Skip to content

Move four-head nanoGPT results to dedicated experiment #125

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
charlesmartin14 merged 1 commit into from
Aug 30, 2026
Merged

Move four-head nanoGPT results to dedicated experiment #125

Show file tree
Hide file tree
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
10 changes: 10 additions & 0 deletions baseline/experiments/nanogpt_four_head_2026_08_27_baseline/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,10 @@
# Four-head nanoGPT baseline — 2026-08-27

This experiment archives the completed five-seed AdamW versus MuonClip four-attention-head baseline.

- [Open rendered report](https://raw.githack.com/CalculatedContent/rg_optimizers/main/baseline/experiments/nanogpt_four_head_2026_08_27_baseline/results/five_seed/report.html)
- [Archived HTML](results/five_seed/report.html)
- [Exact protocol](results/five_seed/protocol.yaml)
- [Plot assets](results/five_seed/plots/)

The campaign used seeds 1337, 2027, 4099, 31415, and 271828. The report contains five-seed aggregate performance, WeightWatcher, and optimizer-diagnostic plots.
0 ...results/four_head_5seed/plots/adamw_D.png → ...eline/results/five_seed/plots/adamw_D.png
View file
Open in desktop
File renamed without changes
0 ...s/four_head_5seed/plots/adamw_ERG_gap.png → ...results/five_seed/plots/adamw_ERG_gap.png
View file
Open in desktop
File renamed without changes
0 ...ead_5seed/plots/adamw_alpha_clip_xmax.png → ...five_seed/plots/adamw_alpha_clip_xmax.png
View file
Open in desktop
File renamed without changes
0 ...ed/plots/adamw_alpha_clip_xmax_zoomed.png → ...ed/plots/adamw_alpha_clip_xmax_zoomed.png
View file
Open in desktop
File renamed without changes
0 ...four_head_5seed/plots/adamw_alpha_raw.png → ...sults/five_seed/plots/adamw_alpha_raw.png
View file
Open in desktop
File renamed without changes
0 ...ad_5seed/plots/adamw_alpha_raw_zoomed.png → ...ive_seed/plots/adamw_alpha_raw_zoomed.png
View file
Open in desktop
File renamed without changes
0 ...four_head_5seed/plots/adamw_num_traps.png → ...sults/five_seed/plots/adamw_num_traps.png
View file
Open in desktop
File renamed without changes
0 ..._head_5seed/plots/adamw_rand_distance.png → ...s/five_seed/plots/adamw_rand_distance.png
View file
Open in desktop
File renamed without changes
0 ...r_head_5seed/plots/grad_norm_pre_clip.png → ...ts/five_seed/plots/grad_norm_pre_clip.png
View file
Open in desktop
File renamed without changes
0 ...lts/four_head_5seed/plots/muon_clip_D.png → ...e/results/five_seed/plots/muon_clip_D.png
View file
Open in desktop
File renamed without changes
0 ...ur_head_5seed/plots/muon_clip_ERG_gap.png → ...lts/five_seed/plots/muon_clip_ERG_gap.png
View file
Open in desktop
File renamed without changes
0 ...5seed/plots/muon_clip_alpha_clip_xmax.png → ..._seed/plots/muon_clip_alpha_clip_xmax.png
View file
Open in desktop
File renamed without changes
0 ...lots/muon_clip_alpha_clip_xmax_zoomed.png → ...lots/muon_clip_alpha_clip_xmax_zoomed.png
View file
Open in desktop
File renamed without changes
0 ..._head_5seed/plots/muon_clip_alpha_raw.png → ...s/five_seed/plots/muon_clip_alpha_raw.png
View file
Open in desktop
File renamed without changes
0 ...seed/plots/muon_clip_alpha_raw_zoomed.png → ...seed/plots/muon_clip_alpha_raw_zoomed.png
View file
Open in desktop
File renamed without changes
0 ..._head_5seed/plots/muon_clip_num_traps.png → ...s/five_seed/plots/muon_clip_num_traps.png
View file
Open in desktop
File renamed without changes
0 ...d_5seed/plots/muon_clip_rand_distance.png → ...ve_seed/plots/muon_clip_rand_distance.png
View file
Open in desktop
File renamed without changes
0 ...ults/four_head_5seed/plots/train_loss.png → ...ne/results/five_seed/plots/train_loss.png
View file
Open in desktop
File renamed without changes
0 ...ad_5seed/plots/update_to_weight_ratio.png → ...ive_seed/plots/update_to_weight_ratio.png
View file
Open in desktop
File renamed without changes
0 ...ts/four_head_5seed/plots/val_accuracy.png → .../results/five_seed/plots/val_accuracy.png
View file
Open in desktop
File renamed without changes
0 ...ad_5seed/plots/val_generalization_gap.png → ...ive_seed/plots/val_generalization_gap.png
View file
Open in desktop
File renamed without changes
0 ...esults/four_head_5seed/plots/val_loss.png → ...line/results/five_seed/plots/val_loss.png
View file
Open in desktop
File renamed without changes
0 .../four_head_5seed/plots/val_perplexity.png → ...esults/five_seed/plots/val_perplexity.png
View file
Open in desktop
File renamed without changes
0 ...lts/four_head_5seed/plots/weight_norm.png → ...e/results/five_seed/plots/weight_norm.png
View file
Open in desktop
File renamed without changes
0 ...ine/results/four_head_5seed/protocol.yaml → ..._baseline/results/five_seed/protocol.yaml
View file
Open in desktop
File renamed without changes.
0 ...eline/results/four_head_5seed/report.html → ...27_baseline/results/five_seed/report.html
View file
Open in desktop
File renamed without changes.
13 changes: 7 additions & 6 deletions baseline/experiments/nanogpt_one_head_2026_08_21_baseline/results/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,8 +1,9 @@
# Archived nanoGPT baseline results
# One-head nanoGPT baseline results

These reports archive the completed five-seed AdamW and MuonClip baselines. The rendered links are pinned to commit `19e3c280d55beb89cdf68e23288dab78ccb0607d`, including every referenced plot asset.
This directory contains the completed five-seed AdamW and MuonClip one-head baseline.

| Experiment | Rendered report | Archived files |
|---|---|---|
| One attention head, five seeds | [Open rendered report](https://raw.githack.com/CalculatedContent/rg_optimizers/19e3c280d55beb89cdf68e23288dab78ccb0607d/baseline/experiments/nanogpt_one_head_2026_08_21_baseline/results/one_head_5seed/report.html) | [HTML](one_head_5seed/report.html) · [Protocol](one_head_5seed/protocol.yaml) |
| Four attention heads, five seeds | [Open rendered report](https://raw.githack.com/CalculatedContent/rg_optimizers/19e3c280d55beb89cdf68e23288dab78ccb0607d/baseline/experiments/nanogpt_one_head_2026_08_21_baseline/results/four_head_5seed/report.html) | [HTML](four_head_5seed/report.html) · [Protocol](four_head_5seed/protocol.yaml) |
- [Open rendered report](https://raw.githack.com/CalculatedContent/rg_optimizers/main/baseline/experiments/nanogpt_one_head_2026_08_21_baseline/results/one_head_5seed/report.html)
- [Archived HTML](one_head_5seed/report.html)
- [Exact protocol](one_head_5seed/protocol.yaml)

The four-head baseline is archived separately under [`nanogpt_four_head_2026_08_27_baseline`](../../nanogpt_four_head_2026_08_27_baseline/).
Loading