Pinned Loading

1. distributed-transformer distributed-transformer Public

   Distributed decoder-only Transformer training with PyTorch DDP and FSDP

   Python

2. flash-attention-cuda flash-attention-cuda Public

   Flash attention implementation Minimal CUDA implementation of Flash Attention with tiled computation and online softmax. Educational implementation based on Dao et al., 2022.

   Cuda 21

3. llm-inference-benchmarking llm-inference-benchmarking Public

   Reproducible LLM inference batching benchmarks with TTFT, ITL, throughput, GPU memory, and KV-cache metrics.

   Python