LICENSE
MANIFEST.in
THIRDPARTYNOTICES
pyproject.toml
setup.py
requirements/common.txt
requirements/cuda-thirdparty.txt
requirements/cuda.txt
requirements/rocm-thirdparty.txt
requirements/rocm.txt
tokenspeed_kernel/__init__.py
tokenspeed_kernel/_triton.py
tokenspeed_kernel/platform.py
tokenspeed_kernel/profiling.py
tokenspeed_kernel/registry.py
tokenspeed_kernel/selection.py
tokenspeed_kernel/signature.py
tokenspeed_kernel/torch_compile.py
tokenspeed_kernel.egg-info/PKG-INFO
tokenspeed_kernel.egg-info/SOURCES.txt
tokenspeed_kernel.egg-info/dependency_links.txt
tokenspeed_kernel.egg-info/requires.txt
tokenspeed_kernel.egg-info/top_level.txt
tokenspeed_kernel/benchmark/__init__.py
tokenspeed_kernel/benchmark/__main__.py
tokenspeed_kernel/benchmark/cli.py
tokenspeed_kernel/benchmark/config.py
tokenspeed_kernel/benchmark/report.py
tokenspeed_kernel/benchmark/result.py
tokenspeed_kernel/benchmark/runner.py
tokenspeed_kernel/benchmark/throughput.py
tokenspeed_kernel/numerics/__init__.py
tokenspeed_kernel/numerics/__main__.py
tokenspeed_kernel/numerics/cli.py
tokenspeed_kernel/numerics/comparison.py
tokenspeed_kernel/numerics/gemm.py
tokenspeed_kernel/numerics/inputs.py
tokenspeed_kernel/numerics/moe.py
tokenspeed_kernel/numerics/quantize.py
tokenspeed_kernel/numerics/tolerance.py
tokenspeed_kernel/numerics/verify.py
tokenspeed_kernel/numerics/reference/__init__.py
tokenspeed_kernel/numerics/reference/gemm.py
tokenspeed_kernel/numerics/reference/quantize.py
tokenspeed_kernel/ops/__init__.py
tokenspeed_kernel/ops/tuning.py
tokenspeed_kernel/ops/activation/__init__.py
tokenspeed_kernel/ops/activation/cuda.py
tokenspeed_kernel/ops/activation/flashinfer.py
tokenspeed_kernel/ops/activation/triton.py
tokenspeed_kernel/ops/attention/__init__.py
tokenspeed_kernel/ops/attention/gdn_utils.py
tokenspeed_kernel/ops/attention/msa.py
tokenspeed_kernel/ops/attention/msa_score.py
tokenspeed_kernel/ops/attention/score_mods.py
tokenspeed_kernel/ops/attention/cuda/__init__.py
tokenspeed_kernel/ops/attention/cuda/deepseek_v4.py
tokenspeed_kernel/ops/attention/cuda/dsa_topk.py
tokenspeed_kernel/ops/attention/cute_dsl/__init__.py
tokenspeed_kernel/ops/attention/cute_dsl/cute_utils.py
tokenspeed_kernel/ops/attention/cute_dsl/dsa_topk.py
tokenspeed_kernel/ops/attention/cute_dsl/minimax_index_decode_score.py
tokenspeed_kernel/ops/attention/cute_dsl/rel_mha/__init__.py
tokenspeed_kernel/ops/attention/cute_dsl/rel_mha/cu_blocks_kernels.py
tokenspeed_kernel/ops/attention/cute_dsl/rel_mha/flash_fwd_combine.py
tokenspeed_kernel/ops/attention/cute_dsl/rel_mha/flash_fwd_sm100_bias.py
tokenspeed_kernel/ops/attention/cute_dsl/rel_mha/flash_fwd_sm100_bias_decode.py
tokenspeed_kernel/ops/attention/cute_dsl/rel_mha/fmha_bias_helper.py
tokenspeed_kernel/ops/attention/cute_dsl/rel_mha/rel_decode.py
tokenspeed_kernel/ops/attention/cute_dsl/rel_mha/rel_decode_v2.py
tokenspeed_kernel/ops/attention/cute_dsl/rel_mha/rel_extend.py
tokenspeed_kernel/ops/attention/cute_dsl/rel_mha/shearing_bias.py
tokenspeed_kernel/ops/attention/deep_gemm/__init__.py
tokenspeed_kernel/ops/attention/flash_attn/__init__.py
tokenspeed_kernel/ops/attention/flash_mla/__init__.py
tokenspeed_kernel/ops/attention/flashinfer/__init__.py
tokenspeed_kernel/ops/attention/flashinfer/dsa_topk.py
tokenspeed_kernel/ops/attention/flashinfer/gated_delta_rule.py
tokenspeed_kernel/ops/attention/gluon/__init__.py
tokenspeed_kernel/ops/attention/tokenspeed_mla/__init__.py
tokenspeed_kernel/ops/attention/triton/__init__.py
tokenspeed_kernel/ops/attention/triton/context.py
tokenspeed_kernel/ops/attention/triton/deepseek_v4.py
tokenspeed_kernel/ops/attention/triton/dsa.py
tokenspeed_kernel/ops/attention/triton/dsa_topk.py
tokenspeed_kernel/ops/attention/triton/gated_delta_rule.py
tokenspeed_kernel/ops/attention/triton/gdn_qkv_split.py
tokenspeed_kernel/ops/attention/triton/merge_state.py
tokenspeed_kernel/ops/attention/triton/mha_decode.py
tokenspeed_kernel/ops/attention/triton/mha_prefill.py
tokenspeed_kernel/ops/attention/triton/minimax_indexer.py
tokenspeed_kernel/ops/attention/triton/minimax_sparse_attention.py
tokenspeed_kernel/ops/attention/triton/mla_decode.py
tokenspeed_kernel/ops/attention/triton/mla_prefill.py
tokenspeed_kernel/ops/attention/triton/qkv_rotary.py
tokenspeed_kernel/ops/attention/triton/rel_mha.py
tokenspeed_kernel/ops/attention/triton/linear/__init__.py
tokenspeed_kernel/ops/attention/triton/linear/chunk.py
tokenspeed_kernel/ops/attention/triton/linear/chunk_delta_h.py
tokenspeed_kernel/ops/attention/triton/linear/chunk_o.py
tokenspeed_kernel/ops/attention/triton/linear/chunk_scaled_dot_kkt.py
tokenspeed_kernel/ops/attention/triton/linear/cumsum.py
tokenspeed_kernel/ops/attention/triton/linear/index.py
tokenspeed_kernel/ops/attention/triton/linear/l2norm.py
tokenspeed_kernel/ops/attention/triton/linear/op.py
tokenspeed_kernel/ops/attention/triton/linear/solve_tril.py
tokenspeed_kernel/ops/attention/triton/linear/utils.py
tokenspeed_kernel/ops/attention/triton/linear/wy_fast.py
tokenspeed_kernel/ops/communication/__init__.py
tokenspeed_kernel/ops/communication/deep_ep.py
tokenspeed_kernel/ops/communication/flashinfer.py
tokenspeed_kernel/ops/communication/iris.py
tokenspeed_kernel/ops/communication/nccl.py
tokenspeed_kernel/ops/communication/triton.py
tokenspeed_kernel/ops/communication/trtllm.py
tokenspeed_kernel/ops/conv/__init__.py
tokenspeed_kernel/ops/conv/triton.py
tokenspeed_kernel/ops/embedding/__init__.py
tokenspeed_kernel/ops/embedding/cuda.py
tokenspeed_kernel/ops/embedding/flashinfer.py
tokenspeed_kernel/ops/embedding/triton.py
tokenspeed_kernel/ops/gemm/__init__.py
tokenspeed_kernel/ops/gemm/cuda.py
tokenspeed_kernel/ops/gemm/cute_dsl.py
tokenspeed_kernel/ops/gemm/deep_gemm.py
tokenspeed_kernel/ops/gemm/flashinfer.py
tokenspeed_kernel/ops/gemm/fp8_utils.py
tokenspeed_kernel/ops/gemm/gluon.py
tokenspeed_kernel/ops/gemm/triton.py
tokenspeed_kernel/ops/gemm/trtllm.py
tokenspeed_kernel/ops/kvcache/__init__.py
tokenspeed_kernel/ops/kvcache/cuda.py
tokenspeed_kernel/ops/kvcache/triton.py
tokenspeed_kernel/ops/layernorm/__init__.py
tokenspeed_kernel/ops/layernorm/cuda.py
tokenspeed_kernel/ops/layernorm/flashinfer.py
tokenspeed_kernel/ops/layernorm/triton.py
tokenspeed_kernel/ops/moe/__init__.py
tokenspeed_kernel/ops/moe/cuda.py
tokenspeed_kernel/ops/moe/flashinfer/__init__.py
tokenspeed_kernel/ops/moe/flashinfer/cutedsl_deepep_nvfp4.py
tokenspeed_kernel/ops/moe/flashinfer/cutlass_fp8.py
tokenspeed_kernel/ops/moe/flashinfer/cutlass_nvfp4.py
tokenspeed_kernel/ops/moe/flashinfer/cutlass_unquant.py
tokenspeed_kernel/ops/moe/flashinfer/trtllm_fp8.py
tokenspeed_kernel/ops/moe/flashinfer/trtllm_mxfp4.py
tokenspeed_kernel/ops/moe/flashinfer/trtllm_mxint4.py
tokenspeed_kernel/ops/moe/flashinfer/trtllm_nvfp4.py
tokenspeed_kernel/ops/moe/flashinfer/trtllm_unquant.py
tokenspeed_kernel/ops/moe/gluon/__init__.py
tokenspeed_kernel/ops/moe/gluon/bf16.py
tokenspeed_kernel/ops/moe/gluon/mxfp4.py
tokenspeed_kernel/ops/moe/triton/__init__.py
tokenspeed_kernel/ops/moe/triton/fp8.py
tokenspeed_kernel/ops/moe/triton/inkling_topk.py
tokenspeed_kernel/ops/moe/triton/mxfp4.py
tokenspeed_kernel/ops/quantization/__init__.py
tokenspeed_kernel/ops/quantization/cuda.py
tokenspeed_kernel/ops/quantization/flashinfer.py
tokenspeed_kernel/ops/quantization/triton.py
tokenspeed_kernel/ops/quantization/trtllm.py
tokenspeed_kernel/ops/sampling/__init__.py
tokenspeed_kernel/ops/sampling/cuda.py
tokenspeed_kernel/ops/sampling/cute_dsl.py
tokenspeed_kernel/ops/sampling/flashinfer.py
tokenspeed_kernel/ops/sampling/gluon/__init__.py
tokenspeed_kernel/ops/sampling/triton/__init__.py
tokenspeed_kernel/ops/sampling/triton/common.py
tokenspeed_kernel/ops/sampling/triton/generic.py
tokenspeed_kernel/ops/sampling/triton/gumbel.py
tokenspeed_kernel/ops/sampling/triton/logprobs.py
tokenspeed_kernel/ops/sampling/triton/min_p.py
tokenspeed_kernel/ops/sampling/triton/penalties.py
tokenspeed_kernel/ops/sampling/triton/probability.py
tokenspeed_kernel/ops/sampling/triton/topk_topp.py
tokenspeed_kernel/ops/sampling/triton/topp.py
tokenspeed_kernel/ops/sampling/triton/verify.py
tokenspeed_kernel/ops/transform/__init__.py
tokenspeed_kernel/ops/transform/weak_ref.py
tokenspeed_kernel/ops/transform/faster_hadamard_transform/__init__.py
tokenspeed_kernel/ops/transform/triton/__init__.py
tokenspeed_kernel/plugins/__init__.py
tokenspeed_kernel/plugins/__main__.py
tokenspeed_kernel/plugins/cli.py
tokenspeed_kernel/thirdparty/__init__.py
tokenspeed_kernel/thirdparty/cuda/__init__.py
tokenspeed_kernel/thirdparty/cuda/activation.py
tokenspeed_kernel/thirdparty/cuda/cuda_ipc.py
tokenspeed_kernel/thirdparty/cuda/deepseek_v4_attention.py
tokenspeed_kernel/thirdparty/cuda/dsv3_gemm.py
tokenspeed_kernel/thirdparty/cuda/flashinfer_softmax.py
tokenspeed_kernel/thirdparty/cuda/fused_topk_topp.py
tokenspeed_kernel/thirdparty/cuda/kvcacheio.py
tokenspeed_kernel/thirdparty/cuda/lm_head_gemm.py
tokenspeed_kernel/thirdparty/cuda/marlin.py
tokenspeed_kernel/thirdparty/cuda/merge_state.py
tokenspeed_kernel/thirdparty/cuda/minimax_m3_fused.py
tokenspeed_kernel/thirdparty/cuda/moe.py
tokenspeed_kernel/thirdparty/cuda/rmsnorm.py
tokenspeed_kernel/thirdparty/cuda/rope.py
tokenspeed_kernel/thirdparty/cuda/routing.py
tokenspeed_kernel/thirdparty/cuda/sampling_chain.py
tokenspeed_kernel/thirdparty/cuda/trtllm.py
tokenspeed_kernel/thirdparty/cute_dsl/__init__.py
tokenspeed_kernel/thirdparty/cute_dsl/argmax.py
tokenspeed_kernel/thirdparty/cute_dsl/nvfp4_gemm_swiglu_nvfp4_quant.py
tokenspeed_kernel/thirdparty/cute_dsl/topk/__init__.py
tokenspeed_kernel/thirdparty/cute_dsl/topk/runner.py
tokenspeed_kernel/thirdparty/cute_dsl/topk/utils.py
tokenspeed_kernel/thirdparty/cute_dsl/topk/top_k/__init__.py
tokenspeed_kernel/thirdparty/cute_dsl/topk/top_k/block_scan.py
tokenspeed_kernel/thirdparty/cute_dsl/topk/top_k/filtered_top_k_varlen_util.py
tokenspeed_kernel/thirdparty/cute_dsl/topk/top_k/single_pass_multi_cta_radix_topk.py
tokenspeed_kernel/thirdparty/cute_dsl/topk/top_k/single_pass_multi_cta_radix_topk_cluster.py
tokenspeed_kernel/thirdparty/deep_gemm/__init__.py
tokenspeed_kernel/thirdparty/deep_gemm/warmup.py
tokenspeed_kernel/thirdparty/deep_gemm/utils/__init__.py
tokenspeed_kernel/thirdparty/deep_gemm/utils/layout.py
tokenspeed_kernel/thirdparty/msa/README.md
tokenspeed_kernel/thirdparty/msa/__init__.py
tokenspeed_kernel/thirdparty/msa/api.py
tokenspeed_kernel/thirdparty/msa/jit.py
tokenspeed_kernel/thirdparty/msa/sparse.py
tokenspeed_kernel/thirdparty/msa/sparse_fmha_adapter.py
tokenspeed_kernel/thirdparty/msa/csrc/fmha_sm100_inst.jinja
tokenspeed_kernel/thirdparty/msa/csrc/fmha_sm100_params.h
tokenspeed_kernel/thirdparty/msa/csrc/fmha_sm100_plan.cu
tokenspeed_kernel/thirdparty/msa/csrc/fmha_sm100_plan.h
tokenspeed_kernel/thirdparty/msa/csrc/fmha_sm100_reduction.cu
tokenspeed_kernel/thirdparty/msa/csrc/fmha_sm100_reduction.h
tokenspeed_kernel/thirdparty/msa/csrc/fmha_sm100_variant_run.cu.jinja
tokenspeed_kernel/thirdparty/msa/csrc/gmem_bounds_check.h
tokenspeed_kernel/thirdparty/msa/csrc/sparse_topk_select.cu
tokenspeed_kernel/thirdparty/msa/csrc/tvm_ffi_utils.h
tokenspeed_kernel/thirdparty/msa/csrc/include/allocator.h
tokenspeed_kernel/thirdparty/msa/csrc/include/cutlass_utils.cuh
tokenspeed_kernel/thirdparty/msa/csrc/include/exception.h
tokenspeed_kernel/thirdparty/msa/csrc/include/fmha.hpp
tokenspeed_kernel/thirdparty/msa/csrc/include/fmha_common.hpp
tokenspeed_kernel/thirdparty/msa/csrc/include/fmha_cutlass_sm100.cuh
tokenspeed_kernel/thirdparty/msa/csrc/include/fmha_fusion.hpp
tokenspeed_kernel/thirdparty/msa/csrc/include/fmha_options.hpp
tokenspeed_kernel/thirdparty/msa/csrc/include/fmha_tile_scheduler.hpp
tokenspeed_kernel/thirdparty/msa/csrc/include/gather_tensor.hpp
tokenspeed_kernel/thirdparty/msa/csrc/include/gpu_trace.h
tokenspeed_kernel/thirdparty/msa/csrc/include/mask.cuh
tokenspeed_kernel/thirdparty/msa/csrc/include/plan.cuh
tokenspeed_kernel/thirdparty/msa/csrc/include/pow_2.hpp
tokenspeed_kernel/thirdparty/msa/csrc/include/sm100_fmha_fwd_epilogue_tma_warpspecialized.hpp
tokenspeed_kernel/thirdparty/msa/csrc/include/sm100_fmha_fwd_kernel_tma_warpspecialized.hpp
tokenspeed_kernel/thirdparty/msa/csrc/include/sm100_fmha_fwd_mainloop_tma_warpspecialized.hpp
tokenspeed_kernel/thirdparty/msa/csrc/include/sm100_fmha_load_tma_warpspecialized.hpp
tokenspeed_kernel/thirdparty/msa/csrc/include/sm100_fmha_reduction.hpp
tokenspeed_kernel/thirdparty/msa/csrc/include/sparse_topk_select.cuh
tokenspeed_kernel/thirdparty/msa/csrc/include/utils.cuh
tokenspeed_kernel/thirdparty/msa/csrc/include/vec_dtypes.cuh
tokenspeed_kernel/thirdparty/msa/cute/README.md
tokenspeed_kernel/thirdparty/msa/cute/fp4_indexer_interface.py
tokenspeed_kernel/thirdparty/msa/cute/interface.py
tokenspeed_kernel/thirdparty/msa/cute/quantize.py
tokenspeed_kernel/thirdparty/msa/cute/requirements.txt
tokenspeed_kernel/thirdparty/msa/cute/sparse_index_utils.py
tokenspeed_kernel/thirdparty/msa/cute/src/__init__.py
tokenspeed_kernel/thirdparty/msa/cute/src/common/__init__.py
tokenspeed_kernel/thirdparty/msa/cute/src/common/aot_cache.py
tokenspeed_kernel/thirdparty/msa/cute/src/common/barrier.py
tokenspeed_kernel/thirdparty/msa/cute/src/common/blackwell_helpers.py
tokenspeed_kernel/thirdparty/msa/cute/src/common/block_info.py
tokenspeed_kernel/thirdparty/msa/cute/src/common/copy_utils.py
tokenspeed_kernel/thirdparty/msa/cute/src/common/cute_dsl_utils.py
tokenspeed_kernel/thirdparty/msa/cute/src/common/fast_math.py
tokenspeed_kernel/thirdparty/msa/cute/src/common/mask.py
tokenspeed_kernel/thirdparty/msa/cute/src/common/mma_sm100_desc.py
tokenspeed_kernel/thirdparty/msa/cute/src/common/named_barrier.py
tokenspeed_kernel/thirdparty/msa/cute/src/common/pack_gqa.py
tokenspeed_kernel/thirdparty/msa/cute/src/common/paged_kv.py
tokenspeed_kernel/thirdparty/msa/cute/src/common/pipeline.py
tokenspeed_kernel/thirdparty/msa/cute/src/common/seqlen_info.py
tokenspeed_kernel/thirdparty/msa/cute/src/common/softmax.py
tokenspeed_kernel/thirdparty/msa/cute/src/common/tile_scheduler.py
tokenspeed_kernel/thirdparty/msa/cute/src/common/tma_utils.py
tokenspeed_kernel/thirdparty/msa/cute/src/common/utils.py
tokenspeed_kernel/thirdparty/msa/cute/src/sm100/__init__.py
tokenspeed_kernel/thirdparty/msa/cute/src/sm100/decode_schedule.py
tokenspeed_kernel/thirdparty/msa/cute/src/sm100/fp4_indexer.py
tokenspeed_kernel/thirdparty/msa/cute/src/sm100/prepare_k2q_csr.py
tokenspeed_kernel/thirdparty/msa/cute/src/sm100/prepare_scheduler.py
tokenspeed_kernel/thirdparty/msa/cute/src/sm100/build_k2q_csr/__init__.py
tokenspeed_kernel/thirdparty/msa/cute/src/sm100/build_k2q_csr/build_k2q_csr.cu
tokenspeed_kernel/thirdparty/msa/cute/src/sm100/fwd/__init__.py
tokenspeed_kernel/thirdparty/msa/cute/src/sm100/fwd/atten_fwd.py
tokenspeed_kernel/thirdparty/msa/cute/src/sm100/fwd/atten_fwd_nvfp4_kv.py
tokenspeed_kernel/thirdparty/msa/cute/src/sm100/fwd/combine.py
tokenspeed_kernel/thirdparty/msa/cute/src/sm100/fwd_decode/__init__.py
tokenspeed_kernel/thirdparty/msa/cute/src/sm100/fwd_decode/atten_fwd.py
tokenspeed_kernel/thirdparty/msa/cute/src/sm100/fwd_decode/combine.py
tokenspeed_kernel/thirdparty/msa/cute/src/sm100/fwd_decode/tile_scheduler.py
tokenspeed_kernel/thirdparty/msa/cute/src/sm100/fwd_decode/build_decode_schedule/__init__.py
tokenspeed_kernel/thirdparty/msa/cute/src/sm100/fwd_decode/build_decode_schedule/build_decode_schedule.cu
tokenspeed_kernel/thirdparty/triton/__init__.py
tokenspeed_kernel/thirdparty/trtllm/__init__.py