3 people 929230a6f0 Fix GLM-OCR MTP multimodal embeddings and positions (#39088) ...

Co-authored-by: Xinyuan Tong <115166877+JustinTong0323@users.noreply.github.com>
Co-authored-by: Xinyuan Tong <xinyuantong.cs@gmail.com>

2026-09-18 21:10:04 -07:00

..

cpp_ngram

[Kernel] Phase 4 batch-2: migrate JIT operator groups into kernels.ops (no shims) (RFC #29630) (#32015)

2026-07-22 17:49:51 +08:00

dspark_components

Read process groups through the runtime context (#40068)

2026-09-18 17:40:32 -07:00

adaptive_runtime_state.py

Allow custom policy for adaptive speculative decoding (#37274)

2026-09-01 23:15:21 -07:00

adaptive_spec_params.py

GLM-5.3-Flash support (#36507)

2026-09-06 02:27:59 -07:00

base_spec_worker.py

[Unified Cache][7/N] Support MTP, EAGLE, and DSpark draft KV caches in the external linker (#37914)

2026-09-13 18:22:06 +08:00

decoupled_spec_io.py

[Spec][1/N] Decoupled speculative decoding: IPC protocol + cross-process request id + server flags (#27634)

2026-06-23 17:04:45 -07:00

dflash_disaggregation.py

[NPU] Support DFlash speculative decoding for MiMo-V2.5-Pro (mxfp4) (#37565)

2026-09-12 11:59:23 +08:00

dflash_info_v2.py

[mem_cache] Move kv_committed_len into ReqKvInfo (#37078)

2026-08-29 22:42:25 -07:00

dflash_info.py

dsv4.1: remaining model and runtime integration (#38798)

2026-09-18 02:55:30 -07:00

dflash_utils.py

[Kernel] Move CUDA and ROCm speculative kernels to JIT (#40033)

2026-09-17 17:27:26 -07:00

dflash_worker_v2.py

Read process groups through the runtime context (#40068)

2026-09-18 17:40:32 -07:00

domino_utils.py

feat: support TP>1 Domino rollout for DFlash V2 (#37069)

2026-09-11 18:51:11 -07:00

draft_utils.py

[Qwen 3.8 Next] Remove unused tokenwise QSA implementation and tests (#38960)

2026-09-12 01:32:01 -07:00

draft_worker_common.py

DFLASH support added for XPU (#32798)

2026-09-11 10:39:37 +08:00

dspark_disaggregation.py

spec: size the speculative buffers from the bags, not the startup record (#35024)

2026-08-17 16:16:56 -07:00

eagle_disaggregation.py

spec: size the speculative buffers from the bags, not the startup record (#35024)

2026-08-17 16:16:56 -07:00

eagle_draft_cuda_graph_runner.py

Fix GLM-OCR MTP multimodal embeddings and positions (#39088)

2026-09-18 21:10:04 -07:00

eagle_draft_extend_cuda_graph_runner.py

Fix GLM-OCR MTP multimodal embeddings and positions (#39088)

2026-09-18 21:10:04 -07:00

eagle_info.py

Use runtime token widths for Triton speculative verification (#39859)

2026-09-18 11:07:55 -07:00

eagle_utils.py

Read process groups through the runtime context (#40068)

2026-09-18 17:40:32 -07:00

eagle_worker_common.py

Pipeline parallelism x speculative decoding (EAGLE/MTP) compatibility (#30775)

2026-09-16 21:41:54 -07:00

eagle_worker_v2.py

Fix GLM-OCR MTP multimodal embeddings and positions (#39088)

2026-09-18 21:10:04 -07:00

external_corpus_manager.py

[Spec][Ngram] Return token counts in list_external_corpora API (#22471)

2026-04-10 21:50:02 -07:00

frozen_kv_mtp_cuda_graph_runner.py

[Refactor] Generalize attention graph variants in the decode runner (#38993)

2026-09-11 00:45:58 -07:00

frozen_kv_mtp_info.py

[Spec] Converge DP-attention spec width scaling onto num_tokens_per_req (#31244)

2026-07-15 17:57:08 -07:00

frozen_kv_mtp_utils.py

[sp] Make attention-TP sequence sharding a per-forward batch property (#37546)

2026-09-04 02:11:03 -07:00

frozen_kv_mtp_worker_v2.py

[CI][RFC] Replace black-jupyter with ruff-format (#37210)

2026-09-02 19:46:08 -07:00

multi_layer_eagle_draft_extend_cuda_graph_runner.py

[Refactor] Generalize attention graph variants in the decode runner (#38993)

2026-09-11 00:45:58 -07:00

multi_layer_eagle_utils.py

add fill_draft_extend_prepare_buffers_native for NPU (#32427)

2026-07-26 20:42:14 +08:00

multi_layer_eagle_worker_v2.py

[Spec] Stage Inkling MTP draft metadata before verify (#38169)

2026-09-10 15:32:19 -07:00

ngram_info.py

[Spec] Relay ngram accept tokens through the FutureMap (#35198)

2026-08-17 14:21:07 -07:00

ngram_worker.py

[Kernel] Move CUDA and ROCm speculative kernels to JIT (#40033)

2026-09-17 17:27:26 -07:00

pp_spec_relay.py

Pipeline parallelism x speculative decoding (EAGLE/MTP) compatibility (#30775)

2026-09-16 21:41:54 -07:00

ragged_verify.py

[Kimi] Support kimi-k3 (#32541)

2026-08-04 13:22:49 -07:00

spec_info.py

[Spec] Add explicit prefill shared-read capability for plugins (#39502)

2026-09-18 13:11:01 -07:00

spec_registry.py

[Spec] Add explicit prefill shared-read capability for plugins (#39502)

2026-09-18 13:11:01 -07:00

spec_tp_sync.py

[Spec] Fix Dspark and Dflash state divergence across TP rank (#33614)

2026-08-29 17:27:43 -07:00

spec_utils.py

[Kernel] Move CUDA and ROCm speculative kernels to JIT (#40033)

2026-09-17 17:27:26 -07:00

standalone_worker_v2.py

[Kernel] Move CUDA and ROCm speculative kernels to JIT (#40033)

2026-09-17 17:27:26 -07:00

uno_cuda_graph_runner.py

[Refactor] Generalize attention graph variants in the decode runner (#38993)

2026-09-11 00:45:58 -07:00

uno_info.py

[Speculative Decoding] Add native UNO serving support (#37667)

2026-09-03 20:08:41 +08:00

uno_lora.py

[Speculative Decoding] Add native UNO serving support (#37667)

2026-09-03 20:08:41 +08:00

uno_tree.py

[Speculative Decoding] Add native UNO serving support (#37667)

2026-09-03 20:08:41 +08:00

uno_utils.py

[Speculative Decoding] Add native UNO serving support (#37667)

2026-09-03 20:08:41 +08:00

uno_validation.py

[Speculative Decoding] Add native UNO serving support (#37667)

2026-09-03 20:08:41 +08:00

uno_worker_v2.py

[Speculative Decoding] Add native UNO serving support (#37667)

2026-09-03 20:08:41 +08:00