Sep 7

1. 15:36

   Calibrate duration from three timed runs; set verified 44k token ceiling; document memory findings Running

   multimodalart HF Staff
   Verified signature

   0581e9d
2. 15:29

   Use expandable_segments instead of cudaMallocAsync; log eviction effect

   multimodalart HF Staff
   Verified signature

   d209439
3. 15:20

   Evict connectors between stages; log VRAM to calibrate the token ceiling

   multimodalart HF Staff
   Verified signature

   c2f1d45
4. 15:14

   Calibrate ZeroGPU duration from the measured run; raise token ceiling to 46k

   multimodalart HF Staff
   Verified signature

   9f89609
5. 15:07

   Evict Gemma after prompt encode; free stage-1 latents; token budget guard; Gradio 6 theme/css on launch()

   multimodalart HF Staff
   Verified signature

   a4c64ba
6. 14:52

   MSR reference-to-video demo on LTX-2.5

   multimodalart HF Staff
   Verified signature

   6f29667
7. 14:39

   initial commit

   multimodalart HF Staff
   Verified signature

   ebbee14