EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

[Feature] optimize sp & qwen3 next support sp. (#3225)

Browse Source 
This PR will accomplish the following tasks: 
**optimize SP**
In the old version implementation, the first layer was all_reduce, which
used rms to split chunks. We changed it to perform reduce_scatter on the
embedding side, replace one all_reduce operation and one chunk with one
reduce_scatter operation.
**Support qwen3 next**
Since Qwen3 Next includes a linear attention module, the prefix name of
this module cannot take effect directly.


- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0

---------

Signed-off-by: weijinqian_v1 <weijinqian@huawei.com>
Co-authored-by: weijinqian_v1 <weijinqian@huawei.com>
This commit is contained in:
weijinqian0
2025-10-13 23:02:12 +08:00
committed by GitHub
parent 31682961af
commit 6972df5951
10 changed files with 140 additions and 193 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
tests/ut/ops/test_layernorm.py
View File

@@ -9,12 +9,6 @@ from tests.ut.base import PytestBase
from vllm_ascend.quantization.w8a8 import AscendW8A8LinearMethod from vllm_ascend.quantization.w8a8 import AscendW8A8LinearMethod
def mock_maybe_chunk_residual(x, residual):
if x.size(0) != residual.size(0):
return residual[:4]
return residual
def mock_rms_norm(x, weight, eps): def mock_rms_norm(x, weight, eps):
return x + 1, None return x + 1, None
@@ -36,8 +30,6 @@ class TestAscendRMSNorm(PytestBase):
@pytest.fixture(autouse=True) @pytest.fixture(autouse=True)
def context(self, mocker: MockerFixture): def context(self, mocker: MockerFixture):
mocker.patch("torch.ops.vllm.maybe_chunk_residual",
side_effect=mock_maybe_chunk_residual)
mocker.patch("torch_npu.npu_rms_norm", side_effect=mock_rms_norm) mocker.patch("torch_npu.npu_rms_norm", side_effect=mock_rms_norm)
mocker.patch("torch_npu.npu_add_rms_norm", mocker.patch("torch_npu.npu_add_rms_norm",
side_effect=mock_add_rms_norm) side_effect=mock_add_rms_norm)
@@ -66,21 +58,6 @@ class TestAscendRMSNorm(PytestBase):
assert torch.allclose(x_out, x_out_expected) assert torch.allclose(x_out, x_out_expected)
# Test case for flashcomm_v1 scenario
def test_forward_oot_with_flashcomm_v1(self):
layer = RMSNorm(hidden_size=512, eps=1e-05)
x = torch.randn(4, 512, dtype=torch.bfloat16)
residual = torch.randn(16, 512, dtype=torch.bfloat16)
x_out, residual_out = layer.forward_oot(x, residual)
x_out_expected = 2 * x
residual_out_expected = 2 * residual[:4]
assert residual_out.size(0) == 4
assert torch.allclose(x_out, x_out_expected)
assert torch.allclose(residual_out, residual_out_expected)
# Test case for addrmsnorm + w8a8 quant fusion # Test case for addrmsnorm + w8a8 quant fusion
def test_forward_oot_with_quant_fusion(self, mocker: MockerFixture): def test_forward_oot_with_quant_fusion(self, mocker: MockerFixture):
mock_is_310p = mocker.patch("vllm_ascend.utils.is_310p") mock_is_310p = mocker.patch("vllm_ascend.utils.is_310p")

20
tests/ut/ops/test_vocab_parallel_embedding.py
View File

@@ -104,9 +104,8 @@ class TestCustomVocabParallelEmbedding(unittest.TestCase):
input_ = torch.tensor([1, 2, 3]) input_ = torch.tensor([1, 2, 3])
with patch( with patch("torch.ops.vllm.maybe_pad_and_reduce",
"vllm_ascend.ops.vocab_parallel_embedding.tensor_model_parallel_all_reduce", side_effect=lambda x: x) as mock_reduce_tp1:
side_effect=lambda x: x) as mock_reduce_tp1:
output = layer.forward(input_) output = layer.forward(input_)
# Should just pass through without masking # Should just pass through without masking
@@ -123,9 +122,8 @@ class TestCustomVocabParallelEmbedding(unittest.TestCase):
input_ = torch.tensor([15, 35]) # one org vocab, one added vocab input_ = torch.tensor([15, 35]) # one org vocab, one added vocab
with patch( with patch("torch.ops.vllm.maybe_pad_and_reduce",
"vllm_ascend.ops.vocab_parallel_embedding.tensor_model_parallel_all_reduce", side_effect=lambda x: x) as mock_reduce_tp:
side_effect=lambda x: x) as mock_reduce_tp:
# Call the forward method # Call the forward method
output = layer.forward(input_) output = layer.forward(input_)
@@ -150,9 +148,8 @@ class TestCustomVocabParallelEmbedding(unittest.TestCase):
return_value=mock_output.clone()) return_value=mock_output.clone())
# Patch tensor_model_parallel_all_reduce to mock its behavior # Patch tensor_model_parallel_all_reduce to mock its behavior
with patch( with patch("torch.ops.vllm.maybe_pad_and_reduce",
"vllm_ascend.ops.vocab_parallel_embedding.tensor_model_parallel_all_reduce", side_effect=lambda x: x):
side_effect=lambda x: x):
# Call the forward method # Call the forward method
output = layer.forward(input_) output = layer.forward(input_)
# Check that invalid positions (0, 2, 4) were zeroed out # Check that invalid positions (0, 2, 4) were zeroed out
@@ -176,9 +173,8 @@ class TestCustomVocabParallelEmbedding(unittest.TestCase):
for input_, expected_shape in test_cases: for input_, expected_shape in test_cases:
with self.subTest(input=input_): with self.subTest(input=input_):
with patch( with patch("torch.ops.vllm.maybe_pad_and_reduce",
"vllm_ascend.ops.vocab_parallel_embedding.tensor_model_parallel_all_reduce", side_effect=lambda x: x):
side_effect=lambda x: x):
# Call the forward method # Call the forward method
output = layer.forward(input_) output = layer.forward(input_)
self.assertEqual(output.shape, expected_shape) self.assertEqual(output.shape, expected_shape)

5
tests/ut/torchair/models/test_torchair_deepseek_v2.py
View File

@@ -276,10 +276,7 @@ def test_torchair_deepseek_v2_mla_attention(mock_rms_norm, mock_distributed,
@patch("torch_npu.npu_add_rms_norm") @patch("torch_npu.npu_add_rms_norm")
@patch("torch_npu.npu_rms_norm") @patch("torch_npu.npu_rms_norm")
@patch("torch.ops.vllm.maybe_wait_prefetch_done", side_effect=lambda x: None) @patch("torch.ops.vllm.maybe_wait_prefetch_done", side_effect=lambda x: None)
@patch("torch.ops.vllm.maybe_chunk_residual", def test_torchair_deepseek_v2_decoder_layer(mock_maybe_wait_prefetch_done,
side_effect=lambda x, residual: residual)
def test_torchair_deepseek_v2_decoder_layer(mock_maybe_chunk_residual,
mock_maybe_wait_prefetch_done,
mock_rms_norm, mock_add_norm, mock_rms_norm, mock_add_norm,
mock_distributed, base_config, mock_distributed, base_config,
vllm_config, mock_forward_context): vllm_config, mock_forward_context):

1
vllm_ascend/ops/layernorm.py
View File

@@ -64,7 +64,6 @@ class AscendRMSNorm(RMSNorm):
import torch_npu import torch_npu
if residual is not None: if residual is not None:
residual = torch.ops.vllm.maybe_chunk_residual(x, residual)
assert x.size(0) == residual.size(0) assert x.size(0) == residual.size(0)
x, residual = _addrmsnorm_forward_oot( x, residual = _addrmsnorm_forward_oot(
self, x, residual, self.next_need_quant_fusion_linear) self, x, residual, self.next_need_quant_fusion_linear)

19
vllm_ascend/ops/linear.py
View File

@@ -34,8 +34,7 @@ from vllm.model_executor.layers.quantization.base_config import \
QuantizationConfig QuantizationConfig
from vllm.model_executor.utils import set_weight_attrs from vllm.model_executor.utils import set_weight_attrs
from vllm_ascend.ops.linear_op import (get_column_parallel_op, from vllm_ascend.ops.linear_op import get_parallel_op
get_row_parallel_op)
# TODO(realliujiaxu): Remove this class after linear of vllm supports custom comm group # TODO(realliujiaxu): Remove this class after linear of vllm supports custom comm group
@@ -100,8 +99,8 @@ class AscendQKVParallelLinear(QKVParallelLinear):
return_bias: bool = True, return_bias: bool = True,
disable_tp: bool = False, disable_tp: bool = False,
): ):
self.custom_op, _, tp_size = get_column_parallel_op( self.custom_op, _, tp_size = get_parallel_op(disable_tp, prefix, self,
disable_tp, prefix, self) "column")
# TODO(realliujiaxu): Replace the initialization code below with super().__init__ after linear of vllm supports custom comm group # TODO(realliujiaxu): Replace the initialization code below with super().__init__ after linear of vllm supports custom comm group
self.hidden_size = hidden_size self.hidden_size = hidden_size
self.head_size = head_size self.head_size = head_size
@@ -173,8 +172,8 @@ class AscendMergedColumnParallelLinear(MergedColumnParallelLinear):
return_bias: bool = True, return_bias: bool = True,
disable_tp: bool = False, disable_tp: bool = False,
): ):
self.custom_op, self.tp_rank, self.tp_size = get_column_parallel_op( self.custom_op, self.tp_rank, self.tp_size = get_parallel_op(
disable_tp, prefix, self) disable_tp, prefix, self, "column")
# TODO(realliujiaxu): Replace the initialization code below with super().__init__ after linear of vllm supports custom comm group # TODO(realliujiaxu): Replace the initialization code below with super().__init__ after linear of vllm supports custom comm group
self.output_sizes = output_sizes self.output_sizes = output_sizes
assert all(output_size % self.tp_size == 0 assert all(output_size % self.tp_size == 0
@@ -222,8 +221,8 @@ class AscendRowParallelLinear(RowParallelLinear):
return_bias: bool = True, return_bias: bool = True,
disable_tp: bool = False, disable_tp: bool = False,
): ):
self.custom_op, self.tp_rank, self.tp_size = get_row_parallel_op( self.custom_op, self.tp_rank, self.tp_size = get_parallel_op(
disable_tp, prefix, self) disable_tp, prefix, self, "row")
# TODO(realliujiaxu): Replace the initialization code below with super().__init__ after linear of vllm supports custom comm group # TODO(realliujiaxu): Replace the initialization code below with super().__init__ after linear of vllm supports custom comm group
# Divide the weight matrix along the first dimension. # Divide the weight matrix along the first dimension.
self.input_size_per_partition = divide(input_size, self.tp_size) self.input_size_per_partition = divide(input_size, self.tp_size)
@@ -304,8 +303,8 @@ class AscendColumnParallelLinear(ColumnParallelLinear):
return_bias: bool = True, return_bias: bool = True,
disable_tp: bool = False, disable_tp: bool = False,
): ):
self.custom_op, self.tp_rank, self.tp_size = get_column_parallel_op( self.custom_op, self.tp_rank, self.tp_size = get_parallel_op(
disable_tp, prefix, self) disable_tp, prefix, self, "column")
# TODO(realliujiaxu): Replace the initialization code below with super().__init__ after linear of vllm supports custom comm group # TODO(realliujiaxu): Replace the initialization code below with super().__init__ after linear of vllm supports custom comm group
self.input_size_per_partition = input_size self.input_size_per_partition = input_size
self.output_size_per_partition = divide(output_size, self.tp_size) self.output_size_per_partition = divide(output_size, self.tp_size)

187
vllm_ascend/ops/linear_op.py
View File

@@ -20,13 +20,12 @@ Current class inheritance structure:
CustomTensorParallelOp CustomTensorParallelOp
├── CustomColumnParallelOp ├── CustomColumnParallelOp
│ ├── MLPColumnParallelOp │ ├── MLPColumnParallelOp
│ ├── DenseOptimMergedColumnParallelOp │ ├── SequenceColumnParallelOp
│ └── DenseOptimQKVParallelOp
└── CustomRowParallelOp └── CustomRowParallelOp
├── MLPRowParallelOp ├── MLPRowParallelOp
├── OProjRowParallelOp ├── OProjRowParallelOp
├── MatmulAllreduceRowParallelOp ├── MatmulAllreduceRowParallelOp
└── DenseOptimRowParallelOp └── SequenceRowParallelOp
How to extend a new linear op? Taking column parallel op as an example: How to extend a new linear op? Taking column parallel op as an example:
1. Inherit from CustomColumnParallelOp and create a new class MyColumnParallelOp 1. Inherit from CustomColumnParallelOp and create a new class MyColumnParallelOp
@@ -36,7 +35,7 @@ How to extend a new linear op? Taking column parallel op as an example:
Row parallel op follows a similar approach - inherit from RowColumnParallelOp and register the new class in get_row_parallel_op. Row parallel op follows a similar approach - inherit from RowColumnParallelOp and register the new class in get_row_parallel_op.
""" """
from typing import Optional, Tuple, Union from typing import Optional, Union
import torch import torch
import torch.distributed as dist import torch.distributed as dist
@@ -153,69 +152,6 @@ class MLPColumnParallelOp(CustomColumnParallelOp):
return output, output_bias return output, output_bias
class SequenceMergedColumnParallelOp(CustomColumnParallelOp):
def apply_impl(
self, input_: torch.Tensor
) -> Union[torch.Tensor, tuple[torch.Tensor, Optional[Parameter]]]:
"""Linear layer with column parallelism.
Implemented multiple optimization projects for dense models, such as FlashComm and
communication-computation fusion.
"""
bias = self.bias if not self.skip_bias_add else None
# Matrix multiply.
assert self.quant_method is not None
input_ = torch.ops.vllm.maybe_all_gather_and_maybe_unpad(input_, True)
output_parallel = self.quant_method.apply(self.layer, input_, bias)
if self.gather_output:
# All-gather across the partitions.
output = self.comm_group.all_gather(output_parallel)
else:
output = output_parallel
output_bias = self.bias if self.skip_bias_add else None
return output, output_bias
class SequenceQKVParallelOp(CustomColumnParallelOp):
def __init__(self, layer, prefix):
super().__init__(layer)
self.prefix = prefix
def apply_impl(
self, input_: torch.Tensor
) -> Union[torch.Tensor, tuple[torch.Tensor, Optional[Parameter]]]:
"""Linear layer with column parallelism.
Implemented multiple optimization projects for dense models, such as FlashComm and
communication-computation fusion.
"""
bias = self.bias if not self.skip_bias_add else None
# Matrix multiply.
assert self.quant_method is not None
layer_num = self.prefix.split('.')[2]
input_ = torch.ops.vllm.maybe_all_gather_and_maybe_unpad(
input_, layer_num != '0')
output_parallel = self.quant_method.apply(self.layer, input_, bias)
if self.gather_output:
# All-gather across the partitions.
output = self.comm_group.all_gather(output_parallel)
else:
output = output_parallel
output_bias = self.bias if self.skip_bias_add else None
return output, output_bias
class MLPRowParallelOp(CustomRowParallelOp): class MLPRowParallelOp(CustomRowParallelOp):
def __init__(self, layer): def __init__(self, layer):
@@ -364,11 +300,35 @@ class MatmulAllreduceRowParallelOp(CustomRowParallelOp):
self.weight_t = self.layer.weight.t() self.weight_t = self.layer.weight.t()
class SequenceRowParallelOp(CustomRowParallelOp): class SequenceColumnParallelOp(CustomColumnParallelOp):
def __init__(self, layer, prefix): def apply_impl(
super().__init__(layer) self, input_: torch.Tensor
self.prefix = prefix ) -> Union[torch.Tensor, tuple[torch.Tensor, Optional[Parameter]]]:
"""Linear layer with column parallelism.
Implemented multiple optimization projects for dense models, such as FlashComm and
communication-computation fusion.
"""
bias = self.bias if not self.skip_bias_add else None
# Matrix multiply.
assert self.quant_method is not None
input_ = torch.ops.vllm.maybe_all_gather_and_maybe_unpad(input_, True)
output_parallel = self.quant_method.apply(self.layer, input_, bias)
if self.gather_output:
# All-gather across the partitions.
output = self.comm_group.all_gather(output_parallel)
else:
output = output_parallel
output_bias = self.bias if self.skip_bias_add else None
return output, output_bias
class SequenceRowParallelOp(CustomRowParallelOp):
def apply_impl( def apply_impl(
self, input_: torch.Tensor self, input_: torch.Tensor
@@ -408,50 +368,55 @@ class SequenceRowParallelOp(CustomRowParallelOp):
self.reduce_results = self.layer.reduce_results self.reduce_results = self.layer.reduce_results
def get_column_parallel_op( def _get_column_parallel_op(
disable_tp, prefix, layer prefix, layer
) -> Tuple[Optional[Union[MLPColumnParallelOp, SequenceMergedColumnParallelOp, ) -> Optional[Union[MLPColumnParallelOp, SequenceColumnParallelOp]]:
SequenceQKVParallelOp]], int, int]: if mlp_tp_enable() and "gate_up_proj" in prefix:
return MLPColumnParallelOp(layer)
if enable_sp():
if "shared_expert" in prefix:
return None
if "gate_up_proj" in prefix:
return SequenceColumnParallelOp(layer)
if "in_proj" in prefix:
return SequenceColumnParallelOp(layer)
if "qkv_proj" in prefix or "conv1d" in prefix:
return SequenceColumnParallelOp(layer)
return None
def _get_row_parallel_op(
prefix, layer
) -> Optional[Union[MLPRowParallelOp, OProjRowParallelOp,
MatmulAllreduceRowParallelOp, SequenceRowParallelOp]]:
if "down_proj" in prefix and mlp_tp_enable():
return MLPRowParallelOp(layer)
if "o_proj" in prefix and oproj_tp_enable():
return OProjRowParallelOp(layer)
if matmul_allreduce_enable():
return MatmulAllreduceRowParallelOp(layer)
if enable_sp():
if "shared_expert" in prefix:
return None
if "o_proj" in prefix or "out_proj" in prefix or "down_proj" in prefix:
return SequenceRowParallelOp(layer)
return None
def get_parallel_op(disable_tp, prefix, layer, direct):
if disable_tp: if disable_tp:
return None, 0, 1 return None, 0, 1
custom_op: Optional[Union[MLPColumnParallelOp, SequenceColumnParallelOp,
custom_op: Optional[Union[ MLPRowParallelOp, OProjRowParallelOp,
MLPColumnParallelOp,
SequenceMergedColumnParallelOp,
SequenceQKVParallelOp,
]] = None
if "gate_up_proj" in prefix and mlp_tp_enable():
custom_op = MLPColumnParallelOp(layer)
elif "gate_up_proj" in prefix and enable_sp():
custom_op = SequenceMergedColumnParallelOp(layer)
elif enable_sp():
custom_op = SequenceQKVParallelOp(layer, prefix)
if custom_op is not None:
return custom_op, custom_op.tp_rank, custom_op.tp_size
return None, get_tp_group().rank_in_group, get_tp_group().world_size
def get_row_parallel_op(
disable_tp, prefix, layer
) -> Tuple[Optional[Union[MLPRowParallelOp, OProjRowParallelOp,
MatmulAllreduceRowParallelOp,
SequenceRowParallelOp]], int, int]:
if disable_tp:
return None, 0, 1
custom_op: Optional[Union[MLPRowParallelOp, OProjRowParallelOp,
MatmulAllreduceRowParallelOp, MatmulAllreduceRowParallelOp,
SequenceRowParallelOp]] = None SequenceRowParallelOp]] = None
if "down_proj" in prefix and mlp_tp_enable(): if direct == "row":
custom_op = MLPRowParallelOp(layer) custom_op = _get_row_parallel_op(prefix, layer)
elif "o_proj" in prefix and oproj_tp_enable():
custom_op = OProjRowParallelOp(layer) if direct == "column":
elif matmul_allreduce_enable(): custom_op = _get_column_parallel_op(prefix, layer)
custom_op = MatmulAllreduceRowParallelOp(layer)
elif enable_sp():
custom_op = SequenceRowParallelOp(layer, prefix)
if custom_op is not None: if custom_op is not None:
return custom_op, custom_op.tp_rank, custom_op.tp_size return custom_op, custom_op.tp_rank, custom_op.tp_size

31
vllm_ascend/ops/register_custom_ops.py
View File

@@ -1,9 +1,7 @@
import torch import torch
import torch.nn.functional as F import torch.nn.functional as F
import torch_npu import torch_npu
from vllm.distributed import (get_tensor_model_parallel_rank, from vllm.distributed import (tensor_model_parallel_all_gather,
get_tensor_model_parallel_world_size,
tensor_model_parallel_all_gather,
tensor_model_parallel_all_reduce, tensor_model_parallel_all_reduce,
tensor_model_parallel_reduce_scatter) tensor_model_parallel_reduce_scatter)
from vllm.forward_context import get_forward_context from vllm.forward_context import get_forward_context
@@ -15,27 +13,6 @@ from vllm_ascend.ops.weight_prefetch import maybe_npu_prefetch
from vllm_ascend.utils import npu_stream_switch, prefetch_stream from vllm_ascend.utils import npu_stream_switch, prefetch_stream
def _maybe_chunk_residual_impl(x: torch.Tensor,
residual: torch.Tensor) -> torch.Tensor:
try:
forward_context = get_forward_context()
except AssertionError:
return residual
if x.size(0) != residual.size(0):
sp_enabled = forward_context.sp_enabled
assert sp_enabled is True, ("Currently, this situation only occurs "
"when sp is enabled")
pad_size = forward_context.pad_size
if pad_size > 0:
residual = F.pad(residual, (0, 0, 0, pad_size))
tp_size = get_tensor_model_parallel_world_size()
tp_rank = get_tensor_model_parallel_rank()
residual = torch.chunk(residual, tp_size, dim=0)[tp_rank]
return residual
def _maybe_all_gather_and_maybe_unpad_impl(x: torch.Tensor, def _maybe_all_gather_and_maybe_unpad_impl(x: torch.Tensor,
label: bool) -> torch.Tensor: label: bool) -> torch.Tensor:
try: try:
@@ -187,12 +164,6 @@ def _maybe_all_reduce_tensor_model_parallel_impl(
return tensor_model_parallel_all_reduce(final_hidden_states) return tensor_model_parallel_all_reduce(final_hidden_states)
direct_register_custom_op(op_name="maybe_chunk_residual",
op_func=_maybe_chunk_residual_impl,
fake_impl=lambda x, residual: residual,
mutates_args=[],
dispatch_key="PrivateUse1")
direct_register_custom_op(op_name="maybe_all_gather_and_maybe_unpad", direct_register_custom_op(op_name="maybe_all_gather_and_maybe_unpad",
op_func=_maybe_all_gather_and_maybe_unpad_impl, op_func=_maybe_all_gather_and_maybe_unpad_impl,
fake_impl=lambda x, label: x, fake_impl=lambda x, label: x,

4
vllm_ascend/ops/vocab_parallel_embedding.py
View File

@@ -20,7 +20,7 @@ from typing import Optional, Tuple
import torch import torch
from torch import nn from torch import nn
from torch.nn.parameter import Parameter from torch.nn.parameter import Parameter
from vllm.distributed import divide, tensor_model_parallel_all_reduce from vllm.distributed import divide
from vllm.distributed.parallel_state import get_tp_group from vllm.distributed.parallel_state import get_tp_group
from vllm.model_executor.layers.logits_processor import LogitsProcessor from vllm.model_executor.layers.logits_processor import LogitsProcessor
from vllm.model_executor.layers.quantization.base_config import ( from vllm.model_executor.layers.quantization.base_config import (
@@ -163,7 +163,7 @@ class AscendVocabParallelEmbedding(VocabParallelEmbedding):
if self.tp_size > 1: if self.tp_size > 1:
output_parallel.masked_fill_(input_mask.unsqueeze(-1), 0) output_parallel.masked_fill_(input_mask.unsqueeze(-1), 0)
# Reduce across all the model parallel GPUs. # Reduce across all the model parallel GPUs.
output = tensor_model_parallel_all_reduce(output_parallel) output = torch.ops.vllm.maybe_pad_and_reduce(output_parallel)
return output return output

38
vllm_ascend/torchair/ops/torchair_vocab_parallel_embedding.py Normal file
View File

@@ -0,0 +1,38 @@
#
# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
# This file is a part of the vllm-ascend project.
#
# 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 vllm.distributed import tensor_model_parallel_all_reduce
def vocab_embedding_forward(self, input_):
if self.tp_size > 1:
# Build the mask.
masked_input, input_mask = self._get_masked_input_and_mask(
input_, self.shard_indices.org_vocab_start_index,
self.shard_indices.org_vocab_end_index,
self.shard_indices.num_org_vocab_padding,
self.shard_indices.added_vocab_start_index,
self.shard_indices.added_vocab_end_index)
else:
masked_input = input_
# Get the embeddings.
output_parallel = self.quant_method.embedding(self, masked_input.long())
# Mask the output embedding.
if self.tp_size > 1:
output_parallel.masked_fill_(input_mask.unsqueeze(-1), 0)
# Reduce across all the model parallel GPUs.
output = tensor_model_parallel_all_reduce(output_parallel)
return output

5
vllm_ascend/torchair/utils.py
View File

@@ -208,11 +208,15 @@ def torchair_ops_patch():
from vllm_ascend.ops.layernorm import AscendRMSNorm from vllm_ascend.ops.layernorm import AscendRMSNorm
from vllm_ascend.ops.rotary_embedding import ( from vllm_ascend.ops.rotary_embedding import (
AscendDeepseekScalingRotaryEmbedding, AscendRotaryEmbedding) AscendDeepseekScalingRotaryEmbedding, AscendRotaryEmbedding)
from vllm_ascend.ops.vocab_parallel_embedding import \
AscendVocabParallelEmbedding
from vllm_ascend.torchair.ops import (torchair_activation, from vllm_ascend.torchair.ops import (torchair_activation,
torchair_layernorm) torchair_layernorm)
from vllm_ascend.torchair.ops.torchair_rotary_embedding import ( from vllm_ascend.torchair.ops.torchair_rotary_embedding import (
deepseek_rope_init_func, native_rope_deepseek_forward, deepseek_rope_init_func, native_rope_deepseek_forward,
qwen_rope_init_func, rope_forward) qwen_rope_init_func, rope_forward)
from vllm_ascend.torchair.ops.torchair_vocab_parallel_embedding import \
vocab_embedding_forward
AscendRotaryEmbedding.__init__ = qwen_rope_init_func # type: ignore[method-assign] AscendRotaryEmbedding.__init__ = qwen_rope_init_func # type: ignore[method-assign]
AscendRotaryEmbedding.forward_oot = rope_forward # type: ignore[method-assign] AscendRotaryEmbedding.forward_oot = rope_forward # type: ignore[method-assign]
@@ -222,3 +226,4 @@ def torchair_ops_patch():
AscendRMSNorm.forward_oot = torchair_layernorm.torchair_rmsnorm_forward_oot # type: ignore[method-assign] AscendRMSNorm.forward_oot = torchair_layernorm.torchair_rmsnorm_forward_oot # type: ignore[method-assign]
AscendSiluAndMul.forward_oot = torchair_activation.torchair_silu_and_mul_forward_oot # type: ignore[method-assign] AscendSiluAndMul.forward_oot = torchair_activation.torchair_silu_and_mul_forward_oot # type: ignore[method-assign]
AscendVocabParallelEmbedding.forward = vocab_embedding_forward # type: ignore[method-assign]