[ElmoPA]

ImageNet training can now require a complete chain of native CPU, corpus, decoder, target-runtime and GPU smoke evidence before full submission. The smoke exercises the exact complete MoT with image/class conditioning, validates active and inactive parameters, all-rank optimizer and RNG state, scheduled validation and sampling, and resource capacity.

The image adapter and five source-data producer components remain byte-identical to the parent. This change adds operational admission and instrumentation; it preserves the model, data transform, noise and loss definitions used for this experiment.

Validation: 580 focused local CPU tests pass, including Python 3.11/3.12 Pillow profiles and a real small Lightning fixture. An independent review verified the sealed publication and 254 preserved scientific files. Production corpus preparation is still running; target-runtime and four-H100 ImageNet validation are not yet certified.

[ElmoPA]

Warning

This pull request is not mergeable via GitHub because a downstack PR is open. Once all requirements are satisfied, merge this PR as a stack on Graphite.
Learn more

• Bind QK full-v5 recovery ops to published service #326
• Add exact QK full-v5 recovery history admission #325
• Add verified ImageNet smoke and full training admission #323 👈 (View in Graphite)
• Add native ImageNet data and generation path for the joint MoT model #322
• Add optional QK normalization, RMSNorm and SwiGLU to MoT experts #321
• Add Elmo continuous objectives and verified interruption recovery #320
• Add paired modality ablations for joint image forecast quality #265 : 1 other dependent PR (#276 )
• Train joint image, proprio and action MoT with smaller vector experts #264
• Add modular pixel video training and resilient log mirroring #262
• Optimize MoT attention and enable Flash4 BF16 defaults #261
• Add continuous-time MoT world model and offline data pipeline #260
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.