Skip to content

W6: Nautilus n_batch scan — SEARCHES_NAUTILUS_N_BATCH override + A100 array submits (#163) - #167

Draft
Jammy2211 wants to merge 2 commits into
mainfrom
feature/nautilus-nbatch-scan
Draft

W6: Nautilus n_batch scan — SEARCHES_NAUTILUS_N_BATCH override + A100 array submits (#163)#167
Jammy2211 wants to merge 2 commits into
mainfrom
feature/nautilus-nbatch-scan

Conversation

@Jammy2211

Copy link
Copy Markdown
Contributor

Summary

  • vmap_batch_for_cell honours SEARCHES_NAUTILUS_N_BATCH (mirrors the SEARCHES_NSS_* overrides); _sampler_config_dict goes through the same function so the JSON records the value run. n_batch is not an af.Nautilus identifier field — the submits carry it in --config-name (hpc_a100_fp64_nbatch${NB}) so results and output/ dirs stay disjoint.
  • Two array submits: imaging/mge/hst × {64, 128, 256, 512, 1000}, imaging/delaunay/hst × {16, 64, 256}.

Status

Submitted on RAL 2026-08-24 as 339842[0-4] and 339843[0-2] (PENDING behind another user's 8-GPU allocation). Draft until the rows are harvested and the read-out (ms/eval vs n_batch, ESS, logZ stability) is written into phase_02_nss_mainline/RESULTS.md.

Closes #163 on harvest.

🤖 Generated with Claude Code

https://claude.ai/code/session_01EDABYoH6giHXhFJUks8yd6

Jammy2211 and others added 2 commits August 24, 2026 14:50
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

W6 · Nautilus n_batch scan on MGE + Delaunay (A100): is the 4× per-eval overhead a config knob?

1 participant