87a19d2d00
来源:
1. Chranos/ixformer (GitHub) → ixformer_sdk/ (230 files, 70K lines)
- inference/functions/vllm.py: vllm_moe_topk_softmax 完整实现 (2033 lines)
- inference/functions/moe.py: MoE ops 完整实现 (1380 lines)
- contrib/vllm_flash_attn/: FA2 Python 接口 (1018 lines)
- contrib/tgi/fused_moe.py: TGI fused MoE (429 lines)
- csrc/include/ixformer/: C++ kernel headers + cmake
2. Deep-Spark/xllm (GitHub) → upstream_ref/xllm_latest/ (+15 files)
- npu_torch/qwen3_5_decoder_layer_impl.cpp/.h
- npu_torch/qwen3_5_gated_delta_net.cpp/.h
- npu_torch/qwen3_next_*.cpp/.h (6 files)
- npu_torch/attention.cpp/.h + fused_moe.cpp/.h + CMakeLists.txt
- models/llm/qwen3_5.h + qwen3_5_mtp.h + qwen3_next.h
- models/vlm/qwen3_5.h
调用链完整性:
ixformer_sdk/inference/functions/vllm.py
→ ops.infer.moe_topk_softmax() (C++ 层)
→ 这就是 base 镜像 libixformer.so 里的实现
upstream_ref/xllm_latest/core/layers/ilu/fused_moe.cpp
→ ixformer::infer::topk_softmax() (直接 C++ 调用)
→ ixformer::infer::group_gemm() → 完整 7-step MoE pipeline
58 lines
2.0 KiB
Python
58 lines
2.0 KiB
Python
from typing import Callable, Dict, List, Union
|
|
from torch.nn import Module
|
|
|
|
from ixformer.train.speedformer.policy.utils import SubModuleReplacementDescription
|
|
from ixformer.train.speedformer.policy.replacer import Replacer
|
|
|
|
from ixformer.train.speedformer.layers.normalization import APEXFusedRMSNorm, IXFFusedRMSNorm
|
|
from ixformer.train.speedformer.layers.chatglm.attention import ChatglmFlashAttention
|
|
from ixformer.train.speedformer.layers.chatglm.methods import ChatGLMModel_forward
|
|
|
|
|
|
class ChatglmReplacer(Replacer):
|
|
def __init__(self):
|
|
self.policy = {}
|
|
|
|
def module_policy(self) -> Dict[str | Module, List[SubModuleReplacementDescription]]:
|
|
self.append_or_create_submodule_replacement(
|
|
description=[
|
|
SubModuleReplacementDescription(
|
|
suffix="final_layernorm",
|
|
target_module=APEXFusedRMSNorm,
|
|
kwargs={}
|
|
),
|
|
],
|
|
target_key="GLMTransformer"
|
|
)
|
|
self.append_or_create_submodule_replacement(
|
|
description=[
|
|
SubModuleReplacementDescription(
|
|
suffix="input_layernorm",
|
|
target_module=APEXFusedRMSNorm,
|
|
kwargs={}
|
|
),
|
|
SubModuleReplacementDescription(
|
|
suffix="post_attention_layernorm",
|
|
target_module=APEXFusedRMSNorm,
|
|
kwargs={}
|
|
),
|
|
],
|
|
target_key="GLMBlock"
|
|
)
|
|
self.append_or_create_submodule_replacement(
|
|
description=[
|
|
SubModuleReplacementDescription(
|
|
suffix="self_attention",
|
|
target_module=ChatglmFlashAttention,
|
|
kwargs={}
|
|
),
|
|
],
|
|
target_key="GLMBlock"
|
|
)
|
|
self.append_or_create_method_replacement(
|
|
description=[
|
|
{"forward": ChatGLMModel_forward()}
|
|
],
|
|
target_key="ChatGLMModel"
|
|
)
|