From abe9ff94bbe96f01307e401ff197dcb60b76a515 Mon Sep 17 00:00:00 2001 From: randaller Date: Sun, 19 Mar 2023 16:47:16 +0300 Subject: [PATCH] Update README.md --- README.md | 2 ++ 1 file changed, 2 insertions(+) diff --git a/README.md b/README.md index 1db135a..927b36b 100644 --- a/README.md +++ b/README.md @@ -157,6 +157,8 @@ Confirming that 30B model is able to generate SQL code: https://github.com/randa ## Hugging Face 🤗 version (inference & training) +### Inference + Thanks to Yam Peleg, we now have *"No overengineering bullshit"* version. You do not need to download torrent or merge weights, as model shards and tokenizer will be downloaded from HF automatically at the first run. They will be cached in [C:\Users\USERNAME\\.cache\huggingface\hub] folder under Windows, so do not forget to clean up to 250 Gb after experiments.