From 1897d2a9fac730aa5f5a8e4ef337f7a0d312aa8a Mon Sep 17 00:00:00 2001 From: "Dr. Dominik Lindorfer" Date: Thu, 10 Aug 2023 11:34:41 +0200 Subject: [PATCH] Update README.md --- README.md | 1 + 1 file changed, 1 insertion(+) diff --git a/README.md b/README.md index 0f02916ff..f3ae240fd 100755 --- a/README.md +++ b/README.md @@ -199,6 +199,7 @@ The SQL-LLaMA models are fine-tuned using HuggingFace's Trainer an the following | Weight decay | 0.1 | 0.1 | 0.1 | 0.1 | | Warm-Up-Ratio | - | - | - | - | +**SQL-LLaMA 13B with 5 and 10 Epoch Training are now available too! (Same other parameters as SQL-LLaMA-13B)** Please note that both SQL-LLaMA-small 7B & 13B use the same LIMA training strategy proposed in Ref. [7] except that no dropout has been used and a cosine LR scheduler was employed.