| license: bigscience-bloom-rail-1.0 | |
| This is a custom version of the original [BLOOM weights](https://huggingface.co/bigscience/bloom) to make it fast to use with the [DeepSpeed-Inference](https://www.deepspeed.ai/tutorials/inference-tutorial/) engine which uses Tensor Parallelism. In this repo the tensors are split into 8 shards to target 8 GPUs. | |
| The full BLOOM documentation is [here](https://huggingface.co/bigscience/bloom). | |
| To use the weights in repo, you can adapt to your needs the scripts found [here](https://github.com/bigscience-workshop/Megatron-DeepSpeed/tree/main/scripts/inference) (XXX: they are going to migrate soon to HF Transformers code base, so will need to update the link once moved). | |