Popular repositories Loading

1. stallscope stallscope Public

   Early-warning bottleneck profiler for GPU training nodes: GPU + RDMA fabric telemetry with job-level classification

   Python 10 1

2. nccl-doctor nccl-doctor Public

   Evidence-based NCCL failure diagnosis, topology linting, and tuning advice for GPU clusters.

   Python 3

3. torchguard torchguard Public

   Zero-code-change PyTorch training profiler: verdicts instead of traces, under 2% overhead

   Python 1

4. datapath-doctor datapath-doctor Public

   10-rule diagnostic engine for the ML training storage/input path, pinpoints disk I/O saturation, network filesystem latency, CPU-bound decode, prefetch underrun, and checkpoint stalls before they m…

   Python 1

5. infra-for-ai infra-for-ai Public

   ml infra network, gpu networking dev, networking for ml

   HTML

6. gpu-networking-examples gpu-networking-examples Public

   GPU networking quick examples for NCCL and DPDK

   Cuda