FastWAM experiment checkpoints (raw, undocumented)

Raw research checkpoints from internal experiments built on top of FastWAM (Fast-WAM, arXiv 2603.16666, MIT) and Wan2.2-TI2V-5B (Apache-2.0).

  • This is a storage dump, not a release: no documentation, no evaluation numbers, no support, and no guarantee that any checkpoint is useful or even correct.
  • Directory names are experiment names; the layout mirrors the working tree (runs/<experiment>/..., runs90/..., checkpoints/...). Only weight files (.pt, .pth, .ckpt, .safetensors, .bin) are included.
  • Licensing: weights are derived from the two base models above, so their licenses apply. Some runs were additionally trained with supervision derived from third-party teacher models; that teacher's license was not checked โ€” verify it before any reuse.
  • .pt files are Python pickles: only load them if you trust the source.

arc4rc/ckpt_geoft โ€” NON-COMMERCIAL

Fine-tune of the 4RC model (https://github.com/Luo-Yihang/4RC). It is a derivative of S-Lab-licensed weights: non-commercial use only, redistribution requires the notice and conditions in arc4rc/ckpt_geoft/LICENSE_4RC_S-Lab-1.0.txt (reproduced there in full). For commercial use contact the 4RC authors.

Downloads last month

-

Downloads are not tracked for this model. How to track
Video Preview
loading