dylanyunlong/project_6 - project_6 - Gitea: Git with a cup of tea

dylanyunlong/project_6

Files

1 line

139 B

Python

Raw Normal View History

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 02:31:56 +00:00

from .llama_decoder_layer_overlap import LlamaDecoderLayerOverlapProtocol, LlamaDecoderLayerOverlapDefault, create_vllm_llama_decoder_layer