diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s0__20260602-165934-gcfb-smoke-L256-s0/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s0__20260602-165934-gcfb-smoke-L256-s0/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..fc501233e279d1a02f62b57df9e5c1aba6680a48 --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s0__20260602-165934-gcfb-smoke-L256-s0/config.yaml @@ -0,0 +1,77 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 2000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 0 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: gcfb_smoke_L256_s0 +wandb_entity: amsks +wandb_group: gcfb_smoke_L256 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 256 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +goal_cond: false +fixed_b: none +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s0__20260602-165934-gcfb-smoke-baseline-s0/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s0__20260602-165934-gcfb-smoke-baseline-s0/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b068a35a44116137a3e9febb46ca685d33cf9395 --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s0__20260602-165934-gcfb-smoke-baseline-s0/config.yaml @@ -0,0 +1,77 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 2000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 0 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: gcfb_smoke_baseline_s0 +wandb_entity: amsks +wandb_group: gcfb_smoke_baseline +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +goal_cond: false +fixed_b: none +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s0__20260602-165934-gcfb-smoke-v1-s0/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s0__20260602-165934-gcfb-smoke-v1-s0/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9dd5ca8c3e1dc27705af2e60fb5fb40f880c91f5 --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s0__20260602-165934-gcfb-smoke-v1-s0/config.yaml @@ -0,0 +1,77 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 2000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 0 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: gcfb_smoke_v1_s0 +wandb_entity: amsks +wandb_group: gcfb_smoke_v1 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +goal_cond: true +fixed_b: none +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s0__20260602-170331-gcfb-smoke-v2recheck-s0/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s0__20260602-170331-gcfb-smoke-v2recheck-s0/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a2d0fcc4d095adf02520b950180956d980e0ff12 --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s0__20260602-170331-gcfb-smoke-v2recheck-s0/config.yaml @@ -0,0 +1,77 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 2000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 0 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: gcfb_smoke_v2recheck_s0 +wandb_entity: amsks +wandb_group: gcfb_smoke_v2recheck +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +goal_cond: false +fixed_b: cube_xyz +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s0__20260602-203933-fb-state-v1-s0/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s0__20260602-203933-fb-state-v1-s0/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..32b33c45b1f32f0e8bda2f23e520245037f90e27 --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s0__20260602-203933-fb-state-v1-s0/config.yaml @@ -0,0 +1,77 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 0 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_v1_s0 +wandb_entity: amsks +wandb_group: fb_state_v1 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +goal_cond: true +fixed_b: none +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s0__20260603-020044-fb-state-v2-s0/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s0__20260603-020044-fb-state-v2-s0/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9bd7a98aba084e2ce0dc6526e8d32abd148d93e9 --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s0__20260603-020044-fb-state-v2-s0/config.yaml @@ -0,0 +1,77 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 0 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_v2_s0 +wandb_entity: amsks +wandb_group: fb_state_v2 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +goal_cond: false +fixed_b: cube_xyz +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s0__20260603-064553-fb-state-L256-s0/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s0__20260603-064553-fb-state-L256-s0/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..fafb9311ea3f273867efbfbc5f47d95d51c75074 --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s0__20260603-064553-fb-state-L256-s0/config.yaml @@ -0,0 +1,75 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 0 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_L256_s0 +wandb_entity: amsks +wandb_group: fb_state_L256 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s1__20260602-203933-fb-state-v1-s1/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s1__20260602-203933-fb-state-v1-s1/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a36c285fe4f298ef170e2c003bbb39220fc11a9e --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s1__20260602-203933-fb-state-v1-s1/config.yaml @@ -0,0 +1,77 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 1 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_v1_s1 +wandb_entity: amsks +wandb_group: fb_state_v1 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +goal_cond: true +fixed_b: none +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s1__20260603-020044-fb-state-v2-s1/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s1__20260603-020044-fb-state-v2-s1/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..df3e1bba08a43bc9bac327c4d3753ce14882cd4e --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s1__20260603-020044-fb-state-v2-s1/config.yaml @@ -0,0 +1,77 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 1 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_v2_s1 +wandb_entity: amsks +wandb_group: fb_state_v2 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +goal_cond: false +fixed_b: cube_xyz +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s1__20260603-064553-fb-state-L256-s1/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s1__20260603-064553-fb-state-L256-s1/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..1cb8f779565e149ef033209b1df657e0819130a6 --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s1__20260603-064553-fb-state-L256-s1/config.yaml @@ -0,0 +1,75 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 1 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_L256_s1 +wandb_entity: amsks +wandb_group: fb_state_L256 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s2__20260602-203933-fb-state-v1-s2/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s2__20260602-203933-fb-state-v1-s2/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7abdaf1fc8636b7d408e8317c58d7f2ba75dcfb9 --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s2__20260602-203933-fb-state-v1-s2/config.yaml @@ -0,0 +1,77 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 2 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_v1_s2 +wandb_entity: amsks +wandb_group: fb_state_v1 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +goal_cond: true +fixed_b: none +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s2__20260603-020044-fb-state-v2-s2/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s2__20260603-020044-fb-state-v2-s2/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4e525a2385bab77015c1b55a789a69c428dc798b --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s2__20260603-020044-fb-state-v2-s2/config.yaml @@ -0,0 +1,77 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 2 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_v2_s2 +wandb_entity: amsks +wandb_group: fb_state_v2 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +goal_cond: false +fixed_b: cube_xyz +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s2__20260603-064553-fb-state-L256-s2/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s2__20260603-064553-fb-state-L256-s2/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7293544ebe063dc80491cbfb935b229e154367ad --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s2__20260603-064553-fb-state-L256-s2/config.yaml @@ -0,0 +1,75 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 2 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_L256_s2 +wandb_entity: amsks +wandb_group: fb_state_L256 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s3__20260602-203933-fb-state-v1-s3/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s3__20260602-203933-fb-state-v1-s3/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..bfd9471ca1dcf3b3c95e117e1a871cb8f090d114 --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s3__20260602-203933-fb-state-v1-s3/config.yaml @@ -0,0 +1,77 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 3 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_v1_s3 +wandb_entity: amsks +wandb_group: fb_state_v1 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +goal_cond: true +fixed_b: none +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s3__20260603-020044-fb-state-v2-s3/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s3__20260603-020044-fb-state-v2-s3/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..bae5108b64c1f9f3386b898fdce4609d655fb794 --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s3__20260603-020044-fb-state-v2-s3/config.yaml @@ -0,0 +1,77 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 3 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_v2_s3 +wandb_entity: amsks +wandb_group: fb_state_v2 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +goal_cond: false +fixed_b: cube_xyz +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s3__20260603-064553-fb-state-L256-s3/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s3__20260603-064553-fb-state-L256-s3/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..777ac48de794b27405fd81c6ca4801015c56af18 --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s3__20260603-064553-fb-state-L256-s3/config.yaml @@ -0,0 +1,75 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 3 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_L256_s3 +wandb_entity: amsks +wandb_group: fb_state_L256 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s4__20260602-203933-fb-state-v1-s4/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s4__20260602-203933-fb-state-v1-s4/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..bf1866f5fb083676b43224bb8dbfec4d6c336dcf --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s4__20260602-203933-fb-state-v1-s4/config.yaml @@ -0,0 +1,77 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 4 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_v1_s4 +wandb_entity: amsks +wandb_group: fb_state_v1 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +goal_cond: true +fixed_b: none +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s4__20260603-020044-fb-state-v2-s4/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s4__20260603-020044-fb-state-v2-s4/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0f3aa19d2b3242e0a3fcbc24117ff2d9a0c0a7e4 --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s4__20260603-020044-fb-state-v2-s4/config.yaml @@ -0,0 +1,77 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 4 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_v2_s4 +wandb_entity: amsks +wandb_group: fb_state_v2 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +goal_cond: false +fixed_b: cube_xyz +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s4__20260603-064553-fb-state-L256-s4/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s4__20260603-064553-fb-state-L256-s4/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4098cc8253feabf78597c2ad72b2649635aef300 --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s4__20260603-064553-fb-state-L256-s4/config.yaml @@ -0,0 +1,75 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 4 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_L256_s4 +wandb_entity: amsks +wandb_group: fb_state_L256 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s5__20260602-203933-fb-state-v1-s5/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s5__20260602-203933-fb-state-v1-s5/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..257723af117b7fc4de79354344f83f1e337fe2c3 --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s5__20260602-203933-fb-state-v1-s5/config.yaml @@ -0,0 +1,77 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 5 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_v1_s5 +wandb_entity: amsks +wandb_group: fb_state_v1 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +goal_cond: true +fixed_b: none +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s5__20260603-020044-fb-state-v2-s5/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s5__20260603-020044-fb-state-v2-s5/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7d573bac4a5ad3e98df394659def896b5d27ffce --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s5__20260603-020044-fb-state-v2-s5/config.yaml @@ -0,0 +1,77 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 5 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_v2_s5 +wandb_entity: amsks +wandb_group: fb_state_v2 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +goal_cond: false +fixed_b: cube_xyz +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s5__20260603-064553-fb-state-L256-s5/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s5__20260603-064553-fb-state-L256-s5/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f4d82955472a780e58ab8b475b1d433ad61de751 --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s5__20260603-064553-fb-state-L256-s5/config.yaml @@ -0,0 +1,75 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 5 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_L256_s5 +wandb_entity: amsks +wandb_group: fb_state_L256 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s6__20260602-203933-fb-state-v1-s6/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s6__20260602-203933-fb-state-v1-s6/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..372ac06efc9e4875dd152642cf18d6563c511b2a --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s6__20260602-203933-fb-state-v1-s6/config.yaml @@ -0,0 +1,77 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 6 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_v1_s6 +wandb_entity: amsks +wandb_group: fb_state_v1 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +goal_cond: true +fixed_b: none +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s6__20260603-020044-fb-state-v2-s6/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s6__20260603-020044-fb-state-v2-s6/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..cda1ba1bb9c8f067323aa80c61242fe24c70d6b0 --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s6__20260603-020044-fb-state-v2-s6/config.yaml @@ -0,0 +1,77 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 6 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_v2_s6 +wandb_entity: amsks +wandb_group: fb_state_v2 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +goal_cond: false +fixed_b: cube_xyz +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s6__20260603-064553-fb-state-L256-s6/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s6__20260603-064553-fb-state-L256-s6/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ab369b67834637880eab7176d0076cae77354829 --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s6__20260603-064553-fb-state-L256-s6/config.yaml @@ -0,0 +1,75 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 6 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_L256_s6 +wandb_entity: amsks +wandb_group: fb_state_L256 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s7__20260602-203933-fb-state-v1-s7/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s7__20260602-203933-fb-state-v1-s7/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f05fc9702e3cd20a9e854878b8d234fa4c7d852f --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s7__20260602-203933-fb-state-v1-s7/config.yaml @@ -0,0 +1,77 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 7 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_v1_s7 +wandb_entity: amsks +wandb_group: fb_state_v1 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +goal_cond: true +fixed_b: none +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s7__20260603-020044-fb-state-v2-s7/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s7__20260603-020044-fb-state-v2-s7/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..94754f62281801cec2c6b4749616f8904427173a --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s7__20260603-020044-fb-state-v2-s7/config.yaml @@ -0,0 +1,77 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 7 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_v2_s7 +wandb_entity: amsks +wandb_group: fb_state_v2 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +goal_cond: false +fixed_b: cube_xyz +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s7__20260603-064553-fb-state-L256-s7/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s7__20260603-064553-fb-state-L256-s7/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..98d1b207d768419147798cacaa93413c145fefa3 --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s7__20260603-064553-fb-state-L256-s7/config.yaml @@ -0,0 +1,75 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 7 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_L256_s7 +wandb_entity: amsks +wandb_group: fb_state_L256 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s8__20260602-203933-fb-state-v1-s8/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s8__20260602-203933-fb-state-v1-s8/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..76076ac01ab5f91a927fe810ff5bc89fb1f99187 --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s8__20260602-203933-fb-state-v1-s8/config.yaml @@ -0,0 +1,77 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 8 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_v1_s8 +wandb_entity: amsks +wandb_group: fb_state_v1 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +goal_cond: true +fixed_b: none +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s8__20260603-020044-fb-state-v2-s8/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s8__20260603-020044-fb-state-v2-s8/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..10806755da345434e6e1d52f9db08cfa9e320ad3 --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s8__20260603-020044-fb-state-v2-s8/config.yaml @@ -0,0 +1,77 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 8 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_v2_s8 +wandb_entity: amsks +wandb_group: fb_state_v2 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +goal_cond: false +fixed_b: cube_xyz +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s8__20260603-064553-fb-state-L256-s8/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s8__20260603-064553-fb-state-L256-s8/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b457a29ce36852ab673cd795d623335b3406d1ca --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s8__20260603-064553-fb-state-L256-s8/config.yaml @@ -0,0 +1,75 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 8 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_L256_s8 +wandb_entity: amsks +wandb_group: fb_state_L256 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s9__20260602-203933-fb-state-v1-s9/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s9__20260602-203933-fb-state-v1-s9/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..217e75db0105efd05e0a8d3defe369e48d5a765a --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s9__20260602-203933-fb-state-v1-s9/config.yaml @@ -0,0 +1,77 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 9 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_v1_s9 +wandb_entity: amsks +wandb_group: fb_state_v1 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +goal_cond: true +fixed_b: none +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s9__20260603-020044-fb-state-v2-s9/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s9__20260603-020044-fb-state-v2-s9/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e8bb3f42884c4b3afb30cc0145b242f166ab026f --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s9__20260603-020044-fb-state-v2-s9/config.yaml @@ -0,0 +1,77 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 9 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_v2_s9 +wandb_entity: amsks +wandb_group: fb_state_v2 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +goal_cond: false +fixed_b: cube_xyz +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s9__20260603-064553-fb-state-L256-s9/config.yaml b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s9__20260603-064553-fb-state-L256-s9/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..22d63f897517e348d688224ed6106ffae47e0abe --- /dev/null +++ b/checkpoints/torch/fb-flowbc/cube-single-play-v0-state__s9__20260603-064553-fb-state-L256-s9/config.yaml @@ -0,0 +1,75 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 9 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: fb_state_L256_s9 +wandb_entity: amsks +wandb_group: fb_state_L256 +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: fb_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 1000 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 diff --git a/checkpoints/torch/psm-flowbc/cube-single-play-v0-pixels__s0__20260603-070822-psm-smoke-pixel-s0/config.yaml b/checkpoints/torch/psm-flowbc/cube-single-play-v0-pixels__s0__20260603-070822-psm-smoke-pixel-s0/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7b0f60d65248916e53f2835cef3bf1a4e8828ebc --- /dev/null +++ b/checkpoints/torch/psm-flowbc/cube-single-play-v0-pixels__s0__20260603-070822-psm-smoke-pixel-s0/config.yaml @@ -0,0 +1,75 @@ +domain: cube-single-play-v0 +obs_type: pixels +data_path: /dev/shm/factored-fb/datasets_visual +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 2000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 0 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: psm_smoke_pixel_s0 +wandb_entity: amsks +wandb_group: psm_smoke_pixel +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: psm_flowbc +z_dim: 50 +norm_z: true +max_log_seed: 16 +phi_input: s +norm_obs: false +discount: 0.99 +phi: + hidden_dim: 256 + hidden_layers: 2 + norm: true + batch_norm: false +sf: + hidden_dim: 1024 + hidden_layers: 1 + embedding_layers: 2 + num_parallel: 2 +actor: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + name: noise_conditioned_actor +lr_sf: 0.0001 +lr_phi: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +target_tau: 0.01 +ortho_coef: 1.0 +mix_ratio: 0.5 +pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.5 +stddev: 0.2 +stddev_clip: 0.3 +amp: false +bc_coeff: 3.0 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 +frame_stack: 3 +obs_normalizer: + name: RGBNormalizerConfig +rgb_encoder: + name: drq + feature_dim: 256 +augmentator: + name: random_shifts + pad: 2 diff --git a/checkpoints/torch/psm-flowbc/cube-single-play-v0-pixels__s0__20260604-052341-psm-pixel-s0/config.yaml b/checkpoints/torch/psm-flowbc/cube-single-play-v0-pixels__s0__20260604-052341-psm-pixel-s0/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e588f5a1979b12a41452e17090d59e8858f5fc80 --- /dev/null +++ b/checkpoints/torch/psm-flowbc/cube-single-play-v0-pixels__s0__20260604-052341-psm-pixel-s0/config.yaml @@ -0,0 +1,75 @@ +domain: cube-single-play-v0 +obs_type: pixels +data_path: /dev/shm/factored-fb/datasets_pixel +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 500000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 0 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: psm_pixel_s0 +wandb_entity: amsks +wandb_group: psm_pixel +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: psm_flowbc +z_dim: 50 +norm_z: true +max_log_seed: 16 +phi_input: s +norm_obs: false +discount: 0.99 +phi: + hidden_dim: 256 + hidden_layers: 2 + norm: true + batch_norm: false +sf: + hidden_dim: 1024 + hidden_layers: 1 + embedding_layers: 2 + num_parallel: 2 +actor: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + name: noise_conditioned_actor +lr_sf: 0.0001 +lr_phi: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +target_tau: 0.01 +ortho_coef: 1.0 +mix_ratio: 0.5 +pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.5 +stddev: 0.2 +stddev_clip: 0.3 +amp: false +bc_coeff: 3.0 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 +frame_stack: 3 +obs_normalizer: + name: RGBNormalizerConfig +rgb_encoder: + name: drq + feature_dim: 256 +augmentator: + name: random_shifts + pad: 2 diff --git a/checkpoints/torch/psm-flowbc/cube-single-play-v0-pixels__s1__20260604-052341-psm-pixel-s1/config.yaml b/checkpoints/torch/psm-flowbc/cube-single-play-v0-pixels__s1__20260604-052341-psm-pixel-s1/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..6d5494fffc4a830844df01c442e77258ad8c00c1 --- /dev/null +++ b/checkpoints/torch/psm-flowbc/cube-single-play-v0-pixels__s1__20260604-052341-psm-pixel-s1/config.yaml @@ -0,0 +1,75 @@ +domain: cube-single-play-v0 +obs_type: pixels +data_path: /dev/shm/factored-fb/datasets_pixel +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 500000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 1 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: psm_pixel_s1 +wandb_entity: amsks +wandb_group: psm_pixel +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: psm_flowbc +z_dim: 50 +norm_z: true +max_log_seed: 16 +phi_input: s +norm_obs: false +discount: 0.99 +phi: + hidden_dim: 256 + hidden_layers: 2 + norm: true + batch_norm: false +sf: + hidden_dim: 1024 + hidden_layers: 1 + embedding_layers: 2 + num_parallel: 2 +actor: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + name: noise_conditioned_actor +lr_sf: 0.0001 +lr_phi: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +target_tau: 0.01 +ortho_coef: 1.0 +mix_ratio: 0.5 +pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.5 +stddev: 0.2 +stddev_clip: 0.3 +amp: false +bc_coeff: 3.0 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 +frame_stack: 3 +obs_normalizer: + name: RGBNormalizerConfig +rgb_encoder: + name: drq + feature_dim: 256 +augmentator: + name: random_shifts + pad: 2 diff --git a/checkpoints/torch/psm-flowbc/cube-single-play-v0-pixels__s2__20260604-052341-psm-pixel-s2/config.yaml b/checkpoints/torch/psm-flowbc/cube-single-play-v0-pixels__s2__20260604-052341-psm-pixel-s2/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e6946b9e3e950d351fe2411d8dac312ac973da69 --- /dev/null +++ b/checkpoints/torch/psm-flowbc/cube-single-play-v0-pixels__s2__20260604-052341-psm-pixel-s2/config.yaml @@ -0,0 +1,75 @@ +domain: cube-single-play-v0 +obs_type: pixels +data_path: /dev/shm/factored-fb/datasets_pixel +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 500000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 2 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: psm_pixel_s2 +wandb_entity: amsks +wandb_group: psm_pixel +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: psm_flowbc +z_dim: 50 +norm_z: true +max_log_seed: 16 +phi_input: s +norm_obs: false +discount: 0.99 +phi: + hidden_dim: 256 + hidden_layers: 2 + norm: true + batch_norm: false +sf: + hidden_dim: 1024 + hidden_layers: 1 + embedding_layers: 2 + num_parallel: 2 +actor: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + name: noise_conditioned_actor +lr_sf: 0.0001 +lr_phi: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +target_tau: 0.01 +ortho_coef: 1.0 +mix_ratio: 0.5 +pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.5 +stddev: 0.2 +stddev_clip: 0.3 +amp: false +bc_coeff: 3.0 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 +frame_stack: 3 +obs_normalizer: + name: RGBNormalizerConfig +rgb_encoder: + name: drq + feature_dim: 256 +augmentator: + name: random_shifts + pad: 2 diff --git a/checkpoints/torch/psm-flowbc/cube-single-play-v0-pixels__s3__20260604-052341-psm-pixel-s3/config.yaml b/checkpoints/torch/psm-flowbc/cube-single-play-v0-pixels__s3__20260604-052341-psm-pixel-s3/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..cc8623e6ebb6dba5096634692a0f7d993aa2bdf9 --- /dev/null +++ b/checkpoints/torch/psm-flowbc/cube-single-play-v0-pixels__s3__20260604-052341-psm-pixel-s3/config.yaml @@ -0,0 +1,75 @@ +domain: cube-single-play-v0 +obs_type: pixels +data_path: /dev/shm/factored-fb/datasets_pixel +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 500000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 3 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: psm_pixel_s3 +wandb_entity: amsks +wandb_group: psm_pixel +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: psm_flowbc +z_dim: 50 +norm_z: true +max_log_seed: 16 +phi_input: s +norm_obs: false +discount: 0.99 +phi: + hidden_dim: 256 + hidden_layers: 2 + norm: true + batch_norm: false +sf: + hidden_dim: 1024 + hidden_layers: 1 + embedding_layers: 2 + num_parallel: 2 +actor: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + name: noise_conditioned_actor +lr_sf: 0.0001 +lr_phi: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +target_tau: 0.01 +ortho_coef: 1.0 +mix_ratio: 0.5 +pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.5 +stddev: 0.2 +stddev_clip: 0.3 +amp: false +bc_coeff: 3.0 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 +frame_stack: 3 +obs_normalizer: + name: RGBNormalizerConfig +rgb_encoder: + name: drq + feature_dim: 256 +augmentator: + name: random_shifts + pad: 2 diff --git a/checkpoints/torch/psm-flowbc/cube-single-play-v0-pixels__s4__20260604-052341-psm-pixel-s4/config.yaml b/checkpoints/torch/psm-flowbc/cube-single-play-v0-pixels__s4__20260604-052341-psm-pixel-s4/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..224b568e0e75193b882dc277f0b57894df0cee56 --- /dev/null +++ b/checkpoints/torch/psm-flowbc/cube-single-play-v0-pixels__s4__20260604-052341-psm-pixel-s4/config.yaml @@ -0,0 +1,75 @@ +domain: cube-single-play-v0 +obs_type: pixels +data_path: /dev/shm/factored-fb/datasets_pixel +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 500000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 4 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: psm_pixel_s4 +wandb_entity: amsks +wandb_group: psm_pixel +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: psm_flowbc +z_dim: 50 +norm_z: true +max_log_seed: 16 +phi_input: s +norm_obs: false +discount: 0.99 +phi: + hidden_dim: 256 + hidden_layers: 2 + norm: true + batch_norm: false +sf: + hidden_dim: 1024 + hidden_layers: 1 + embedding_layers: 2 + num_parallel: 2 +actor: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + name: noise_conditioned_actor +lr_sf: 0.0001 +lr_phi: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +target_tau: 0.01 +ortho_coef: 1.0 +mix_ratio: 0.5 +pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.5 +stddev: 0.2 +stddev_clip: 0.3 +amp: false +bc_coeff: 3.0 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 +frame_stack: 3 +obs_normalizer: + name: RGBNormalizerConfig +rgb_encoder: + name: drq + feature_dim: 256 +augmentator: + name: random_shifts + pad: 2 diff --git a/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s0__20260603-054406-psm-smoke-state-s0/config.yaml b/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s0__20260603-054406-psm-smoke-state-s0/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..fbaf8099d5fca966ace45b2d4d348066c4763868 --- /dev/null +++ b/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s0__20260603-054406-psm-smoke-state-s0/config.yaml @@ -0,0 +1,67 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 2000 +batch_size: 1024 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 0 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: psm_smoke_state_s0 +wandb_entity: amsks +wandb_group: psm_smoke_state +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: psm_flowbc +z_dim: 128 +norm_z: true +max_log_seed: 16 +phi_input: s +norm_obs: false +discount: 0.98 +phi: + hidden_dim: 256 + hidden_layers: 2 + norm: true + batch_norm: false +sf: + hidden_dim: 1024 + hidden_layers: 1 + embedding_layers: 2 + num_parallel: 2 +actor: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + name: noise_conditioned_actor +lr_sf: 0.0001 +lr_phi: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +target_tau: 0.01 +ortho_coef: 1.0 +mix_ratio: 0.5 +pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.5 +stddev: 0.2 +stddev_clip: 0.3 +amp: false +bc_coeff: 3.0 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 +lr_b: 0.0001 diff --git a/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s0__20260603-071856-psm-flowbc-state-s0/config.yaml b/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s0__20260603-071856-psm-flowbc-state-s0/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e435e04c89429597e7117f7949514b6721970af5 --- /dev/null +++ b/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s0__20260603-071856-psm-flowbc-state-s0/config.yaml @@ -0,0 +1,67 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 1024 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 0 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: psm_flowbc_state_s0 +wandb_entity: amsks +wandb_group: psm_flowbc_state +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: psm_flowbc +z_dim: 128 +norm_z: true +max_log_seed: 16 +phi_input: s +norm_obs: false +discount: 0.98 +phi: + hidden_dim: 256 + hidden_layers: 2 + norm: true + batch_norm: false +sf: + hidden_dim: 1024 + hidden_layers: 1 + embedding_layers: 2 + num_parallel: 2 +actor: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + name: noise_conditioned_actor +lr_sf: 0.0001 +lr_phi: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +target_tau: 0.01 +ortho_coef: 1.0 +mix_ratio: 0.5 +pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.5 +stddev: 0.2 +stddev_clip: 0.3 +amp: false +bc_coeff: 3.0 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 +lr_b: 0.0001 diff --git a/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s1__20260603-071856-psm-flowbc-state-s1/config.yaml b/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s1__20260603-071856-psm-flowbc-state-s1/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..13d78de4e63ba1d96d79a034b31edf35f0b43825 --- /dev/null +++ b/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s1__20260603-071856-psm-flowbc-state-s1/config.yaml @@ -0,0 +1,67 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 1024 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 1 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: psm_flowbc_state_s1 +wandb_entity: amsks +wandb_group: psm_flowbc_state +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: psm_flowbc +z_dim: 128 +norm_z: true +max_log_seed: 16 +phi_input: s +norm_obs: false +discount: 0.98 +phi: + hidden_dim: 256 + hidden_layers: 2 + norm: true + batch_norm: false +sf: + hidden_dim: 1024 + hidden_layers: 1 + embedding_layers: 2 + num_parallel: 2 +actor: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + name: noise_conditioned_actor +lr_sf: 0.0001 +lr_phi: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +target_tau: 0.01 +ortho_coef: 1.0 +mix_ratio: 0.5 +pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.5 +stddev: 0.2 +stddev_clip: 0.3 +amp: false +bc_coeff: 3.0 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 +lr_b: 0.0001 diff --git a/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s2__20260603-071856-psm-flowbc-state-s2/config.yaml b/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s2__20260603-071856-psm-flowbc-state-s2/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..99df48d7b62e142087c71b991c411ea7d0b90d06 --- /dev/null +++ b/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s2__20260603-071856-psm-flowbc-state-s2/config.yaml @@ -0,0 +1,67 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 1024 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 2 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: psm_flowbc_state_s2 +wandb_entity: amsks +wandb_group: psm_flowbc_state +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: psm_flowbc +z_dim: 128 +norm_z: true +max_log_seed: 16 +phi_input: s +norm_obs: false +discount: 0.98 +phi: + hidden_dim: 256 + hidden_layers: 2 + norm: true + batch_norm: false +sf: + hidden_dim: 1024 + hidden_layers: 1 + embedding_layers: 2 + num_parallel: 2 +actor: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + name: noise_conditioned_actor +lr_sf: 0.0001 +lr_phi: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +target_tau: 0.01 +ortho_coef: 1.0 +mix_ratio: 0.5 +pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.5 +stddev: 0.2 +stddev_clip: 0.3 +amp: false +bc_coeff: 3.0 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 +lr_b: 0.0001 diff --git a/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s3__20260603-071856-psm-flowbc-state-s3/config.yaml b/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s3__20260603-071856-psm-flowbc-state-s3/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..29e4c961a3daee9ca472ed0a38058b5faa13a76d --- /dev/null +++ b/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s3__20260603-071856-psm-flowbc-state-s3/config.yaml @@ -0,0 +1,67 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 1024 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 3 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: psm_flowbc_state_s3 +wandb_entity: amsks +wandb_group: psm_flowbc_state +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: psm_flowbc +z_dim: 128 +norm_z: true +max_log_seed: 16 +phi_input: s +norm_obs: false +discount: 0.98 +phi: + hidden_dim: 256 + hidden_layers: 2 + norm: true + batch_norm: false +sf: + hidden_dim: 1024 + hidden_layers: 1 + embedding_layers: 2 + num_parallel: 2 +actor: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + name: noise_conditioned_actor +lr_sf: 0.0001 +lr_phi: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +target_tau: 0.01 +ortho_coef: 1.0 +mix_ratio: 0.5 +pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.5 +stddev: 0.2 +stddev_clip: 0.3 +amp: false +bc_coeff: 3.0 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 +lr_b: 0.0001 diff --git a/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s4__20260603-071856-psm-flowbc-state-s4/config.yaml b/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s4__20260603-071856-psm-flowbc-state-s4/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..380c461a405504a3c4d20a3bce572612eb3594f3 --- /dev/null +++ b/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s4__20260603-071856-psm-flowbc-state-s4/config.yaml @@ -0,0 +1,67 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 1024 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 4 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: psm_flowbc_state_s4 +wandb_entity: amsks +wandb_group: psm_flowbc_state +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: psm_flowbc +z_dim: 128 +norm_z: true +max_log_seed: 16 +phi_input: s +norm_obs: false +discount: 0.98 +phi: + hidden_dim: 256 + hidden_layers: 2 + norm: true + batch_norm: false +sf: + hidden_dim: 1024 + hidden_layers: 1 + embedding_layers: 2 + num_parallel: 2 +actor: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + name: noise_conditioned_actor +lr_sf: 0.0001 +lr_phi: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +target_tau: 0.01 +ortho_coef: 1.0 +mix_ratio: 0.5 +pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.5 +stddev: 0.2 +stddev_clip: 0.3 +amp: false +bc_coeff: 3.0 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 +lr_b: 0.0001 diff --git a/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s5__20260603-071856-psm-flowbc-state-s5/config.yaml b/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s5__20260603-071856-psm-flowbc-state-s5/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a5caeddab6f56a3d984ac6415d53be4b9acb35a2 --- /dev/null +++ b/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s5__20260603-071856-psm-flowbc-state-s5/config.yaml @@ -0,0 +1,67 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 1024 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 5 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: psm_flowbc_state_s5 +wandb_entity: amsks +wandb_group: psm_flowbc_state +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: psm_flowbc +z_dim: 128 +norm_z: true +max_log_seed: 16 +phi_input: s +norm_obs: false +discount: 0.98 +phi: + hidden_dim: 256 + hidden_layers: 2 + norm: true + batch_norm: false +sf: + hidden_dim: 1024 + hidden_layers: 1 + embedding_layers: 2 + num_parallel: 2 +actor: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + name: noise_conditioned_actor +lr_sf: 0.0001 +lr_phi: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +target_tau: 0.01 +ortho_coef: 1.0 +mix_ratio: 0.5 +pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.5 +stddev: 0.2 +stddev_clip: 0.3 +amp: false +bc_coeff: 3.0 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 +lr_b: 0.0001 diff --git a/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s6__20260603-071856-psm-flowbc-state-s6/config.yaml b/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s6__20260603-071856-psm-flowbc-state-s6/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b6f6d84743c1de6cb8d46134b5e5789dd768bb18 --- /dev/null +++ b/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s6__20260603-071856-psm-flowbc-state-s6/config.yaml @@ -0,0 +1,67 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 1024 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 6 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: psm_flowbc_state_s6 +wandb_entity: amsks +wandb_group: psm_flowbc_state +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: psm_flowbc +z_dim: 128 +norm_z: true +max_log_seed: 16 +phi_input: s +norm_obs: false +discount: 0.98 +phi: + hidden_dim: 256 + hidden_layers: 2 + norm: true + batch_norm: false +sf: + hidden_dim: 1024 + hidden_layers: 1 + embedding_layers: 2 + num_parallel: 2 +actor: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + name: noise_conditioned_actor +lr_sf: 0.0001 +lr_phi: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +target_tau: 0.01 +ortho_coef: 1.0 +mix_ratio: 0.5 +pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.5 +stddev: 0.2 +stddev_clip: 0.3 +amp: false +bc_coeff: 3.0 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 +lr_b: 0.0001 diff --git a/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s7__20260603-071856-psm-flowbc-state-s7/config.yaml b/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s7__20260603-071856-psm-flowbc-state-s7/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7cc296257a7704ba46564f46dcccc1b9652d7f51 --- /dev/null +++ b/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s7__20260603-071856-psm-flowbc-state-s7/config.yaml @@ -0,0 +1,67 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 1024 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 7 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: psm_flowbc_state_s7 +wandb_entity: amsks +wandb_group: psm_flowbc_state +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: psm_flowbc +z_dim: 128 +norm_z: true +max_log_seed: 16 +phi_input: s +norm_obs: false +discount: 0.98 +phi: + hidden_dim: 256 + hidden_layers: 2 + norm: true + batch_norm: false +sf: + hidden_dim: 1024 + hidden_layers: 1 + embedding_layers: 2 + num_parallel: 2 +actor: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + name: noise_conditioned_actor +lr_sf: 0.0001 +lr_phi: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +target_tau: 0.01 +ortho_coef: 1.0 +mix_ratio: 0.5 +pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.5 +stddev: 0.2 +stddev_clip: 0.3 +amp: false +bc_coeff: 3.0 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 +lr_b: 0.0001 diff --git a/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s8__20260603-071856-psm-flowbc-state-s8/config.yaml b/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s8__20260603-071856-psm-flowbc-state-s8/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..681ca773d510d1c82cc5ffaae763b62016d92bf5 --- /dev/null +++ b/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s8__20260603-071856-psm-flowbc-state-s8/config.yaml @@ -0,0 +1,67 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 1024 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 8 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: psm_flowbc_state_s8 +wandb_entity: amsks +wandb_group: psm_flowbc_state +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: psm_flowbc +z_dim: 128 +norm_z: true +max_log_seed: 16 +phi_input: s +norm_obs: false +discount: 0.98 +phi: + hidden_dim: 256 + hidden_layers: 2 + norm: true + batch_norm: false +sf: + hidden_dim: 1024 + hidden_layers: 1 + embedding_layers: 2 + num_parallel: 2 +actor: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + name: noise_conditioned_actor +lr_sf: 0.0001 +lr_phi: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +target_tau: 0.01 +ortho_coef: 1.0 +mix_ratio: 0.5 +pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.5 +stddev: 0.2 +stddev_clip: 0.3 +amp: false +bc_coeff: 3.0 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 +lr_b: 0.0001 diff --git a/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s9__20260603-071856-psm-flowbc-state-s9/config.yaml b/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s9__20260603-071856-psm-flowbc-state-s9/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3f2f1c7e64cb08bb4bbe0133d1877e5f959608bb --- /dev/null +++ b/checkpoints/torch/psm-flowbc/cube-single-play-v0-state__s9__20260603-071856-psm-flowbc-state-s9/config.yaml @@ -0,0 +1,67 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 1024 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 9 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: psm_flowbc_state_s9 +wandb_entity: amsks +wandb_group: psm_flowbc_state +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: psm_flowbc +z_dim: 128 +norm_z: true +max_log_seed: 16 +phi_input: s +norm_obs: false +discount: 0.98 +phi: + hidden_dim: 256 + hidden_layers: 2 + norm: true + batch_norm: false +sf: + hidden_dim: 1024 + hidden_layers: 1 + embedding_layers: 2 + num_parallel: 2 +actor: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + name: noise_conditioned_actor +lr_sf: 0.0001 +lr_phi: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +target_tau: 0.01 +ortho_coef: 1.0 +mix_ratio: 0.5 +pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.5 +stddev: 0.2 +stddev_clip: 0.3 +amp: false +bc_coeff: 3.0 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 +lr_b: 0.0001 diff --git a/checkpoints/torch/rldp-flowbc/cube-single-play-v0-pixels__s0__20260604-052341-rldp-pixel-s0/config.yaml b/checkpoints/torch/rldp-flowbc/cube-single-play-v0-pixels__s0__20260604-052341-rldp-pixel-s0/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..33d5fc7488302b81b30ee9f5038b988152a03217 --- /dev/null +++ b/checkpoints/torch/rldp-flowbc/cube-single-play-v0-pixels__s0__20260604-052341-rldp-pixel-s0/config.yaml @@ -0,0 +1,90 @@ +domain: cube-single-play-v0 +obs_type: pixels +data_path: /dev/shm/factored-fb/datasets_pixel +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 500000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 0 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: rldp_pixel_s0 +wandb_entity: amsks +wandb_group: rldp_pixel +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: rldp_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 0.01 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 +horizon: 3 +predictor: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +frame_stack: 3 +obs_normalizer: + name: RGBNormalizerConfig +rgb_encoder: + name: drq + feature_dim: 256 +augmentator: + name: random_shifts + pad: 2 diff --git a/checkpoints/torch/rldp-flowbc/cube-single-play-v0-pixels__s1__20260604-052341-rldp-pixel-s1/config.yaml b/checkpoints/torch/rldp-flowbc/cube-single-play-v0-pixels__s1__20260604-052341-rldp-pixel-s1/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5832bf4bc31ae19f58e71f85c4da40a953021455 --- /dev/null +++ b/checkpoints/torch/rldp-flowbc/cube-single-play-v0-pixels__s1__20260604-052341-rldp-pixel-s1/config.yaml @@ -0,0 +1,90 @@ +domain: cube-single-play-v0 +obs_type: pixels +data_path: /dev/shm/factored-fb/datasets_pixel +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 500000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 1 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: rldp_pixel_s1 +wandb_entity: amsks +wandb_group: rldp_pixel +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: rldp_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 0.01 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 +horizon: 3 +predictor: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +frame_stack: 3 +obs_normalizer: + name: RGBNormalizerConfig +rgb_encoder: + name: drq + feature_dim: 256 +augmentator: + name: random_shifts + pad: 2 diff --git a/checkpoints/torch/rldp-flowbc/cube-single-play-v0-pixels__s2__20260604-052341-rldp-pixel-s2/config.yaml b/checkpoints/torch/rldp-flowbc/cube-single-play-v0-pixels__s2__20260604-052341-rldp-pixel-s2/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..58762c29e22a30be70f121143af456cd579a61b6 --- /dev/null +++ b/checkpoints/torch/rldp-flowbc/cube-single-play-v0-pixels__s2__20260604-052341-rldp-pixel-s2/config.yaml @@ -0,0 +1,90 @@ +domain: cube-single-play-v0 +obs_type: pixels +data_path: /dev/shm/factored-fb/datasets_pixel +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 500000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 2 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: rldp_pixel_s2 +wandb_entity: amsks +wandb_group: rldp_pixel +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: rldp_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 0.01 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 +horizon: 3 +predictor: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +frame_stack: 3 +obs_normalizer: + name: RGBNormalizerConfig +rgb_encoder: + name: drq + feature_dim: 256 +augmentator: + name: random_shifts + pad: 2 diff --git a/checkpoints/torch/rldp-flowbc/cube-single-play-v0-pixels__s3__20260604-052341-rldp-pixel-s3/config.yaml b/checkpoints/torch/rldp-flowbc/cube-single-play-v0-pixels__s3__20260604-052341-rldp-pixel-s3/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8f940805036f6549d3ef261b4adfa8dafdcaadc9 --- /dev/null +++ b/checkpoints/torch/rldp-flowbc/cube-single-play-v0-pixels__s3__20260604-052341-rldp-pixel-s3/config.yaml @@ -0,0 +1,90 @@ +domain: cube-single-play-v0 +obs_type: pixels +data_path: /dev/shm/factored-fb/datasets_pixel +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 500000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 3 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: rldp_pixel_s3 +wandb_entity: amsks +wandb_group: rldp_pixel +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: rldp_flowbc +z_dim: 50 +L_dim: 50 +actor_encode_obs: false +discount: 0.99 +lr_f: 0.0001 +lr_b: 0.0001 +lr_actor: 0.0001 +weight_decay: 0.0 +clip_grad_norm: 0.0 +ortho_coef: 0.01 +train_goal_ratio: 0.5 +fb_pessimism_penalty: 0.0 +actor_pessimism_penalty: 0.0 +actor_std: 0.2 +stddev_clip: 0.3 +f_target_tau: 0.005 +b_target_tau: 0.005 +bc_coeff: 3.0 +q_loss_coef: 0.0 +reweight_alpha: 0.0 +reweight_clip: 10.0 +reweight_density_path: analysis/cube_density.npz +weight_diag: false +weight_z: false +onestep: false +amp: false +forward: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +backward: + hidden_dim: 512 + hidden_layers: 4 + norm: true +left_encoder: + hidden_dim: 512 + hidden_layers: 4 + norm: true +actor: + name: noise_conditioned_actor + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 +flow_steps: 10 +lr_actor_vf: 0.0003 +actor_vf: + hidden_dim: 512 + hidden_layers: 4 +horizon: 3 +predictor: + hidden_dim: 512 + hidden_layers: 2 + embedding_layers: 2 + num_parallel: 2 +frame_stack: 3 +obs_normalizer: + name: RGBNormalizerConfig +rgb_encoder: + name: drq + feature_dim: 256 +augmentator: + name: random_shifts + pad: 2 diff --git a/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s1__20260529-002709-tdmpc2-state-10seeds-s1/config.yaml b/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s1__20260529-002709-tdmpc2-state-10seeds-s1/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..59995702bcea3775a6ac78ece211acc2e5b0ce1a --- /dev/null +++ b/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s1__20260529-002709-tdmpc2-state-10seeds-s1/config.yaml @@ -0,0 +1,41 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 1 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: tdmpc2_state_10seeds_s1 +wandb_entity: amsks +wandb_group: tdmpc2_state_10seeds +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: tdmpc2 +horizon: 3 +tdmpc2: + random_goal_ratio: 0.5 + success_threshold: 0.04 + latent_dim: 512 + mlp_dim: 512 + num_q: 5 + horizon: 3 + num_samples: 512 + num_elites: 64 + num_pi_trajs: 24 + iterations: 6 + episode_length: 200 +bc_coeff: 3.0 +lr_b: 0.0001 diff --git a/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s2__20260529-002709-tdmpc2-state-10seeds-s2/config.yaml b/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s2__20260529-002709-tdmpc2-state-10seeds-s2/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..fd5a74e33deaa4d6078e44b4a41580f99bf77e9f --- /dev/null +++ b/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s2__20260529-002709-tdmpc2-state-10seeds-s2/config.yaml @@ -0,0 +1,41 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 2 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: tdmpc2_state_10seeds_s2 +wandb_entity: amsks +wandb_group: tdmpc2_state_10seeds +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: tdmpc2 +horizon: 3 +tdmpc2: + random_goal_ratio: 0.5 + success_threshold: 0.04 + latent_dim: 512 + mlp_dim: 512 + num_q: 5 + horizon: 3 + num_samples: 512 + num_elites: 64 + num_pi_trajs: 24 + iterations: 6 + episode_length: 200 +bc_coeff: 3.0 +lr_b: 0.0001 diff --git a/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s3__20260529-002709-tdmpc2-state-10seeds-s3/config.yaml b/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s3__20260529-002709-tdmpc2-state-10seeds-s3/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0c44fee54552053bc04ad9e56e8a76f52b0fbcc9 --- /dev/null +++ b/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s3__20260529-002709-tdmpc2-state-10seeds-s3/config.yaml @@ -0,0 +1,41 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 3 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: tdmpc2_state_10seeds_s3 +wandb_entity: amsks +wandb_group: tdmpc2_state_10seeds +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: tdmpc2 +horizon: 3 +tdmpc2: + random_goal_ratio: 0.5 + success_threshold: 0.04 + latent_dim: 512 + mlp_dim: 512 + num_q: 5 + horizon: 3 + num_samples: 512 + num_elites: 64 + num_pi_trajs: 24 + iterations: 6 + episode_length: 200 +bc_coeff: 3.0 +lr_b: 0.0001 diff --git a/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s4__20260529-002709-tdmpc2-state-10seeds-s4/config.yaml b/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s4__20260529-002709-tdmpc2-state-10seeds-s4/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ad5c72b40b80db4cd971d820f48640cfcce804ae --- /dev/null +++ b/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s4__20260529-002709-tdmpc2-state-10seeds-s4/config.yaml @@ -0,0 +1,41 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 4 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: tdmpc2_state_10seeds_s4 +wandb_entity: amsks +wandb_group: tdmpc2_state_10seeds +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: tdmpc2 +horizon: 3 +tdmpc2: + random_goal_ratio: 0.5 + success_threshold: 0.04 + latent_dim: 512 + mlp_dim: 512 + num_q: 5 + horizon: 3 + num_samples: 512 + num_elites: 64 + num_pi_trajs: 24 + iterations: 6 + episode_length: 200 +bc_coeff: 3.0 +lr_b: 0.0001 diff --git a/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s5__20260529-002709-tdmpc2-state-10seeds-s5/config.yaml b/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s5__20260529-002709-tdmpc2-state-10seeds-s5/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3d3b9a2242440b35673ed0fb51ec09edd25d9841 --- /dev/null +++ b/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s5__20260529-002709-tdmpc2-state-10seeds-s5/config.yaml @@ -0,0 +1,41 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 5 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: tdmpc2_state_10seeds_s5 +wandb_entity: amsks +wandb_group: tdmpc2_state_10seeds +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: tdmpc2 +horizon: 3 +tdmpc2: + random_goal_ratio: 0.5 + success_threshold: 0.04 + latent_dim: 512 + mlp_dim: 512 + num_q: 5 + horizon: 3 + num_samples: 512 + num_elites: 64 + num_pi_trajs: 24 + iterations: 6 + episode_length: 200 +bc_coeff: 3.0 +lr_b: 0.0001 diff --git a/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s6__20260529-002709-tdmpc2-state-10seeds-s6/config.yaml b/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s6__20260529-002709-tdmpc2-state-10seeds-s6/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ff4810283795f8630f842f317df46e22611bbc27 --- /dev/null +++ b/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s6__20260529-002709-tdmpc2-state-10seeds-s6/config.yaml @@ -0,0 +1,41 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 6 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: tdmpc2_state_10seeds_s6 +wandb_entity: amsks +wandb_group: tdmpc2_state_10seeds +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: tdmpc2 +horizon: 3 +tdmpc2: + random_goal_ratio: 0.5 + success_threshold: 0.04 + latent_dim: 512 + mlp_dim: 512 + num_q: 5 + horizon: 3 + num_samples: 512 + num_elites: 64 + num_pi_trajs: 24 + iterations: 6 + episode_length: 200 +bc_coeff: 3.0 +lr_b: 0.0001 diff --git a/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s7__20260529-002709-tdmpc2-state-10seeds-s7/config.yaml b/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s7__20260529-002709-tdmpc2-state-10seeds-s7/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..271925ae34f01dd512f4a2769e0de07b7c90c163 --- /dev/null +++ b/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s7__20260529-002709-tdmpc2-state-10seeds-s7/config.yaml @@ -0,0 +1,41 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 7 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: tdmpc2_state_10seeds_s7 +wandb_entity: amsks +wandb_group: tdmpc2_state_10seeds +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: tdmpc2 +horizon: 3 +tdmpc2: + random_goal_ratio: 0.5 + success_threshold: 0.04 + latent_dim: 512 + mlp_dim: 512 + num_q: 5 + horizon: 3 + num_samples: 512 + num_elites: 64 + num_pi_trajs: 24 + iterations: 6 + episode_length: 200 +bc_coeff: 3.0 +lr_b: 0.0001 diff --git a/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s8__20260529-002709-tdmpc2-state-10seeds-s8/config.yaml b/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s8__20260529-002709-tdmpc2-state-10seeds-s8/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..6fb80498940b52d30846ec656fdb2a1d77c80b4d --- /dev/null +++ b/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s8__20260529-002709-tdmpc2-state-10seeds-s8/config.yaml @@ -0,0 +1,41 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 8 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: tdmpc2_state_10seeds_s8 +wandb_entity: amsks +wandb_group: tdmpc2_state_10seeds +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: tdmpc2 +horizon: 3 +tdmpc2: + random_goal_ratio: 0.5 + success_threshold: 0.04 + latent_dim: 512 + mlp_dim: 512 + num_q: 5 + horizon: 3 + num_samples: 512 + num_elites: 64 + num_pi_trajs: 24 + iterations: 6 + episode_length: 200 +bc_coeff: 3.0 +lr_b: 0.0001 diff --git a/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s9__20260529-002709-tdmpc2-state-10seeds-s9/config.yaml b/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s9__20260529-002709-tdmpc2-state-10seeds-s9/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..6d7e95324a7857bac042824925cdc9f6a5ba3d43 --- /dev/null +++ b/checkpoints/torch/tdmpc2/cube-single-play-v0-state__s9__20260529-002709-tdmpc2-state-10seeds-s9/config.yaml @@ -0,0 +1,41 @@ +domain: cube-single-play-v0 +obs_type: state +data_path: /dev/shm/factored-fb/datasets +n_transitions: null +load_n_episodes: 1000 +num_train_steps: 1000000 +batch_size: 256 +eval_every: 50000 +log_every: 1000 +save_every: 200000 +seed: 9 +device: cuda +eval_n_episodes: 10 +eval_relabel_size: 10000 +eval_shift_reward: 1.0 +save_eval_videos: false +use_wandb: true +wandb_project: factored-fb +wandb_run_name: tdmpc2_state_10seeds_s9 +wandb_entity: amsks +wandb_group: tdmpc2_state_10seeds +wandb_tags: [] +wandb_save_code: true +save_dir: ${hydra:runtime.output_dir}/checkpoints +eval_videos_dir: ${hydra:runtime.output_dir}/eval_videos +agent: tdmpc2 +horizon: 3 +tdmpc2: + random_goal_ratio: 0.5 + success_threshold: 0.04 + latent_dim: 512 + mlp_dim: 512 + num_q: 5 + horizon: 3 + num_samples: 512 + num_elites: 64 + num_pi_trajs: 24 + iterations: 6 + episode_length: 200 +bc_coeff: 3.0 +lr_b: 0.0001