Downloads · 30 days
0
coeva_cycle2_acopd_80
THU-SIGS-EILAB/coeva_cycle2_acopd_80
coeva_cycle2_acopd_80 is a machine learning model from THU-SIGS-EILAB. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
This directory contains the student model after AC-OPD training step 80 in the Cycle2 CoEvo experiment.
Access
Public
Updated Sep 23, 2026
Repo size
8.6 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.pt8.6 GB · 100%
At a glance
- Access
- Public
- Created
- Sep 23, 2026
- Updated
- Sep 23, 2026
- SHA
- cd6a89af
Model card
From the Hugging Face model README
CoEvo Cycle2 AC-OPD Student — Step 80
This directory contains the student model after AC-OPD training step 80 in the Cycle2 CoEvo experiment.
full_weights.pt: consolidated student weights.physical-intelligence/libero/norm_stats.json: LIBERO normalization statistics.training_config.yaml: source Hydra training configuration.training_resolved_config.yaml: resolved TensorBoard configuration from the run.evaluation/student_step_80_nozv_eval_result.json: LIBERO-10 evaluation result.evaluation/student_step_80_nozv_resolved_config.yaml: exact no-ZV evaluation configuration.full_weights.pt.sha256: checksum for the consolidated weights.
The step-80 student was evaluated without Z/V/action priors, RECAP, or online
Cosmos, using seed 42 and 500 LIBERO-10 trajectories. Its success_once rate
was 67.2%.
Source checkpoint:
logs/coevo/cycle2/acopd_cosmos_rollout_teacher_nozv_student_4gpu/
The original FSDP distributed shards are intentionally not duplicated here;
the consolidated full_weights.pt is the deployable model artifact.
- Created
- Sep 23, 2026
- Updated
- Sep 23, 2026