feat(train): add cudnn_deterministic option for reproducible training (#3102)

Add a `cudnn_deterministic` flag to `TrainPipelineConfig` (default: False)
that sets `torch.backends.cudnn.deterministic = True` and disables benchmark
mode, eliminating CUDA floating-point non-determinism at the cost of ~10-20%
training speed. When False (default) the existing benchmark=True behaviour
is preserved.

src/lerobot/configs/train.py

@@ -50,6 +50,9 @@ class TrainPipelineConfig(HubMixin):
     # `seed` is used for training (eg: model initialization, dataset shuffling)
     # AND for the evaluation environments.
     seed: int | None = 1000
+    # Set to True to use deterministic cuDNN algorithms for reproducibility.
+    # This disables cudnn.benchmark and may reduce training speed by ~10-20%.
+    cudnn_deterministic: bool = False
     # Number of workers for the dataloader.
     num_workers: int = 4
     batch_size: int = 8