feat(doc): update thinking and chat_template notes (#3114) [skip ci]
* feat: update thinking and chat_template notes * fix: grammar
This commit is contained in:
@@ -106,6 +106,16 @@ See [Nanobit/text-tools-2k-test](https://huggingface.co/datasets/Nanobit/text-to
|
||||
|
||||
Refer to [our docs](https://docs.axolotl.ai/docs/dataset-formats/conversation.html#using-tool-use) for more info.
|
||||
|
||||
### Thinking and chat_template masking conflict
|
||||
|
||||
OpenAI’s Harmony template hides `thinking` in all non-final turns, which conflicts with Axolotl’s `chat_template` masking.
|
||||
|
||||
If your dataset has `thinking` content mid-turn, there are two paths we recommend:
|
||||
|
||||
- Train only on the last turn. This can be accomplished via chat_template's [train on last doc](https://docs.axolotl.ai/docs/dataset-formats/conversation.html#training-on-last-message).
|
||||
|
||||
- Adjust your dataset to only have `thinking` content in the last turn.
|
||||
|
||||
### TIPS
|
||||
|
||||
- Read more on how to load your own dataset at [docs](https://docs.axolotl.ai/docs/dataset_loading.html).
|
||||
|
||||
Reference in New Issue
Block a user