Skip to content

Add description for 16B model tokenizer for deepseek-v3 model #1530

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
tianyu-l merged 1 commit into from
Aug 5, 2025
Merged

Add description for 16B model tokenizer for deepseek-v3 model #1530

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions torchtitan/models/deepseek_v3/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,6 +16,9 @@ python scripts/download_tokenizer.py --repo_id deepseek-ai/DeepSeek-V3
python scripts/download_tokenizer.py --repo_id deepseek-ai/deepseek-moe-16b-base
```

> **Note:** We are reusing the tokenizer from deepseek-ai/deepseek-moe-16b-base to help users test and run the 16B model. This is not the official tokenizer for the DeepSeek-V3-16B model. The DeepSeek-V3 model has a different architecture from the deepseek-moe models (different attention implementation, MoE router implementation, etc.), making it not feasible to load deepseek-moe-16b model weights into DeepSeek-V3-16B.


## Training

### Debug Training
Expand Down
Loading