Download — Bloom Model ((full))

If you don't have the hardware to host the model, you can use Petals to load the model in a distributed, BitTorrent-style network. Deployment Best Practices

Loading the full 176B model in 16-bit precision requires roughly 350GB of VRAM. This typically necessitates a cluster of 8x NVIDIA A100 (80GB) GPUs. download bloom model

If you lack enterprise-grade hardware, you can download 8-bit or 4-bit quantized versions. These can run on significantly less VRAM (approx. 175GB or 90GB respectively). If you don't have the hardware to host