Dataset is now directly tokenized to avoid further post-processing of the trainer; locCustomPromptTemplate to specify a custom Jinja template; showChatTemplate for debugging;

2024-12-26 21:42:59 +01:00
parent cb9e62dbac
commit 63192f2bc0
5 changed files with 26 additions and 6 deletions
--- a/testProject/llama3_prompt_min.txt
+++ b/testProject/llama3_prompt_min.txt
@@ -0,0 +1,3 @@
+{% set loop_messages = messages %}{% for message in loop_messages %}{% set content = '<|start_header_id|>' + message['role'] + '<|end_header_id|>
+
+'+ message['content'] | trim + '<|eot_id|>' %}{{ content }}{% endfor %}