Search results

[RFC]: Investigate whether HF-native MTP support can simplify our MTP finetune + stitch pathway

RFC

Status: Open.

#844 In vllm-project/speculators;

· rahul-tuli opened on Jul 22, 2026

[RFC]: Config-file-first CLI for scripts/train.py

RFC

Status: Open.

#833 In vllm-project/speculators;

· rahul-tuli opened on Jul 21, 2026

[Feature] Support gradient checkpointing through the Hugging Face API

Status: Open.

#831 In vllm-project/speculators;

· jessiewei7 opened on Jul 21, 2026

[RFC]: Bounded asynchronous hidden-state fan-out to independent consumers

Status: Open.

#810 In vllm-project/speculators;

· heiheiha798 opened on Jul 18, 2026

[FEATURE] add support for speculative speculative decoding

Status: Open.

#809 In vllm-project/speculators;

· JINO-ROHIT opened on Jul 18, 2026

[RFC] Add minimal, evidence-informed AGENTS.md to Speculators

Status: Open.

#802 In vllm-project/speculators;

· orestis-z opened on Jul 17, 2026

[Bug]: DFlash training performs worse on Qwen3.5 than on Qwen3

bugSomething isn't workingSomething isn't working

Status: Open.

#797 In vllm-project/speculators;

· chenfengt opened on Jul 16, 2026

[Bug]: does it support training mistral architectures

bugSomething isn't workingSomething isn't working

Status: Open.

#796 In vllm-project/speculators;

· JINO-ROHIT opened on Jul 16, 2026

RFC: Add DeLS-Spec (Decoupled Long-Short Contexts) speculator

Status: Open.

#782 In vllm-project/speculators;

· orestis-z opened on Jul 13, 2026

Q3 Roadmap

keep-open

RFC

ROADMAP

Status: Open.

#781 In vllm-project/speculators;

· dsikka opened on Jul 13, 2026

RFC: Add PARD-2 speculative decoding (CAT optimization + dual-mode)

RFC

Status: Open.

#779 In vllm-project/speculators;

· orestis-z opened on Jul 13, 2026

[Bug]: Finetuned Decoder Performs Worse Than Z-Lab Baseline on Finetuning Dataset

bugSomething isn't workingSomething isn't working

Status: Open.

#761 In vllm-project/speculators;

· huaxuan250 opened on Jul 9, 2026