1Cat-vLLM

architecture

[Core] Share SM70 AWQ and FP8 MoE stages and capture engine policy

2026-10-09 10:54:32 +08:00

data

[Doc] Preserve the rejected IQ3 register-pair screen

2026-10-09 09:47:08 +08:00

minimax_h3

…

native_media

…

arch_overview.md

…

attention_backends.md

…

cuda_graphs_multimodal.md

…

cuda_graphs.md

…

custom_op.md

…

dbo.md

…

debug_vllm_compile.md

…

early_graph_model_inputs.md

[Doc] Record same-engine early graph input comparison

2026-10-05 12:56:44 +08:00

environment_metadata.md

…

flashnext_gguf_graph_parity.md

[Doc] Audit newly merged dense GGUF capabilities

2026-10-05 15:38:15 +08:00

flashnext_mtp4_round12_screens_20261008.md

[Doc] Preserve Flash-Next screens and prototype validation limits

2026-10-09 08:53:04 +08:00

fused_moe_modular_kernel.md

…

fusions.md

…

gguf_collective_norm.md

[Doc] Record matched GGUF collective norm model results

2026-10-07 02:02:32 +08:00

gguf_dense_segments.md

[Doc] Attribute installed GGUF target and shared-stream overlap

2026-10-06 02:08:02 +08:00

gguf_dflash2_iq3s.md

[Doc] Record integrated GGUF DFlash2 round measurements

2026-10-05 21:26:34 +08:00

gguf_dflash2_projection_ledger.md

[Doc] Record matched GGUF QPN projection batch performance

2026-10-06 03:01:33 +08:00

gguf_dflash2_tp_sizes.md

[Core] Share DFlash attention and norm boundaries across TP sizes (#1025 )

2026-10-07 15:10:21 +08:00

gguf_dmv_projections.md

[Doc] Separate projection increments from graph-context costs

2026-10-06 23:05:38 +08:00

gguf_dmv_tp2.md

[Kernel] Share GGUF projection planes with TP2 shards (#1026 )

2026-10-07 16:38:06 +08:00

gguf_dp4a_small_m.md

[Doc] Record integer expert round latency and paired quality

2026-10-05 18:38:33 +08:00

gguf_fp16_projection.md

[Doc] Record primary GGUF floating row verification

2026-10-05 15:21:24 +08:00

gguf_fp16_short_context_splits.md

[Kernel] Improve FP16 grouped verification at short contexts (#1022 )

2026-10-07 12:07:00 +08:00

gguf_head_layout_preparation.md

[Doc] Record final Flash-Next graph and acceptance comparison

2026-10-05 15:33:19 +08:00

gguf_iq2_signed_nibbles.md

[Doc] Qualify IQ2 projection gains in complete rounds

2026-10-07 07:18:55 +08:00

gguf_iq3_gated_pair_dispatch.md

[Doc] Record GGUF pair model compilation and natural output

2026-10-05 14:29:38 +08:00

gguf_iq3_xxs_pair_reader.md

[Doc] Record installed IQ3_XXS model graph and natural output

2026-10-05 16:07:21 +08:00

gguf_iq3_xxs_pair_records.md

[Core] Prepare source-sized IQ3_XXS pair records

2026-10-05 14:45:45 +08:00

gguf_iq3xxs_joint_experts.md

[Doc] Round the retained expert bank size consistently

2026-10-05 11:44:32 +08:00

gguf_iq4_native_reader.md

[Kernel] Preserve IQ4_XS source bytes and nested scales (#950 )

2026-10-05 11:11:01 +08:00

gguf_lm_head_sm70.md

[Kernel] Route measured GGUF LM heads through canonical SM70 projections

2026-10-04 19:08:28 +08:00

gguf_lut4_q8_experts.md

[Doc] Record matched IQ4 model quality and partition screening

2026-10-06 05:47:14 +08:00

gguf_mixed_projection_output.md

[Doc] Qualify mixed output with cache-exceeding weights

2026-10-05 12:16:31 +08:00

gguf_native_iq1_m_pair.md

[Doc] Record complete mixed-pair coverage and installed checks

2026-10-05 19:47:55 +08:00

gguf_native_iq2_s_pair.md

[Kernel] Admit measured IQ2_S mixed gated pairs

2026-10-05 18:07:04 +08:00

gguf_native_iq2_xs_pair.md

[Doc] Record installed IQ2_XS route checks

2026-10-05 19:43:01 +08:00

gguf_native_iq2_xxs_pair.md

[Doc] Record installed IQ2_XXS route checks

2026-10-05 19:43:30 +08:00

gguf_native_iq3xxs_iq4_pair.md

[Doc] Record pure-pair admission and rejected IQ4 route

2026-10-06 01:11:53 +08:00

gguf_native_mixed_pair.md

[Doc] Record primary V100 mixed pair measurements

2026-10-05 15:20:08 +08:00

gguf_native_q2_k_pair.md

[Kernel] Admit measured Q2_K IQ3_S gated pair

2026-10-05 19:38:57 +08:00

gguf_native_q4_iq3_pair.md

[Kernel] Admit measured Q4_K and IQ3_S gated pairs

2026-10-05 18:36:18 +08:00

gguf_native_q4_k_pair.md

[Doc] Record primary admitted-pair regression and byte coverage

2026-10-05 17:27:02 +08:00

gguf_native_sm70.md

…

gguf_ordered_ple_rows.md

[Doc] Validate ordered PLE row decoding and graph cost

2026-10-05 15:39:53 +08:00

gguf_packed_embeddings.md

[Doc] Record packed GGUF embedding correctness and storage

2026-10-08 00:33:39 +08:00

gguf_packed_ple.md

[Doc] Record installed cascade and real packed row checks

2026-10-05 00:59:52 +08:00

gguf_pinned_ple_decode.md

[Kernel] Read packed GGUF PLE decode rows from pinned TP shards (#996 )

2026-10-06 04:00:12 +08:00

gguf_projection_clock_diagnostics.md

[Kernel] Add device warp timing for GGUF projection diagnostics (#1018 )

2026-10-07 10:49:56 +08:00

gguf_q4_k_pair_records.md

[Kernel] Preserve original Q4_K bytes in pair records

2026-10-05 15:50:51 +08:00

gguf_q8_intermediate.md

[Doc] Record installed-backend GGUF intermediate validation

2026-10-05 23:27:00 +08:00

gguf_qkv_profile_calibration.md

[Kernel] Fuse measured three-format GGUF QKV projections (#1017 )

2026-10-07 09:24:37 +08:00

gguf_qpn_attention_inputs.md

[Doc] Record installed GGUF attention projection gates

2026-10-06 04:24:15 +08:00

gguf_qpn_canonical_streams.md

[Doc] Retain canonical QPN screening and reject slower routes

2026-10-06 04:34:03 +08:00

gguf_qpn_linear.md

[Doc] Close compiled Q2_K down projection validation

2026-10-06 00:58:42 +08:00

gguf_qpn_output_batch.md

[Doc] Add same-clock head reference and reporting checks

2026-10-06 05:24:52 +08:00

gguf_qpn_projection_batch.md

[Doc] Attribute the post-QPN graph-linked projection round

2026-10-06 03:57:20 +08:00

gguf_qpn_qkvz.md

[Doc] Record installed compiled GDN input checks

2026-10-06 02:09:01 +08:00

gguf_qpn_register_pair.md

[Doc] Preserve the rejected IQ3 register-pair screen

2026-10-09 09:47:08 +08:00

gguf_qpn_small_outputs.md

[Doc] Record qualified QPN output projection results

2026-10-06 04:03:45 +08:00

gguf_qwen35_moe_adapter.md

[Model] Add native Qwen3.5 MoE GGUF adapter

2026-10-05 01:03:34 +08:00

gguf_qwen38_iq3s_layer_types.csv

[Doc] Normalize GGUF inventory line endings

2026-10-05 11:45:49 +08:00

gguf_qwen38_iq3s_source_inventory.md

[Doc] Include all GGUF layer type shares

2026-10-05 11:45:27 +08:00

gguf_qwen38_mixed_pair_coverage.md

[Doc] Record complete mixed-pair coverage and installed checks

2026-10-05 19:47:55 +08:00

gguf_tp2_256k_memory.md

[Doc] Record dual V100 GGUF DFlash2 256K execution

2026-10-08 00:35:04 +08:00

gguf_tp4_verify_cost.md

[Doc] Measure TP4 GGUF DFlash2 verification cost

2026-10-08 03:07:13 +08:00

gguf_turbomind_bitplanes.md

[Doc] Record installed GGUF bit-plane wheel verification

2026-10-03 16:52:13 +08:00

gguf_turbomind_codebook_words.md

[Doc] Record installed lattice word-copy validation

2026-10-04 04:01:31 +08:00

gguf_turbomind_iq3_packet.md

[Bench] Identify loaded GGUF core and retain rejected IQ3 evidence

2026-10-04 03:27:23 +08:00

gguf_turbomind_lattice_banks.md

[Doc] Record installed IQ3 vector codebook results

2026-10-04 04:51:00 +08:00

gguf_turbomind_lattice_decode.md

[Doc] Record matched lattice capture measurements

2026-10-04 01:38:55 +08:00

gguf_turbomind_lattice_grouped.md

[Doc] Record installed GGUF grouped dispatch measurements

2026-10-03 23:48:52 +08:00

gguf_turbomind_lattice.md

[Doc] Record installed GGUF lattice validation

2026-10-03 21:56:00 +08:00

gguf_turbomind_lut4.md

[Doc] Record LUT4 artifact validation on current main

2026-10-03 18:59:41 +08:00

gguf_turbomind_model_moe.md

[Doc] Record installed canonical expert integration checks

2026-10-05 04:28:33 +08:00

gguf_turbomind_model.md

[Doc] Record corrected GGUF model quality and matched C4 repeat

2026-10-04 11:05:25 +08:00

gguf_turbomind_prefill.md

[Doc] Record prefill and LUT4 integration checks

2026-10-03 19:20:20 +08:00

gguf_turbomind_u2.md

[Kernel] Complete grouped GGUF operator comparisons and artifact checks

2026-10-03 14:21:00 +08:00

gguf_turbomind.md

…

gguf_v100_control.md

[Model] Integrate standalone Qwen4Exp GGUF mapping on main

2026-10-05 00:47:24 +08:00

grouped_cpu_offload.md

…

h3_encoder_prepared_cache.md

[Core] Reuse exact DiT checkpoint tensors before layout preparation

2026-10-04 02:08:28 +08:00

huggingface_integration.md

…

hybrid_kv_cache_manager.md

…

io_processor_plugins.md

…

logits_processors.md

…

lora_resolver_plugins.md

…

metrics.md

…

mixed_prefill_latency.md

[Doc] Record the final mixed-load replay-window measurements

2026-10-04 02:01:26 +08:00

mm_processing.md

…

model_runner_v2.md

…

moe_kernel_features.md

…

multiprocessing.md

…

nixl_kv_cache_lease.md

…

omni_workflow_coverage.md

…

optimization_levels.md

…

p2p_nccl_connector.md

…

paged_attention.md

…

ple_mapped_row_gather.md

[Kernel] Prefetch and gather disk-mapped PLE rows on CPU (#885 )

2026-10-05 12:18:47 +08:00

ple_metadata_request_indices.md

[Doc] Record metadata transfer hits in the full model

2026-10-05 15:33:20 +08:00

ple_split_mainline_audit.md

…

plugin_system.md

…

prefix_caching.md

…

qwen4exp_ple_tier_cascade.md

[Bugfix] Admit PLE offload with decode context parallelism

2026-10-04 03:11:55 +08:00

qwen4exp_qsa_dcp2.md

…

qwen36_27b_awq_mtp_target_verifier_fastpaths.md

…

safetensors_direct_io.md

…

sm70_awq_exact_m5_batched_gemv.md

…

sm70_awq_fp8_mtp_experts.md

…

sm70_awq_long_prefill_exact_dense.md

…

sm70_awq_qpn_m1.md

…

sm70_awq_small_n_hmma_operator.md

…

sm70_compact_candidate_gather.md

[Doc] Record compact candidate gather measurements

2026-10-08 06:56:23 +08:00

sm70_concurrency_integration_audit.md

…

sm70_deepseek_v4_dspark_acceptance.md

…

sm70_deepseek_v4_dspark_n7_latest_main.md

…

sm70_deepseek_v4_dspark_optimization.md

…

sm70_deepseek_v4_dspark_verifier_20ms_long_context.md

…

sm70_deepseek_v4_flash_bringup.md

…

sm70_deepseek_v4_fp16_gemv.md

…

sm70_deepseek_v4_mhc_decode.md

…

sm70_deepseek_v4_mxfp4_active_expert.md

…

sm70_deepseek_v4_mxfp4_grouped_decode.md

…

sm70_deepseek_v4_performance_integration.md

…

sm70_deepseek_v4_sparse_mla_qk_dsplit.md

…

sm70_deepseek_v4_sparse_mla_splitk.md

…

sm70_deepseek_v4_tp8_graph_allreduce.md

…

sm70_dflash2_acceptance_20260909.md

…

sm70_dflash2_acceptance_rd.md

…

sm70_dflash2_attention_resources_20260910.md

…

sm70_dflash2_batch_attention_gdn.md

…

sm70_dflash2_batch_latency_20260926.md

…

sm70_dflash2_context_cost_20260909.md

…

sm70_dflash2_fastpath_numerics.md

…

sm70_dflash2_fp8_verify_20ms.md

…

sm70_dflash2_fp32_defaults.md

…

sm70_dflash2_long_batch_20260927.md

…

sm70_dflash2_long_decode_followup_20260927.md

…

sm70_dflash2_long_verify_curve_20260909.md

…

sm70_dflash2_long_verify_decay.md

…

sm70_dflash2_lookup_augmented.md

…

sm70_dflash2_m48_scaling_20260927.md

…

sm70_dflash2_ngram_hybrid.md

…

sm70_dflash2_nvfp4_17ms.md

…

sm70_dflash2_nvfp4_prefill_promotion.md

…

sm70_dflash2_policy.md

…

sm70_dflash2_prefill_closure.md

…

sm70_dflash2_quality_audit.md

…

sm70_dflash2_retired_candidate_order.md

…

sm70_dflash2_shared_nvfp4.md

…

sm70_dflash2_tail_graphs_20260911.md

…

sm70_dflash2_target_graph_20ms.md

…

sm70_dflash2_tp2_verifier.md

…

sm70_dflash2_verifier_route_audit.md

…

sm70_dflash_ddtree_27b_awq_plan.md

…

sm70_dflash_page832.md

[Doc] Record qualified DFlash window split speed and quality

2026-10-07 04:43:45 +08:00

sm70_dflash_page1648.md

[Core] Keep DFlash split windows with E4M3 hybrid pages (#1024 )

2026-10-07 12:53:06 +08:00

sm70_dflash_smallm.md

[Kernel] Add packed small-M DFlash2 projections on SM70

2026-10-04 13:11:35 +08:00

sm70_e4m3_grouped_fp32.md

…

sm70_environment_surface.md

…

sm70_exact_auxiliary_fusion.md

[Kernel] Isolate unused MoE and auxiliary policies from compile caches

2026-10-03 19:42:44 +08:00

sm70_fa2_d256_prefill_pipeline.md

…

sm70_flash_next_batch_defaults.md

[Bugfix] Align FP32 router preparation with its runtime precision guard

2026-10-05 01:34:19 +08:00

sm70_flash_next_mtp4_15ms.md

[Kernel] Enable qualified full-vocabulary Flash-Next MTP4 defaults (#821 )

2026-10-05 19:59:24 +08:00

sm70_flash_next_mtp4_256k_eval_20260927.csv

…

sm70_flash_next_mtp4_256k_eval_20260927.md

…

sm70_flash_next_mtp4_batch_gdn.md

…

sm70_flash_next_mtp4_default_profile.md

…

sm70_flash_next_mtp4_trace_reconciliation.md

…

sm70_flash_next_prefix_prefill.md

…

sm70_flash_v37_prefill.md

…

sm70_flash_v100_fp8_kv_long_context.md

…

sm70_flash_v100_prefill_operator_optimization.md

…

sm70_flashinfer_native_backend.md

…

sm70_flashnext_gguf_mtp4_operator_routes.md

[Benchmark] Record reproducible Flash-Next GGUF MTP4 composition

2026-10-05 06:13:17 +08:00

sm70_flashnext_gguf_mtp4.md

[Doc] Record the repeated Flash-Next C4 ring comparison

2026-10-04 17:25:47 +08:00

sm70_fp8_kernel_integration.md

…

sm70_fp8_long_prefill_exact_dense.md

…

sm70_fp8_policy_cleanup.md

…

sm70_fp16_gemv_silu_schedule.md

…

sm70_fp16_grouped_verify.md

[Doc] Record packaged FP16 grouped backend validation

2026-10-04 20:02:28 +08:00

sm70_gdn_ba_verify.md

[Kernel] Fuse SM70 verifier b/a and channel FP8 projections

2026-10-04 13:33:18 +08:00

sm70_gdn_preprocess.md

[Doc] Record five-row GDN preprocessing correctness and timing

2026-10-05 02:40:34 +08:00

sm70_gdn_projection_copy_audit.md

…

sm70_gdn_projection_tails.md

[Doc] Record default GDN tail-copy route and timing

2026-10-05 05:17:04 +08:00

sm70_gdn_verify_tile.md

[Kernel] Tile single-request SM70 GDN verification columns

2026-10-04 12:51:02 +08:00

sm70_gemma_long_prefill_fusion.md

…

sm70_gguf_canonical_head.md

[Doc] Record packed LM-head method measurements

2026-10-04 21:32:01 +08:00

sm70_gguf_expert_down_dispatch.md

[Doc] Record installed GGUF down dispatch validation

2026-10-05 04:33:07 +08:00

sm70_gguf_fp16_dense_routes.md

[Doc] Record extended GDN row dispatch checks

2026-10-04 21:31:28 +08:00

sm70_gguf_joint_expert_dispatch.md

[Doc] Record joint expert model dispatch qualification

2026-10-05 06:11:25 +08:00

sm70_gguf_joint_original_experts.md

[Doc] Qualify standalone joint experts with cold-bank timings

2026-10-05 06:10:46 +08:00

sm70_gguf_small_grouped_down.md

[Kernel] Admit measured small grouped down vector batches

2026-10-05 01:12:07 +08:00

sm70_glm53_flash_nvfp4.md

…

sm70_hc_batch_topology.md

[Doc] Record M20 HC calibration and bounded admission

2026-10-05 02:15:38 +08:00

sm70_hc_ll_shards.md

[Doc] Record packaged HC ABBA and model limits

2026-10-06 02:06:15 +08:00

sm70_hc_weight_prefetch_screen.md

[Benchmark] Record reproducible Flash-Next GGUF MTP4 composition

2026-10-05 06:13:17 +08:00

sm70_hmma_pipeline_search.md

…

sm70_linear_kernel_integration.md

…

sm70_loaded_route_reporting.md

…

sm70_m8_exact_decode.md

[Kernel] Reuse M8 gate operands and TP4 norm packets (#1041 )

2026-10-08 08:50:20 +08:00

sm70_mamba_state_grid_decoupling.md

…

sm70_memory_arena.md

…

sm70_memory_defaults.md

…

sm70_memory_workspace_tuning.md

…

sm70_mixed_batch_decode_rows.md

[Perf][SM70] Keep resident decode rows on fast routes inside mixed prefill batches

2026-10-03 03:18:46 -07:00

sm70_mrv2_sampling_contract_repairs.md

…

sm70_mtp_gdn_output.md

[Bench] Record direct MTP recurrent output copy savings

2026-10-05 03:40:28 +08:00

sm70_mtp_strided_qkv.md

[Bench] Record strided MTP QKV timing and exact state validation

2026-10-05 03:25:52 +08:00

sm70_nvfp4_bundled_scale.md

[Kernel] Record bundled NVFP4 serving validation

2026-10-07 00:49:24 +08:00

sm70_nvfp4_turbomind_operator_optimization.md

…

sm70_ple_input_prepare.md

[Bench] Preserve teacher histories across chained comparisons

2026-10-06 02:40:47 +08:00

sm70_ple_ngram_batch.md

[Doc] Record exact PLE batch indices and graph operator savings

2026-10-05 00:10:00 +08:00

sm70_prefill_score_overflow.md

[Bugfix] Recover overflowing SM70 compact score tiles

2026-10-03 21:12:09 +08:00

sm70_prefill_tail_intermediate.md

[Doc] Record long-context tail peak recovery

2026-10-04 10:05:19 +08:00

sm70_prefix_anchored_swa.md

…

sm70_prefix_peak_repair.md

[Bugfix] Repair missed SM70 prefill prefix maxima

2026-10-03 20:38:28 +08:00

sm70_qsa_page4_allocation_invariance.md

…

sm70_quantized_batch_reuse.md

…

sm70_quasar_dflash2_15ms.md

…

sm70_quasar_dflash2_operator_audit.md

…

sm70_quasar_dflash2_quality_speed_recovery.md

…

sm70_quasar_dflash2_resource_audit_20260909.md

…

sm70_quasar_e4m3_fp32_logits.md

…

sm70_quasar_nvfp4_dflash2_acceptance.md

…

sm70_quasar_tp2_tp4_quality.md

…

sm70_qwen36_27b_awq_mtp4_optimization.md

…

sm70_qwen38_130_acceptance.md

…

sm70_qwen38_awq_active_grouped_decode.md

…

sm70_qwen38_concurrency40.md

…

sm70_qwen38_default_fastpath.md

…

sm70_qwen38_distribution_acceptance.md

[Kernel] Enable qualified full-vocabulary Flash-Next MTP4 defaults (#821 )

2026-10-05 19:59:24 +08:00

sm70_qwen38_exact_decode_followup.md

…

sm70_qwen38_flash_next_nvfp4.md

…

sm70_qwen38_fp8_prefill_5500.md

…

sm70_qwen38_fp8_prefill_decay.md

…

sm70_qwen38_nomtp_concurrency.md

…

sm70_qwen38_nvfp4_decode.md

…

sm70_qwen38_nvfp4_quality_repair.md

…

sm70_qwen38_ple_disk_offload.md

…

sm70_qwen38_projection_w13_followup.md

…

sm70_qwen38_qpn8_decode.md

…

sm70_qwen38_reproducible_baseline.md

…

sm70_qwen38_step0_kernel_budget.md

[Perf] Add decode boundary and PLE budget measurements

2026-10-04 11:30:32 +08:00

sm70_release151_compile_cache.md

…

sm70_sampling_input_guards.md

…

sm70_sm120_long_context_attention_parity.md

…

sm70_small_moe_routing.md

[Doc] Record exact expert restoration and graph timing

2026-10-04 21:52:40 +08:00

sm70_sparse_bmm_capabilities.md

…

sm70_tile_runtime_exploration.md

…

sm70_tp4_mtp4_multigpu_scaling.md

…

sm70_tp4_nomtp_long_context_decode.md

…

sm70_tp4_push_gemma_norm.md

[Kernel] Fuse SM70 TP4 push allreduce and Gemma norm

2026-10-04 12:04:13 +08:00

sm70_tp_quality_audit.md

…

sm70_tp_shape_coverage.md

…

sm70_v100_migration_control.md

[Core] Share SM70 AWQ and FP8 MoE stages and capture engine policy

2026-10-09 10:54:32 +08:00

speculative_draft_load_config.md

[Doc] Record installed draft loader regression checks

2026-10-05 00:44:42 +08:00

torch_compile_multimodal.md

…

torch_compile.md

…

vllm_ir.md

…