💭 [Data] Fix DeepSeek-R1 case #3522
Merged
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
…tes to full conversations
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
|
Nice, just a quick question, this issue occurs with DeepSeek-v3 but not with DeepSeek-v3-0528, right? |
|
@qgallouedec checking! |
|
@qgallouedec yes correct no issue with DeepSeek-v3-0528 on main |
qgallouedec
reviewed
Jun 4, 2025
Comment on lines
-141
to
-154
| # Ensure that the prompt is the initial part of the prompt-completion string | ||
| if "prompt" in example: | ||
| error_message = ( | ||
| "The chat template applied to the prompt + completion does not start with the chat template applied to " | ||
| "the prompt alone. This can indicate that the chat template is not supported by TRL." | ||
| "\n**Prompt**:\n{}\n\n**Prompt + Completion**:\n{}" | ||
| ) | ||
| if "chosen" in example and not prompt_chosen.startswith(prompt): | ||
| raise ValueError(error_message.format(prompt, prompt_chosen)) | ||
| if "rejected" in example and not prompt_rejected.startswith(prompt): | ||
| raise ValueError(error_message.format(prompt, prompt_rejected)) | ||
| if "completion" in example and not prompt_completion.startswith(prompt): | ||
| raise ValueError(error_message.format(prompt, prompt_completion)) | ||
|
|
qgallouedec
changed the title
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Outdated suggestions cannot be applied.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.
What does this PR do?
fixes the issue that:
add_generation_prompt=True, DeepSeek's tokenizer would add<think>tokens<think>tokens were addedprompt_completion[len(prompt):], but since the strings didn't match at the beginning, it would fail thestartswithcheckWe now add
add_generation_prompt=Falseto all calls that apply chat templates to full conversations (prompt + completion/chosen/rejected) to ensure consistency.Fixes #3490
Before submitting
Pull Request section?
to it if that's the case.
Who can review?
Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.