Upload folder using huggingface_hub

checkpoint-100.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:222fe8d576c9f94d02ae80838368838d51c899200f66d898530df864ee698f99
+size 1086628090

checkpoint-prism-100.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1415a83c870155f3e7544858261aa610c2c4cd0acfa614883d0d66849f835f21
+size 1086624506

events.out.tfevents.1768103766.xzhang-vla-finetune-h100-8gpu-group-worker-dwrdd.3494898.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9500c3b3775d6f015fea813faf14010274b744f1049fa6432baacf897ec915a3
+size 14259198

log.txt

@@ -0,0 +1,102 @@
+{"train_lr": 0.00012442254388666513, "train_loss": 0.3106695828571558, "epoch": 0}
+{"train_lr": 0.00037442254388666657, "train_loss": 0.1765579505204583, "epoch": 1}
+{"train_lr": 0.0004999895847251151, "train_loss": 0.1204350220322297, "epoch": 2}
+{"train_lr": 0.0004999267965555358, "train_loss": 0.10512903698047145, "epoch": 3}
+{"train_lr": 0.000499801090651457, "train_loss": 0.09721263018323434, "epoch": 4}
+{"train_lr": 0.000499612498658695, "train_loss": 0.09171448527308411, "epoch": 5}
+{"train_lr": 0.0004993610680542452, "train_loss": 0.08695411841962446, "epoch": 6}
+{"train_lr": 0.0004990468621344369, "train_loss": 0.08087445833769888, "epoch": 7}
+{"train_lr": 0.000498669959998947, "train_loss": 0.07634419156898135, "epoch": 8}
+{"train_lr": 0.0004982304565308317, "train_loss": 0.07051568671852732, "epoch": 9}
+{"train_lr": 0.0004977284623727689, "train_loss": 0.0674788464112583, "epoch": 10}
+{"train_lr": 0.000497164103899089, "train_loss": 0.0654696337283541, "epoch": 11}
+{"train_lr": 0.0004965375231840242, "train_loss": 0.06279148746162543, "epoch": 12}
+{"train_lr": 0.000495848877965922, "train_loss": 0.060323238908664575, "epoch": 13}
+{"train_lr": 0.000495098341607544, "train_loss": 0.05694042379746439, "epoch": 14}
+{"train_lr": 0.0004942861030523532, "train_loss": 0.05254918934498011, "epoch": 15}
+{"train_lr": 0.0004934123667771247, "train_loss": 0.048871026988278055, "epoch": 16}
+{"train_lr": 0.0004924773527402652, "train_loss": 0.046795259160964865, "epoch": 17}
+{"train_lr": 0.0004914812963265581, "train_loss": 0.04460288662668364, "epoch": 18}
+{"train_lr": 0.0004904244482878885, "train_loss": 0.042538205515167356, "epoch": 19}
+{"train_lr": 0.0004893070746800906, "train_loss": 0.0405509257690194, "epoch": 20}
+{"train_lr": 0.0004881294567959938, "train_loss": 0.039287820584866386, "epoch": 21}
+{"train_lr": 0.0004868918910945998, "train_loss": 0.03803709820202604, "epoch": 22}
+{"train_lr": 0.00048559468912647095, "train_loss": 0.03727623393734321, "epoch": 23}
+{"train_lr": 0.00048423817745526197, "train_loss": 0.03635868519914033, "epoch": 24}
+{"train_lr": 0.00048282269757551324, "train_loss": 0.035470682053193155, "epoch": 25}
+{"train_lr": 0.00048134860582673604, "train_loss": 0.03436020502226452, "epoch": 26}
+{"train_lr": 0.00047981627330363384, "train_loss": 0.03362559669345335, "epoch": 27}
+{"train_lr": 0.0004782260857627244, "train_loss": 0.033024097851675184, "epoch": 28}
+{"train_lr": 0.00047657844352522205, "train_loss": 0.03246270484458953, "epoch": 29}
+{"train_lr": 0.0004748737613762465, "train_loss": 0.03193189644536263, "epoch": 30}
+{"train_lr": 0.0004731124684604214, "train_loss": 0.031190929833294356, "epoch": 31}
+{"train_lr": 0.0004712950081737899, "train_loss": 0.03056241925207776, "epoch": 32}
+{"train_lr": 0.0004694218380523076, "train_loss": 0.029931797038955647, "epoch": 33}
+{"train_lr": 0.00046749342965647715, "train_loss": 0.02938416024297777, "epoch": 34}
+{"train_lr": 0.00046551026845287634, "train_loss": 0.02895656854657426, "epoch": 35}
+{"train_lr": 0.0004634728536916469, "train_loss": 0.02855706407095833, "epoch": 36}
+{"train_lr": 0.00046138169828110727, "train_loss": 0.028285302986864293, "epoch": 37}
+{"train_lr": 0.00045923732865845644, "train_loss": 0.02793090866068992, "epoch": 38}
+{"train_lr": 0.0004570402846572667, "train_loss": 0.02750447166595731, "epoch": 39}
+{"train_lr": 0.0004547911193715973, "train_loss": 0.02719697173304648, "epoch": 40}
+{"train_lr": 0.0004524903990167991, "train_loss": 0.026828367616214727, "epoch": 41}
+{"train_lr": 0.0004501387027869346, "train_loss": 0.02656354377338646, "epoch": 42}
+{"train_lr": 0.0004477366227089122, "train_loss": 0.026073855690526584, "epoch": 43}
+{"train_lr": 0.00044528476349357706, "train_loss": 0.02558653131257146, "epoch": 44}
+{"train_lr": 0.0004427837423833779, "train_loss": 0.025155968964189517, "epoch": 45}
+{"train_lr": 0.0004402341889969815, "train_loss": 0.024923711238267036, "epoch": 46}
+{"train_lr": 0.00043763674517082824, "train_loss": 0.024778983413970626, "epoch": 47}
+{"train_lr": 0.00043499206479752557, "train_loss": 0.02437653219389572, "epoch": 48}
+{"train_lr": 0.0004323008136611902, "train_loss": 0.024127835892176, "epoch": 49}
+{"train_lr": 0.0004295636692699243, "train_loss": 0.024061144793673307, "epoch": 50}
+{"train_lr": 0.0004267813206851613, "train_loss": 0.02371762236895402, "epoch": 51}
+{"train_lr": 0.00042395446834832116, "train_loss": 0.023586711529009177, "epoch": 52}
+{"train_lr": 0.0004210838239043303, "train_loss": 0.02339996546571059, "epoch": 53}
+{"train_lr": 0.000418170110022581, "train_loss": 0.023331272883557557, "epoch": 54}
+{"train_lr": 0.0004152140602149747, "train_loss": 0.022939835333343955, "epoch": 55}
+{"train_lr": 0.0004122164186512228, "train_loss": 0.022742653930062685, "epoch": 56}
+{"train_lr": 0.0004091779399715766, "train_loss": 0.022589488850228434, "epoch": 57}
+{"train_lr": 0.00040609938909678126, "train_loss": 0.02226874454806062, "epoch": 58}
+{"train_lr": 0.0004029815410355584, "train_loss": 0.022061608113099807, "epoch": 59}
+{"train_lr": 0.0003998251806894967, "train_loss": 0.0218637260706577, "epoch": 60}
+{"train_lr": 0.0003966311026554213, "train_loss": 0.021659375753381765, "epoch": 61}
+{"train_lr": 0.00039340011102539246, "train_loss": 0.021473534835900046, "epoch": 62}
+{"train_lr": 0.00039013301918427843, "train_loss": 0.02136045050270073, "epoch": 63}
+{"train_lr": 0.0003868306496049915, "train_loss": 0.021258544771603508, "epoch": 64}
+{"train_lr": 0.00038349383364139636, "train_loss": 0.021133298709823525, "epoch": 65}
+{"train_lr": 0.00038012341131907234, "train_loss": 0.021022808593492766, "epoch": 66}
+{"train_lr": 0.0003767202311238196, "train_loss": 0.02089578813532913, "epoch": 67}
+{"train_lr": 0.00037328514978804834, "train_loss": 0.020739652609188294, "epoch": 68}
+{"train_lr": 0.0003698190320751148, "train_loss": 0.020612976758250082, "epoch": 69}
+{"train_lr": 0.0003663227505616289, "train_loss": 0.020530629719294605, "epoch": 70}
+{"train_lr": 0.0003627971854177555, "train_loss": 0.020418202951371936, "epoch": 71}
+{"train_lr": 0.0003592432241856427, "train_loss": 0.020347804001564475, "epoch": 72}
+{"train_lr": 0.00035566176155605976, "train_loss": 0.020233220371425876, "epoch": 73}
+{"train_lr": 0.00035205369914305997, "train_loss": 0.020152423528576206, "epoch": 74}
+{"train_lr": 0.000348419945257054, "train_loss": 0.020014205569804946, "epoch": 75}
+{"train_lr": 0.0003447614146761572, "train_loss": 0.019932750324309353, "epoch": 76}
+{"train_lr": 0.00034107902841587126, "train_loss": 0.01980366973392981, "epoch": 77}
+{"train_lr": 0.00033737371349725365, "train_loss": 0.019748475697988082, "epoch": 78}
+{"train_lr": 0.00033364640271351683, "train_loss": 0.019675734331181676, "epoch": 79}
+{"train_lr": 0.00032989803439521125, "train_loss": 0.019610411182588324, "epoch": 80}
+{"train_lr": 0.00032612955217403125, "train_loss": 0.019501670381825986, "epoch": 81}
+{"train_lr": 0.0003223419047452072, "train_loss": 0.019465307419461952, "epoch": 82}
+{"train_lr": 0.00031853604562872363, "train_loss": 0.019317056300060628, "epoch": 83}
+{"train_lr": 0.000314712932929276, "train_loss": 0.01924570535176884, "epoch": 84}
+{"train_lr": 0.00031087352909502683, "train_loss": 0.01921790678825727, "epoch": 85}
+{"train_lr": 0.0003070188006753687, "train_loss": 0.0191203878184208, "epoch": 86}
+{"train_lr": 0.0003031497180775688, "train_loss": 0.018995062414310645, "epoch": 87}
+{"train_lr": 0.00029926725532246237, "train_loss": 0.018894653228115295, "epoch": 88}
+{"train_lr": 0.000295372389799301, "train_loss": 0.018844347112270145, "epoch": 89}
+{"train_lr": 0.0002914661020196496, "train_loss": 0.018712236197917165, "epoch": 90}
+{"train_lr": 0.00028754937537056586, "train_loss": 0.018601788093821667, "epoch": 91}
+{"train_lr": 0.00028362319586703824, "train_loss": 0.018510111748926545, "epoch": 92}
+{"train_lr": 0.0002796885519037529, "train_loss": 0.018389944572524148, "epoch": 93}
+{"train_lr": 0.0002757464340062982, "train_loss": 0.01826450743363705, "epoch": 94}
+{"train_lr": 0.0002717978345817676, "train_loss": 0.018135870946037018, "epoch": 95}
+{"train_lr": 0.00026784374766894995, "train_loss": 0.017987648956262522, "epoch": 96}
+{"train_lr": 0.0002638851686880748, "train_loss": 0.017831332910776397, "epoch": 97}
+{"train_lr": 0.0002599230941902473, "train_loss": 0.017715244356683084, "epoch": 98}
+{"train_lr": 0.0002559585216065196, "train_loss": 0.017702003495481893, "epoch": 99}
+{"train_lr": 0.0002519924489968411, "train_loss": 0.017629278209084092, "epoch": 100}
+{"train_lr": 0.0002480258747987811, "train_loss": 0.017556594441993915, "epoch": 101}

results_s0_evals20_tt_window_obs_checkpoint-100.csv

@@ -0,0 +1,7 @@
+n_eval,success_rate,task_name
+20,0.55,MemFruitInSinkRightFar
+20,0.6,MemFruitInSinkLeftFar
+20,0.25,MemPutKBreadInMicrowave
+20,0.35,MemWashAndReturnRight
+20,0.25,MemWashAndReturnLeft
+20,0.6,MemHeatPot

results_s0_evals20_tt_window_obs_checkpoint-50.csv

@@ -0,0 +1,7 @@
+n_eval,success_rate,task_name
+20,0.4,MemFruitInSinkRightFar
+20,0.35,MemFruitInSinkLeftFar
+20,0.45,MemPutKBreadInMicrowave
+20,0.3,MemWashAndReturnRight
+20,0.35,MemWashAndReturnLeft
+20,0.7,MemHeatPot

results_s0_evals20_tt_window_obs_checkpoint-60.csv

@@ -0,0 +1,7 @@
+n_eval,success_rate,task_name
+20,0.2,MemFruitInSinkRightFar
+20,0.55,MemFruitInSinkLeftFar
+20,0.1,MemPutKBreadInMicrowave
+20,0.35,MemWashAndReturnRight
+20,0.5,MemWashAndReturnLeft
+20,0.8,MemHeatPot

results_s0_evals20_tt_window_obs_checkpoint-70.csv

@@ -0,0 +1,7 @@
+n_eval,success_rate,task_name
+20,0.35,MemFruitInSinkRightFar
+20,0.5,MemFruitInSinkLeftFar
+20,0.05,MemPutKBreadInMicrowave
+20,0.3,MemWashAndReturnRight
+20,0.35,MemWashAndReturnLeft
+20,0.7,MemHeatPot

results_s0_evals20_tt_window_obs_checkpoint-80.csv

@@ -0,0 +1,7 @@
+n_eval,success_rate,task_name
+20,0.4,MemFruitInSinkRightFar
+20,0.4,MemFruitInSinkLeftFar
+20,0.2,MemPutKBreadInMicrowave
+20,0.4,MemWashAndReturnRight
+20,0.3,MemWashAndReturnLeft
+20,0.75,MemHeatPot

results_s0_evals20_tt_window_obs_checkpoint-90.csv

@@ -0,0 +1,7 @@
+n_eval,success_rate,task_name
+20,0.4,MemFruitInSinkRightFar
+20,0.4,MemFruitInSinkLeftFar
+20,0.2,MemPutKBreadInMicrowave
+20,0.2,MemWashAndReturnRight
+20,0.4,MemWashAndReturnLeft
+20,0.8,MemHeatPot

run-prism.yaml

@@ -0,0 +1,122 @@
+dataset_cfg:
+  action_noise: 0.0
+  dataset_json:
+  - /storage/nfs/xzhang/experiments/nopool_exps/exp_ds8_libero_1x_task_robocasa_mem_four_GBS256_cfgtask_robocasa_mem_four_sl2048_s2_crossmae_block_sigmoid_g1_nopool_compress/exp_ds8_libero_1x_task_robocasa_mem_four_GBS256_cfgtask_robocasa_mem_four_sl2048_s2_crossmae_block_sigmoid_g1_nopool_compress/task_robocasa_mem_four.json
+  dataset_val_json:
+  - ''
+  load_in_mem: false
+  n_examples_only: -1
+  non_overlapping:
+  - 2
+  num_repeat_traj:
+  - 256
+  proprio_noise: 0.005
+  sample_each_state: false
+  task_name: ''
+  task_names: null
+  train_on_exploration: false
+  use_dali: true
+dataset_json_contents:
+- action_keys:
+  - actions
+  dataset_path:
+  - memory/MemPutKBreadInMicrowave/2025-07-25-10-40-55/demo_im128_notp.hdf5
+  - memory/MemPutKBreadInMicrowave/2025-07-25-13-25-05/demo_im128_notp.hdf5
+  - memory/MemPutKBreadInMicrowave/2025-07-25-14-50-59/demo_im128_notp.hdf5
+  - memory/MemFruitInSinkLeftFar/2025-07-13-15-44-32/demo_im128_notp.hdf5
+  - memory/MemFruitInSinkRightFar/2025-07-15-15-07-57/demo_im128_notp.hdf5
+  - memory/MemHeatPot/2025-07-24-22-26-20/demo_im128_notp.hdf5
+  - memory/MemHeatPot/2025-07-25-15-56-53/demo_im128_notp.hdf5
+  - memory/MemWashAndReturnLeft/2025-07-25-00-12-14/demo_im128_notp.hdf5
+  - memory/MemWashAndReturnRight/2025-07-25-00-45-53/demo_im128_notp.hdf5
+  group_method: dinov2
+  image_keys:
+  - obs/robot0_agentview_center_image
+  - obs/robot0_eye_in_hand_image
+  low_dim_keys: []
+  max_traj_len: 1024
+  proprio_keys:
+  - obs/robot0_joint_pos_cos
+  - obs/robot0_gripper_qpos
+device: cuda
+dist_on_itp: false
+dist_url: env://
+load_config: null
+logging_cfg:
+  log_dir: /storage/nfs/xzhang/experiments/nopool_exps/exp_ds8_libero_1x_task_robocasa_mem_four_GBS256_cfgtask_robocasa_mem_four_sl2048_s2_crossmae_block_sigmoid_g1_nopool_compress/exp_ds8_libero_1x_task_robocasa_mem_four_GBS256_cfgtask_robocasa_mem_four_sl2048_s2_crossmae_block_sigmoid_g1_nopool_compress
+  log_name: exp_ds8_libero_1x_task_robocasa_mem_four_GBS256_cfgtask_robocasa_mem_four_sl2048_s2_crossmae_block_sigmoid_g1_nopool_compress
+  output_dir: /storage/nfs/xzhang/experiments/nopool_exps/exp_ds8_libero_1x_task_robocasa_mem_four_GBS256_cfgtask_robocasa_mem_four_sl2048_s2_crossmae_block_sigmoid_g1_nopool_compress/exp_ds8_libero_1x_task_robocasa_mem_four_GBS256_cfgtask_robocasa_mem_four_sl2048_s2_crossmae_block_sigmoid_g1_nopool_compress
+model_cfg:
+  policy_cfg:
+    attn_drop: 0.0
+    attn_scale_factor: null
+    block_attn_ind: null
+    block_pattern_start_offset: 0
+    compress_full_attn: true
+    compressor_latent_len: 1
+    decoder_pred_head: mlp
+    full_attn_inds: null
+    gate_full_attn_layers: true
+    gating_type: sigmoid_g1
+    lora_rank: 4
+    model_version: v2
+    phase: pretrain
+    pool_block_inds: null
+    position_language_embedding: start
+    pretrained_path: null
+    remove_action: false
+    remove_proprio: false
+    scratch_llama_config: config/model_config/libero_1x.json
+    tt_window_obs: false
+    use_block_attention: true
+    use_linear_attention: false
+    use_lstm: false
+    use_mamba: false
+    use_transformer_xl: false
+    xl_chunk_factor: 32
+    xl_mem_len: 32
+  vision_encoder_cfg:
+    freeze_all: false
+    vision_encoder: /storage/nfs/rshah/datasets/robocasa/datasets/crossmae_rtx/cross-mae-rtx-vitb.pth
+    vision_lora: false
+    vision_lora_rank: 8
+    vision_nonpretrained: false
+    vision_unfreeze_all: false
+    vision_unfreeze_last_n: 0
+optimizer_cfg:
+  blr: 0.001
+  lr: 0.0005
+  min_lr: 0.0
+  warmup_epochs: 2.0
+  weight_decay: 0.01
+shared_cfg:
+  batch_size: 2
+  block_finetune: ''
+  downsample_obs: 8
+  enable_gradient_checkpointing: false
+  gripper_img: true
+  has_base_action: true
+  is_bimanual: false
+  k_ptp: 0
+  num_cameras: 2
+  num_pred_steps: 32
+  pad_to_max_length: true
+  resume: null
+  save_every: 10
+  seed: 2
+  seq_length: 2048
+  split_epoch: 1
+  start_epoch: 0
+  use_language_conditioning: true
+train: true
+trainer_cfg:
+  accum_iter: 32
+  break_after_n_epochs: 101
+  compile_model: true
+  epochs: 200
+  num_workers: 72
+  pin_memory: true
+  val_every: 10
+  wandb_project: icrt_reruns
+  wandb_watch: true
+world_size: 1

run.yaml

@@ -0,0 +1,223 @@
+!!python/object:icrt.util.args.ExperimentConfig
+dataset_cfg: !!python/object:icrt.util.args.DatasetConfig
+  action_noise: 0.0
+  action_traj_noise: false
+  dataset_fraction: 1.0
+  dataset_json:
+  - /storage/nfs/xzhang/experiments/nopool_exps/exp_ds8_libero_1x_task_robocasa_mem_four_GBS256_cfgtask_robocasa_mem_four_sl2048_s2_crossmae_block_sigmoid_g1_nopool_compress/exp_ds8_libero_1x_task_robocasa_mem_four_GBS256_cfgtask_robocasa_mem_four_sl2048_s2_crossmae_block_sigmoid_g1_nopool_compress/task_robocasa_mem_four.json
+  dataset_val_json:
+  - ''
+  goal_conditioned: false
+  load_in_mem: false
+  low_dim_noise: 0.005
+  min_subsample_length: 64
+  n_examples_only: -1
+  non_overlapping:
+  - 2
+  num_repeat_traj:
+  - 256
+  num_weighted_steps: 30
+  only_k: 10
+  proprio_noise: 0.005
+  random_patch_masking: false
+  rebalance_indices: false
+  rebalance_tasks: true
+  sample_each_state: false
+  shuffle_repeat_traj: true
+  shuffle_rnd_prob: 0.0
+  sim_score_th: -1.0
+  skip_step: false
+  sort_by_closeness: false
+  sort_by_lang: true
+  start_from_beginning: false
+  subsample_traj: false
+  task_barrier: true
+  task_name: ''
+  task_names: null
+  toy_max_distractors: 1
+  toy_max_tasks: 5
+  toy_task_id: 1
+  toy_train_dataset_size: 1000000
+  toy_train_seed: 42
+  toy_val_dataset_size: 500000
+  toy_val_seed: 84
+  train_on_exploration: false
+  trim_num: -1
+  trim_set: ''
+  use_dali: true
+  val_only_k: 5
+  video_conditioned: false
+  vision_aug: true
+  weight_by_dataset:
+  - 1
+dataset_json_contents:
+- action_keys:
+  - actions
+  dataset_path:
+  - memory/MemPutKBreadInMicrowave/2025-07-25-10-40-55/demo_im128_notp.hdf5
+  - memory/MemPutKBreadInMicrowave/2025-07-25-13-25-05/demo_im128_notp.hdf5
+  - memory/MemPutKBreadInMicrowave/2025-07-25-14-50-59/demo_im128_notp.hdf5
+  - memory/MemFruitInSinkLeftFar/2025-07-13-15-44-32/demo_im128_notp.hdf5
+  - memory/MemFruitInSinkRightFar/2025-07-15-15-07-57/demo_im128_notp.hdf5
+  - memory/MemHeatPot/2025-07-24-22-26-20/demo_im128_notp.hdf5
+  - memory/MemHeatPot/2025-07-25-15-56-53/demo_im128_notp.hdf5
+  - memory/MemWashAndReturnLeft/2025-07-25-00-12-14/demo_im128_notp.hdf5
+  - memory/MemWashAndReturnRight/2025-07-25-00-45-53/demo_im128_notp.hdf5
+  group_method: dinov2
+  image_keys:
+  - obs/robot0_agentview_center_image
+  - obs/robot0_eye_in_hand_image
+  low_dim_keys: []
+  max_traj_len: 1024
+  proprio_keys:
+  - obs/robot0_joint_pos_cos
+  - obs/robot0_gripper_qpos
+device: cuda
+dist_on_itp: false
+dist_url: env://
+load_config: null
+local_rank: -1
+logging_cfg: !!python/object:icrt.util.args.LoggingConfig
+  log_dir: /storage/nfs/xzhang/experiments/nopool_exps/exp_ds8_libero_1x_task_robocasa_mem_four_GBS256_cfgtask_robocasa_mem_four_sl2048_s2_crossmae_block_sigmoid_g1_nopool_compress/exp_ds8_libero_1x_task_robocasa_mem_four_GBS256_cfgtask_robocasa_mem_four_sl2048_s2_crossmae_block_sigmoid_g1_nopool_compress
+  log_name: exp_ds8_libero_1x_task_robocasa_mem_four_GBS256_cfgtask_robocasa_mem_four_sl2048_s2_crossmae_block_sigmoid_g1_nopool_compress
+  output_dir: /storage/nfs/xzhang/experiments/nopool_exps/exp_ds8_libero_1x_task_robocasa_mem_four_GBS256_cfgtask_robocasa_mem_four_sl2048_s2_crossmae_block_sigmoid_g1_nopool_compress/exp_ds8_libero_1x_task_robocasa_mem_four_GBS256_cfgtask_robocasa_mem_four_sl2048_s2_crossmae_block_sigmoid_g1_nopool_compress
+model_cfg: !!python/object:icrt.util.args.ModelConfig
+  policy_cfg: !!python/object:icrt.util.args.PolicyConfig
+    adapter_mlp_ratio: 4.0
+    adapter_num_heads: 8
+    add_sink_token_to_full_attn: false
+    attn_drop: 0.0
+    attn_latent_len: 1
+    attn_scale_factor: null
+    block_attn_ind: null
+    block_pattern_start_offset: 0
+    compress_full_attn: true
+    compress_include_prev_action: false
+    compressor_latent_len: 1
+    decoder_cfg_path: config/model_config/action_head.json
+    decoder_pred_head: mlp
+    full_attn_inds: null
+    gate_full_attn_layers: true
+    gating_type: sigmoid_g1
+    gr00t_add_dit_lora: false
+    gr00t_add_full_attn: true
+    gr00t_add_lora_to_vision_projector: false
+    gr00t_compute_dtype: bfloat16
+    gr00t_full_attn_idx:
+    - 3
+    - 7
+    - 11
+    - 15
+    gr00t_pretrained_model: true
+    gr00t_tune_diffusion_model: true
+    gr00t_tune_llm: false
+    gr00t_tune_projector: true
+    gr00t_tune_vision_projector: false
+    gr00t_tune_visual: false
+    hist_attn_latent_len: 1
+    hist_blind_gating: false
+    hist_feature_gating: false
+    interpolation_factor: null
+    llama_ckpt_dir: /home/rutavms/.llama/checkpoints/Llama3.2-1B-Instruct
+    load_llama: true
+    lora_layer_idxs: null
+    lora_rank: 4
+    loss_w_action: 1.0
+    model_version: v2
+    multikv_attn_pool: false
+    no_prompt_loss: true
+    num_inference_diffusion_steps: null
+    num_train_diffusion_steps: null
+    phase: pretrain
+    pool_block_inds: null
+    position_language_embedding: start
+    pretrained_path: null
+    prop_act_norm: null
+    remove_action: false
+    remove_proprio: false
+    scale_loss: 1.0
+    scratch_llama_config: config/model_config/libero_1x.json
+    separate_camera_adapter: true
+    skip_obs_pooling: true
+    step_weight: 1.0
+    tt_window_obs: false
+    use_block_attention: true
+    use_linear_attention: false
+    use_lstm: false
+    use_mamba: false
+    use_reshape_block_attn: false
+    use_transformer_xl: false
+    use_yarn: null
+    vision_projection_layer: false
+    xl_chunk_factor: 32
+    xl_mem_len: 32
+  vision_encoder_cfg: !!python/object:icrt.util.args.VisionEncoderConfig
+    bbox_vision_encoder_cfg: config/model_config/bbox_vit_128_16.json
+    freeze_all: false
+    vision_encoder: /storage/nfs/rshah/datasets/robocasa/datasets/crossmae_rtx/cross-mae-rtx-vitb.pth
+    vision_lora: false
+    vision_lora_rank: 8
+    vision_nonpretrained: false
+    vision_unfreeze_all: false
+    vision_unfreeze_last_n: 0
+optimizer_cfg: !!python/object:icrt.util.args.OptimizerConfig
+  blr: 0.001
+  lr: 0.0005
+  min_lr: 0.0
+  warmup_epochs: 2.0
+  weight_decay: 0.01
+shared_cfg: !!python/object:icrt.util.args.SharedConfig
+  batch_size: 2
+  bbox_noise: 0.015625
+  block_finetune: ''
+  downsample_obs: 8
+  drop_bbox_prob: 0.0
+  dropping_prompt_rnd: false
+  enable_gradient_checkpointing: false
+  gr00t_embodiment_str: single_panda_omron
+  gr00t_metadata_path: metadata.json
+  gr00t_model_path: nvidia/GR00T-N1.5-3B
+  gripper_img: true
+  has_base_action: true
+  hist_seq_length: 2048
+  is_bimanual: false
+  k_ptp: 0
+  max_bbox_per_img: 15
+  no_img: false
+  no_img_low_dim: 42
+  normalize_io: false
+  num_cameras: 2
+  num_pred_steps: 32
+  num_stages: 1
+  only_first_obs: false
+  pad_to_max_length: true
+  resume: null
+  rot_6d: false
+  rot_euler: false
+  save_every: 10
+  seed: 2
+  seq_length: 2048
+  split_epoch: 1
+  start_epoch: 0
+  state_center_coords: false
+  tokenizer_path: physical-intelligence/fast
+  use_bboxes: false
+  use_delta_action: false
+  use_fast_tokenizer: false
+  use_gr00t: false
+  use_history: false
+  use_language_conditioning: true
+  use_toy_vision_dataset: false
+  use_vqvae_tokenizer: false
+train: true
+trainer_cfg: !!python/object:icrt.util.args.TrainerConfig
+  accum_iter: 32
+  break_after_n_epochs: 101
+  compile_model: true
+  epochs: 200
+  num_workers: 72
+  pin_memory: true
+  val_every: 10
+  wandb_project: icrt_reruns
+  wandb_watch: true
+world_size: 1

run_id.txt

@@ -0,0 +1 @@
+j6hocd88

task_robocasa_mem_four.json

@@ -0,0 +1,27 @@
+{
+    "dataset_path": [
+        "memory/MemPutKBreadInMicrowave/2025-07-25-10-40-55/demo_im128_notp.hdf5",
+        "memory/MemPutKBreadInMicrowave/2025-07-25-13-25-05/demo_im128_notp.hdf5",
+        "memory/MemPutKBreadInMicrowave/2025-07-25-14-50-59/demo_im128_notp.hdf5",
+        "memory/MemFruitInSinkLeftFar/2025-07-13-15-44-32/demo_im128_notp.hdf5",
+        "memory/MemFruitInSinkRightFar/2025-07-15-15-07-57/demo_im128_notp.hdf5",
+        "memory/MemHeatPot/2025-07-24-22-26-20/demo_im128_notp.hdf5",
+        "memory/MemHeatPot/2025-07-25-15-56-53/demo_im128_notp.hdf5",
+        "memory/MemWashAndReturnLeft/2025-07-25-00-12-14/demo_im128_notp.hdf5",
+        "memory/MemWashAndReturnRight/2025-07-25-00-45-53/demo_im128_notp.hdf5"
+    ],
+    "image_keys": [
+        "obs/robot0_agentview_center_image",
+        "obs/robot0_eye_in_hand_image"
+    ],
+    "max_traj_len": 1024,
+    "low_dim_keys": [],
+    "group_method": "dinov2",
+    "proprio_keys": [
+        "obs/robot0_joint_pos_cos",
+        "obs/robot0_gripper_qpos"
+    ],
+    "action_keys": [
+        "actions"
+    ]
+}