9d1452c74d
### What this PR does / why we need it?
This PR introduces a new fused Triton kernel,
`split_qkv_tp_rmsnorm_rope` for Minimax-m2.5.
The implementation includes two Triton kernels:
1. `_split_qkv_and_compute_local_qk_var_kernel`: Splits the QKV input
and computes the local variance for RMSNorm.
2. `_apply_global_rmsnorm_kernel`: Applies global RMSNorm (considering
TP all-reduce for variance) and Neox-style RoPE.
### Does this PR introduce _any_ user-facing change?
Does not.
### How was this patch tested?
```python
pytest tests/e2e/nightly/single_node/ops/singlecard_ops/triton/test_split_qkv_tp_rmsnorm_rope.py
```
### Test Data
A3 TP16
基线
| data | TTFT(ms) | TPOT(ms) | TPS |
|------------|---------:|---------:|-------:|
| 4k/1k@bs1 | 267.55 | 25.5 | 38.85 |
| 4k/1k@bs4 | 542.4 | 26.51 | 148.06 |
测试线
| data | TTFT(ms) | TPOT(ms) | TPS |
|------------|---------:|---------:|-------:|
| 4k/1k@bs1 | 234.64 | 20.96 | 47.24 |
| 4k/1k@bs4 | 508.36 | 22.16 | 176.69 |
- vLLM version: v0.17.0
- vLLM main:
4034c3d32e
Signed-off-by: xutianyi <xutianyi5@huawei.com>
Co-authored-by: xutianyi <xutianyi5@huawei.com>
54 lines
2.3 KiB
Python
54 lines
2.3 KiB
Python
#
|
|
# Copyright (c) 2025 Huawei Technologies Co., Ltd. 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.
|
|
# This file is a part of the vllm-ascend project.
|
|
#
|
|
|
|
import torch
|
|
from vllm.triton_utils import HAS_TRITON
|
|
|
|
import vllm_ascend.ops.fused_moe.fused_moe # noqa
|
|
import vllm_ascend.ops.layernorm # noqa
|
|
import vllm_ascend.ops.register_custom_ops # noqa
|
|
|
|
if HAS_TRITON:
|
|
import vllm_ascend.ops.triton.linearnorm.split_qkv_rmsnorm_rope # noqa
|
|
import vllm_ascend.ops.triton.linearnorm.split_qkv_rmsnorm_mrope
|
|
import vllm_ascend.ops.triton.linearnorm.split_qkv_tp_rmsnorm_rope
|
|
|
|
import vllm_ascend.ops.vocab_parallel_embedding # noqa
|
|
from vllm_ascend.ops.activation import AscendQuickGELU, AscendSiluAndMul
|
|
from vllm_ascend.ops.rotary_embedding import AscendDeepseekScalingRotaryEmbedding, AscendRotaryEmbedding
|
|
|
|
|
|
class dummyFusionOp:
|
|
default = None
|
|
|
|
def __init__(self, name=""):
|
|
self.name = name
|
|
|
|
|
|
def register_dummy_fusion_op() -> None:
|
|
torch.ops._C_ascend.rms_norm = dummyFusionOp(name="rms_norm")
|
|
torch.ops._C_ascend.fused_add_rms_norm = dummyFusionOp(name="fused_add_rms_norm")
|
|
torch.ops._C_ascend.static_scaled_fp8_quant = dummyFusionOp(name="static_scaled_fp8_quant")
|
|
torch.ops._C_ascend.dynamic_scaled_fp8_quant = dummyFusionOp(name="dynamic_scaled_fp8_quant")
|
|
torch.ops._C_ascend.dynamic_per_token_scaled_fp8_quant = dummyFusionOp(name="dynamic_per_token_scaled_fp8_quant")
|
|
torch.ops._C_ascend.rms_norm_static_fp8_quant = dummyFusionOp(name="rms_norm_static_fp8_quant")
|
|
torch.ops._C_ascend.fused_add_rms_norm_static_fp8_quant = dummyFusionOp(name="fused_add_rms_norm_static_fp8_quant")
|
|
torch.ops._C_ascend.rms_norm_dynamic_per_token_quant = dummyFusionOp(name="rms_norm_dynamic_per_token_quant")
|
|
|
|
|
|
__all__ = ["AscendQuickGELU", "AscendSiluAndMul", "AscendRotaryEmbedding", "AscendDeepseekScalingRotaryEmbedding"]
|