Add model weights

checkpoints/post_ckpt.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:33d897704f1d358b0c28dd244684587f48fc5cae72eb626c6ba471c69cf88958
+size 33260216370

config.json

@@ -0,0 +1,60 @@
+{
+  "action_dim": 7,
+  "action_tokenizer_exist": false,
+  "class_dropout_prob": 0.0,
+  "data_root_dir": "/media/liuzhuoyang/data/rlbench/rlds/12tasks_selected_keyframe_pointcloud_0814",
+  "future_action_window_size": 0,
+  "hf_token": ".hf_token",
+  "image_aug": false,
+  "is_resume": false,
+  "llm_vision_layers": 8,
+  "load_all_data_for_training": true,
+  "past_action_window_size": 0,
+  "pretrained_checkpoint": "/media/liuzhuoyang/new_vla/Rec_Diff_beta/exp/exp_12tasks_selected_keyframe_pointcloud_0814_Pretrainrtx0818e1_FreezeVistrue_Window0_Difftrue_Recfalse_Contrastive_Vislayer8_1024_0403_0820/checkpoints/step-005415-epoch-300-loss=1.4414.pt",
+  "recon_image": true,
+  "recon_pointcloud": true,
+  "repeated_diffusion_steps": 4,
+  "resume_epoch": null,
+  "resume_step": null,
+  "run_id": "exp_12tasks_selected_keyframe_pointcloud_0814_Pretraindiff300_FreezeVistrue_Window0_Difftrue_PCtrue_Rectrue_Contrastive_Vislayer8_1024_0403_0822",
+  "run_id_note": null,
+  "run_root_dir": "/media/liuzhuoyang/new_vla/Rec_Diff_beta/exp",
+  "save_interval": 100,
+  "seed": 42,
+  "trackers": [
+    "jsonl",
+    "wandb"
+  ],
+  "use_contrastive": true,
+  "use_diff": true,
+  "use_ema": false,
+  "use_pointcloud": true,
+  "use_reconstruction": true,
+  "use_roi": false,
+  "vla": {
+    "base_vlm": "prism-dinosiglip-224px+7b",
+    "data_mix": "rlbench",
+    "enable_gradient_checkpointing": true,
+    "enable_mixed_precision_training": true,
+    "epochs": 300,
+    "expected_world_size": 32,
+    "freeze_llm_backbone": false,
+    "freeze_vision_tower": true,
+    "global_batch_size": 256,
+    "learning_rate": 2e-05,
+    "lr_scheduler_type": "constant",
+    "max_grad_norm": 1.0,
+    "max_steps": null,
+    "per_device_batch_size": 8,
+    "reduce_in_full_precision": true,
+    "shuffle_buffer_size": 10000,
+    "train_strategy": "fsdp-full-shard",
+    "type": "prism-dinosiglip-224px+oxe+diffusion",
+    "unfreeze_last_llm_layer": false,
+    "vla_id": "prism-dinosiglip-224px+oxe+diffusion",
+    "warmup_ratio": 0.0,
+    "weight_decay": 0.0
+  },
+  "wandb_entity": "liumail2023-peking-university",
+  "wandb_project": "one_model_vla_sft"
+}

config.yaml

@@ -0,0 +1,56 @@
+action_dim: 7
+action_tokenizer_exist: false
+class_dropout_prob: 0.0
+data_root_dir: /media/liuzhuoyang/data/rlbench/rlds/12tasks_selected_keyframe_pointcloud_0814
+future_action_window_size: 0
+hf_token: .hf_token
+image_aug: false
+is_resume: false
+llm_vision_layers: 8
+load_all_data_for_training: true
+past_action_window_size: 0
+pretrained_checkpoint: /media/liuzhuoyang/new_vla/Rec_Diff_beta/exp/exp_12tasks_selected_keyframe_pointcloud_0814_Pretrainrtx0818e1_FreezeVistrue_Window0_Difftrue_Recfalse_Contrastive_Vislayer8_1024_0403_0820/checkpoints/step-005415-epoch-300-loss=1.4414.pt
+recon_image: true
+recon_pointcloud: true
+repeated_diffusion_steps: 4
+resume_epoch: null
+resume_step: null
+run_id: exp_12tasks_selected_keyframe_pointcloud_0814_Pretraindiff300_FreezeVistrue_Window0_Difftrue_PCtrue_Rectrue_Contrastive_Vislayer8_1024_0403_0822
+run_id_note: null
+run_root_dir: /media/liuzhuoyang/new_vla/Rec_Diff_beta/exp
+save_interval: 100
+seed: 42
+trackers:
+- jsonl
+- wandb
+use_contrastive: true
+use_diff: true
+use_ema: false
+use_pointcloud: true
+use_reconstruction: true
+use_roi: false
+vla:
+  base_vlm: prism-dinosiglip-224px+7b
+  data_mix: rlbench
+  enable_gradient_checkpointing: true
+  enable_mixed_precision_training: true
+  epochs: 300
+  expected_world_size: 32
+  freeze_llm_backbone: false
+  freeze_vision_tower: true
+  global_batch_size: 256
+  learning_rate: 2.0e-05
+  lr_scheduler_type: constant
+  max_grad_norm: 1.0
+  max_steps: null
+  per_device_batch_size: 8
+  reduce_in_full_precision: true
+  shuffle_buffer_size: 10000
+  train_strategy: fsdp-full-shard
+  type: prism-dinosiglip-224px+oxe+diffusion
+  unfreeze_last_llm_layer: false
+  vla_id: prism-dinosiglip-224px+oxe+diffusion
+  warmup_ratio: 0.0
+  weight_decay: 0.0
+wandb_entity: liumail2023-peking-university
+wandb_project: one_model_vla_sft

dataset_statistics.json

@@ -0,0 +1,136 @@
+{
+  "rlbench": {
+    "action": {
+      "mean": [
+        -0.010306132026016712,
+        -0.0055046346969902515,
+        -0.11197133362293243,
+        -0.0019046454690396786,
+        -0.15908075869083405,
+        0.1675104796886444,
+        0.39728033542633057
+      ],
+      "std": [
+        0.08434142917394638,
+        0.14875024557113647,
+        0.1996437907218933,
+        2.636392831802368,
+        0.6083221435546875,
+        1.8673105239868164,
+        0.48934081196784973
+      ],
+      "max": [
+        0.24629077315330505,
+        0.4493105411529541,
+        0.2484073042869568,
+        3.141592264175415,
+        1.570451259613037,
+        3.141460657119751,
+        1.0
+      ],
+      "min": [
+        -0.4184693694114685,
+        -0.4492553770542145,
+        -0.6604752540588379,
+        -3.1415884494781494,
+        -1.5704164505004883,
+        -3.1412742137908936,
+        0.0
+      ],
+      "q01": [
+        -0.2869752472639084,
+        -0.4469184121489525,
+        -0.6007548034191131,
+        -3.141468975543976,
+        -1.5699386596679688,
+        -3.1044607901573182,
+        0.0
+      ],
+      "q99": [
+        0.14981620118021965,
+        0.44710047245025636,
+        0.24729809522628785,
+        3.141501989364624,
+        1.5696296310424804,
+        3.110466537475586,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        false
+      ]
+    },
+    "proprio": {
+      "mean": [
+        0.22678296267986298,
+        0.011608771048486233,
+        1.124215841293335,
+        -0.7795729637145996,
+        -0.1190786063671112,
+        -0.1414777636528015,
+        0.5020920634269714
+      ],
+      "std": [
+        0.09569656103849411,
+        0.17490601539611816,
+        0.2254764586687088,
+        2.6655027866363525,
+        0.5839682221412659,
+        2.2652320861816406,
+        0.49999505281448364
+      ],
+      "max": [
+        0.5282712578773499,
+        0.42229560017585754,
+        1.4721031188964844,
+        3.141592502593994,
+        1.568900227546692,
+        3.1415927410125732,
+        1.0
+      ],
+      "min": [
+        -0.14719118177890778,
+        -0.43890005350112915,
+        0.7954218983650208,
+        -3.1415927410125732,
+        -1.5704164505004883,
+        -3.1415927410125732,
+        0.0
+      ],
+      "q01": [
+        -0.05181949622929096,
+        -0.34075534880161285,
+        0.7963145178556442,
+        -3.1415915489196777,
+        -1.5699386596679688,
+        -3.141591787338257,
+        0.0
+      ],
+      "q99": [
+        0.44086600691080097,
+        0.3493700134754181,
+        1.4720294535160066,
+        3.1415910720825195,
+        1.5673939752578736,
+        3.141592025756836,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        false
+      ]
+    },
+    "num_transitions": 4780,
+    "num_trajectories": 1200
+  }
+}