Tensor Parallelism Python Packages

kaggle-vllm

Run upstream vLLM on Kaggle dual NVIDIA Tesla T4 GPUs with a validated CUDA runtime, safe staging, NCCL tensor parallelism, and reproducible compatibility checks.

petals

🌸 Run LLMs at home, BitTorrent-style. Fine-tuning and inference up to 10x faster than offloading

torchshard

Slicing a PyTorch Tensor Into Parallel Shards

test-petals

🌸 Run LLMs at home, BitTorrent-style. Fine-tuning and inference up to 10x faster than offloading