iFaz/eqm-pusht-seed3-half

EQM Policy — eqm_pusht_seed3

Trained with LeRobot.
Date: 2026-08-21 06:16
Policy type: eqm | Device: cuda

---

📦 Dataset

Parameter	Value
dataset.repo_id	lerobot/pusht

---

🏋️ Training Config

Parameter	Value
steps	20000
batch_size	8
eval_freq	0
save_freq	10000
num_workers	4
seed	3
eval.n_episodes	1
eval.batch_size	1
eval.use_async_envs	True

---

📐 Policy Architecture

Parameter	Value
use_context_aware_encoder	True
context_aware_encoder_name	facebook/dinov2-with-registers-small
context_aware_embodied_pooling	cls_mean_patch
context_aware_num_register_tokens	4
context_aware_freeze_backbone	False
enable_world_model	True
jepa_encoder_name	facebook/vjepa2-vitl-fpc64-256
freeze_jepa_encoder	True
jepa_tubelet_size	2
num_video_frames	8
predictor_depth	12
predictor_num_heads	8
predictor_mlp_ratio	4.0
num_action_tokens_per_timestep	4
world_model_loss_weight	0.6
num_predictor_views	2

---

🎯 Eval Config

Parameter	Value
env.type	pusht
env.task	PushT-v0
eval.n_episodes	100
eval.batch_size	4
eval.use_async_envs	False
policy.path	/kaggle/working/outputs/train/pusht_seed3/checkpoints/last/pretrained_model

---

📊 Eval Results

Metric	Value
Episodes	100
Success rate	1.0%
Avg sum reward	33.35
Avg max reward	0.28
Eval time (s)	552.5

---

Citation

@misc{cadene2024lerobot,
  author = {Cadene, Remi and Alibert, Simon and others},
  title  = {LeRobot},
  year   = {2024},
  url    = {https://github.com/huggingface/lerobot}
}