jgrusewski c20d810067 fix: supervised GPU smoke test dimension mismatches ...

- Liquid: remove seq_len from input (forward_loss uses 2D [batch, input_size])
- Mamba2: single sample input matching d_model=32
- xLSTM: remove seq_len from input (same as Liquid)
- Diffusion: create checkpoint dir for models using dir-based saves

All 8 supervised GPU smoke tests pass.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

2026-03-21 09:14:53 +01:00

16 KiB

Raw Blame History

View Raw