Downloads · 30 days
0
giakhuyendihoc/pi05-l18-libero-nolora-fullvlm-baseline
pi05-l18-libero-nolora-fullvlm-baseline is a robotics model from giakhuyendihoc. Use it for the robotics task on the model card, and read the license before you ship it in a product.
30k-step LIBERO finetuned checkpoint for the PI0.5 L18 no-LoRA fullVLM baseline used for PI-Mamba comparison.
Downloads · 30 days
0
Access
Public
Updated Jul 15, 2026
Parameters
3.6B
7.5 GB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors7.5 GB · 100%
How the weights are stored.
BF163.5B · 97%
From the Hugging Face model README
30k-step LIBERO finetuned checkpoint for the PI0.5 L18 no-LoRA fullVLM baseline used for PI-Mamba comparison.
pi05_libero_full_5090icra_nolora_fullvlm_pi05_original_l18_cosine5e5_5e6_b16_acc1_seed4230000pi05_base_pytorch5e-5 to 5e-6 over 30k stepsFiles included:
model.safetensorsmetadata.ptassets/physical-intelligence/libero/norm_stats.json| Spatial | Object | Goal | LIBERO-10 | Average |
|---|---|---|---|---|
| 97.6 | 98.8 | 96.6 | 93.2 | 96.5 |
| Metric | Value |
|---|---|
| Action expert core | 311.46M |
| Train time | 14.20 h |
| Best train step time | 1.70 s/step |
| GFLOPs/query | 4690.08 |
| Prefix/context | 40.76 ms |
| Denoise | 160.26 ms |
| Denoise per step | 16.03 ms |
| Median model latency | 228.59 ms |
| Query Hz | 4.37 |
| Action Hz | 21.87 |
Use with the matching OpenPI code and config:
--config-name pi05_libero_full_5090
--checkpoint-dir /path/to/checkpoint
The checkpoint includes the LIBERO norm stats under assets/physical-intelligence/libero.