Add files using upload-large-folder tool

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/AdjustToasterOvenTemperature/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.26
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/AdjustWaterTemperature/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.52
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/CloseBlenderLid/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.02
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/CloseFridge/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.9
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/CloseToasterOvenDoor/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.8
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/CoffeeSetupMug/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.44
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/NavigateKitchen/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.4
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/OpenCabinet/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.5
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/OpenDrawer/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.44
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/OpenElectricKettleLid/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.62
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/OpenStandMixerHead/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.92
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/PickPlaceCabinetToCounter/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.46
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/PickPlaceCounterToCabinet/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.6
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/PickPlaceCounterToStove/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.44
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/PickPlaceDrawerToCounter/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.38
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/PickPlaceSinkToCounter/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.68
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/PickPlaceToasterToCounter/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.88
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/SlideDishwasherRack/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.72
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/TurnOffMicrowave/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.28
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/TurnOffSinkFaucet/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.86
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/TurnOffStove/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.28
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/TurnOnElectricKettle/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.76
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/TurnOnMicrowave/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.26
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/TurnOnSinkFaucet/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.62
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/TurnOnStove/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.38
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/evals/pretrain/TurnSinkSpout/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.9
+}

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/mgd_phase3_aux_fixed_0p5.yaml

@@ -0,0 +1,31 @@
+experiment_name: mgd_phase3_aux_fixed_0p5
+policy_type: groot_mgd
+base_model_path: ${HOME}/groot_robocasa/robocasa_v2/MGD_Processing_line_Only/mgd_token_mask_ratio_${sweep.mgd_token_mask_ratio}/phase2
+output_root: ${HOME}/groot_robocasa/robocasa_v2/processing_line_only/loss_strategy_ablation/mgd
+sweep:
+  loss_strategy:
+  - aux_fixed_0p5
+  mgd_token_mask_ratio:
+  - 0.3
+dataset:
+  dataset_soup: my_atomic26_human
+training:
+  num_gpus: 2
+  batch_size: 64
+model: null
+phases:
+- name: phase3_fm_mgd_fixed_0p5
+  max_steps: 30000
+  save_steps: 0
+  trainable:
+    tune_llm: false
+    tune_visual: false
+    tune_projector: true
+    tune_diffusion_model: true
+  losses:
+    mgd_enabled: true
+    mgd_fm_loss_weight: 1.0
+    mgd_loss_weight: 0.5
+    mgd_token_mask_ratio: ${sweep.mgd_token_mask_ratio}
+    mgd_use_cosine_loss: true
+    mgd_use_mse_loss: false

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/resolved_config.yaml

@@ -0,0 +1,34 @@
+experiment_name: mgd_phase3_aux_fixed_0p5
+policy_type: groot_mgd
+base_model_path: ${HOME}/groot_robocasa/robocasa_v2/MGD_Processing_line_Only/mgd_token_mask_ratio_0.3/phase2
+output_root: ${HOME}/groot_robocasa/robocasa_v2/processing_line_only/loss_strategy_ablation/mgd
+sweep:
+  loss_strategy:
+  - aux_fixed_0p5
+  mgd_token_mask_ratio:
+  - 0.3
+dataset:
+  dataset_soup: my_atomic26_human
+training:
+  num_gpus: 2
+  batch_size: 64
+model: null
+phases:
+- name: phase3_fm_mgd_fixed_0p5
+  max_steps: 30000
+  save_steps: 0
+  trainable:
+    tune_llm: false
+    tune_visual: false
+    tune_projector: true
+    tune_diffusion_model: true
+  losses:
+    mgd_enabled: true
+    mgd_fm_loss_weight: 1.0
+    mgd_loss_weight: 0.5
+    mgd_token_mask_ratio: 0.3
+    mgd_use_cosine_loss: true
+    mgd_use_mse_loss: false
+resolved_sweep:
+  loss_strategy: aux_fixed_0p5
+  mgd_token_mask_ratio: 0.3

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/selected_eval_config.yaml

@@ -0,0 +1,34 @@
+experiment_name: mgd_phase3_aux_fixed_0p5
+policy_type: groot_mgd
+base_model_path: ${HOME}/groot_robocasa/robocasa_v2/MGD_Processing_line_Only/mgd_token_mask_ratio_0.3/phase2
+output_root: ${HOME}/groot_robocasa/robocasa_v2/processing_line_only/loss_strategy_ablation/mgd
+sweep:
+  loss_strategy:
+  - aux_fixed_0p5
+  mgd_token_mask_ratio:
+  - 0.3
+dataset:
+  dataset_soup: my_atomic26_human
+training:
+  num_gpus: 2
+  batch_size: 64
+model: null
+phases:
+- name: phase3_fm_mgd_fixed_0p5
+  max_steps: 30000
+  save_steps: 0
+  trainable:
+    tune_llm: false
+    tune_visual: false
+    tune_projector: true
+    tune_diffusion_model: true
+  losses:
+    mgd_enabled: true
+    mgd_fm_loss_weight: 1.0
+    mgd_loss_weight: 0.5
+    mgd_token_mask_ratio: 0.3
+    mgd_use_cosine_loss: true
+    mgd_use_mse_loss: false
+resolved_sweep:
+  loss_strategy: aux_fixed_0p5
+  mgd_token_mask_ratio: 0.3

loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/summary.txt

@@ -0,0 +1,56 @@
+Eval metadata
+- total_eval_time: 3h 13m 46s (11625.87s)
+- model_path: /home/seonho/groot_robocasa/robocasa_v2/processing_line_only/loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3
+- checkpoint_name: phase3
+- checkpoint_parent: /home/seonho/groot_robocasa/robocasa_v2/processing_line_only/loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3
+- video_dir: /home/seonho/groot_robocasa/Experiments/loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3
+- task_set: my_atomic26
+- split: pretrain
+- n_envs: 10
+- n_episodes: 50
+- seed: 42
+- command: clvla/benchmarks/robocasa_v2/my_scripts/run_eval.py --model_path /home/seonho/groot_robocasa/robocasa_v2/processing_line_only/loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3 --video_dir /home/seonho/groot_robocasa/Experiments/loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3 --task_set my_atomic26 --split pretrain --n_envs 10 --n_episodes 50 --port 4179 --seed 42
+- config_yaml_candidates:
+  - /home/seonho/groot_robocasa/robocasa_v2/processing_line_only/loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/mgd_phase3_aux_fixed_0p5.yaml
+  - /home/seonho/groot_robocasa/robocasa_v2/processing_line_only/loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/resolved_config.yaml
+- selected_config_yaml: /home/seonho/groot_robocasa/robocasa_v2/processing_line_only/loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/resolved_config.yaml
+- copied_config_yamls:
+  - /home/seonho/groot_robocasa/Experiments/loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/mgd_phase3_aux_fixed_0p5.yaml
+  - /home/seonho/groot_robocasa/Experiments/loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/resolved_config.yaml
+  - /home/seonho/groot_robocasa/Experiments/loss_strategy_ablation/mgd/loss_strategy_aux_fixed_0p5__mgd_token_mask_ratio_0.3/phase3/selected_eval_config.yaml
+
+Stats for task group: MY_ATOMIC26
+CloseBlenderLid: 2 | None
+CloseFridge: 90 | None
+CloseToasterOvenDoor: 80 | None
+CoffeeSetupMug: 44 | None
+NavigateKitchen: 40 | None
+OpenCabinet: 50 | None
+OpenDrawer: 44 | None
+OpenStandMixerHead: 92 | None
+PickPlaceCounterToCabinet: 60 | None
+PickPlaceCounterToStove: 44 | None
+PickPlaceDrawerToCounter: 38 | None
+PickPlaceSinkToCounter: 68 | None
+PickPlaceToasterToCounter: 88 | None
+PickPlaceCabinetToCounter: 46 | None
+TurnOnElectricKettle: 76 | None
+TurnOnMicrowave: 26 | None
+TurnOffMicrowave: 28 | None
+OpenElectricKettleLid: 62 | None
+SlideDishwasherRack: 72 | None
+TurnOnSinkFaucet: 62 | None
+TurnSinkSpout: 90 | None
+TurnOffSinkFaucet: 86 | None
+AdjustWaterTemperature: 52 | None
+TurnOffStove: 28 | None
+TurnOnStove: 38 | None
+AdjustToasterOvenTemperature: 26 | None
+AVG: 55.1 | nan
+
+Eval completeness check
+- split: pretrain
+- expected tasks: 26
+- completed tasks: 26/26
+- expected episodes per task: 50
+CHECK PASSED

loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3/evals/pretrain/CloseBlenderLid/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.06
+}

loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3/evals/pretrain/CloseToasterOvenDoor/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.74
+}

loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3/evals/pretrain/CoffeeSetupMug/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.46
+}

loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3/evals/pretrain/NavigateKitchen/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.34
+}

loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3/evals/pretrain/OpenCabinet/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.5
+}

loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3/evals/pretrain/OpenDrawer/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.54
+}

loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3/evals/pretrain/OpenElectricKettleLid/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.58
+}

loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3/evals/pretrain/PickPlaceCabinetToCounter/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.4
+}

loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3/evals/pretrain/PickPlaceCounterToCabinet/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.5
+}

loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3/evals/pretrain/PickPlaceDrawerToCounter/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.34
+}

loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3/evals/pretrain/PickPlaceSinkToCounter/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.66
+}

loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3/evals/pretrain/PickPlaceToasterToCounter/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.84
+}

loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3/evals/pretrain/TurnOffStove/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.3
+}

loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3/evals/pretrain/TurnOnElectricKettle/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.62
+}

loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3/evals/pretrain/TurnOnMicrowave/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.34
+}

loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3/evals/pretrain/TurnSinkSpout/stats.json

@@ -0,0 +1,4 @@
+{
+    "num_episodes": 50,
+    "success_rate": 0.86
+}

loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3/resolved_config.yaml

@@ -0,0 +1,34 @@
+experiment_name: rkd_phase3_aux_fixed_0p5
+policy_type: groot_rkd
+base_model_path: ${HOME}/groot_robocasa/robocasa_v2/RKD_Processing_line_Only/rkd_temp_0.1/phase2
+output_root: ${HOME}/groot_robocasa/robocasa_v2/processing_line_only/loss_strategy_ablation/rkd
+sweep:
+  loss_strategy:
+  - aux_fixed_0p5
+  rkd_temp:
+  - 0.1
+dataset:
+  dataset_soup: my_atomic26_human
+training:
+  num_gpus: 2
+  batch_size: 64
+model: null
+phases:
+- name: phase3_fm_rkd_fixed_0p5
+  max_steps: 30000
+  save_steps: 0
+  trainable:
+    tune_llm: false
+    tune_visual: false
+    tune_projector: true
+    tune_diffusion_model: true
+  losses:
+    rkd_enabled: true
+    rkd_fm_loss_weight: 1.0
+    rkd_loss_weight: 0.5
+    rkd_action_temp: 0.1
+    rkd_vlm_temp: 0.1
+    rkd_relation_mode: token_pair_mean
+resolved_sweep:
+  loss_strategy: aux_fixed_0p5
+  rkd_temp: 0.1

loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3/rkd_phase3_aux_fixed_0p5.yaml

@@ -0,0 +1,31 @@
+experiment_name: rkd_phase3_aux_fixed_0p5
+policy_type: groot_rkd
+base_model_path: ${HOME}/groot_robocasa/robocasa_v2/RKD_Processing_line_Only/rkd_temp_${sweep.rkd_temp}/phase2
+output_root: ${HOME}/groot_robocasa/robocasa_v2/processing_line_only/loss_strategy_ablation/rkd
+sweep:
+  loss_strategy:
+  - aux_fixed_0p5
+  rkd_temp:
+  - 0.1
+dataset:
+  dataset_soup: my_atomic26_human
+training:
+  num_gpus: 2
+  batch_size: 64
+model: null
+phases:
+- name: phase3_fm_rkd_fixed_0p5
+  max_steps: 30000
+  save_steps: 0
+  trainable:
+    tune_llm: false
+    tune_visual: false
+    tune_projector: true
+    tune_diffusion_model: true
+  losses:
+    rkd_enabled: true
+    rkd_fm_loss_weight: 1.0
+    rkd_loss_weight: 0.5
+    rkd_action_temp: ${sweep.rkd_temp}
+    rkd_vlm_temp: ${sweep.rkd_temp}
+    rkd_relation_mode: token_pair_mean

loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3/selected_eval_config.yaml

@@ -0,0 +1,34 @@
+experiment_name: rkd_phase3_aux_fixed_0p5
+policy_type: groot_rkd
+base_model_path: ${HOME}/groot_robocasa/robocasa_v2/RKD_Processing_line_Only/rkd_temp_0.1/phase2
+output_root: ${HOME}/groot_robocasa/robocasa_v2/processing_line_only/loss_strategy_ablation/rkd
+sweep:
+  loss_strategy:
+  - aux_fixed_0p5
+  rkd_temp:
+  - 0.1
+dataset:
+  dataset_soup: my_atomic26_human
+training:
+  num_gpus: 2
+  batch_size: 64
+model: null
+phases:
+- name: phase3_fm_rkd_fixed_0p5
+  max_steps: 30000
+  save_steps: 0
+  trainable:
+    tune_llm: false
+    tune_visual: false
+    tune_projector: true
+    tune_diffusion_model: true
+  losses:
+    rkd_enabled: true
+    rkd_fm_loss_weight: 1.0
+    rkd_loss_weight: 0.5
+    rkd_action_temp: 0.1
+    rkd_vlm_temp: 0.1
+    rkd_relation_mode: token_pair_mean
+resolved_sweep:
+  loss_strategy: aux_fixed_0p5
+  rkd_temp: 0.1

loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3/summary.txt

@@ -0,0 +1,56 @@
+Eval metadata
+- total_eval_time: 2h 34m 33s (9272.66s)
+- model_path: /home/seonho/groot_robocasa/robocasa_v2/processing_line_only/loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3
+- checkpoint_name: phase3
+- checkpoint_parent: /home/seonho/groot_robocasa/robocasa_v2/processing_line_only/loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1
+- video_dir: /home/seonho/groot_robocasa/Experiments/loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3
+- task_set: my_atomic26
+- split: pretrain
+- n_envs: 10
+- n_episodes: 50
+- seed: 42
+- command: clvla/benchmarks/robocasa_v2/my_scripts/run_eval.py --model_path /home/seonho/groot_robocasa/robocasa_v2/processing_line_only/loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3 --video_dir /home/seonho/groot_robocasa/Experiments/loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3 --task_set my_atomic26 --split pretrain --n_envs 10 --n_episodes 50 --port 6809 --seed 42
+- config_yaml_candidates:
+  - /home/seonho/groot_robocasa/robocasa_v2/processing_line_only/loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/resolved_config.yaml
+  - /home/seonho/groot_robocasa/robocasa_v2/processing_line_only/loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/rkd_phase3_aux_fixed_0p5.yaml
+- selected_config_yaml: /home/seonho/groot_robocasa/robocasa_v2/processing_line_only/loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/resolved_config.yaml
+- copied_config_yamls:
+  - /home/seonho/groot_robocasa/Experiments/loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3/resolved_config.yaml
+  - /home/seonho/groot_robocasa/Experiments/loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3/rkd_phase3_aux_fixed_0p5.yaml
+  - /home/seonho/groot_robocasa/Experiments/loss_strategy_ablation/rkd/loss_strategy_aux_fixed_0p5__rkd_temp_0.1/phase3/selected_eval_config.yaml
+
+Stats for task group: MY_ATOMIC26
+CloseBlenderLid: 6 | None
+CloseFridge: 88 | None
+CloseToasterOvenDoor: 74 | None
+CoffeeSetupMug: 46 | None
+NavigateKitchen: 34 | None
+OpenCabinet: 50 | None
+OpenDrawer: 54 | None
+OpenStandMixerHead: 84 | None
+PickPlaceCounterToCabinet: 50 | None
+PickPlaceCounterToStove: 54 | None
+PickPlaceDrawerToCounter: 34 | None
+PickPlaceSinkToCounter: 66 | None
+PickPlaceToasterToCounter: 84 | None
+PickPlaceCabinetToCounter: 40 | None
+TurnOnElectricKettle: 62 | None
+TurnOnMicrowave: 34 | None
+TurnOffMicrowave: 32 | None
+OpenElectricKettleLid: 58 | None
+SlideDishwasherRack: 70 | None
+TurnOnSinkFaucet: 68 | None
+TurnSinkSpout: 86 | None
+TurnOffSinkFaucet: 94 | None
+AdjustWaterTemperature: 48 | None
+TurnOffStove: 30 | None
+TurnOnStove: 38 | None
+AdjustToasterOvenTemperature: 34 | None
+AVG: 54.5 | nan
+
+Eval completeness check
+- split: pretrain
+- expected tasks: 26
+- completed tasks: 26/26
+- expected episodes per task: 50
+CHECK PASSED