Apr 27

1. 06:08

   v3: patch_size=4 (64 tokens), 2 core layers, iters [2,3,4], ~16min total training

   asdf98
   Verified signature

   65af9bd
2. 06:08

   PERF: force fp32 for FFT, JIT scan, fix RMSNorm, cache Manhattan

   asdf98
   Verified signature

   eb07d9d
3. 05:55

   Fix: force_download=True to bypass HF cache, add assertion checks

   asdf98
   Verified signature

   213365a
4. 05:49

   v2: Use pre-trained SD-VAE, fix all bugs, pre-cache everything, massive speedup

   asdf98
   Verified signature

   2aee9f4
5. 05:49

   v2: Use pre-trained SD-VAE, fix all bugs, pre-cache everything, massive speedup

   asdf98
   Verified signature

   a8d6acc
6. 05:01

   Perf fix: cache Manhattan dist, optimize scan, pre-cache CLIP, fix deprecated AMP API

   asdf98
   Verified signature

   99a5f54
7. 05:01

   Perf fix: cache Manhattan dist, optimize scan, pre-cache CLIP, fix deprecated AMP API

   asdf98
   Verified signature

   3e17403
8. 04:49

   Update README with Colab notebook link

   asdf98
   Verified signature

   ce0086e
9. 04:48

   Add complete Colab/Kaggle training notebook

   asdf98
   Verified signature

   cb1106e
10. 04:37

    Add ARCHITECTURE.md

    asdf98
    Verified signature

    91e9e63
11. 04:37

    Add test_iris.py

    asdf98
    Verified signature

    89579fd
12. 04:37

    Add train_iris.py

    asdf98
    Verified signature

    5a5cffa
13. 04:37

    Add iris_model.py

    asdf98
    Verified signature

    774e194
14. 04:36

    Add IRIS architecture - novel mobile-first image generation

    asdf98
    Verified signature

    f69abc9
15. 04:35

    initial commit

    asdf98
    Verified signature

    2e6e763