Are our runs reasonable?

Controlled understanding of our ability to scale.

Life's too short to sweep everything.

Need to figure out how to get fine-tuning work. Testing on RTT single session to kick off.

- On RTT data alone, factor 1 is the only one that achieves competence. - With Maze data augmented (i.e. 30K base + 3K added, likely not a representational quality increase), RTT Factor 2 becomes feasible.

Goal: Understand scaling when it probably works

Does scaling help us improve 5K Pitt trials?

Hm, BPS clouds picture, let's just look at loss.

To scale well we should follow good practices.

Masking more increases throughput (and perf/flop), but at what cost?

Factorized is objectively much more efficient but a flat model has more eventual potential/is more agnostic. Per Kaiming's spatiotemporal paper, this might be better at scaled throughputs.

4 appears better for BPS, loss is too noisy.