Pepijn dbe01b0444 fix(profiling): fix pi0 cuBLAS error and pi05 OOM on 22GB GPU ...

- Move cudnn_deterministic to per-spec train_args instead of hardcoding
  it for all models. cuBLAS deterministic mode triggers internal errors
  on Gemma-based models (pi0, pi05) during backward pass.
- Enable use_amp=true for pi0, pi0_fast, and pi05 to reduce memory
  footprint from fp32 (~16GB weights alone) to bf16, fitting within
  22GB GPU budget with room for activations and gradients.
- Small models (act, diffusion, multi_task_dit) still use deterministic
  mode for reproducible profiling results.

Made-with: Cursor

2026-04-16 15:34:17 +02:00

..

extract_task_descriptions.py

feat(ci): benchmark smoke tests with isolated Docker images (LIBERO + MetaWorld) (#3319)

2026-04-13 21:24:01 +02:00

parse_eval_metrics.py

feat(ci): benchmark smoke tests with isolated Docker images (LIBERO + MetaWorld) (#3319)

2026-04-13 21:24:01 +02:00

run_model_profiling.py

fix(profiling): fix pi0 cuBLAS error and pi05 OOM on 22GB GPU

2026-04-16 15:34:17 +02:00