Back to Topics Directory
Topic Hub

#speculative-decoding (5 Repositories)

Ranked open-source repositories tagged with #speculative-decoding, scored by pull request acceptance likelihood and maintainer engagement velocity.

Topic Avg Merge Rate

37.5%

Avg Review Latency

4.9h

Filter by language

5 repositories tagged #speculative-decoding

B TierRust 328

avifenesh/memra

Rust + CUDA inference engine for NVIDIA RTX PRO 6000 Blackwell and RTX 5090. Serves safetensors and GGUF over an OpenAI-compatible API, with per-device tuned defaults and speculative decode gated byte-identical to plain decode. Hosted instance: inference.tiyuvta.ai

73.1%
Merge Rate
2h
First Review
100%
1st-Timers
0
Maintainers
A TierRust 669 1 GFIs

Avarok-Cybersecurity/atlas

Pure Rust Inference Engine

51.9%
Merge Rate
20h
First Review
38%
1st-Timers
27
Maintainers
A TierPython 195

syv-ai/qwen38-27b-rtx3090

Qwen3.8-27B on a single RTX 3090 with vLLM: ~1,000 tok/s at 64 concurrent (int8 tensor-core GEMMs, fp16 DeltaNet state), ~114 tok/s single-user at default sampling / ~124 greedy (MTP drafts, own-output draft vocab, calibrated int4 lm_head, split-KV verify attention), 150k-262k context; patches, requant scripts, benchmarks

62.5%
Merge Rate
3h
First Review
50%
1st-Timers
20
Maintainers
D TierPython 104

Nathanw1014/strix-halo-llamacpp

Performance-tuned llama.cpp for AMD Strix Halo (gfx1151): FA + MoE-prefill fixes with a bundled current Mesa driver. Vulkan and HIP; portable dir, Docker, and distrobox.

0.0%
Merge Rate
-
First Review
0%
1st-Timers
0
Maintainers
D TierPython 136

julianmb/q38rocm

Qwen 3.8 27B ROCmFP4 on AMD Strix Halo (Ryzen AI Max+ 395). Up to 36 tok/s via MTP Speculation, TurboQuant & Mesa RADV Wave64.

0.0%
Merge Rate
-
First Review
0%
1st-Timers
0
Maintainers
Best Speculative-decoding Open Source Repositories & C-Rank™ | GetMerged