π₀.₅ with a recurrent memory mechanism, fine-tuned from lerobot/pi05_base on a real-robot memory task. A bank of 64 learnable memory tokens is carried across timesteps inside an episode and updated end-to-end with truncated backpropagation through time, so the policy can condition on an observation that is no longer visible in the current frame.

This is the port of the memory mechanism from μVLA (a fork of OpenVLA-OFT) onto π₀.₅ as implemented in LeRobot. Code: avanturist322/muVLA-pi0.5.

Sibling checkpoint: the same recipe with a shorter truncation window is at avanturist/pi05-mem-so101-pick-cube-color-m64-k2-tbptt. Read the Comparability section before treating the pair as an ablation over K.

Dataset: tttonyalpha/so101_pick_cube_300_task3 (LeRobot v3).

The cue is visible only early in the episode and is gone by the time the grasp has to be committed, which is what makes a recurrent state useful here rather than decorative.

Fine-tuned from lerobot/pi05_base on 8 GPUs (80 GB each) with DDP.

Final training loss 0.0312 (mean of the last 10 logged values: 0.0404); the run started at 1.910. train_config.json and training_summary.json in this repository are the verbatim artifacts the run wrote, including the DDP sharding check.

memory_module.pt is deliberately redundant with model.safetensors: the safetensors path goes through π₀.₅'s key remapper with strict=False, so a silent rename would leave the memory at its random initialization — which looks like "memory does not help" rather than like a bug. Loading the standalone file with strict=True makes that failure loud.

This checkpoint does not carry normalize_inputs / unnormalize_outputs buffers. The loader takes normalization from the dataset directory, because π₀.₅ discretizes the normalized state into 256 bins inside the text prompt — evaluating with the wrong quantiles feeds the model a corrupted prompt and produces plausible-looking but wrong actions, with no error raised.

So you need the dataset alongside the weights. dataset_stats.json here is a copy of meta/stats.json for reference and reproducibility, but the loader reads the dataset directory, not this file.

load_eval_policy reads memory_meta.json, so memory does not have to be re-declared on the command line: the checkpoint states whether memory is on and how it is configured.

Inference regime defaults to receding horizon for memory checkpoints (n_action_steps=1, requery every environment step, memory advances once per step exactly as in training). Pass receding_horizon=False for the open-loop chunked regime.

The two checkpoints are not batch-matched, so the difference between them is not a clean ablation over the truncation length:

The K=2 run saw twice the frames and took four times as many optimizer steps. Attributing its lower loss to the shorter TBPTT window would be wrong.

No in-environment success rates. Both checkpoints were trained but not yet evaluated on the physical SO-101 or in simulation, so this card reports training loss only. Training loss on a 300-episode imitation dataset is a weak proxy for task success and should not be read as one.

The memory mechanism this ports:

The base model is π₀.₅ (arXiv:2504.16054) as implemented in LeRobot. Weights inherit the Gemma Terms of Use from lerobot/pi05_base.