cuLA
CUDA kernels for linear attention variants, written in CuTe DSL and CUTLASS C++.
파일 탐색기
- build-release.yml
- lint.yml
- pull_request_template.md
- bench_chunk_delta_h.py
- bench_fused_kda_conv_mtp.py
- bench_fwd_o.py
- bench_intracard_cp.py
- bench_intracard_cp_sm90.py
- bench_kda.py
- bench_kda_bwd_wy_dqkg_sm100.py
- bench_kda_chunk_intra.py
- bench_kda_decode.py
- bench_kda_decode_mtp.py
- bench_kda_fwd_bwd_e2e.py
- bench_kda_packed_decode.py
- bench_kda_sm90_cp.py
- bench_kda_sm90_prefill.py
- bench_la_decode_mtp.py
- bench_la_decode_vs_fla.py
- bench_la_kvbuffer.py
- bench_lightning_attn_prefill.py
- bench_linear_attn.py
- bench_recompute_wu.py
- utils.py
- kda_sm100.cu
- kda_sm90.cu
- pybind.cu
- pybind_sm100.cu
- pybind_sm90.cu
- fwd_helpers.hpp
- kda_config.hpp
- kda_fwd_common.cuh
- kda_fwd_intra_kernel_sm100.hpp
- kda_fwd_intra_mainloop_sm100.hpp
- kda_fwd_recomp_w_u_kernel_sm100.hpp
- kda_fwd_recomp_w_u_mainloop_sm100.hpp
- kda_fwd_sm100.cu
- sm100_umma_ext.hpp
- tile_scheduler.hpp
- common.hpp
- load_predicated.hpp
- load_tma.hpp
- mainloop_kda_fwd.hpp
- named_barriers.hpp
- store_tma.hpp
- device_universal.hpp
- builder_kda_fwd.hpp
- kernel_kda_fwd.hpp
- options.hpp
- tile_scheduler.hpp
- common.hpp
- debug.hpp
- math.hpp
- math_order_barrier.hpp
- type_traits.hpp
- unused.hpp
- kda_fwd_sm90.cu
- kda_fwd_sm90_safe_gate.cu
- prefill_kernel.hpp
- prefill_kernel_kda_fwd_sm90.cuh
- common.h
- cute_ext.hpp
- static_switch.h
- helpers.cuh
- intrinsics.cuh
- collective_inverse.hpp
- helpers.cuh
- helpers.cuh
- common.h
- device.cuh
- host.h
- kerutils.cuh
- cutlass
- __init__.py
- flashkda.py
- fully_fused.py
- __init__.py
- auto_route.py
- chunk.py
- chunk_bwd.py
- chunk_fwd.py
- chunk_intra.py
- cp_context.py
- cp_h_boundary.py
- flashkda.py
- gate_l2norm_fused.py
- hopper_fused_fwd.py
- hopper_fused_fwd_opt.py
- l2norm_qk_fused.py
- README.md
- wy_intra.py
- wy_recompute.py
- __init__.py
- la_decode_mtp.py
- la_state_update_kvbuffer.py
- la_verify_kvbuffer.py
- __init__.py
- linear_attn_prototype.py
- __init__.py
- cute.py
- mtp.py
- mtp_conv.py
- mtp_conv_kvbuffer.py
- mtp_kvbuffer.py
- reference_fla.py
- __init__.py
- __init__.py
- chunk_delta_h.py
- merge.py
- pre_scan.py
- __init__.py
- bwd_wy_dqkg.py
- delta_h.py
- fwd_o.py
- policy.py
- __init__.py
- driver.py
- merge.py
- plan.py
- pre_scan.py
- __init__.py
- _common.py
- fwd.py
- k1.py
- k2.py
- __init__.py
- cp_mode.py
- __init__.py
- prefill_kernel.py
- schedule.py
- __init__.py
- decode.py
- prefill.py
- prefill_sm100.py
- prefill_sm90.py
- __init__.py
- ptx.py
- __init__.py
- _cutedsl_compat.py
- _mlir_compat.py
- inv.py
- ptx.py
- __init__.py
- backends.py
- cudac.py
- utils.py
- chunk_delta_h_pipeline.md
- cuLA-logo.png
- fwd_o_pipeline.md
- lightning_attn_pipeline.md
- lightning_attn_sm90.md
- build_wheel.sh
- modal_validate.py
- conftest.py
- lightning_attn_reference.py
- test_compare_with_fla.py
- test_cutedsl_compat.py
- test_intracard_cp_sm90.py
- test_kda_conv_mtp.py
- test_kda_conv_mtp_kvbuffer.py
- test_kda_decode.py
- test_kda_decode_mtp.py
- test_kda_packed_decode.py
- test_kda_sm100_chunk_vs_fla.py
- test_kda_sm100_chunk_vs_naive.py
- test_kda_sm100_intracard_cp.py
- test_kda_sm100_output.py
- test_kda_sm100_recurrence.py
- test_kda_sm90_intracard_cp.py
- test_kda_sm90_prefill_vs_fla.py
- test_la_decode_mtp.py
- test_la_kvbuffer.py
- test_lightning_attn_prefill_dispatch.py
- test_lightning_attn_prefill_sm90.py
- test_lightning_decode.py
- test_lightning_decode_pool.py
- test_lightning_sm100_prefill.py
- test_ptx_umma_masked.py
- test_ptx_umma_ws.py
- flash-linear-attention
- .clang-format
- .gitignore
- .gitmodules
- .pre-commit-config.yaml
- BENCHMARK_GB200.md
- BENCHMARK_GB200_CUDA_130.md
- BENCHMARK_GB300.md
- BENCHMARK_H200.md
- BENCHMARK_KDA_DECODE_GB200.md
- BENCHMARK_KDA_DECODE_H203E.md
- CITATION.cff
- LICENSE
- MANIFEST.in
- pyproject.toml
- README.md
- RECOMMENDED_CODING_STYLE.md
- REPO_LAYOUT.md
- setup.py
- USAGE.md
# CDN으로 사용하기
jsDelivrjsDelivr는 공개 GitHub 리포지토리를 별도 설정 없이 CDN으로 즉시 서빙합니다. 버전과 파일을 고르면 웹페이지에 바로 붙일 수 있는 링크와 예시 코드가 만들어집니다.
명령어 용어집
이 문서에서 사용된 명령어를 모아봤습니다. 낯선 명령어가 있다면 펼쳐서 확인해보세요.
git clone
설명 보기 ▼
git clone
Clone an existing repository.
git clone {{remote_repository_location}} {{path/to/directory}}
Clone an existing repository into a new directory (the default directory is the repository name):
git clone --recursive {{remote_repository_location}}
Clone an existing repository and its submodules:
git clone {{[-n|--no-checkout]}} {{remote_repository_location}}
Clone only the `.git` directory of an existing repository:
git submodule
설명 보기 ▼
git submodule
Inspect, update, and manage submodules.
git submodule
View existing submodules and the checked-out commit for each one:
git submodule update --init --recursive
Install a repository's submodules (listed in `.gitmodules`):
git submodule add {{repository_url}}
Add a Git repository as a submodule of the current one:
nvcc
설명 보기 ▼
nvcc
NVIDIA CUDA 컴파일러 드라이버.
더 많은 정보: <https://docs.nvidia.com/cuda/cuda-compiler-driver-nvcc/>.
nvcc {{path/to/source.cu}} {{[-o|--output-file]}} {{path/to/executable}}
CUDA 프로그램 컴파일:
nvcc {{path/to/source.cu}} {{[-o|--output-file]}} {{path/to/executable}} {{[-g|--debug]}} {{[-G|--device-debug]}}
디버그 정보 생성:
nvcc {{path/to/source.cu}} {{[-o|--output-file]}} {{path/to/executable}} {{[-I|--include-path]}} {{path/to/includes}} {{[-L|--library-path]}} {{path/to/library}} {{[-l|--library]}} {{library_name}}
다른 경로에서 라이브러리 포함:
pip install
설명 보기 ▼
pip install
Install Python packages.
pip install {{package1 package2 ...}}
Install one or more packages:
pip install {{package1 package2 ...}} {{[-U|--upgrade]}}
Upgrade all specified packages to the latest version, installing any that are not already present:
pip install {{package}}=={{version}}
Install a specific version of a package:
python
설명 보기 ▼
python
Python 언어 인터프리터.
더 많은 정보: <https://docs.python.org/using/cmdline.html>.
python
REPL(대화형 셸) 시작:
python {{path/to/file.py}}
특정 Python 파일 실행:
python -i {{path/to/file.py}}
특정 Python 파일 실행 후 REPL 시작:
TAG
설명 보기 ▼
TAG
Edit tags on Mac OS X files (10.9 Mavericks and above).
tag {{[-a|--add]}} {{tag_name1,tag_name2,...}} {{path/to/file}}
Add tags to a file:
tag {{[-r|--remove]}} {{tag_name}} {{path/to/file}}
Remove a tag:
tag {{[-r|--remove]}} \* {{path/to/file}}
Remove all tags from a file:
