text-generation-inference

952b450a - Using HF_HOME instead of CACHE to get token read in addition to models. (#2288)