[Train-on-what; human simulator] #85
Merged
+52
−23
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
joschu
approved these changes
Nov 13, 2025
Collaborator
joschu
left a comment
joschu
left a comment
There was a problem hiding this comment.
This looks good. In case you haven't yet, could you do viz_sft_dataset with a couple different renderers and train_on_what settings and look at the outputted colorized text and make sure it looks right?
Collaborator
|
In particular, try both the role_colon renderer (nontrivial ac_tail) and another special token renderer. |
Contributor
Author
Thanks for checking! yeah it works. 
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Adding options to train on user-specified messages.
Ran
python -m tinker_cookbook.recipes.chat_sl.train model_name=Qwen/Qwen3-8B-Base \ dataset=no_robots \ learning_rate=5e-4 \ batch_size=64 \ lora_rank=64 \ eval_every=20 \ save_every=20 \ wandb_project=cookbook_sland I can reproduce the performance.