giakhuyendihoc/pimamba-l18-depth4-libero-nolora-fullvlm

PI-Mamba L18 Depth4 LIBERO no-LoRA fullVLM

30k-step LIBERO finetuned checkpoint for PI-Mamba L18 depth4 no-LoRA fullVLM.

Checkpoint

• OpenPI config: pi05_mamba_tiny_ctxd18_gated_fullvlm_libero_5090
• Experiment: icra_nolora_fullvlm_pimamba_l18_depth4_cosine5e5_5e6_b16_acc1_seed42
• Step: 30000
• Initializer: pi05_base_pytorch
• Adaptation: full VLM / full vision / full PI-Mamba action path, no LoRA
• Context: full L18 VLM prefix, ctxD K=64, pooled gated conditioning
• Mamba action mixer: depth 4, width 512
• Training: batch 16, grad accumulation 1, seed 42
• LR schedule: cosine from 5e-5 to 5e-6 over 30k steps
• LIBERO eval: seed 7, 50 trials per task, 10 denoising steps

Files included:

• model.safetensors
• metadata.pt
• assets/physical-intelligence/libero/norm_stats.json

LIBERO Results

Spatial	Object	Goal	LIBERO-10	Average
99.2	99.8	98.8	95.8	98.4

Efficiency

Metric	Value
Action expert core	29.74M
Train time	13.78 h
Best train step time	1.40 s/step
GFLOPs/query	4478.26
Prefix + ctxD	46.61 ms
Denoise	56.40 ms
Denoise per step	5.64 ms
Median model latency	107.61 ms
Query Hz	9.29
Action Hz	46.47

Use

Use with the matching PI-Mamba/OpenPI code and config:

--config-name pi05_mamba_tiny_ctxd18_gated_fullvlm_libero_5090
--checkpoint-dir /path/to/checkpoint

The checkpoint includes the LIBERO norm stats under assets/physical-intelligence/libero.