Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,10 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+results_step-240000-epoch-05-loss=0.0910_1/log/log3.log filter=lfs diff=lfs merge=lfs -text
+results_step-240000-epoch-05-loss=0.0910_1/log/log7.log filter=lfs diff=lfs merge=lfs -text
+results_step-240000-epoch-05-loss=0.0910_2/log/log3.log filter=lfs diff=lfs merge=lfs -text
+results_step-240000-epoch-05-loss=0.0910_2/log/log7.log filter=lfs diff=lfs merge=lfs -text
+results_step-240000-epoch-05-loss=0.0910_3/log/log3.log filter=lfs diff=lfs merge=lfs -text
+results_step-240000-epoch-05-loss=0.0910_3/log/log7.log filter=lfs diff=lfs merge=lfs -text
+sys12_meta_query_action_only_sync_pretraining_v2_query_64_mlp_lora_reimplement_transformer_4_50_single_node_bs128_2--image_augstage1.jsonl filter=lfs diff=lfs merge=lfs -text

checkpoints/step-240000-epoch-05-loss=0.0910.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9248ea07056100a0b3780dcd71b5c99d9efdf3c63635a7295a1ecad2a1d5d25e
+size 11398385050

config.json

@@ -0,0 +1,57 @@
+{
+  "action_dim": 7,
+  "data_root_dir": "/mnt/petrelfs/yangshuai1/rep/InstructVLA_official/cache",
+  "debug": false,
+  "disable_instruction": false,
+  "fix_system1": false,
+  "future_action_window_size": 15,
+  "hf_token": ".hf_token",
+  "image_aug": true,
+  "is_resume": true,
+  "load_all_data_for_training": true,
+  "num_of_meta_query": 64,
+  "past_action_window_size": 0,
+  "pretrained_checkpoint": "/mnt/petrelfs/yangshuai1/rep/InstructVLA_official/outputs/code_reimp/sys12_meta_query_action_only_sync_pretraining_v2_query_64_mlp_lora_reimplement_transformer_4_50_single_node_bs128_2--image_augstage1/checkpoints/step-180000-epoch-03-loss=0.1214.pt",
+  "repeated_diffusion_steps": 4,
+  "resume_epoch": 3,
+  "resume_step": 180000,
+  "run_id": "sys12_meta_query_action_only_sync_pretraining_v2_query_64_mlp_lora_reimplement_transformer_4_50_single_node_bs128_2--image_augstage1",
+  "run_id_note": null,
+  "run_root_dir": "outputs/code_reimp",
+  "save_interval": 20000,
+  "seed": 42,
+  "stage": "stage1",
+  "trackers": [
+    "jsonl",
+    "wandb"
+  ],
+  "use_mm": false,
+  "vla": {
+    "action_tokenizer": "extra_action_tokenizer",
+    "base_vlm": "ckpt/Eagle2-2B",
+    "data_mix": "bridge_rt_1",
+    "enable_gradient_checkpointing": true,
+    "enable_mixed_precision_training": true,
+    "epochs": 100,
+    "expected_world_size": 8,
+    "freeze_llm_backbone": false,
+    "freeze_vision_backbone": false,
+    "global_batch_size": 128,
+    "learning_rate": 5e-05,
+    "lr_scheduler_type": "constant",
+    "max_grad_norm": 1.0,
+    "max_steps": null,
+    "per_device_batch_size": 16,
+    "reduce_in_full_precision": true,
+    "shuffle_buffer_size": 250000,
+    "train_strategy": "fsdp-full-shard",
+    "type": "prism-qwen25-dinosiglip-224px+0_5b",
+    "unfreeze_last_llm_layer": false,
+    "vla_id": "prism-qwen25-dinosiglip-224px+0_5b",
+    "warmup_ratio": 0.0,
+    "weight_decay": 0.0
+  },
+  "wandb_entity": "shuaiyang2003",
+  "wandb_project": "dual_sys_code_clean",
+  "with_pointing": false
+}

config.yaml

@@ -0,0 +1,53 @@
+action_dim: 7
+data_root_dir: /mnt/petrelfs/yangshuai1/rep/InstructVLA_official/cache
+debug: false
+disable_instruction: false
+fix_system1: false
+future_action_window_size: 15
+hf_token: .hf_token
+image_aug: true
+is_resume: true
+load_all_data_for_training: true
+num_of_meta_query: 64
+past_action_window_size: 0
+pretrained_checkpoint: /mnt/petrelfs/yangshuai1/rep/InstructVLA_official/outputs/code_reimp/sys12_meta_query_action_only_sync_pretraining_v2_query_64_mlp_lora_reimplement_transformer_4_50_single_node_bs128_2--image_augstage1/checkpoints/step-180000-epoch-03-loss=0.1214.pt
+repeated_diffusion_steps: 4
+resume_epoch: 3
+resume_step: 180000
+run_id: sys12_meta_query_action_only_sync_pretraining_v2_query_64_mlp_lora_reimplement_transformer_4_50_single_node_bs128_2--image_augstage1
+run_id_note: null
+run_root_dir: outputs/code_reimp
+save_interval: 20000
+seed: 42
+stage: stage1
+trackers:
+- jsonl
+- wandb
+use_mm: false
+vla:
+  action_tokenizer: extra_action_tokenizer
+  base_vlm: ckpt/Eagle2-2B
+  data_mix: bridge_rt_1
+  enable_gradient_checkpointing: true
+  enable_mixed_precision_training: true
+  epochs: 100
+  expected_world_size: 8
+  freeze_llm_backbone: false
+  freeze_vision_backbone: false
+  global_batch_size: 128
+  learning_rate: 5.0e-05
+  lr_scheduler_type: constant
+  max_grad_norm: 1.0
+  max_steps: null
+  per_device_batch_size: 16
+  reduce_in_full_precision: true
+  shuffle_buffer_size: 250000
+  train_strategy: fsdp-full-shard
+  type: prism-qwen25-dinosiglip-224px+0_5b
+  unfreeze_last_llm_layer: false
+  vla_id: prism-qwen25-dinosiglip-224px+0_5b
+  warmup_ratio: 0.0
+  weight_decay: 0.0
+wandb_entity: shuaiyang2003
+wandb_project: dual_sys_code_clean
+with_pointing: false

dataset_statistics.json

@@ -0,0 +1,264 @@
+{
+  "bridge_dataset": {
+    "action": {
+      "mean": [
+        0.0002334193413844332,
+        0.0001300490548601374,
+        -0.0001276246621273458,
+        -0.00015565502690151334,
+        -0.0004039333143737167,
+        0.0002355769247515127,
+        0.5764579772949219
+      ],
+      "std": [
+        0.009765916503965855,
+        0.013689138926565647,
+        0.012667354196310043,
+        0.02853417582809925,
+        0.0306379534304142,
+        0.07691461592912674,
+        0.49737000465393066
+      ],
+      "max": [
+        0.41691166162490845,
+        0.25864794850349426,
+        0.21218234300613403,
+        3.122201919555664,
+        1.8618112802505493,
+        6.280478477478027,
+        1.0
+      ],
+      "min": [
+        -0.4007510244846344,
+        -0.13874775171279907,
+        -0.22553899884223938,
+        -3.2010786533355713,
+        -1.8618112802505493,
+        -6.279075622558594,
+        0.0
+      ],
+      "q01": [
+        -0.02872725307941437,
+        -0.04170349963009357,
+        -0.026093858778476715,
+        -0.08092105075716972,
+        -0.09288699507713317,
+        -0.20718276381492615,
+        0.0
+      ],
+      "q99": [
+        0.028309678435325586,
+        0.040855254605412394,
+        0.040161586627364146,
+        0.08192047759890528,
+        0.07792850524187081,
+        0.20382574498653397,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        false
+      ]
+    },
+    "proprio": {
+      "mean": [
+        0.3094092905521393,
+        0.030575834214687347,
+        0.06454048305749893,
+        0.006824127864092588,
+        -0.07762698084115982,
+        0.10757855325937271,
+        0.0,
+        0.7084035277366638
+      ],
+      "std": [
+        0.060528464615345,
+        0.09188621491193771,
+        0.05159851908683777,
+        0.13182717561721802,
+        0.17031119763851166,
+        0.5767312049865723,
+        0.0,
+        0.3519800305366516
+      ],
+      "max": [
+        0.5862360596656799,
+        0.4034728705883026,
+        0.36494991183280945,
+        1.514088749885559,
+        1.570796251296997,
+        3.1415255069732666,
+        0.0,
+        1.1154625415802002
+      ],
+      "min": [
+        -0.04167502000927925,
+        -0.3945816159248352,
+        -0.15537554025650024,
+        -3.141592502593994,
+        -1.4992541074752808,
+        -3.14153790473938,
+        0.0,
+        0.04637829214334488
+      ],
+      "q01": [
+        0.17111587673425674,
+        -0.16998695254325866,
+        -0.05544630073010921,
+        -0.366876106262207,
+        -0.5443069756031036,
+        -1.3536006283760071,
+        0.0,
+        0.052190229296684265
+      ],
+      "q99": [
+        0.45320980012416834,
+        0.23518154799938193,
+        0.1951873075962065,
+        0.3806115746498103,
+        0.2789784955978382,
+        1.8410426235198971,
+        0.0,
+        1.0105689764022827
+      ]
+    },
+    "num_transitions": 2135463,
+    "num_trajectories": 60064
+  },
+  "fractal20220817_data": {
+    "action": {
+      "mean": [
+        0.006987507455050945,
+        0.0062658516690135,
+        -0.012625164352357388,
+        0.04333285614848137,
+        -0.005756272468715906,
+        0.0009130411199294031,
+        0.5354204773902893
+      ],
+      "std": [
+        0.06921109557151794,
+        0.05970890820026398,
+        0.0735311210155487,
+        0.15610598027706146,
+        0.1316441297531128,
+        0.1459376960992813,
+        0.49711623787879944
+      ],
+      "max": [
+        2.9984593391418457,
+        22.09052848815918,
+        2.7507524490356445,
+        1.570636510848999,
+        1.5321086645126343,
+        1.5691522359848022,
+        1.0
+      ],
+      "min": [
+        -2.0204520225524902,
+        -5.497899532318115,
+        -2.031663417816162,
+        -1.569917917251587,
+        -1.569892168045044,
+        -1.570419430732727,
+        0.0
+      ],
+      "q01": [
+        -0.22453527510166169,
+        -0.14820013284683228,
+        -0.231589707583189,
+        -0.3517994859814644,
+        -0.4193011274933815,
+        -0.43643461108207704,
+        0.0
+      ],
+      "q99": [
+        0.17824687153100965,
+        0.14938379630446405,
+        0.21842354819178575,
+        0.5892666035890578,
+        0.35272657424211445,
+        0.44796681255102094,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        false
+      ]
+    },
+    "proprio": {
+      "mean": [
+        0.5598955750465393,
+        -0.08333974331617355,
+        0.7771074175834656,
+        -0.2480376660823822,
+        0.4951733648777008,
+        0.09266174584627151,
+        0.20975501835346222,
+        0.42613404989242554
+      ],
+      "std": [
+        0.12432783097028732,
+        0.11558859050273895,
+        0.24595840275287628,
+        0.5126973986625671,
+        0.5218129754066467,
+        0.16630405187606812,
+        0.2754833996295929,
+        0.45545175671577454
+      ],
+      "max": [
+        1.0534898042678833,
+        0.48018959164619446,
+        1.6896663904190063,
+        0.9999993443489075,
+        0.9999874830245972,
+        0.9554369449615479,
+        0.9914546012878418,
+        1.0
+      ],
+      "min": [
+        -0.4436439275741577,
+        -0.9970501065254211,
+        -0.006579156965017319,
+        -0.8643477559089661,
+        -0.7079970240592957,
+        -0.7688722014427185,
+        -0.4999994933605194,
+        0.0
+      ],
+      "q01": [
+        0.32481380939483645,
+        -0.28334290891885755,
+        0.14107070609927178,
+        -0.686474204659462,
+        -0.6808923494815826,
+        -0.36045596331357954,
+        -0.454380963742733,
+        0.0
+      ],
+      "q99": [
+        0.8750156319141384,
+        0.21247054174542404,
+        1.0727112340927123,
+        0.9377871316671368,
+        0.9563051050901409,
+        0.45990042358636823,
+        0.7216041100025177,
+        1.0
+      ]
+    },
+    "num_transitions": 3786400,
+    "num_trajectories": 87212
+  }
+}

results_step-240000-epoch-05-loss=0.0910_1/log/final_result.log

@@ -0,0 +1,114 @@
+***Pick coke can results***
+--------------------
+horizontal sim variant avg success {'CogACT-Base': 0.9644444444444444}
+vertical sim variant avg success {'CogACT-Base': 0.9600000000000001}
+standing sim variant avg success {'CogACT-Base': 0.9022222222222224}
+avg_orientation_sim_variant_results [0.9422222222222224]
+--------------------
+Orientation horizontal, ckpt CogACT-Base all robot arm visual matching success: [1.0, 0.88, 0.88, 0.92]
+Orientation vertical, ckpt CogACT-Base all robot arm visual matching success: [0.92, 0.88, 0.96, 0.96]
+Orientation standing, ckpt CogACT-Base all robot arm visual matching success: [0.84, 0.84, 0.84, 0.84]
+horizontal visual matching sim success {'CogACT-Base': 0.9199999999999999}
+vertical visual matching sim success {'CogACT-Base': 0.9299999999999999}
+standing visual matching sim success {'CogACT-Base': 0.84}
+avg_orientation_sim_visual_matching_results [0.8966666666666666]
+********************
+
+
+
+***Move Near results***
+--------------------
+sim variant avg success {'CogACT-Base': 0.7645833333333333}
+--------------------
+Ckpt CogACT-Base all robot arm visual matching success: [0.7166666666666667, 0.6166666666666667, 0.6833333333333333, 0.6666666666666666]
+sim visual matching success {'CogACT-Base': 0.6708333333333333}
+********************
+
+
+
+***Drawer results***
+--------------------
+open sim variant avg success {'CogACT-Base': 0.5873015873015872}
+close sim variant avg success {'CogACT-Base': 0.6560846560846562}
+avg_sim_variant_results [0.6216931216931216]
+--------------------
+Drawer task open, ckpt CogACT-Base all robot arm visual matching success: [0.5555555555555556, 0.5277777777777778, 0.41666666666666663, 0.5185185185185185]
+Drawer task close, ckpt CogACT-Base all robot arm visual matching success: [0.75, 0.75, 0.5555555555555556, 0.7407407407407408]
+open visual matching sim success {'CogACT-Base': 0.49999999999999994}
+close visual matching sim success {'CogACT-Base': 0.6851851851851852}
+avg_sim_visual_matching_results [0.5925925925925926]
+********************
+
+
+
+***Drawer results***
+--------------------
+put_apple_into_top_drawer sim variant avg success {'CogACT-Base': 0.3968253968253968}
+avg_sim_variant_results [0.3968253968253968]
+--------------------
+Drawer task put_apple_into_top_drawer, ckpt CogACT-Base all robot arm visual matching success: [0.32407407407407407, 0.3209876543209876, 0.31481481481481477, 0.3333333333333333]
+put_apple_into_top_drawer visual matching sim success {'CogACT-Base': 0.32407407407407407}
+avg_sim_visual_matching_results [0.32407407407407407]
+********************
+
+
+
+***Bridge Put On Env results***
+********** Results for put_spoon_on_tablecloth **********
+sim visual matching partial success {'CogACT-Base': 0.7916666666666666}
+sim visual matching success {'CogACT-Base': 0.5}
+********************
+
+
+
+********** Results for put_carrot_on_plate **********
+sim visual matching partial success {'CogACT-Base': 0.5416666666666666}
+sim visual matching success {'CogACT-Base': 0.5}
+********************
+
+
+
+********** Results for stack_green_block_on_yellow_block **********
+sim visual matching partial success {'CogACT-Base': 0.4166666666666667}
+sim visual matching success {'CogACT-Base': 0.08333333333333333}
+********************
+
+
+
+********** Results for put_eggplant_in_basket **********
+sim visual matching partial success {'CogACT-Base': 1.0}
+sim visual matching success {'CogACT-Base': 1.0}
+********************
+
+
+
+
+google_var:
+94.2
+94.2
+76.5
+76.5
+62.2
+62.2
+39.7
+39.7
+
+google_matching
+89.7
+89.7
+67.1
+67.1
+59.3
+59.3
+32.4
+32.4
+
+widowx_matching
+50.0
+50.0
+50.0
+50.0
+8.3
+8.3
+100.0
+100.0

results_step-240000-epoch-05-loss=0.0910_1/log/log3.log

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:435efdea2aacce6285c3f1e125b70aa1aa75efb01ae333f954aa652675db564c
+size 13953553

results_step-240000-epoch-05-loss=0.0910_1/log/log7.log

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6179126e5e0e2b90cd25760d6492ac99ac0a12947e512390b6c9988e5ab9392
+size 26761819

results_step-240000-epoch-05-loss=0.0910_2/log/final_result.log

@@ -0,0 +1,114 @@
+***Pick coke can results***
+--------------------
+horizontal sim variant avg success {'CogACT-Base': 0.9555555555555555}
+vertical sim variant avg success {'CogACT-Base': 0.9600000000000001}
+standing sim variant avg success {'CogACT-Base': 0.9022222222222224}
+avg_orientation_sim_variant_results [0.9392592592592593]
+--------------------
+Orientation horizontal, ckpt CogACT-Base all robot arm visual matching success: [0.92, 0.88, 0.88, 0.96]
+Orientation vertical, ckpt CogACT-Base all robot arm visual matching success: [0.88, 0.84, 0.96, 0.88]
+Orientation standing, ckpt CogACT-Base all robot arm visual matching success: [0.92, 0.88, 0.84, 0.92]
+horizontal visual matching sim success {'CogACT-Base': 0.91}
+vertical visual matching sim success {'CogACT-Base': 0.8899999999999999}
+standing visual matching sim success {'CogACT-Base': 0.89}
+avg_orientation_sim_visual_matching_results [0.8966666666666666]
+********************
+
+
+
+***Move Near results***
+--------------------
+sim variant avg success {'CogACT-Base': 0.7666666666666667}
+--------------------
+Ckpt CogACT-Base all robot arm visual matching success: [0.6833333333333333, 0.6833333333333333, 0.7, 0.7166666666666667]
+sim visual matching success {'CogACT-Base': 0.6958333333333333}
+********************
+
+
+
+***Drawer results***
+--------------------
+open sim variant avg success {'CogACT-Base': 0.5396825396825398}
+close sim variant avg success {'CogACT-Base': 0.6931216931216931}
+avg_sim_variant_results [0.6164021164021165]
+--------------------
+Drawer task open, ckpt CogACT-Base all robot arm visual matching success: [0.611111111111111, 0.5833333333333333, 0.41666666666666663, 0.5555555555555555]
+Drawer task close, ckpt CogACT-Base all robot arm visual matching success: [0.861111111111111, 0.7222222222222221, 0.6666666666666665, 0.8148148148148148]
+open visual matching sim success {'CogACT-Base': 0.537037037037037}
+close visual matching sim success {'CogACT-Base': 0.75}
+avg_sim_visual_matching_results [0.6435185185185185]
+********************
+
+
+
+***Drawer results***
+--------------------
+put_apple_into_top_drawer sim variant avg success {'CogACT-Base': 0.3492063492063492}
+avg_sim_variant_results [0.3492063492063492]
+--------------------
+Drawer task put_apple_into_top_drawer, ckpt CogACT-Base all robot arm visual matching success: [0.2962962962962963, 0.30864197530864196, 0.31481481481481477, 0.2962962962962963]
+put_apple_into_top_drawer visual matching sim success {'CogACT-Base': 0.2962962962962963}
+avg_sim_visual_matching_results [0.2962962962962963]
+********************
+
+
+
+***Bridge Put On Env results***
+********** Results for put_spoon_on_tablecloth **********
+sim visual matching partial success {'CogACT-Base': 0.9583333333333334}
+sim visual matching success {'CogACT-Base': 0.6666666666666666}
+********************
+
+
+
+********** Results for put_carrot_on_plate **********
+sim visual matching partial success {'CogACT-Base': 0.6666666666666666}
+sim visual matching success {'CogACT-Base': 0.4583333333333333}
+********************
+
+
+
+********** Results for stack_green_block_on_yellow_block **********
+sim visual matching partial success {'CogACT-Base': 0.5416666666666666}
+sim visual matching success {'CogACT-Base': 0.041666666666666664}
+********************
+
+
+
+********** Results for put_eggplant_in_basket **********
+sim visual matching partial success {'CogACT-Base': 0.9583333333333334}
+sim visual matching success {'CogACT-Base': 0.9166666666666666}
+********************
+
+
+
+
+google_var:
+93.9
+93.9
+76.7
+76.7
+61.6
+61.6
+34.9
+34.9
+
+google_matching
+89.7
+89.7
+69.6
+69.6
+64.4
+64.4
+29.6
+29.6
+
+widowx_matching
+66.7
+66.7
+45.8
+45.8
+4.2
+4.2
+91.7
+91.7

results_step-240000-epoch-05-loss=0.0910_2/log/log3.log

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e51dc8d0522974bf2a585fc8308dff74276d7282564e6955d1f9fe8fa8d7d491
+size 13952635

results_step-240000-epoch-05-loss=0.0910_2/log/log7.log

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:871b30c8e6d5a0777f216c0df35f7eb06f08d4da9b70b8f2d70b46ecaf72968d
+size 26758058

results_step-240000-epoch-05-loss=0.0910_3/log/final_result.log

@@ -0,0 +1,114 @@
+***Pick coke can results***
+--------------------
+horizontal sim variant avg success {'CogACT-Base': 0.9333333333333333}
+vertical sim variant avg success {'CogACT-Base': 0.9688888888888888}
+standing sim variant avg success {'CogACT-Base': 0.9111111111111112}
+avg_orientation_sim_variant_results [0.9377777777777778]
+--------------------
+Orientation horizontal, ckpt CogACT-Base all robot arm visual matching success: [0.96, 0.72, 0.76, 0.88]
+Orientation vertical, ckpt CogACT-Base all robot arm visual matching success: [1.0, 0.96, 0.88, 1.0]
+Orientation standing, ckpt CogACT-Base all robot arm visual matching success: [0.88, 0.8, 0.76, 0.8]
+horizontal visual matching sim success {'CogACT-Base': 0.83}
+vertical visual matching sim success {'CogACT-Base': 0.96}
+standing visual matching sim success {'CogACT-Base': 0.81}
+avg_orientation_sim_visual_matching_results [0.8666666666666667]
+********************
+
+
+
+***Move Near results***
+--------------------
+sim variant avg success {'CogACT-Base': 0.7749999999999999}
+--------------------
+Ckpt CogACT-Base all robot arm visual matching success: [0.7166666666666667, 0.5666666666666667, 0.6166666666666667, 0.7166666666666667]
+sim visual matching success {'CogACT-Base': 0.6541666666666667}
+********************
+
+
+
+***Drawer results***
+--------------------
+open sim variant avg success {'CogACT-Base': 0.5873015873015873}
+close sim variant avg success {'CogACT-Base': 0.7037037037037037}
+avg_sim_variant_results [0.6455026455026456]
+--------------------
+Drawer task open, ckpt CogACT-Base all robot arm visual matching success: [0.5555555555555556, 0.5555555555555556, 0.5277777777777778, 0.48148148148148145]
+Drawer task close, ckpt CogACT-Base all robot arm visual matching success: [0.7222222222222222, 0.8333333333333333, 0.49999999999999994, 0.7037037037037037]
+open visual matching sim success {'CogACT-Base': 0.5462962962962963}
+close visual matching sim success {'CogACT-Base': 0.6851851851851851}
+avg_sim_visual_matching_results [0.6157407407407407]
+********************
+
+
+
+***Drawer results***
+--------------------
+put_apple_into_top_drawer sim variant avg success {'CogACT-Base': 0.43386243386243384}
+avg_sim_variant_results [0.43386243386243384]
+--------------------
+Drawer task put_apple_into_top_drawer, ckpt CogACT-Base all robot arm visual matching success: [0.33333333333333337, 0.32098765432098764, 0.2962962962962963, 0.2222222222222222]
+put_apple_into_top_drawer visual matching sim success {'CogACT-Base': 0.33333333333333337}
+avg_sim_visual_matching_results [0.33333333333333337]
+********************
+
+
+
+***Bridge Put On Env results***
+********** Results for put_spoon_on_tablecloth **********
+sim visual matching partial success {'CogACT-Base': 0.8333333333333334}
+sim visual matching success {'CogACT-Base': 0.7083333333333334}
+********************
+
+
+
+********** Results for put_carrot_on_plate **********
+sim visual matching partial success {'CogACT-Base': 0.625}
+sim visual matching success {'CogACT-Base': 0.5}
+********************
+
+
+
+********** Results for stack_green_block_on_yellow_block **********
+sim visual matching partial success {'CogACT-Base': 0.5416666666666666}
+sim visual matching success {'CogACT-Base': 0.125}
+********************
+
+
+
+********** Results for put_eggplant_in_basket **********
+sim visual matching partial success {'CogACT-Base': 0.9583333333333334}
+sim visual matching success {'CogACT-Base': 0.9583333333333334}
+********************
+
+
+
+
+google_var:
+93.8
+93.8
+77.5
+77.5
+64.6
+64.6
+43.4
+43.4
+
+google_matching
+86.7
+86.7
+65.4
+65.4
+61.6
+61.6
+33.3
+33.3
+
+widowx_matching
+70.8
+70.8
+50.0
+50.0
+12.5
+12.5
+95.8
+95.8

results_step-240000-epoch-05-loss=0.0910_3/log/log3.log

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4df2cba4960fc3755d2888cb080dbe926b15741ab218fe5fc7c9e0a94c8cbedb
+size 13953631

results_step-240000-epoch-05-loss=0.0910_3/log/log7.log

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d737720240837dd5ce73f8b79e0194c3d07ca1df7e9fa073dfadbe14f5d755ff
+size 26761982

run-metrics.jsonl

@@ -0,0 +1 @@
+{"hparams": {"action_dim": 7, "data_root_dir": "/mnt/petrelfs/yangshuai1/rep/InstructVLA_official/cache", "debug": false, "disable_instruction": false, "fix_system1": false, "future_action_window_size": 15, "hf_token": ".hf_token", "image_aug": true, "is_resume": true, "load_all_data_for_training": true, "num_of_meta_query": 64, "past_action_window_size": 0, "pretrained_checkpoint": "/mnt/petrelfs/yangshuai1/rep/InstructVLA_official/outputs/code_reimp/sys12_meta_query_action_only_sync_pretraining_v2_query_64_mlp_lora_reimplement_transformer_4_50_single_node_bs128_2--image_augstage1/checkpoints/step-180000-epoch-03-loss=0.1214.pt", "repeated_diffusion_steps": 4, "resume_epoch": 3, "resume_step": 180000, "run_id": "sys12_meta_query_action_only_sync_pretraining_v2_query_64_mlp_lora_reimplement_transformer_4_50_single_node_bs128_2--image_augstage1", "run_id_note": null, "run_root_dir": "outputs/code_reimp", "save_interval": 20000, "seed": 42, "stage": "stage1", "trackers": ["jsonl", "wandb"], "use_mm": false, "vla": {"action_tokenizer": "extra_action_tokenizer", "base_vlm": "ckpt/Eagle2-2B", "data_mix": "bridge_rt_1", "enable_gradient_checkpointing": true, "enable_mixed_precision_training": true, "epochs": 100, "expected_world_size": 8, "freeze_llm_backbone": false, "freeze_vision_backbone": false, "global_batch_size": 128, "learning_rate": 5e-05, "lr_scheduler_type": "constant", "max_grad_norm": 1.0, "max_steps": null, "per_device_batch_size": 16, "reduce_in_full_precision": true, "shuffle_buffer_size": 250000, "train_strategy": "fsdp-full-shard", "type": "prism-qwen25-dinosiglip-224px+0_5b", "unfreeze_last_llm_layer": false, "vla_id": "prism-qwen25-dinosiglip-224px+0_5b", "warmup_ratio": 0.0, "weight_decay": 0.0}, "wandb_entity": "shuaiyang2003", "wandb_project": "dual_sys_code_clean", "with_pointing": false}, "run_id": "sys12_meta_query_action_only_sync_pretraining_v2_query_64_mlp_lora_reimplement_transformer_4_50_single_node_bs128_2--image_augstage1"}

sys12_meta_query_action_only_sync_pretraining_v2_query_64_mlp_lora_reimplement_transformer_4_50_single_node_bs128_2--image_augstage1.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e1603b9ec92ce156f49e36c028bdb6cddd084e536c649ce416d0ae6f4bc6f533
+size 92516492