heatdrop.ai

A model-agnostic Python toolkit that handles the boring parts of computer vision: annotations, dataset juggling, and tracking.

A PyTorch implementation of "Attention Is All You Need" that scales from 13M to multi-billion parameter models.

OpenMed packages clinical entity extraction and HIPAA-grade de-identification into models small enough for Apple Silicon and impatient DevOps teams.

Companion notebooks for an 800-page quantitative finance textbook, from linear regression to deep reinforcement learning trading agents.

A notebook-based workout plan for PyTorch fluency, from linear regression up to building LLM components from scratch.

Reproducible world-model research is usually a pile of glue scripts; this library tries to make it a single import.

A C-based RL library claiming to train super-human models in seconds, built from scratch by a research team that also sells professional environments.

MLX-VLM crams speculative decoding, continuous batching, and KV cache quantization into a Mac-native toolkit for running multimodal models locally.

Chinese open-source community Datawhale built a from-zero embodied-AI course that gets you from `print('hello')` to fine-tuning SmolVLA and Pi0.

A grab-bag node pack whose Set/Get rewrite might finally tame your worst workflow tangles.

A single Python framework that pretrains DINOv2/v3 on unlabeled data, then fine-tunes and distills for detection and segmentation tasks.

A curated, Qualcomm-optimized model catalog that handles the messy translation from PyTorch to on-device NPU binaries.

A Python bridge that turns messy geospatial data—streets, transit feeds, building footprints—into PyTorch Geometric tensors without the usual hand-rolled pain.

Curated technical deep-dives covering everything from NVLink signal integrity to Kubernetes GPU scheduling and Huawei NPU porting.

Pre-trained voice activity detection that runs on a CPU thread in under a millisecond, no API keys or telemetry attached.

Megatron-LM splits into a reference training stack and a composable core for anyone who needs to squeeze every FLOP from a GPU cluster.

A world-model RL algorithm that reportedly needs no tuning across Atari, Minecraft, and continuous control.

A pedagogical autograd engine so small you can read it on your coffee break, yet complete enough to train a real MLP.

MNN is a battle-tested inference framework designed to squeeze large language models and vision transformers onto mobile CPUs and GPUs without calling home.

TransformerLens lets you intercept, cache, and surgically edit the hidden activations of 50+ language models as they run.