Files
project_6/ixformer_sdk/inference/functions/__init__.py
project6-dev 87a19d2d00 feat(CRITICAL): 从 GitHub 扫描搬运 ixformer SDK + xllm 完整 GDN/MoE 代码
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
2026-08-11 02:32:06 +00:00

45 lines
1.1 KiB
Python

from .act_and_mul import *
from .act_bias_mm import *
from .add import *
from .bert import *
from .bnb_dequant import *
from .bnb_double_quant import *
from .bnb_mm_dequant import *
from .bnb_qgemm import *
from .bnb_quant import *
from .bnb_rowcol_absmax import *
from .conv2d import *
from .cross_entropy_loss import *
from .flash_attn import *
from .flash_attn_lib import *
from .fused_rope import *
from .gemv import *
from .groupnorm import *
from .i8w8o32 import *
from .layernorm import *
from .lightllm import *
from .linalg import *
from .linear import *
from .lmdeploy import *
from .marlin import *
from .matmul import *
from .mla_fused import *
from .mm import *
from .moe import *
from .overlap_comm import *
from .paged_attention import *
from .quantized_linear import *
from .residual_bias import *
from .rms_norm import *
from .scaled_dot_product_attention import *
from .smoothquant import *
from .softmax import *
from .store_kv_cache import *
from .t5 import *
from .tgi import *
from .vllm import *
from .w8a8 import *
from .w8a16 import *
from .wi4a16 import *
from .wui4a16 import *