d3c93fba5c
### What this PR does / why we need it?
* **Unify execution paths:** Consolidates the quantized and
non-quantized execution paths into a single `fused_experts` function,
removing duplicated logic and making the control flow clearer and easier
to maintain.
* **W8A8 dynamic quantization:** Adds support for W8A8 dynamic
quantization inside the unified MoE kernel. Communication routines are
updated to correctly handle dynamic quantization scales for activations.
* **Weight pre-processing:** Prae-transpose the `w13` and `w2` weight
matrices (as implemented in PR #2025) so that quantized and
non-quantized models follow the same code path for the MoE gating,
up-projection, and down-projection operations.
* **All-to-all communication:** Adds an `all-to-all` collective
communication pattern. For large token counts on modern hardware,
`all-to-all` is more efficient than the previous `all-gather` strategy.
However, `all-to-all` is not really captured and replayed due to
multiple D2H operations which will trigger synchronization, and thus
raise error when capture graphs. We only use `all-to-all` when fallback
to `compiled_graph_for_general_shape`.
* **Dynamic communication selection:** The model runner now selects the
optimal MoE communication method (`mc2`, `allgather`, or `alltoall`) at
runtime based on token count and the Ascend SoC version.
* **Limitation:** `all-gather` is not yet supported for quantized
models, which means there is still something left to do on A2.
### Does this PR introduce _any_ user-facing change?
None.
### How was this patch tested?
No further test cases needed.
- vLLM version: v0.10.1.1
- vLLM main:
d660c98c1b
---------
Signed-off-by: Yizhou Liu <liu_yizhou@outlook.com>
111 lines
3.2 KiB
Python
111 lines
3.2 KiB
Python
#
|
|
# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
|
|
# Copyright 2023 The vLLM team.
|
|
#
|
|
# 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.
|
|
# Adapted from vllm/tests/basic_correctness/test_basic_correctness.py
|
|
#
|
|
"""Compare the short outputs of HF and vLLM when using greedy sampling.
|
|
|
|
Run `pytest tests/e2e/multicard/test_qwen3_moe.py`.
|
|
"""
|
|
|
|
import os
|
|
|
|
from modelscope import snapshot_download # type: ignore
|
|
|
|
from tests.e2e.conftest import VllmRunner
|
|
|
|
|
|
def test_models_distributed_Qwen3_MOE_TP2():
|
|
example_prompts = [
|
|
"Hello, my name is",
|
|
]
|
|
dtype = "half"
|
|
max_tokens = 5
|
|
with VllmRunner(
|
|
"Qwen/Qwen3-30B-A3B",
|
|
dtype=dtype,
|
|
tensor_parallel_size=2,
|
|
distributed_executor_backend="mp",
|
|
) as vllm_model:
|
|
vllm_model.generate_greedy(example_prompts, max_tokens)
|
|
|
|
|
|
def test_models_distributed_Qwen3_MOE_TP2_WITH_EP():
|
|
example_prompts = [
|
|
"Hello, my name is",
|
|
]
|
|
dtype = "half"
|
|
max_tokens = 5
|
|
with VllmRunner(
|
|
"Qwen/Qwen3-30B-A3B",
|
|
dtype=dtype,
|
|
tensor_parallel_size=2,
|
|
enable_expert_parallel=True,
|
|
distributed_executor_backend="mp",
|
|
enforce_eager=False,
|
|
) as vllm_model:
|
|
vllm_model.generate_greedy(example_prompts, max_tokens)
|
|
|
|
|
|
def test_models_distributed_Qwen3_MOE_W8A8():
|
|
example_prompts = [
|
|
"Hello, my name is",
|
|
]
|
|
dtype = "auto"
|
|
max_tokens = 5
|
|
with VllmRunner(
|
|
snapshot_download("vllm-ascend/Qwen3-30B-A3B-W8A8"),
|
|
max_model_len=8192,
|
|
dtype=dtype,
|
|
tensor_parallel_size=2,
|
|
quantization="ascend",
|
|
enforce_eager=True,
|
|
) as vllm_model:
|
|
vllm_model.generate_greedy(example_prompts, max_tokens)
|
|
|
|
|
|
def test_models_distributed_Qwen3_MOE_TP2_WITH_ACLGRAPH_AIV():
|
|
os.environ['HCCL_OP_EXPANSION_MODE'] = 'AIV'
|
|
example_prompts = [
|
|
"Hello, my name is",
|
|
]
|
|
dtype = "auto"
|
|
max_tokens = 5
|
|
with VllmRunner(
|
|
"Qwen/Qwen3-30B-A3B",
|
|
dtype=dtype,
|
|
tensor_parallel_size=2,
|
|
enforce_eager=False,
|
|
) as vllm_model:
|
|
vllm_model.generate_greedy(example_prompts, max_tokens)
|
|
|
|
|
|
def test_models_distributed_Qwen3_MOE_TP2_WITH_ACLGRAPH():
|
|
if 'HCCL_OP_EXPANSION_MODE' in os.environ:
|
|
del os.environ['HCCL_OP_EXPANSION_MODE']
|
|
example_prompts = [
|
|
"Hello, my name is",
|
|
]
|
|
dtype = "auto"
|
|
max_tokens = 5
|
|
with VllmRunner(
|
|
"Qwen/Qwen3-30B-A3B",
|
|
dtype=dtype,
|
|
tensor_parallel_size=2,
|
|
enforce_eager=False,
|
|
) as vllm_model:
|
|
vllm_model.generate_greedy(example_prompts, max_tokens)
|