Yixin DongandLiangsheng Yin 6350042696 feat: Naive support Spec V2 + Constrained Decoding (#13425) ...

Signed-off-by: Ubospica <ubospica@gmail.com>
Co-authored-by: Liangsheng Yin <lsyincs@gmail.com>

2025-11-27 20:31:46 +08:00

..

cpp_ngram

fix missing output_token_logprobs when using ngram speculative decoding (#10702)

2025-11-10 02:04:42 -08:00

base_spec_worker.py

Fix flush cache API for spec v2 (#11918)

2025-10-21 23:01:16 +08:00

draft_utils.py

[Feature] Support DeepSeek MTP on NPU (#11897)

2025-10-30 21:56:56 +08:00

eagle_draft_cuda_graph_runner.py

align code style eagle draft&draft_extend cuda graph runner (#13533)

2025-11-23 01:55:16 +08:00

eagle_draft_extend_cuda_graph_runner.py

align code style eagle draft&draft_extend cuda graph runner (#13533)

2025-11-23 01:55:16 +08:00

eagle_draft_extend_npu_graph_runner.py

[Feature] Spec-Overlap supporting DP-ATTN; PD-Disaggregation; npugraph mode (#12443)

2025-11-15 21:51:07 +08:00

eagle_draft_npu_graph_runner.py

[Feature] Spec-Overlap supporting DP-ATTN; PD-Disaggregation; npugraph mode (#12443)

2025-11-15 21:51:07 +08:00

eagle_info_v2.py

feat: Naive support Spec V2 + Constrained Decoding (#13425)

2025-11-27 20:31:46 +08:00

eagle_info.py

[Spec v2] Remove allocate_lens and enable over-allocation (#13478)

2025-11-22 22:49:10 +08:00

eagle_utils.py

[Feature] Spec-Overlap supporting DP-ATTN; PD-Disaggregation; npugraph mode (#12443)

2025-11-15 21:51:07 +08:00

eagle_worker_v2.py

feat: Naive support Spec V2 + Constrained Decoding (#13425)

2025-11-27 20:31:46 +08:00

eagle_worker.py

Support FlashAttention3 page_size > 1 and topk > 1 case with paged attn and spec decode (#7725)

2025-11-26 11:44:41 +08:00

ngram_info.py

fix spec dec request level metrics (#13754)

2025-11-27 01:09:21 +08:00

ngram_worker.py

fix spec dec request level metrics (#13754)

2025-11-27 01:09:21 +08:00

spec_info.py

Fix Duplicate Classmethod in spec_info.py (#12354)

2025-10-29 13:57:10 -07:00

spec_utils.py

Support FlashAttention3 page_size > 1 and topk > 1 case with paged attn and spec decode (#7725)

2025-11-26 11:44:41 +08:00

standalone_worker.py

Add --speculative-moe-runner-backend server arg (#10183)

2025-11-04 00:20:56 -08:00