feat: update yml chat_template to specify dataset field (#2001) [skip ci]

* feat: update yml chat_template to specify dataset field * feat: replace sharegpt references with chat_template
2024-10-29 21:26:03 +07:00
parent 107b67b852
commit 8c3a727f9d
7 changed files with 26 additions and 13 deletions
--- a/examples/deepseek-v2/qlora-fsdp-2_5.yaml
+++ b/examples/deepseek-v2/qlora-fsdp-2_5.yaml
@@ -16,7 +16,10 @@ chat_template: deepseek_v2
 datasets:
  - path: mlabonne/FineTome-100k
    type: chat_template
-    split: train
+    split: train[:20%]
+    field_messages: conversations
+    message_field_role: from
+    message_field_content: value

 dataset_prepared_path: last_run_prepared
 val_set_size: 0.0