From 8d77dc385e2f404a6f591b2c1f761834838db100 Mon Sep 17 00:00:00 2001 From: Wing Lian Date: Tue, 24 Dec 2024 15:48:35 -0500 Subject: [PATCH] hash for temperature too --- src/axolotl/utils/data/sft.py | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/src/axolotl/utils/data/sft.py b/src/axolotl/utils/data/sft.py index cfc40406e..6850f3c07 100644 --- a/src/axolotl/utils/data/sft.py +++ b/src/axolotl/utils/data/sft.py @@ -176,10 +176,11 @@ def load_tokenized_prepared_datasets( + "@" + str(cfg.group_by_length) + "@" + + str(cfg.kd_temperature or 1.0) + "|".join( sorted( [ - f"{d.path}:{d.type}:{d.shards}:{d.conversation}{d.split}" + f"{d.path}:{d.type}:{d.shards}:{d.conversation}:{d.split}:{d.temperature or 1.0}" for d in cfg_datasets ] )