From 9d98557772e88a8367d164b6426992a439bb5724 Mon Sep 17 00:00:00 2001 From: chris-aeviator Date: Sun, 27 Aug 2023 17:39:10 +0200 Subject: [PATCH] add POC news --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index a1a1d867..674ac631 100644 --- a/README.md +++ b/README.md @@ -9,7 +9,7 @@ TGI is well suited for distributed/ cloud burst/ on-demand workloads, yet HF's f ## Goals - ☑️ loads LLama2 in 4bit on a Pascal GPU (1080, Llama 2 7B) - Support Model loading from wherever you want (HDFS, S3, HTTPS, …) -- Support Adapters (LORA/PEFT) without merging (possibly huge) Checkpoints and uploading them to 🤗 +- Support Adapters (LORA/PEFT) without merging (possibly huge) Checkpoints and uploading them to 🤗 -> POC in https://github.com/ohmytofu-ai/tgi-angryface/commit/aba56c1343aa77ba0a07d14327d3e52736334308 - Support last Gen GPUS (back to Pascal hopefully) - Reduce operational cost by making TGI-😑 an disposable, hot swapable workhorse - Get back to a truyl open source license