reap
nemotron-3.5-transcribe-mini-64e
Automatic speech recognition model, used with nemo.
- 14
- 4k
- other
- 3 wk ago
Qwen3.8-Flash-Next-REAP128-86B-Unity-OpenCode-GGUF
Text generation model, used with gguf.
- 11
- 1.4k
- other
- 3 mo ago
MiniMax-M2.7-REAP-172B-A10B-AutoRound-W4A16
Text generation model, used with transformers.
- 11
- 189
- other
- 7 mo ago
pli-labs
PLI Labs: legal model prunes (REAPs), owned-hardware serving research, evaluation methods, and open-source tools from Proprietary Legal Intelligence.
- 2
- Python
- MIT
- 2 wk ago
deepseek-v4.1-flash-reap-thor
DeepSeek-V4.1-Flash-1017 pruned to 192 of 384 experts per layer, at ~2.85 bpw with Engram read from NVMe, served on one Jetson AGX Thor. Weights on Hugging Face; full pipeline and gates here.
- 1
- Cuda
- MIT
- 3 wk ago
blackwell-16gb-moe
How I fit 35B-parameter MoE models into 16 GB of consumer VRAM: REAP expert pruning + NVFP4 quantization on an RTX 5070 Ti (SM120), served on vLLM, with measured evals.
- 0
- 6 d ago
deepseek-v4-flash-0731-cuda
Pure-CUDA engine and OpenAI-compatible server for DeepSeek-V4-Flash-0731-REAP (MXFP4) on Jetson AGX Thor: lossless DSpark speculative decoding at 28.4 tok/s with a fine-tuned draft head.
- 0
- Cuda
- MIT
- 4 wk ago
qwen-moe-pruning
Shrinking Qwen3.5-35B-A3B on one 16 GB GPU: MoE expert pruning (REAP), router distillation and GGUF quantization, with interactive lessons and every command
- 0
- HTML
- 4 wk ago
xiaomi-2.6-flash-REAP
HOPE expert pruning of MiMo-V2.6-Flash (309B omnimodal MoE) to 128 of 256 experts for one Jetson AGX Thor, calibrated with real image, audio and video. Published; not yet evaluated.
- 0
- Python
- MIT
- 4 wk ago
glm-5.3-reap
GLM-5.3-Flash (321B MoE) pruned to 144 of 288 experts and quantized to NVFP4/GGUF for one Jetson AGX Thor. Layer-streamed pipeline, paired teacher evaluation, and the negative results.
- 0
- Python
- MIT
- 4 wk ago
dspark-cuda-reap-finetune
Pure-CUDA DeepSeek-V4-Flash REAP-180B engine with DSpark speculative decode on Jetson AGX Thor; kernel baseline for the fine-tuned head in deepseek-v4-flash-0731-cuda
- 0
- Cuda
- MIT
- 3 wk ago