huggingface / accelerate
🚀 A simple way to launch, train, and use PyTorch models on almost any device and distributed configuration, automatic mixed precision (including fp8), and easy-to-configure FSDP and DeepSpeed support
huggingface / accelerate
NVIDIA / TransformerEngine
A library for accelerating Transformer models on NVIDIA GPUs, including using 8-bit and 4-bit floating point (FP8 and FP4) precision on Hopper, Ada and Blackwell GPUs, to provide better performance with lower memory utilization in both training and inference.
intel / neural-compressor
SOTA low-bit LLM quantization (INT8/FP8/MXFP8/INT4/MXFP4/NVFP4) & sparsity; leading model compression techniques on PyTorch, TensorFlow, and ONNX Runtime
weicj / vLLM-2080Ti-Definitive
The definitive vLLM runtime for dual RTX 2080 Ti 22GB + NVLink, delivering Qwen 27B local inference with maximum 200+ tok/s single-request decode with support of FP8 weight ( Join Discord :https://discord.gg/VFqVVySdMS )
NVIDIA / cudnn-frontend
cuDNN Frontend is NVIDIA's modern, open-source entry point to the cuDNN library and a growing collection of high-performance open-source kernels.
jpezzulli / sglang-rtxpro6000
Optimized SGLang runtime for Qwen3.8-27B FP8 with DFlash2 and Qwen3.8 Flash-Next NVFP4 with FR-Spec on one NVIDIA RTX PRO 6000 Blackwell 96 GB GPU (SM120): 524K context, HiCache and NIXL.
DominikBucko / qwen38-flash-next-2x3090
Qwen3.8-Flash-Next on 2× RTX 3090: with 128 GB RAM up to 4,191 tok/s prefill · 111.5 tok/s decode (131K prompt), full 256K window at 2,865 tok/s; with 64 GB RAM 3,410 tok/s prefill · 84 tok/s decode. New: opt-in uncensored mode (runtime abliteration, no new weights).
sohu-mptc / FlashRec
FlashRec is a CUDA-graph engine for generative recommendation: wide beam search (3–5 SID steps, n=50–512+) over a trie-constrained catalog, in-process FP8 serving, and ranked beams on /v1/chat/completions.
atbender / deepseek-v41-flash-mac-mini
Run original DeepSeek V4.1 Flash FP4/FP8 weights on a 16 GB M1 Mac mini with bounded SSD streaming in MLX. Recipe, benchmarks, and demo.
joeynyc / MiniMax-H3-DGX-Spark
Measured SM121 compatibility recipe for MiniMax H3 FL2VA on one NVIDIA DGX Spark with vLLM-Omni and online FP8.
tonyd2wild / GLM-5.3-Flash-NVFP4-1M-KV-4x-DGX-Spark
GLM-5.3-Flash (320B MoE) at TP4 across four DGX Sparks, same day as the model drop: 36 tok/s, 1.26M-token FP8 KV pool, 262K context, MTP spec decode. First TP4 glm5_next outside B200 hardware. Full patched-image recipe + the GB10 memory study.
0xBakeer / Qwen3.8-27B-FP8-on-a-single-DGX-Spark
Serving Qwen3.8-27B-FP8 on a single DGX Spark (GB10): 7.88 to 58.5 tok/s single-stream from decode strategy alone, weights untouched. Speculative decoding and prefix caching benchmarked, plus DFlash 2 — the only Qwen3.8-27B build that can serve it under vLLM.
0xBakeer / Qwen3.8-27B-4-bit-on-a-single-DGX-Spark
Serving 4-bit Qwen3.8-27B on a single DGX Spark (GB10): 75 tok/s single-stream, 246 tok/s aggregate at 8-way concurrency. NVFP4 vs MixedInt4-AutoRound vs the FP8 baseline, measured on one harness — including why the quantization advantage collapses to +0.2% by c16.
tonyd2wild / MiMo-V2.6-Flash-DGX-Spark-Recipe
MiMo-V2.6-Flash-RL on 2x NVIDIA DGX Spark (GB10): vLLM TP2 + DFlash + vision/audio, fp8 KV 300K, four fixes for the stock image, measured baselines
knapcio / GLM-5.3-Flash-4x-DGX-Spark-TP4
GLM-5.3-Flash NVFP4 on 4x NVIDIA DGX Spark: vLLM TP4, DFlash2 adaptive draft, measured FP8 vs NVFP4 vs EXL3 vs SGLang
jnardiello / GLM-5.3-Flash-FP8-4-DGX-Spark-Switchless
GLM-5.3-Flash (FP8) + DFlash2 speculative decoding on 4x NVIDIA DGX Spark / ASUS GX10: vLLM TP4, 256K context, switchless ConnectX-7 ring with patched NCCL
Hitheshkaranth / Ornith-1.5_A3B_Model_DGX_Spark_Setup
Ornith-1.5-35B-A3B (NVFP4 + FP8 + MTP) on NVIDIA DGX Spark (GB10)
r0b0tlab / qwen38-27b-nvfp4-sm121-vllm
Qwen3.8-27B NVFP4 (W4A16 shipped recipe) + MTP on NVIDIA DGX Spark GB10/SM121 — vLLM v0.27.2rc0, FP8 KV, 262K context, full reproducibility pack