hunoutl/bloomchat-deepspeed-inference-fp16

This is a copy of the original 🌸 BLOOMChat weights that is more efficient to use with the DeepSpeed-Inference 🚀. In this repo the original tensors are split into 8 shards to target 8 GPUs, this allows the user to run the model with DeepSpeed-inference Tensor Parallelism.

For specific details about the BLOOMChat model itself, please see the original BLOOMChat model card.

This work was performed using AI/HPC resources (Jean Zay supercomputer) from GENCI-IDRIS