← All experiments THE RESEARCH RECORD / shake-small-baseline

Small Shakespeare baseline

What held-out byte perplexity does the standard small AGPT recipe (d64 L2, depth-16 trie, 10 static epochs, rmsprop 3e-3, anc-grad) reach on Shakespeare's 5% tail?

baselineconcludedn/aattentioneval: canonical

Run history: 1 of 1 predate the 2026-09-24 CUDA kernel race fix; 1 of 1 predate the 2026-09-28 exact ancestor backward.

Opened Updated
THE ANSWER SO FAR

Rolling byte PPL 10.435 (3.38 bits/byte) on the tail-heldout split, in 86 s of wall time. This was the first held-out run under the orchestrator.

d64 L2, depth 16, 10 static epochs 10.4353 rolling byte PPL Run record ↗pre-race-fixtruncated-ancestor-gradient
TOPICS
baseline
RELATED
cudax-d124-probe
SUPERSEDED BY
None listed
CODE
Main branch ↗
THE FULL RECORD

Experiment notes

Open directory on GitHub ↗

shake-small-baseline

Status: concluded, n/a (reviewed 2026-09-28; this line previously read "active"). The answer is in the front matter above.

Hypothesis

(fill in)

Scope

(fill in)

Results

<!-- agpt-experiment-table:start -->

Run ID byte_ppl bits/byte word_ppl wall (s)
20260526T033625-d64l2-d16-10ep 10.4353 3.3834 494799.75 86.0
<!-- agpt-experiment-table:end -->

Conclusion

(fill in once enough runs have landed)

View the experiment directory on GitHub ↗Browse all experiments →