LICENSE
MANIFEST.in
THIRDPARTYNOTICES
pyproject.toml
setup.py
requirements/cuda-thirdparty.txt
requirements/cuda.txt
requirements/rocm-thirdparty.txt
requirements/rocm.txt
tokenspeed_kernel/__init__.py
tokenspeed_kernel/_triton.py
tokenspeed_kernel/compile_monitor.py
tokenspeed_kernel/platform.py
tokenspeed_kernel/profiling.py
tokenspeed_kernel/registry.py
tokenspeed_kernel/selection.py
tokenspeed_kernel/signature.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/ci.py
tokenspeed_kernel/benchmark/cli.py
tokenspeed_kernel/benchmark/config.py
tokenspeed_kernel/benchmark/graph.py
tokenspeed_kernel/benchmark/harness.py
tokenspeed_kernel/benchmark/profiler.py
tokenspeed_kernel/benchmark/proton.py
tokenspeed_kernel/benchmark/report.py
tokenspeed_kernel/benchmark/result.py
tokenspeed_kernel/benchmark/runner.py
tokenspeed_kernel/benchmark/throughput.py
tokenspeed_kernel/benchmark/validation.py
tokenspeed_kernel/benchmark/generators/__init__.py
tokenspeed_kernel/benchmark/generators/dsa.py
tokenspeed_kernel/benchmark/generators/gemm.py
tokenspeed_kernel/benchmark/generators/kda.py
tokenspeed_kernel/benchmark/generators/mla.py
tokenspeed_kernel/benchmark/generators/moe.py
tokenspeed_kernel/benchmark/generators/residual.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/cuda.py
tokenspeed_kernel/ops/attention/triton.py
tokenspeed_kernel/ops/attention/_cute_dsl/__init__.py
tokenspeed_kernel/ops/attention/_cute_dsl/utils.py
tokenspeed_kernel/ops/attention/_triton/__init__.py
tokenspeed_kernel/ops/attention/_triton/prefill_state_checkpoints.py
tokenspeed_kernel/ops/attention/dsa/__init__.py
tokenspeed_kernel/ops/attention/dsa/cuda.py
tokenspeed_kernel/ops/attention/dsa/cute_dsl.py
tokenspeed_kernel/ops/attention/dsa/deep_gemm.py
tokenspeed_kernel/ops/attention/dsa/flashinfer.py
tokenspeed_kernel/ops/attention/dsa/gluon.py
tokenspeed_kernel/ops/attention/dsa/triton.py
tokenspeed_kernel/ops/attention/dsa/_cute_dsl/__init__.py
tokenspeed_kernel/ops/attention/dsa/_cute_dsl/deep_select.py
tokenspeed_kernel/ops/attention/dsa/_triton/__init__.py
tokenspeed_kernel/ops/attention/dsa/_triton/index_candidates.py
tokenspeed_kernel/ops/attention/dsa/_triton/topk.py
tokenspeed_kernel/ops/attention/dsv4/__init__.py
tokenspeed_kernel/ops/attention/dsv4/cuda.py
tokenspeed_kernel/ops/attention/dsv4/deep_gemm.py
tokenspeed_kernel/ops/attention/dsv4/gluon.py
tokenspeed_kernel/ops/attention/dsv4/triton.py
tokenspeed_kernel/ops/attention/dsv4/_triton/__init__.py
tokenspeed_kernel/ops/attention/dsv4/_triton/dcp.py
tokenspeed_kernel/ops/attention/dsv4/_triton/indexer.py
tokenspeed_kernel/ops/attention/dsv41/__init__.py
tokenspeed_kernel/ops/attention/dsv41/cute_dsl.py
tokenspeed_kernel/ops/attention/dsv41/deep_gemm.py
tokenspeed_kernel/ops/attention/dsv41/deep_select.py
tokenspeed_kernel/ops/attention/dsv41/flash_mla.py
tokenspeed_kernel/ops/attention/dsv41/gluon.py
tokenspeed_kernel/ops/attention/dsv41/triton.py
tokenspeed_kernel/ops/attention/dsv41/_cute_dsl/__init__.py
tokenspeed_kernel/ops/attention/dsv41/_cute_dsl/sparse_index_scores.py
tokenspeed_kernel/ops/attention/dsv41/_gluon/__init__.py
tokenspeed_kernel/ops/attention/dsv41/_gluon/indexer.py
tokenspeed_kernel/ops/attention/gdn/__init__.py
tokenspeed_kernel/ops/attention/gdn/flashinfer.py
tokenspeed_kernel/ops/attention/gdn/triton.py
tokenspeed_kernel/ops/attention/gdn/_flashinfer/__init__.py
tokenspeed_kernel/ops/attention/gdn/_flashinfer/adapter.py
tokenspeed_kernel/ops/attention/gdn/_flashinfer/pdl.py
tokenspeed_kernel/ops/attention/gdn/_triton/__init__.py
tokenspeed_kernel/ops/attention/gdn/_triton/causal_conv1d_metadata.py
tokenspeed_kernel/ops/attention/gdn/_triton/chunk.py
tokenspeed_kernel/ops/attention/gdn/_triton/chunk_delta_h.py
tokenspeed_kernel/ops/attention/gdn/_triton/chunk_o.py
tokenspeed_kernel/ops/attention/gdn/_triton/chunk_scaled_dot_kkt.py
tokenspeed_kernel/ops/attention/gdn/_triton/cumsum.py
tokenspeed_kernel/ops/attention/gdn/_triton/index.py
tokenspeed_kernel/ops/attention/gdn/_triton/l2norm.py
tokenspeed_kernel/ops/attention/gdn/_triton/op.py
tokenspeed_kernel/ops/attention/gdn/_triton/prefill_state_inputs.py
tokenspeed_kernel/ops/attention/gdn/_triton/qkv_split.py
tokenspeed_kernel/ops/attention/gdn/_triton/solve_tril.py
tokenspeed_kernel/ops/attention/gdn/_triton/utils.py
tokenspeed_kernel/ops/attention/gdn/_triton/wy_fast.py
tokenspeed_kernel/ops/attention/kda/__init__.py
tokenspeed_kernel/ops/attention/kda/_prefill_capacity.py
tokenspeed_kernel/ops/attention/kda/cuda.py
tokenspeed_kernel/ops/attention/kda/cute_dsl.py
tokenspeed_kernel/ops/attention/kda/gluon.py
tokenspeed_kernel/ops/attention/kda/triton.py
tokenspeed_kernel/ops/attention/kda/_triton/__init__.py
tokenspeed_kernel/ops/attention/kda/_triton/capture_payload.py
tokenspeed_kernel/ops/attention/kda/_triton/fla.py
tokenspeed_kernel/ops/attention/kda/_triton/prefill_scan_inputs.py
tokenspeed_kernel/ops/attention/kda/_triton/recurrent.py
tokenspeed_kernel/ops/attention/kda/_triton/state_pages.py
tokenspeed_kernel/ops/attention/kpool/__init__.py
tokenspeed_kernel/ops/attention/kpool/deep_gemm.py
tokenspeed_kernel/ops/attention/kpool/gluon.py
tokenspeed_kernel/ops/attention/kpool/triton.py
tokenspeed_kernel/ops/attention/kpool/_triton/__init__.py
tokenspeed_kernel/ops/attention/kpool/_triton/cache.py
tokenspeed_kernel/ops/attention/kpool/_triton/expand.py
tokenspeed_kernel/ops/attention/kpool/_triton/score.py
tokenspeed_kernel/ops/attention/mamba2/__init__.py
tokenspeed_kernel/ops/attention/mamba2/triton.py
tokenspeed_kernel/ops/attention/mha/__init__.py
tokenspeed_kernel/ops/attention/mha/cuda.py
tokenspeed_kernel/ops/attention/mha/flashinfer.py
tokenspeed_kernel/ops/attention/mha/gluon.py
tokenspeed_kernel/ops/attention/mha/triton.py
tokenspeed_kernel/ops/attention/mha/_triton/__init__.py
tokenspeed_kernel/ops/attention/mha/_triton/context.py
tokenspeed_kernel/ops/attention/mha/_triton/decode.py
tokenspeed_kernel/ops/attention/mha/_triton/prefill.py
tokenspeed_kernel/ops/attention/mha/_triton/qkv_rotary.py
tokenspeed_kernel/ops/attention/mla/__init__.py
tokenspeed_kernel/ops/attention/mla/cuda.py
tokenspeed_kernel/ops/attention/mla/gluon.py
tokenspeed_kernel/ops/attention/mla/tokenspeed_mla.py
tokenspeed_kernel/ops/attention/mla/triton.py
tokenspeed_kernel/ops/attention/mla/_triton/__init__.py
tokenspeed_kernel/ops/attention/mla/_triton/decode.py
tokenspeed_kernel/ops/attention/mla/_triton/page_table.py
tokenspeed_kernel/ops/attention/mla/_triton/prefill.py
tokenspeed_kernel/ops/attention/msa/__init__.py
tokenspeed_kernel/ops/attention/msa/cuda.py
tokenspeed_kernel/ops/attention/msa/cute_dsl.py
tokenspeed_kernel/ops/attention/msa/triton.py
tokenspeed_kernel/ops/attention/msa/_cute_dsl/__init__.py
tokenspeed_kernel/ops/attention/msa/_cute_dsl/decode_score.py
tokenspeed_kernel/ops/attention/msa/_triton/__init__.py
tokenspeed_kernel/ops/attention/msa/_triton/indexer.py
tokenspeed_kernel/ops/attention/prologue/__init__.py
tokenspeed_kernel/ops/attention/prologue/checks.py
tokenspeed_kernel/ops/attention/prologue/composite.py
tokenspeed_kernel/ops/attention/prologue/triton.py
tokenspeed_kernel/ops/attention/prologue/types.py
tokenspeed_kernel/ops/attention/qsa/__init__.py
tokenspeed_kernel/ops/attention/qsa/cute_dsl.py
tokenspeed_kernel/ops/attention/qsa/flashinfer.py
tokenspeed_kernel/ops/attention/qsa/triton.py
tokenspeed_kernel/ops/attention/qsa/_flashinfer/__init__.py
tokenspeed_kernel/ops/attention/qsa/_flashinfer/runner.py
tokenspeed_kernel/ops/attention/rmha/__init__.py
tokenspeed_kernel/ops/attention/rmha/cuda.py
tokenspeed_kernel/ops/attention/rmha/cute_dsl.py
tokenspeed_kernel/ops/attention/rmha/gluon.py
tokenspeed_kernel/ops/attention/rmha/triton.py
tokenspeed_kernel/ops/attention/rmha/_cute_dsl/__init__.py
tokenspeed_kernel/ops/attention/rmha/_cute_dsl/cu_blocks_kernels.py
tokenspeed_kernel/ops/attention/rmha/_cute_dsl/flash_fwd_combine.py
tokenspeed_kernel/ops/attention/rmha/_cute_dsl/flash_fwd_sm100_bias.py
tokenspeed_kernel/ops/attention/rmha/_cute_dsl/flash_fwd_sm100_bias_decode.py
tokenspeed_kernel/ops/attention/rmha/_cute_dsl/fmha_bias_helper.py
tokenspeed_kernel/ops/attention/rmha/_cute_dsl/rel_decode.py
tokenspeed_kernel/ops/attention/rmha/_cute_dsl/rel_decode_v2.py
tokenspeed_kernel/ops/attention/rmha/_cute_dsl/rel_extend.py
tokenspeed_kernel/ops/attention/rmha/_cute_dsl/score_mods.py
tokenspeed_kernel/ops/attention/rmha/_cute_dsl/shearing_bias.py
tokenspeed_kernel/ops/attention/rmha/_triton/__init__.py
tokenspeed_kernel/ops/attention/rmha/_triton/log_scaling.py
tokenspeed_kernel/ops/attention/tree/__init__.py
tokenspeed_kernel/ops/attention/tree/triton.py
tokenspeed_kernel/ops/communication/__init__.py
tokenspeed_kernel/ops/communication/allreduce_fusion.py
tokenspeed_kernel/ops/communication/cuda_lamport.py
tokenspeed_kernel/ops/communication/deep_ep.py
tokenspeed_kernel/ops/communication/fabric.py
tokenspeed_kernel/ops/communication/flashinfer.py
tokenspeed_kernel/ops/communication/iris.py
tokenspeed_kernel/ops/communication/multimem.py
tokenspeed_kernel/ops/communication/nccl.py
tokenspeed_kernel/ops/communication/triton.py
tokenspeed_kernel/ops/communication/trtllm.py
tokenspeed_kernel/ops/communication/_cuda/chunk_a2a.cu
tokenspeed_kernel/ops/communication/_cuda/lamport_a2a.cu
tokenspeed_kernel/ops/conv/__init__.py
tokenspeed_kernel/ops/conv/triton/__init__.py
tokenspeed_kernel/ops/conv/triton/grouped.py
tokenspeed_kernel/ops/conv/triton/sconv.py
tokenspeed_kernel/ops/copy/__init__.py
tokenspeed_kernel/ops/copy/cuda.py
tokenspeed_kernel/ops/embedding/__init__.py
tokenspeed_kernel/ops/embedding/ascend.py
tokenspeed_kernel/ops/embedding/cuda.py
tokenspeed_kernel/ops/embedding/flashinfer.py
tokenspeed_kernel/ops/embedding/host_gather.py
tokenspeed_kernel/ops/embedding/triton.py
tokenspeed_kernel/ops/embedding/triton_host_gather.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/kimi3.py
tokenspeed_kernel/ops/gemm/linear_attnres_partials.py
tokenspeed_kernel/ops/gemm/ll_bf16.py
tokenspeed_kernel/ops/gemm/triton.py
tokenspeed_kernel/ops/gemm/triton_gemv.py
tokenspeed_kernel/ops/gemm/trtllm.py
tokenspeed_kernel/ops/gemm/trtllm_cutedsl.py
tokenspeed_kernel/ops/kvcache/__init__.py
tokenspeed_kernel/ops/kvcache/cuda.py
tokenspeed_kernel/ops/kvcache/host_transfer.py
tokenspeed_kernel/ops/kvcache/per_token_head.py
tokenspeed_kernel/ops/kvcache/triton.py
tokenspeed_kernel/ops/kvcache/triton_cache_placement.py
tokenspeed_kernel/ops/layernorm/__init__.py
tokenspeed_kernel/ops/layernorm/ascend.py
tokenspeed_kernel/ops/layernorm/cuda.py
tokenspeed_kernel/ops/layernorm/flashinfer.py
tokenspeed_kernel/ops/layernorm/triton.py
tokenspeed_kernel/ops/metadata/__init__.py
tokenspeed_kernel/ops/metadata/accepted_frontier.py
tokenspeed_kernel/ops/metadata/ngram_history.py
tokenspeed_kernel/ops/metadata/prep_tape.py
tokenspeed_kernel/ops/metadata/triton_tape.py
tokenspeed_kernel/ops/moe/__init__.py
tokenspeed_kernel/ops/moe/cuda.py
tokenspeed_kernel/ops/moe/dispatch.py
tokenspeed_kernel/ops/moe/latent_decode.py
tokenspeed_kernel/ops/moe/latent_down.py
tokenspeed_kernel/ops/moe/latent_input.py
tokenspeed_kernel/ops/moe/latent_tail.py
tokenspeed_kernel/ops/moe/mega_moe.py
tokenspeed_kernel/ops/moe/multicast_view.py
tokenspeed_kernel/ops/moe/native.py
tokenspeed_kernel/ops/moe/sigmoid_topk.py
tokenspeed_kernel/ops/moe/softmax_topk.py
tokenspeed_kernel/ops/moe/deep_gemm/__init__.py
tokenspeed_kernel/ops/moe/deep_gemm/deepep_fp8.py
tokenspeed_kernel/ops/moe/deep_gemm/mega_moe_mxfp4.py
tokenspeed_kernel/ops/moe/deep_gemm/ue8m0.py
tokenspeed_kernel/ops/moe/deep_gemm/_triton/__init__.py
tokenspeed_kernel/ops/moe/deep_gemm/_triton/mega_moe_stage.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_mxfp4.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/fp8.py
tokenspeed_kernel/ops/moe/gluon/latent_decode.py
tokenspeed_kernel/ops/moe/gluon/latent_input.py
tokenspeed_kernel/ops/moe/gluon/mxfp4.py
tokenspeed_kernel/ops/moe/gluon/petit.py
tokenspeed_kernel/ops/moe/gluon/sigmoid_topk.py
tokenspeed_kernel/ops/moe/gluon/sqrt_softplus_topk.py
tokenspeed_kernel/ops/moe/marlin/__init__.py
tokenspeed_kernel/ops/moe/marlin/deepep_layout.py
tokenspeed_kernel/ops/moe/marlin/deepep_mxfp4.py
tokenspeed_kernel/ops/moe/marlin/mxfp4.py
tokenspeed_kernel/ops/moe/triton/__init__.py
tokenspeed_kernel/ops/moe/triton/_common.py
tokenspeed_kernel/ops/moe/triton/bf16.py
tokenspeed_kernel/ops/moe/triton/decode_sigmoid_topk.py
tokenspeed_kernel/ops/moe/triton/deepep_permute.py
tokenspeed_kernel/ops/moe/triton/fp8.py
tokenspeed_kernel/ops/moe/triton/inkling_topk.py
tokenspeed_kernel/ops/moe/triton/kimi3_sigmoid_topk.py
tokenspeed_kernel/ops/moe/triton/latent_input.py
tokenspeed_kernel/ops/moe/triton/minimax_topk.py
tokenspeed_kernel/ops/moe/triton/mxfp4.py
tokenspeed_kernel/ops/moe/triton/softmax_topk.py
tokenspeed_kernel/ops/moe/triton/sqrt_softplus_topk.py
tokenspeed_kernel/ops/ple/__init__.py
tokenspeed_kernel/ops/ple/triton.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/residual/__init__.py
tokenspeed_kernel/ops/residual/cuda.py
tokenspeed_kernel/ops/residual/cute_fused.py
tokenspeed_kernel/ops/residual/deep_gemm.py
tokenspeed_kernel/ops/residual/gluon.py
tokenspeed_kernel/ops/residual/torch.py
tokenspeed_kernel/ops/residual/triton.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/logprob.py
tokenspeed_kernel/ops/sampling/torch.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/draft_tree.py
tokenspeed_kernel/ops/sampling/triton/dspark_block.py
tokenspeed_kernel/ops/sampling/triton/generic.py
tokenspeed_kernel/ops/sampling/triton/gumbel.py
tokenspeed_kernel/ops/sampling/triton/logprob_topk.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/spec_path.py
tokenspeed_kernel/ops/sampling/triton/topk_topp.py
tokenspeed_kernel/ops/sampling/triton/topp.py
tokenspeed_kernel/ops/sampling/triton/tree_verify.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/deep_select.py
tokenspeed_kernel/thirdparty/cuda/__init__.py
tokenspeed_kernel/thirdparty/cuda/activation.py
tokenspeed_kernel/thirdparty/cuda/attn_res.py
tokenspeed_kernel/thirdparty/cuda/cuda_ipc.py
tokenspeed_kernel/thirdparty/cuda/dsv3_gemm.py
tokenspeed_kernel/thirdparty/cuda/dsv4_attention.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/marlin_moe.py
tokenspeed_kernel/thirdparty/cuda/memcpy_2d.py
tokenspeed_kernel/thirdparty/cuda/merge_state.py
tokenspeed_kernel/thirdparty/cuda/mhc.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/hc_fused.py
tokenspeed_kernel/thirdparty/cute_dsl/nvfp4_gemm_swiglu_nvfp4_quant.py
tokenspeed_kernel/thirdparty/cute_dsl/qsa_sparse.py
tokenspeed_kernel/thirdparty/cute_dsl/latent_moe_tail/__init__.py
tokenspeed_kernel/thirdparty/cute_dsl/latent_moe_tail/allreduce_rmsnorm_reduce_scatter_early_exit.py
tokenspeed_kernel/thirdparty/cute_dsl/latent_moe_tail/fused_add_multicast_gemm.py
tokenspeed_kernel/thirdparty/cute_dsl/latent_moe_tail/fused_add_multicast_skinny_gemm.py
tokenspeed_kernel/thirdparty/cute_dsl/latent_moe_tail/fused_multicast_latent_down_gemm.py
tokenspeed_kernel/thirdparty/cute_dsl/latent_moe_tail/lamport_copy.py
tokenspeed_kernel/thirdparty/cute_dsl/latent_moe_tail/lamport_copy_nvfp4_quant.py
tokenspeed_kernel/thirdparty/cute_dsl/latent_moe_tail/primitives.py
tokenspeed_kernel/thirdparty/cute_dsl/ll_bf16/__init__.py
tokenspeed_kernel/thirdparty/cute_dsl/ll_bf16/_kernel.py
tokenspeed_kernel/thirdparty/cute_dsl/ll_bf16/_splitk_kernel.py
tokenspeed_kernel/thirdparty/cute_dsl/mega_moe/__init__.py
tokenspeed_kernel/thirdparty/cute_dsl/mega_moe/blocked_scale.py
tokenspeed_kernel/thirdparty/cute_dsl/mega_moe/contract.py
tokenspeed_kernel/thirdparty/cute_dsl/mega_moe/custom_ext.py
tokenspeed_kernel/thirdparty/cute_dsl/mega_moe/dynamic_mainloop.py
tokenspeed_kernel/thirdparty/cute_dsl/mega_moe/epilogue_refactor.py
tokenspeed_kernel/thirdparty/cute_dsl/mega_moe/fc1_fc2_fuse_sched.py
tokenspeed_kernel/thirdparty/cute_dsl/mega_moe/flag_batch.py
tokenspeed_kernel/thirdparty/cute_dsl/mega_moe/grid_sync.py
tokenspeed_kernel/thirdparty/cute_dsl/mega_moe/iket_compat.py
tokenspeed_kernel/thirdparty/cute_dsl/mega_moe/kernel_fc12.py
tokenspeed_kernel/thirdparty/cute_dsl/mega_moe/megamoe_constants.py
tokenspeed_kernel/thirdparty/cute_dsl/mega_moe/megamoe_kernel.py
tokenspeed_kernel/thirdparty/cute_dsl/mega_moe/moe_persistent_scheduler.py
tokenspeed_kernel/thirdparty/cute_dsl/mega_moe/moe_utils.py
tokenspeed_kernel/thirdparty/cute_dsl/mega_moe/ptx_helpers.py
tokenspeed_kernel/thirdparty/cute_dsl/mega_moe/runner.py
tokenspeed_kernel/thirdparty/cute_dsl/mega_moe/sf_swizzle.py
tokenspeed_kernel/thirdparty/cute_dsl/mega_moe/sym_buffer.py
tokenspeed_kernel/thirdparty/cute_dsl/mega_moe/token_comm.py
tokenspeed_kernel/thirdparty/cute_dsl/mega_moe/topk_reduce.py
tokenspeed_kernel/thirdparty/cute_dsl/mnnvl_k3_ht/__init__.py
tokenspeed_kernel/thirdparty/cute_dsl/mnnvl_k3_ht/device_kernel.py
tokenspeed_kernel/thirdparty/cute_dsl/mnnvl_k3_ht/primitives.py
tokenspeed_kernel/thirdparty/cute_dsl/mnnvl_k3_ht/protocol.py
tokenspeed_kernel/thirdparty/cute_dsl/skinny_gemm/__init__.py
tokenspeed_kernel/thirdparty/cute_dsl/skinny_gemm/_kernel.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/cutedsl_kda/__init__.py
tokenspeed_kernel/thirdparty/flashinfer/__init__.py
tokenspeed_kernel/thirdparty/flashinfer/allreduce_fusion.py
tokenspeed_kernel/thirdparty/flashinfer/fp8_gemm.py
tokenspeed_kernel/thirdparty/flashinfer/jit.py
tokenspeed_kernel/thirdparty/flashinfer/moe_alltoall.py
tokenspeed_kernel/thirdparty/flashinfer/mxfp8.py
tokenspeed_kernel/thirdparty/flashinfer/trtllm_moe.py
tokenspeed_kernel/thirdparty/gluon_petit/LICENSE.txt
tokenspeed_kernel/thirdparty/gluon_petit/README.md
tokenspeed_kernel/thirdparty/gluon_petit/__init__.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/__init__.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/streams.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/gemm/__init__.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/gemm/rocm/__init__.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/gemm/rocm/cdna4_ops.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/gemm/rocm/intrinsics.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/gemm/rocm/quantization/__init__.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/__init__.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/__init__.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/fused_moe.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/mega_moe.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/mega_moe_config_selector.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/memory_ops.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/quantization.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/comm/__init__.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/comm/barrier.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/mega_moe/__init__.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/mega_moe/mega_moe_two_stage_kernel.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/mega_moe/scheduler.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/mega_moe/workspace.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/mem/__init__.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/mem/bias.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/mem/input_mxfp4_packed.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/mem/weight_mxfp4.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/ops/__init__.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/ops/activation.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/ops/mxfp4_activation.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/ops/op_stages.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/ops/schedule_matmul.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/ops/schedule_tiles.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/ops/stage1_accumulator_lds.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/ops/mega_moe/__init__.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/ops/mega_moe/route_output.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/ops/mega_moe/token_shuffle_common.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/moe/rocm/ops/mega_moe/token_shuffle_direct_push.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/pybind/__init__.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/pybind/bindings.cc
tokenspeed_kernel/thirdparty/gluon_petit/lib/pybind/mega_moe.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/pybind/pybind.h
tokenspeed_kernel/thirdparty/gluon_petit/lib/pybind/vmm_symmetric_heap.cc
tokenspeed_kernel/thirdparty/gluon_petit/lib/pybind/vmm_symmetric_heap.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/tal/__init__.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/tal/device.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/tal/host_device.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/tal/tensor_ops.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/tal/tensor/__init__.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/tal/tensor/layout.py
tokenspeed_kernel/thirdparty/gluon_petit/lib/tal/tensor/stride.py
tokenspeed_kernel/thirdparty/gluon_petit/petit_kernel/__init__.py
tokenspeed_kernel/thirdparty/gluon_petit/petit_kernel/moe_mxfp4.py
tokenspeed_kernel/thirdparty/gluon_petit/petit_kernel/ops.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/trtllm_blockwise/__init__.py
tokenspeed_kernel/thirdparty/trtllm_blockwise/_kernel.py