs = {'X': [x], 'Y': [weight]}
attrs = {
'transpose_X': False,
'transpose_Y': False,
'alpha': 1,
}
tmp = helper.create_variable_for_type_inference(dtype)
helper.append_op(
type='matmul_v2', inputs=inputs, outputs={'Out': tmp}, attrs=attrs
)
if bias is not None:
res = helper.create_variable_for_type_inference(dtype)
helper.append_op(
type='elementwise_add',
inputs={'X': [tmp], 'Y': [bias]},
outputs={'Out': [res]},
attrs={'axis': len(x.shape) - 1},
)
else:
res = tmp
return res
def _set_var_distributed(var):
if var is None:
return
var.is_distributed = True
# NOTE: use current_block and find_var_recursive to support while_loop
startup_block = paddle.static.default_startup_program().current_block()
main_block = paddle.static.default_main_program().current_block()
startup_block._find_var_recursive(var.name).is_distributed = True
main_block._find_var_recursive(var.name).is_distributed = True
def _parallel_linear(
x,
num_rows,
num_cols,
axis,
param_attr,
bias_attr,
gather_out,
inner_rank,
nranks,
split_tensor,
name,
group=None,
):
"""
Parallel Linear
axis the dimension of the parameter of linear layer.
axis = 0: the row dimension
axis = 1: the col dimension
"""
if group is not None and not group.is_member():
return
ring_id = 0 if group is None else group.id
if axis == 0:
if split_tensor:
x = _c_split(x, group=group)
else:
x = _c_identity(x, group=group)
linear = paddle.nn.Linear(
num_rows,
num_cols,
weight_attr=param_attr,
bias_attr=bias_attr,
name=name,
)
# NOTE: npu linear function use matmul_v2 but linear use matmul
linear_function = paddle.nn.functional.linear
linear_out = linear_function(
x,
linear.weight,
# NOTE(wangxi): row split, bias need add after allreduce
None if axis == 0 else linear.bias,
linear.name,
)
_set_var_distributed(linear.weight)
# set is_distributed for splited bias
# if a linear layer is splited by row, each rank would hold a complete bias and they should be the same in each rank.
# if a linear layer is splited by col, the bias would also be split into each rank as its weight
if axis == 1 and linear._bias_attr is not False:
_set_var_distributed(linear.bias)
if not gather_out:
return linear_out
out_shape = list(linear_out.shape)
out_shape[0] *= 1 if axis == 0 else nranks
main_block = paddle.static.default_main_program().current_block()
out = main_block.create_var(
shape=out_shape,
dtype=linear_out.dtype,
type=linear_out.type,
lod_level=linear_out.lod_level,
persistable=False,
is_data=False,
need_check_feed=linear_out.desc.need_check_feed(),
)
if axis == 0:
main_block.append_op(
type='mp_allreduce_sum',
inputs={'X': linear_out},
outputs={'Out': out},
attrs={
'ring_id': ring_id,
'use_calc_stream': True,
},
)
if linear.bias is not None:
out = out + linear.bias
else:
main_block.append_op(
type='c_concat',
inputs={'X': linear_out},
outputs={'Out': out},
attrs={
'rank': inner_rank,
'ring_id': ring_id,
'nranks': nranks,
'use_calc_stream': True,
'use_model_parallel': True,
},
)
return out
def _parallel_embedding(
x,
per_part_embeddings,
origin_size,
param_attr,
inner_rank,
num_partitions,
name,
group=None,
):
"""
Parallel Embedding
"""
if group is not None and not group.is_member():
return
ring_id = 0 if group is None else group.id
helper = LayerHelper("_parallel_embedding", **locals())
per_part_size = per_part_embeddings
rank = inner_rank
vocab_start_index = rank * per_part_size
dtype = helper.get_default_dtype()
size = [per_part_size, origin_size[1]]
weight = helper.create_parameter(
attr=param_attr, shape=size, dtype=dtype, is_bias=False
)
if num_partitions == 1:
return paddle.nn.functional.embedding(
x, weight=weight, padding_idx=None, sparse=False, name=name
)
startup_block = paddle.static.default_startup_program().global_block()
main_block = paddle.static.default_main_program().global_block()
startup_block.vars[weight.name].is_distributed = True
main_block.vars[weight.name].is_distributed = True
output_parallel = _c_lookup_table(
weight,
x,
start_index=vocab_start_index,
vocab_size=origin_size[0],
name=name,
)
out = _mp_allreduce(
output_parallel,
group=group,
use_calc_stream=True,
use_model_parallel=True,
)
return out
def split(
x: Tensor,
size: Size2,
operation: Literal['linear', 'embedding'],
axis: int = 0,
num_partitions: int = 1,
gather_out: bool = True,
weight_attr: ParamAttrLike | None = None,
bias_attr: ParamAttrLike | None = None,
name: str | None = None,
) -> Tensor:
"""
Split the weight of the specified operation into multiple devices
and do the computation in parallel.
Now the following three cases are supported.
Case 1: Parallel Embedding
The weight of the embedding operation is a NxM matrix with N rows and M columns.
With parallel embedding, the weight is split into num_partitions partitions, each
of which is a matrix with (N/num_partitions + 1) rows and M column where the last
row as the padding idx.
Suppose we split the NxM weight into two partitions on device_0 and device_1
respectively. Then, one each device, the final weight has (N/2 + 1) rows with the
index range from 0 to N/2. On device_0, all values in the input within [0, N/2 -1]
keep unchanged and all other values are changed to N/2 which is the padding index and
are mapped to all zeros after embedding. In the same way, on device_1, the value V in the
input within [N/2, N-1] will be changed to (V - N/2), and all other values are changed
to N/2 and are mapped to all zeros after embedding. Finally, the results on the two
devices are sum-reduced.
The Embedding put on single card is as shown below:
.. image:: https://githubraw.cdn.bcebos.com/PaddlePaddle/docs/develop/docs/api/paddle/distributed/img/split_embedding_single.png
:width: 800
:height: 350
:alt: single_embedding
:align: center
Parallel Embedding is shown as below:
.. image:: https://githubraw.cdn.bcebos.com/PaddlePaddle/docs/develop/docs/api/paddle/distributed/img/split_embedding_split.png
:width: 800
:alt: split_embedding
:align: center
Case 2: Row Parallel Linear
The weight of the linear operation is a NxM matrix with N rows and M columns.
With row parallel linear, the weight is split into num_partitions partitions, each
of which is a matrix with N/num_partitions rows and M column.
The linear layer put on single card is shown as below, the input variable is represented by X,
the weight matrix is represented by W and the output variable is O. The linear layer on single card is
simple matrix multiplication operation, O = X * W.
.. image:: https://githubraw.cdn.bcebos.com/PaddlePaddle/docs/develop/docs/api/paddle/distributed/img/split_single.png
:width: 800
:alt: single_linear
:align: center
Row Parallel Linear is shown as below. As the name suggests, Row Parallel Linear splits the weight matrix W into
[[W_row1], [W_row2]] along the row. And accordingly the input is split along the column into [X_col1, X_col2] and multiply their
respective weight matrices. Finally apply AllReduce on the output from each card to get the final output.
.. image:: https://githubraw.cdn.bcebos.com/PaddlePaddle/docs/develop/docs/api/paddle/distributed/img/split_row.png
:width: 800
:alt: split_row
:align: center
Case 3: Column Parallel Linear
The weight of the linear operation is a NxM matrix with N rows and M columns.
With column parallel linear, the weight is split into num_partitions partitions, each
of which is a matrix with N rows and M/num_partitions column.
The linear layer put on single card has been illustrated on case 2 and Column Parallel Linear
is shown as below. The Column Parallel Linear splits the weight matrix W into [W_col1, W_col2] along the column and
these split matrices respectively multiply the input. Finally apply AllGather on the output from each card to get the final output.
.. image:: https://githubraw.cdn.bcebos.com/PaddlePaddle/docs/develop/docs/api/paddle/distributed/img/split_col.png
:width: 800
:alt: split_col
:align: center
As observed, the column parallel linear and row parallel linear can be combined to skip one ALLGATHER communication
operator. Furthermore the Attention and MLP can be combined to improve the performance as shown below.
.. image:: https://githubraw.cdn.bcebos.com/PaddlePaddle/docs/develop/docs/api/paddle/distributed/img/split_col_row.png
:width: 800
:alt: split_col_row
:align: center
Args:
x (Tensor): Input tensor. It's data type should be float16, float32, float64, int32 or int64.
size (list|tuple): A list or tuple with two elements indicating the shape of the weight.
operation (str): The name of the operation. The supported operations are 'linear' and 'embedding'.
axis (int, Optional): Indicate along which axis to split the weight. Default: 0.
num_partitions (int, Optional): How many parts the weight is partitioned. Default: 1.
gather_out (bool, Optional): Whether to gather the output after computation. By default, the output
on each partitions will be gathered after computation. Default: True.
weight_attr (ParamAttr, Optional): The parameter attribute for the learnable
weights(Parameter) of the specified operation. Default: None.
bias_attr (ParamAttr, Optional): The parameter attribute for the bias
of the specified operation. Default: None.
name (str, Optional): The default value is None. Normally there is no need for user to set this
property. Default: None. For more information, please refer to :ref:`api_guide_Name`.
Returns:
Tensor.
Examples:
.. code-block:: pycon
>>> # doctest: +REQUIRES(env:DISTRIBUTED)
>>> import paddle
>>> import paddle.distributed.fleet as fleet
>>> paddle.enable_static()
>>> paddle.set_device(f'gpu:{paddle.distributed.ParallelEnv().dev_id}')
>>> fleet.init(is_collective=True)
>>> data = paddle.randint(0, 8, size=[10, 4])
>>> emb_out = paddle.distributed.split(
... data,
... (8, 8),
... operation="embedding",
... num_partitions=2,
... )
"""
assert isinstance(size, (list, tuple)), (
"The type of size for paddle.distributed.split must be list or tuple."
)
assert len(size) == 2, (
"Number of elements in size of paddle.distributed.split must be two."
)
assert isinstance(operation, str), (
"The type of operation for paddle.distributed.split must be str."
)
supported_operations = [
'linear',
'embedding',
]
assert operation in supported_operations, (
"The operation for "
f"paddle.distributed.split must be one of {supported_operations}."
)
if in_dynamic_mode():
raise ValueError(
"paddle.distributed.split cannot be used in dynamic "
"graph mode, please use ParallelEmbedding, ParallelRowLinear, "
"ParallelColumnLinear instead."
)
else:
from paddle.distributed.fleet import fleet
assert fleet._role_maker, (
"To use paddle.distributed.split, "
"you must call fleet.init() firstly."
)
rank = fleet.worker_index()
nranks = fleet.worker_num()
# rank within a model parallel group
inner_rank = rank % num_partitions
if operation == "embedding":
assert axis == 0, (
"We only support to split the weight of embedding "
"along the first axis now."
)
assert size[0] % num_partitions == 0, (
"The length of the vocabulary must be divisible by num_partitions "
f"but received vocabulary={size[0]} num_partitions={num_partitions}"
)
per_part_size = size[0] // num_partitions
emb_out = _parallel_embedding(
x,
per_part_size,
size,
weight_attr,
inner_rank,
num_partitions,
name,
group=None,
)
return emb_out
else:
should_split = False
if axis == 0:
assert size[0] % num_partitions == 0, (
f"Number of rows of the weight for linear ({size[0]}) must be"
f" divisible by num_partitions ({num_partitions})"
)
per_part_size = size[0] // num_partitions
linear_size = (per_part_size, size[1])
if x.shape[-1] == size[0]:
should_split = True
elif axis == 1:
assert size[1] % num_partitions == 0, (
f"Number of column of the weight for linear ({size[1]}) must be"
f" divisible by num_partitions ({num_partitions})"
)
per_part_size = size[1] // num_partitions
linear_size = (size[0], per_part_size)
else:
raise ValueError(
"The value of axis must be 0 or 1, but the value "
f"given is {axis}."
)
linear_out = _parallel_linear(
x,
linear_size[0],
linear_size[1],
axis,
weight_attr,
bias_attr,
gather_out,
inner_rank,
num_partitions,
should_split,
name=name,
group=None,
)
return linear_out