BC03 4DoF, 1 Sphere, # of demos

# 6D flow, but using SVD to extract actions, learning from >= 4DoF demos.
SVD_POINTWISE_6D_EE2FLOW_SVD = dict(
    obs_type='point_cloud',
    act_type='ee2flow',
    encoder_type='pointnet_svd_pointwise_6d_flow',
    method_flow2act='svd',
    use_consistency_loss=True,
    lambda_consistency=0.1,
    scale_pcl_flow=True,
    scale_pcl_val=250,
    separate_MLPs_R_t=False,
)


Did 1X run locally on 06/02/2022 (home machine).
06/04/2022: running 4X more on cluster

See this report for our results with 100 demos (for comparison).

NOTE: after running + investigating I made a slight error and this one has separate_MLPs_R_t = True whereas we are NOT using this for the others. However, I am convinced this is a minor thing (we can always re-run this after the deadline).
takeshi: 1X run on 05/27/2022, 1X on 05/29/2022, 1X on 05/31/2022.
cluster: 2X runs on 06/02/2022, that gives us the 5 we need.

Results