Support KTO Trainer with Unsloth #1001
Merged
Conversation
This file contains 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
|
Oh thanks so much!! |
danielhanchen
added a commit
that referenced
this pull request
Sep 8, 2024
* Update _utils.py * Update _utils.py * Update _utils.py * Update _utils.py * Update _utils.py * Update tokenizer_utils.py * Update tokenizer_utils.py * Update tokenizer_utils.py * update token retrieval logic (#952) * Fix DPO (#947) * Update _utils.py * Update _utils.py * Update _utils.py * Update _utils.py * Update _utils.py * Update tokenizer_utils.py * Update tokenizer_utils.py * Update tokenizer_utils.py * update hf token retrieval logic --------- Co-authored-by: Daniel Han <danielhanchen@gmail.com> * Update llama.py * get_token * Update README.md * Update gemma2.py * Update rms_layernorm.py * synchronize * Update gemma2.py * Update rms_layernorm.py * Update rms_layernorm.py * Update rms_layernorm.py * layernorm * Update rms_layernorm.py * Update gemma2.py * Update rms_layernorm.py * Update rms_layernorm.py * revert * Gemma * Update rms_layernorm.py * Update rms_layernorm.py * Update rms_layernorm.py * Update rms_layernorm.py * Update rms_layernorm.py * Update rms_layernorm.py * Update rms_layernorm.py * Update rms_layernorm.py * Update rms_layernorm.py * Update rms_layernorm.py * Update rms_layernorm.py * Update rms_layernorm.py * Update rms_layernorm.py * Update rms_layernorm.py * Update rms_layernorm.py * Update rms_layernorm.py * Update rms_layernorm.py * Update rms_layernorm.py * Update rms_layernorm.py * Update rms_layernorm.py * Update rms_layernorm.py * Update rms_layernorm.py * Update rms_layernorm.py * Update gemma2.py * Change UnslothTrainingArguments base class to SFTConfig (#979) * Cohere * Update trainer.py * Cohere * Cohere * New models * Update llama.py * Update llama.py * Update cohere.py * Update llama.py * Update cohere.py * retry * Update fast_lora.py * Update llama.py * Update fast_lora.py * Update llama.py * Update llama.py * Update cross_entropy_loss.py * _apply_lora_mlp * Update _utils.py * Gemma fixes * Update llama.py * Update flex_attention.py * Update llama.py * layernorm * Update llama.py * Update llama.py * Flex Attention * Update gemma2.py * Update __init__.py * Update flex_attention.py * Update flex_attention.py * Update flex_attention.py * Update flex_attention.py * Update flex_attention.py * Update flex_attention.py * Update flex_attention.py * Update flex_attention.py * Update flex_attention.py * Update flex_attention.py * Update flex_attention.py * Update flex_attention.py * Update flex_attention.py * Update flex_attention.py * Update flex_attention.py * Update flex_attention.py * Update flex_attention.py * Update chat_templates.py (#999) fix all misspelled "unsued" to "unused" * Update key from "from" to "user" (#1000) When use [tokenizer.apply_chat_template](https://huggingface.co/docs/transformers/main/en/chat_templating), the key should be "role" rather than "from", this is liknk to [this issue](#994) I don't know it is suitable for all situation, I also can add a dedicated parameter of the key if you think it is better. * Update chat_templates.py * Also patch the KTO trainer (#1001) * flex attention * Update llama.py * Update flex_attention.py * Update flex_attention.py * Update _utils.py * Update _utils.py * Update flex_attention.py * Update gemma2.py * Update gemma2.py --------- Co-authored-by: Hafedh <70411813+not-lain@users.noreply.github.com> Co-authored-by: Tuan Pham <82665400+vTuanpham@users.noreply.github.com> Co-authored-by: Yihao Wang <42559837+AgainstEntropy@users.noreply.github.com> Co-authored-by: Peng <zphu1024@gmail.com> Co-authored-by: Kyle Corbitt <kyle@openpipe.ai>
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.
This patch appears to be both necessary and sufficient to successfully use KTOTrainer with Unsloth!