Skip to content

[draft] New WR(-1.4sec): drop MLP blocks in first layer. Includes WR Changes from PR#109, PR#117 and PR#118 #120

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
ClassicLarry merged 8 commits into from
Oct 15, 2025
Merged

[draft] New WR(-1.4sec): drop MLP blocks in first layer. Includes WR Changes from PR#109, PR#117 and PR#118 #120

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
9d9dc96
adding sparse attention gate
ClassicLarry Aug 23, 2025
12dfed7
Update README.md
ClassicLarry Aug 23, 2025
84cd472
Added FA3 record
varunneal Aug 27, 2025
393a478
Merge pull request #1 from varunneal/master
EmelyanenkoK Sep 2, 2025
8386811
FA3 with flash_attn_varlen_func (#1)
varunneal Sep 4, 2025
383ea51
Merge pull request #2 from varunneal/master
EmelyanenkoK Sep 4, 2025
1d02940
Add skip_mlp_block code and results
Sep 5, 2025
29f3978
Fix senseless removing of 12 (non-existent) block mlp layer
EmelyanenkoK Sep 8, 2025
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
1 change: 1 addition & 0 deletions .gitignore
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,3 +2,4 @@ fineweb10B/
pylog124M/
__pycache__/
logs/
.DS_Store
4 changes: 2 additions & 2 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,8 +38,8 @@ To run the current record, run the following commands.
git clone https://github.com/KellerJordan/modded-nanogpt.git && cd modded-nanogpt
pip install -r requirements.txt
pip install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu126 --upgrade
# downloads only the first 800M training tokens to save time
python data/cached_fineweb10B.py 8
# downloads only the first 900M training tokens to save time
python data/cached_fineweb10B.py 9
./run.sh
```

Expand Down
Loading