242 lines
8.6 KiB
Python
242 lines
8.6 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 paddle import _C_ops, _legacy_C_ops
|
|
from paddle.common_ops_import import check_variable_and_dtype
|
|
from paddle.framework import (
|
|
LayerHelper,
|
|
in_dynamic_mode,
|
|
in_dynamic_or_pir_mode,
|
|
)
|
|
|
|
|
|
def _number_count(numbers, upper_range):
|
|
"""
|
|
calculate the expert count according to the gate index.
|
|
Args:
|
|
numbers (Tensor): Tensor. The input gate index whose data type should be int32 or int64.
|
|
upper_range (int): The number of the experts.
|
|
Returns:
|
|
out (Tensor): The output expert count.
|
|
Examples:
|
|
.. code-block:: pycon
|
|
|
|
>>> # doctest: +REQUIRES(env: DISTRIBUTED)
|
|
>>> import paddle
|
|
>>> from paddle.distributed.models.moe import utils
|
|
>>> numbers = [[0, 2], [0, 2]]
|
|
>>> upper_range = 6
|
|
>>> numbers = paddle.to_tensor(numbers, dtype="int64")
|
|
>>> number_count = utils._number_count(numbers, upper_range)
|
|
>>> print(number_count)
|
|
Tensor(shape=[6], dtype=int64, place=Place(gpu:0), stop_gradient=True,
|
|
[2, 0, 2, 0, 0, 0])
|
|
"""
|
|
if in_dynamic_or_pir_mode():
|
|
return _C_ops.number_count(numbers, upper_range)
|
|
else:
|
|
op_type = 'number_count'
|
|
|
|
helper = LayerHelper(op_type, **locals())
|
|
out = helper.create_variable_for_type_inference(dtype=numbers.dtype)
|
|
|
|
helper.append_op(
|
|
type=op_type,
|
|
inputs={'numbers': numbers},
|
|
outputs={'Out': out},
|
|
attrs={'upper_range': upper_range},
|
|
)
|
|
return out
|
|
|
|
|
|
def _assign_pos(x, cum_count):
|
|
"""
|
|
Assign pos decides which tokens should be fetched belong to
|
|
specially expert orderly.
|
|
|
|
Args:
|
|
x (Tensor): Tensor. Every element in the list must be a Tensor whose data type
|
|
should be float16, float32, float64, int32 or int64.
|
|
cum_count (Tensor): The cumulative sum tokens of counters. Every element in the list must be a Tensor whose
|
|
data type should be int64.
|
|
|
|
Returns:
|
|
out (Tensor): Assemble numbers in the order of counters.
|
|
|
|
Examples:
|
|
.. code-block:: pycon
|
|
|
|
>>> # doctest: +REQUIRES(env: DISTRIBUTED)
|
|
>>> import paddle
|
|
>>> from paddle.distributed.models.moe import utils
|
|
>>> number_count = [2, 0, 2, 0]
|
|
>>> numbers = [[0, 2], [0, 2]]
|
|
>>> number_count = paddle.to_tensor(number_count, dtype="int64")
|
|
>>> numbers = paddle.to_tensor(numbers, dtype="int64")
|
|
>>> num_cum = paddle.cumsum(number_count)
|
|
>>> pos = utils._assign_pos(x=numbers, cum_count=num_cum)
|
|
>>> print(pos)
|
|
Tensor(shape=[4], dtype=int64, place=Place(gpu:0), stop_gradient=True,
|
|
[2, 0, 3, 1])
|
|
"""
|
|
if in_dynamic_or_pir_mode():
|
|
return _C_ops.assign_pos(x, cum_count, cum_count[-1])
|
|
else:
|
|
op_type = 'assign_pos'
|
|
|
|
helper = LayerHelper(op_type, **locals())
|
|
out = helper.create_variable_for_type_inference(dtype=cum_count.dtype)
|
|
|
|
helper.append_op(
|
|
type=op_type,
|
|
inputs={
|
|
'X': [x],
|
|
'cum_count': [cum_count],
|
|
"eff_num_len": [cum_count[-1]],
|
|
},
|
|
outputs={'Out': [out]},
|
|
)
|
|
return out
|
|
|
|
|
|
def _random_routing(topk_idx, topk_value, prob, topk=2):
|
|
r"""
|
|
random routing topk gate idx
|
|
```
|
|
out = topk_idx
|
|
for i in len(topk_idx):
|
|
if topk * value[i][topk-1] < prob[i]:
|
|
out[i][topk-1] = -1
|
|
```
|
|
Args:
|
|
topk_idx: gate idx, shape=(N, topk)
|
|
topk_value: values, shape = topk_idx.shape
|
|
prob: random prob, shape=(topk_idx.shape[0],)
|
|
"""
|
|
if topk == 2:
|
|
if in_dynamic_mode():
|
|
return _legacy_C_ops.random_routing(prob, topk_value, topk_idx)
|
|
else:
|
|
raise RuntimeError("Not supporting static graph mode now")
|
|
else:
|
|
raise RuntimeError("only topk=2 is supported now")
|
|
|
|
|
|
def _limit_by_capacity(expert_count, capacity, n_worker):
|
|
"""
|
|
limit the expert count by capacity.
|
|
Args:
|
|
expert_count (Tensor): Tensor. The input expert count whose data type should be int32 or int64.
|
|
capacity (Tensor): Tensor. The input capacity whose data type should be int32 or int64 and the elements of capacity should be the same with expert_count.numel()/n_work.
|
|
n_work (int): The number of the works.
|
|
Returns:
|
|
out (Tensor): The output expert count limit by capacity.
|
|
Examples:
|
|
.. code-block:: pycon
|
|
|
|
>>> # doctest: +REQUIRES(env: DISTRIBUTED)
|
|
>>> import paddle
|
|
>>> from paddle.distributed.models.moe import utils
|
|
>>> expert_count = [1, 2, 2, 8, 3, 6]
|
|
>>> capacity = [5, 5, 5]
|
|
>>> n_work = 2
|
|
>>> expert_count = paddle.to_tensor(expert_count, dtype="int64")
|
|
>>> capacity = paddle.to_tensor(capacity, dtype="int64")
|
|
>>> out = utils._limit_by_capacity(expert_count, capacity, n_work)
|
|
>>> print(out)
|
|
Tensor(shape=[6], dtype=int64, place=Place(gpu:0), stop_gradient=True,
|
|
[1, 2, 2, 4, 3, 3])
|
|
"""
|
|
if in_dynamic_or_pir_mode():
|
|
return _C_ops.limit_by_capacity(expert_count, capacity, n_worker)
|
|
else:
|
|
op_type = 'limit_by_capacity'
|
|
|
|
helper = LayerHelper(op_type, **locals())
|
|
out = helper.create_variable_for_type_inference(
|
|
dtype=expert_count.dtype
|
|
)
|
|
|
|
helper.append_op(
|
|
type=op_type,
|
|
inputs={'expert_count': expert_count, 'capacity': capacity},
|
|
outputs={'Out': out},
|
|
attrs={'n_worker': n_worker},
|
|
)
|
|
return out
|
|
|
|
|
|
def _prune_gate_by_capacity(gate_idx, expert_count, n_expert, n_worker):
|
|
"""
|
|
prune gate by capacity(only support CUDA)
|
|
|
|
Args:
|
|
gate_idx (Tensor): Represents the gate_id sequence corresponding to the input data with type int32, int64.
|
|
expert_count (Tensor): The quantity value counted on the gate_id sequence of the input data with type int32, int64.
|
|
n_worker(int, optional): The number of workers on the trainer with type int64.
|
|
|
|
Returns:
|
|
new_gate_idx (Tensor): The gate_id sequence corresponding to the new input data after passing through prune.
|
|
|
|
Examples:
|
|
.. code-block:: pycon
|
|
|
|
>>> # doctest: +REQUIRES(env: DISTRIBUTED)
|
|
>>> import paddle
|
|
>>> from paddle.distributed.models.moe import utils
|
|
>>> gate_idx = paddle.to_tensor([1, 3, 3, 3, 3, 2, 1, 1], dtype='int64')
|
|
>>> expert_count = paddle.to_tensor([0, 3, 1, 3, 0, 0, 0, 0], dtype='int64')
|
|
>>> n_worker = 1
|
|
>>> n_expert = 8
|
|
>>> new_gate_id = utils._prune_gate_by_capacity(
|
|
... gate_idx,
|
|
... expert_count,
|
|
... n_expert,
|
|
... n_worker,
|
|
... )
|
|
>>> print(new_gate_id)
|
|
Tensor(shape=[8], dtype=int64, place=Place(gpu:0), stop_gradient=True,
|
|
[1, 3, 3, 3, -1, 2, 1, 1])
|
|
"""
|
|
if in_dynamic_or_pir_mode():
|
|
return _C_ops.prune_gate_by_capacity(
|
|
gate_idx, expert_count, n_expert, n_worker
|
|
)
|
|
else:
|
|
check_variable_and_dtype(
|
|
gate_idx,
|
|
'GateIdx',
|
|
['int32', 'int64'],
|
|
'paddle.distributed.utils.prune_gate_by_capacity',
|
|
)
|
|
check_variable_and_dtype(
|
|
expert_count,
|
|
'ExpertCount',
|
|
['int32', 'int64'],
|
|
'paddle.distributed.utils.prune_gate_by_capacity',
|
|
)
|
|
|
|
helper = LayerHelper('prune_gate_by_capacity', **locals())
|
|
new_gate_idx = helper.create_variable_for_type_inference(
|
|
dtype=gate_idx.dtype
|
|
)
|
|
helper.append_op(
|
|
type='prune_gate_by_capacity',
|
|
inputs={'GateIdx': gate_idx, "ExpertCount": expert_count},
|
|
outputs={'NewGateIdx': new_gate_idx},
|
|
attrs={"n_expert": n_expert, "n_worker": n_worker},
|
|
)
|
|
|
|
return new_gate_idx
|