RoboTwin-pi05-30000-checkpoints

JackieMM/RoboTwin-pi05-30000-checkpoints

RoboTwin-pi05-30000-checkpoints is a machine learning model from JackieMM. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.

This repository contains the 30000-step checkpoints collected from local RoboTwin pi05 training runs.

Downloads · 30 days

Access

Public

Updated Jul 22, 2026

Repo size

862 GB

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

Other1.1 TB · 100%

At a glance

License: apache-2.0
Access: Public
Created: Jul 19, 2026
Updated: Jul 22, 2026
SHA: 4c5a1c2b

Model card

From the Hugging Face model README

RoboTwin pi05 30000-step Checkpoints

This repository contains the 30000-step checkpoints collected from local RoboTwin pi05 training runs.

Only final step-30000 checkpoints are included. Intermediate 5000/10000/15000/20000/25000 checkpoints are intentionally excluded.

Layout

checkpoints/pi05/pi05_base_aloha_lora/<run_name>/30000/ contains the regular pi05 LoRA checkpoints.
checkpoints/pi05_concept/<experiment_group>/<run_name>/30000/ contains the pi05_concept experiment checkpoints.
Each checkpoint directory preserves the original Orbax/JAX layout: _CHECKPOINT_METADATA, assets, params, and train_state.
MANIFEST.tsv maps every uploaded checkpoint back to its original local path.

Counts

Checkpoints: 92
Files: 2463
Total file size: 813.08 GiB

Groups

Family	Group	Checkpoints	Files	Size GiB
`pi05`	`pi05_base_aloha_lora`	12	320	105.51
`pi05_concept`	`pi05_concept_exp_a`	1	26	8.85
`pi05_concept`	`pi05_concept_exp_b`	1	27	8.86
`pi05_concept`	`pi05_concept_exp_c`	1	26	8.84
`pi05_concept`	`pi05_concept_exp_d`	1	26	8.85
`pi05_concept`	`pi05_concept_exp_d_alpha1_final`	1	28	8.87
`pi05_concept`	`pi05_concept_exp_d_l0_alpha1`	1	26	8.88
`pi05_concept`	`pi05_concept_exp_d_l11_alpha1`	1	25	8.89
`pi05_concept`	`pi05_concept_exp_d_l16_alpha1`	1	27	8.91
`pi05_concept`	`pi05_concept_exp_d_l17_alpha1`	1	26	8.92
`pi05_concept`	`pi05_concept_exp_d_l5_alpha1`	1	28	8.91
`pi05_concept`	`pi05_concept_exp_suffix_layer14_adaptive01_h16`	2	55	17.65
`pi05_concept`	`pi05_concept_exp_suffix_layer14_adaptive_h16`	5	130	44.00
`pi05_concept`	`pi05_concept_exp_suffix_layer14_alpha1`	3	84	26.64
`pi05_concept`	`pi05_concept_exp_suffix_layer14_rmsnorm_adaptive01_h16`	1	26	8.86
`pi05_concept`	`pi05_concept_exp_suffix_layer14_rmsnorm_adaptive10_h16`	1	28	8.86
`pi05_concept`	`pi05_concept_exp_suffix_layer14_rmsnorm_alpha015_h16`	10	275	88.37
`pi05_concept`	`pi05_concept_exp_suffix_layer14_rmsnorm_alpha02_h16`	4	108	35.36
`pi05_concept`	`pi05_concept_exp_suffix_layer15_adaptive01_h16`	2	55	17.65
`pi05_concept`	`pi05_concept_exp_suffix_layer15_alpha1`	1	26	8.87
`pi05_concept`	`pi05_concept_exp_suffix_layer15_rmsnorm_adaptive01_h16`	1	26	8.86
`pi05_concept`	`pi05_concept_exp_suffix_layer15_rmsnorm_adaptive10_h16`	1	27	8.86
`pi05_concept`	`pi05_concept_exp_suffix_layer15_rmsnorm_alpha015_h16`	10	268	88.42
`pi05_concept`	`pi05_concept_exp_suffix_layer16_adaptive01_h16`	2	51	17.63
`pi05_concept`	`pi05_concept_exp_suffix_layer16_alpha1`	1	28	8.87
`pi05_concept`	`pi05_concept_exp_suffix_layer16_norm_warmup_alpha1`	1	28	8.90
`pi05_concept`	`pi05_concept_exp_suffix_layer16_rmsnorm_adaptive01_h16`	1	25	8.85
`pi05_concept`	`pi05_concept_exp_suffix_layer16_rmsnorm_alpha015_h16`	9	238	79.54
`pi05_concept`	`pi05_concept_exp_suffix_layer17_adaptive01_h16`	2	51	17.64
`pi05_concept`	`pi05_concept_exp_suffix_layer17_alpha1`	1	28	8.87
`pi05_concept`	`pi05_concept_exp_suffix_layer17_rmsnorm_adaptive01_h16`	1	25	8.85
`pi05_concept`	`pi05_concept_exp_suffix_layer17_rmsnorm_alpha015_h16`	10	268	88.37
`pi05_concept`	`pi05_concept_exp_suffix_layer18_alpha1`	1	28	8.86

Notes

The directory names intentionally keep both the experiment group and original run name so similarly named tasks across pi05 and pi05_concept remain distinguishable.

License: apache-2.0
Created: Jul 19, 2026
Updated: Jul 22, 2026