Python★ 0↓ 6,470/moOct 20, 2026
PyPI · npm71Goodhealth index

explosion/thinc🔮 A refreshing functional take on deep learning, compatible with your favorite libraries
Python★ 2,890↓ 25.7M/moAug 27, 2026
Python★ 27↓ 19.7K/moOct 8, 2026

pjordanandrsn/experts4bit-qloraTrain and serve MoE models that do not fit in VRAM: fused 4-bit experts, QLoRA, CPU/NVMe offload, and fast inference on consumer NVIDIA GPUs.
Python★ 3↓ 2,064/moSep 15, 2026
Python★ 467↓ 188/moOct 21, 2026

acmore/okdevKubernetes-native remote dev environments for AI/ML engineers. Code locally, run on cluster GPUs — live file sync, SSH/tmux into pods, multi-pod PyTorchJob support, and detached job control.
Go · Shell★ 0Sep 10, 2026
Python★ 887Oct 2, 2026

google-ai-edge/LiteRT-CLIA convenient CLI to streamline LiteRT related development workflows, including converting, quantizing, compiling, managing, running, benchmarking and visualizing LiteRT (TFLite) models on various hardwares (CPU / GPU / NPU) across platforms (desktop, mobile or cloud).
Python★ 41↓ 80/moSep 8, 2026

pjordanandrsn/grouped-nf4-gemmTriton kernels for 4-bit MoE inference: grouped NF4/MXFP4 GEMM, INT4 GEMV, FP8 paged attention, and CPU/NVMe expert streaming.
Python★ 2↓ 3,717/moSep 24, 2026
PyPI63Moderatehealth index
Python★ 263↓ 19/moOct 20, 2026
PyPI63Moderatehealth index
C++ · Python★ 797Sep 19, 2026
npm · PyPI63Moderatehealth index

alex-jb/orallexa-ai-trading-agentSelf-tuning multi-agent AI trading system. 8-source signal fusion, Bull/Bear/Judge debate on Claude Opus 4.7, Kelly + ATR position sizing. Python · Kalshi + Polymarket adapters.
Python · TypeScript★ 65Sep 13, 2026
PyPI63Moderatehealth index

c0rmac/metal-linalgQR, symmetric eigendecomposition and SVD for batches of matrices on Apple Silicon GPUs, for MLX and PyTorch. Metal kernels per regime, routed to the GPU or the CPU by policies measured per Mac. C++, Python (pip: metal-linalg, metal-linalg-torch), Swift and C.
C★ 11↓ 3,008/moOct 4, 2026
PyPI63Moderatehealth index
jagmarques/nexusquantTraining-free KV cache compression via E8 lattice VQ. 2-bit KV that preserves retrieval (30/30 NIAH vs TurboQuant 0/30). Calibration-free, 9 architectures validated.
Python★ 25Jul 16, 2026
PyPI63Moderatehealth index
Python★ 169Sep 12, 2026
PyPI62Moderatehealth index
Python★ 1,203↓ 846.4K/moSep 17, 2026
PyPI60Moderatehealth index
Python · Cuda · C++★ 0↓ 9,412/moOct 21, 2026
PyPI60Moderatehealth index
Python★ 0↓ 7,599/moOct 2, 2026
PyPI60Moderatehealth index
Python★ 18.9K↓ 56.8K/moOct 21, 2026
PyPI59Moderatehealth index

d4l3k/rdma4pyHigh-performance RDMA for Python and PyTorch (ibverbs, efa, NVMe-oF)
Python · Cython★ 12↓ 11.6K/moSep 15, 2026
PyPI59Moderatehealth index
Python★ 1,413Oct 20, 2026
PyPI59Moderatehealth index

thuml/depyfdepyf is a tool to help you understand and adapt to PyTorch compiler torch.compile.
Python★ 815↓ 3.6M/moAug 27, 2026
PyPI57Moderatehealth index

queelius/museModel-agnostic multi-modality generation server with plugin discovery: OpenAI-compatible TTS, embeddings, and image generation. Drop a Python file in ~/.muse/models/ to add a backend.
Python★ 1↓ 262/moSep 16, 2026

webdataset/webdatasetA high-performance Python-based I/O system for large (and small) deep learning problems, with strong support for PyTorch.
Python★ 3,171↓ 2.7M/moAug 27, 2026
PyPI53Moderatehealth index
Python★ 247Sep 15, 2026
PyPI53Moderatehealth index
Python · C++ · Cuda★ 1,745↓ 46.8K/moSep 12, 2026
PyPI53Moderatehealth index

veryviolet/echelon3Config-driven PyTorch training framework: assemble nets, data, losses and trainers from YAML
Python★ 0↓ 300/moSep 9, 2026

AI-Hypercomputer/JetStreamJetStream is a throughput and memory optimized engine for LLM inference on XLA devices, starting with TPUs (and GPUs in future -- PRs welcome).
Python★ 457Sep 10, 2026
PyPI51Moderatehealth index
Python★ 1,127Sep 17, 2026
Go · npm51Moderatehealth index

NexusGPU/gpu-goUse your GPUs like NFS. The agent and IDE extension of GPUGo product
Go · TypeScript★ 9Sep 20, 2026