File size: 1,542 Bytes
95f0c27 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 | # EXOKERN Skill v0.1 — Configuration
# Diffusion Policy for Peg Insertion with Force/Torque (trained on v0.1.1 dataset)
model:
architecture: "TemporalUNet1D"
parameters: 71315719
action_dim: 7
obs_horizon: 10
pred_horizon: 16
action_horizon: 8
base_channels: 256
channel_mults: [1, 2, 4]
cond_dim: 256
diffusion:
num_train_steps: 100
num_inference_steps: 16
noise_schedule: "cosine"
training:
epochs: 300
batch_size: 256
learning_rate: 0.0001
weight_decay: 0.0001
lr_schedule: "cosine_annealing"
lr_min: 0.000001
ema_decay: 0.995
grad_clip: 1.0
seeds: [42, 123, 7]
conditions:
full_ft:
obs_dim: 22
description: "Joint states (16) + Force/Torque wrench (6)"
state_components:
- joint_position: 7
- joint_velocity: 7
- joint_torque: 2
- force_xyz: 3
- torque_xyz: 3
no_ft:
obs_dim: 16
description: "Joint states only (16)"
state_components:
- joint_position: 7
- joint_velocity: 7
- joint_torque: 2
normalization:
method: "min_max"
range: [-1, 1]
# Stats are stored in checkpoint["stats"]
dataset:
repo_id: "EXOKERN/contactbench-forge-peginsert-v0.1.1"
total_episodes: 5000
total_frames: 745000
train_ratio: 0.85
val_ratio: 0.15
environment:
simulator: "NVIDIA Isaac Lab (Isaac Sim 4.5)"
env_name: "Isaac-Forge-PegInsert-Direct-v0"
robot: "Franka FR3"
control_mode: "joint_position"
physics_dt: 0.008333 # 120 Hz
control_dt: 0.066667 # 15 Hz (decimation=8)
domain_randomization: true
|