Giancarlo Delfin fe755c8899 [Bugfix][Model Runner V2][Spec Decode] Decouple the draft's gumbel noise stream from the target's (#54282) ...

Signed-off-by: Giancarlo Delfin <gdelfin@inferact.ai>

2026-08-29 09:58:29 -07:00

..

__init__.py

Test: added acceptance length tests (#32030)

2026-01-20 18:55:15 +00:00

test_acceptance_length.py

Enable DeepSeek V4 and GLM-5.1 on SM120 (#43477)

2026-06-22 11:54:14 -07:00

test_adaptive_verification.py

[Spec Decode] Enable adaptive DSpark on SM100 sparse MLA (#52783)

2026-08-25 13:31:15 -04:00

test_backup_token_async_spec.py

[Bugfix][Async] Fix async spec decoding with hybrid models (#38556)

2026-03-31 11:08:54 -04:00

test_dflash2.py

[Test] Fix mock_current_vllm_config missing kernel_config in test_dflash2 (#54132)

2026-08-27 22:28:43 -07:00

test_dflash_causality.py

[Spec Decode] DFlash2: local convolution + candidate selector (#52816)

2026-08-20 22:27:22 -07:00

test_dflash_lookahead.py

Fix DFlash prefix cache corruption due to missing lookahead block (#42971)

2026-06-02 12:06:33 +00:00

test_dflash_prepare_inputs.py

[Spec decode] Support Kimi-K3 DCP with DSpark (#52188)

2026-08-17 13:08:45 -07:00

test_dspark_topk.py

[Spec Decode] Add top-k DSpark Markov projection (#49969)

2026-08-04 12:52:53 +00:00

test_dynamic_sd_cug.py

[Bugfix][Spec Decode] Capture the widest uniform decode batch by default (#50488)

2026-08-28 22:38:13 -07:00

test_dynamic_sd.py

Disable dynamic speculative decoding when DP is enabled (#45963)

2026-07-07 12:52:41 +00:00

test_eagle_draft_attn_metadata.py

[Attention][DCP] Enable FlashInfer MLA for DSpark drafting (#54277)

2026-08-28 21:11:28 -07:00

test_eagle_step_kernel.py

[XPU] Fix spec-decode UTs under tests/v1/spec_decode (#38491)

2026-04-11 01:31:00 +00:00

test_eagle.py

[Bugfix][Spec Decode] Fix probabilistic sampling for parallel drafting (#45956)

2026-06-24 05:36:23 +00:00

test_extract_hidden_states.py

[V1][Spec Decode] Add Dynamic SD (#32374)

2026-06-14 00:14:27 -07:00

test_llm_base_proposer_sampling.py

[New model] Kimi K3 (#50000)

2026-07-30 03:49:58 -07:00

test_llm_base_proposer.py

[BugFix] eagle draft max position embeddings (#49343)

2026-07-29 01:24:47 -07:00

test_max_len.py

[CI] Explicitly tear down speculative decode runners (#49910)

2026-07-27 05:32:46 +00:00

test_mtp_structured_output.py

[Bugfix][Structured Output] Avoid spurious FSM errors after speculative reasoning end (#53046)

2026-08-21 10:01:06 +08:00

test_mtp.py

[V1][Spec Decode] Add Dynamic SD (#32374)

2026-06-14 00:14:27 -07:00

test_ngram.py

[V1][Spec Decode] Add Dynamic SD (#32374)

2026-06-14 00:14:27 -07:00

test_rejection_sampler_utils.py

[Bugfix][Model Runner V2][Spec Decode] Decouple the draft's gumbel noise stream from the target's (#54282)

2026-08-29 09:58:29 -07:00

test_request_acceptance.py

[Frontend][Core][Spec Decode] Per-request acceptance stats in OpenAI API responses (#48915)

2026-08-20 15:42:49 -07:00

test_speculators_correctness.py

[Spec Decode] Add Qwen3 architecture support for EAGLE3 (#43132)

2026-06-21 13:55:26 -07:00

test_speculators_eagle3.py

[Rocm][CI] Fix test_speculator_eagle3 by skipping the CompressedTensorw4a16 Model (#30001)

2025-12-04 07:52:28 +00:00

test_synthetic_rejection_sampler_utils.py

[Feat] Unified Synthetic Acceptance Rate for V1 and V2 (#40662)

2026-04-24 00:48:08 +00:00

test_vocab_mapping.py

[Feature] Universal speculative decoding for heterogeneous vocabularies (TLI) (#38174)

2026-07-02 01:34:20 -07:00