Upload depth v4 checkpoint 40000 configuration

depth/finetune/xmovepick-depth-v4-query-ffn-moe-flash-rerun2.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2609232304/argv.txt

@@ -0,0 +1,69 @@
+/root/workspace/Haoxuan/code/Psi0/scripts/train.py
+finetune_simple_psi0_config
+--seed=292285
+--exp=xmovepick-depth-v4-query-ffn-moe-flash-rerun2
+--train.name=finetune
+--train.output-dir=/root/workspace/Haoxuan/data/Psi0/.runs
+--train.data-parallel=ddp
+--train.mixed-precision=bf16
+--train.train-batch-size=16
+--train.max-checkpoints-to-keep=5
+--train.gradient-accumulation-steps=1
+--train.optimizer-foreach=False
+--train.num-workers=4
+--train.learning-rate=1e-4
+--train.max-training-steps=40000
+--train.warmup-ratio=None
+--train.warmup-steps=1000
+--train.checkpointing-steps=10000
+--train.validation-steps=500
+--train.val-num-batches=20
+--train.max-grad-norm=1.0
+--train.lr-scheduler-type=cosine
+--train.lr-scheduler-kwargs.weight-decay=1e-6
+--train.lr-scheduler-kwargs.betas 0.95 0.999
+--log.report-to=wandb
+--data.root-dir=/root/workspace/Haoxuan/data/Psi0/data/simple-depth
+--data.train-repo-ids=G1WholebodyXMovePickTeleop-v0
+--data.transform.repack.use-depth
+--data.transform.repack.pad-action-dim=36
+--data.transform.repack.pad-state-dim=36
+--data.transform.field.stat-path=meta/stats_psi0.json
+--data.transform.field.stat-action-key=action
+--data.transform.field.stat-state-key=states
+--data.transform.field.action-norm-type=bounds
+--data.transform.field.no-use-norm-mask
+--data.transform.field.normalize-state
+--data.transform.field.pad-action-dim=36
+--data.transform.field.pad-state-dim=36
+--data.transform.model.img-aug
+--data.transform.model.resize.size 180 320
+--data.transform.model.center-crop.size 180 320
+--model.model-name-or-path=/root/workspace/Haoxuan/data/Psi0/cache/checkpoints/psi0/pre.fast.1by1.2601091803.ckpt.ego200k.he30k
+--model.pretrained-action-header-path=/root/workspace/Haoxuan/data/Psi0/cache/checkpoints/psi0/postpre.1by1.pad36.2601131206.ckpt.he30k
+--model.noise-scheduler=flow
+--model.train-diffusion-steps=1000
+--model.n-conditions=0
+--model.action-chunk-size=30
+--model.action-dim=36
+--model.action-exec-horizon=30
+--model.observation-horizon=1
+--model.odim=36
+--model.view-feature-dim=2048
+--model.no-tune-vlm
+--model.no-use-film
+--model.no-combined-temb
+--model.rtc
+--model.max-delay=8
+--model.use-depth
+--model.depth-encoder-path=/root/workspace/Haoxuan/data/Psi0/cache/checkpoints/defm/defm_vit_s14.pth
+--model.depth-encoder-input-size 196 350
+--model.depth-fusion-mode=query_moe
+--model.depth-query-count=8
+--model.depth-query-moe-expert-count=8
+--model.depth-query-moe-top-k=2
+--model.depth-query-moe-expert-dim=384
+--model.depth-query-load-balance-weight=1e-1
+--model.depth-query-orthogonality-weight=1e-1
+--model.depth-module-lr=1e-4
+--model.action-expert-lr=1e-4

depth/finetune/xmovepick-depth-v4-query-ffn-moe-flash-rerun2.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2609232304/envs.txt

@@ -0,0 +1,18 @@
+OMP_NUM_THREADS=8
+HF_HOME=/root/workspace/Haoxuan/data/Psi0/cache/huggingface
+TORCH_HOME=/root/workspace/Haoxuan/data/Psi0/cache/torch
+HF_TOKEN=hf_...fsbP
+HF_LEROBOT_HOME=/root/workspace/Haoxuan/data/Psi0/data/lerobot
+DATA_HOME=/root/workspace/Haoxuan/data/Psi0/data
+UV_CACHE_DIR=/root/workspace/Haoxuan/data/Psi0/cache/uv
+WANDB_API_KEY=
+PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python
+CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
+PSI_ATTN_IMPLEMENTATION=flash_attention_2
+PSI_FORCE_MATH_SDPA=0
+WORLD_SIZE=8
+LOCAL_WORLD_SIZE=8
+RANK=0
+LOCAL_RANK=0
+MASTER_ADDR=127.0.0.1
+MASTER_PORT=29500

depth/finetune/xmovepick-depth-v4-query-ffn-moe-flash-rerun2.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2609232304/run_config.json

@@ -0,0 +1,378 @@
+{
+    "exp": "xmovepick-depth-v4-query-ffn-moe-flash-rerun2",
+    "seed": 292285,
+    "auto_tag_run": false,
+    "eval": false,
+    "debug": false,
+    "timestamp": "2609232304",
+    "log": {
+        "logging_dir": "logs",
+        "report_to": "wandb",
+        "log_freq": 100
+    },
+    "wandb": {
+        "project": "psi",
+        "entity": "",
+        "group": "finetune",
+        "id": "lwwjfcce",
+        "name": "xmovepick-depth-v4-query-ffn-moe-flash-rerun2.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2609232304",
+        "resume": "allow"
+    },
+    "train": {
+        "num_workers": 4,
+        "overfit_single_batch": false,
+        "name": "finetune",
+        "resume_from_checkpoint": null,
+        "skip_resumed_steps": false,
+        "hf_token": ".hf_token",
+        "lora": false,
+        "output_dir": "/root/workspace/Haoxuan/data/Psi0/.runs",
+        "gradient_accumulation_steps": 1,
+        "mixed_precision": "bf16",
+        "max_grad_norm": 1.0,
+        "optimizer_foreach": false,
+        "train_batch_size": 16,
+        "val_batch_size": 16,
+        "val_num_batches": 20,
+        "checkpointing_steps": 10000,
+        "max_checkpoints_to_keep": 5,
+        "validation_steps": 500,
+        "learning_rate": 0.0001,
+        "lr_scheduler_type": "cosine",
+        "lr_scheduler_kwargs": {
+            "betas": [
+                0.95,
+                0.999
+            ],
+            "weight_decay": 1e-6,
+            "eps": 1e-8
+        },
+        "scheduler_specific_kwargs": {},
+        "data_parallel": "ddp",
+        "sharding_strategy": "full-shard",
+        "deepspeed_config": "/root/workspace/Haoxuan/code/Psi0/src/InternVLA/config/deepseeds/zero3.json",
+        "enable_gradient_checkpointing": true,
+        "enable_mixed_precision_training": true,
+        "reduce_in_full_precision": true,
+        "max_training_steps": 40000,
+        "num_train_epochs": null,
+        "warmup_steps": 1000,
+        "warmup_ratio": null
+    },
+    "data": {
+        "transform": {
+            "repack": {
+                "dataset_name": "simple",
+                "num_past_frames": 0,
+                "action_chunk_size": 30,
+                "use_depth": true,
+                "depth_key": "observation.depth.egocentric",
+                "pad_action_dim": 36,
+                "pad_state_dim": 36
+            },
+            "model": {
+                "resize": {
+                    "size": [
+                        180,
+                        320
+                    ]
+                },
+                "center_crop": {
+                    "size": [
+                        180,
+                        320
+                    ]
+                },
+                "color_jitter": {
+                    "brightness": 0.2,
+                    "contrast": [
+                        0.8,
+                        1.2
+                    ],
+                    "saturation": [
+                        0.8,
+                        1.2
+                    ],
+                    "hue": 0.05
+                },
+                "gaussian_noise": {
+                    "mean": 0.0,
+                    "std": 3.0,
+                    "prob_skip": 0.1
+                },
+                "img_aug": true,
+                "adaptive_resize": false,
+                "img_sizes": {
+                    "egodex": [
+                        270,
+                        480
+                    ],
+                    "he": [
+                        240,
+                        320
+                    ]
+                }
+            },
+            "field": {
+                "stat_path": "meta/stats_psi0.json",
+                "action_norm_type": "bounds",
+                "stat_action_key": "action",
+                "stat_state_key": "states",
+                "use_norm_mask": false,
+                "action_norm_masks": [
+                    true,
+                    true,
+                    true,
+                    true,
+                    true,
+                    true,
+                    false
+                ],
+                "action_min": [
+                    0.0,
+                    0.0,
+                    0.0,
+                    0.0,
+                    0.0,
+                    0.0,
+                    0.0,
+                    -0.5,
+                    -0.699999988079071,
+                    -0.699999988079071,
+                    -8.807794347046262e-19,
+                    -8.807794347046262e-19,
+                    -3.4037360342851314e-19,
+                    -8.807794347046262e-19,
+                    -0.7234289646148682,
+                    0.1900009959936142,
+                    -0.3478638231754303,
+                    -0.24381141364574432,
+                    -0.2871176600456238,
+                    -0.46943801641464233,
+                    -0.28036442399024963,
+                    -0.28476205468177795,
+                    -0.2781273424625397,
+                    -0.6339197754859924,
+                    -0.2889905273914337,
+                    -0.2651921808719635,
+                    -0.6309908628463745,
+                    -0.19380050897598267,
+                    -0.12375029176473618,
+                    0.004558231681585312,
+                    -0.299873948097229,
+                    0.7400000095367432,
+                    -5.537859719245743e-17,
+                    0.0,
+                    -0.19462604820728302,
+                    0.0
+                ],
+                "action_max": [
+                    0.0,
+                    0.0,
+                    0.0,
+                    0.0,
+                    0.0,
+                    0.0,
+                    0.0,
+                    3.40585956201392e-19,
+                    4.519914941300404e-19,
+                    4.519914941300404e-19,
+                    1.5,
+                    1.5,
+                    0.6000000238418579,
+                    1.5,
+                    0.23390473425388336,
+                    0.6660271883010864,
+                    0.7131545543670654,
+                    0.6127181053161621,
+                    0.12901827692985535,
+                    0.23185482621192932,
+                    0.3763704001903534,
+                    0.2929672598838806,
+                    -0.1900009959936142,
+                    0.3377506732940674,
+                    0.585555911064148,
+                    0.7287320494651794,
+                    0.2222127765417099,
+                    0.8800510764122009,
+                    0.14049682021141052,
+                    0.20396868884563446,
+                    0.30819037556648254,
+                    0.7400000095367432,
+                    0.5,
+                    0.0,
+                    0.18601855635643005,
+                    0.0
+                ],
+                "state_min": [
+                    7.777451855872641e-7,
+                    -0.00033337774220854044,
+                    -5.059851446276298e-6,
+                    -0.00017079990357160568,
+                    -9.60247780312784e-6,
+                    -0.00018899694259744138,
+                    -7.104898486431921e-6,
+                    -0.5265660881996155,
+                    -0.35556095838546753,
+                    -0.5981534123420715,
+                    -0.00021067277702968568,
+                    -8.959311230682943e-7,
+                    -8.316740604641382e-6,
+                    -1.4921358797437279e-6,
+                    -0.3891601860523224,
+                    0.150077685713768,
+                    -0.3817761540412903,
+                    -0.15058015286922455,
+                    -0.2737894654273987,
+                    -0.4105795621871948,
+                    -0.2672588527202606,
+                    -0.21800069510936737,
+                    -0.26531729102134705,
+                    -0.6129422783851624,
+                    -0.2084185779094696,
+                    -0.25410670042037964,
+                    -0.5689104199409485,
+                    -0.1778532713651657,
+                    -0.09933578222990036,
+                    0.03510970249772072,
+                    -0.30416738986968994,
+                    0.7400000095367432,
+                    0.0,
+                    0.0,
+                    0.0,
+                    0.0
+                ],
+                "state_max": [
+                    3.826543888862943e-6,
+                    0.003359275171533227,
+                    2.544531525927596e-6,
+                    0.00021755567286163568,
+                    3.435341568547301e-6,
+                    0.00019837530271615833,
+                    5.035403773945291e-6,
+                    7.077142640810052e-7,
+                    0.16438177227973938,
+                    6.031086172697542e-7,
+                    0.47953173518180847,
+                    1.3467339277267456,
+                    0.7186622619628906,
+                    1.1524261236190796,
+                    0.2516690492630005,
+                    0.39732399582862854,
+                    0.6492785811424255,
+                    0.6760621070861816,
+                    0.1240595132112503,
+                    0.22742627561092377,
+                    0.3518524467945099,
+                    0.31155094504356384,
+                    -0.15275707840919495,
+                    0.3515203297138214,
+                    0.643481969833374,
+                    0.6658613681793213,
+                    0.2018662989139557,
+                    0.8597758412361145,
+                    0.13490912318229675,
+                    0.25929149985313416,
+                    0.31226828694343567,
+                    0.7400000095367432,
+                    0.0,
+                    0.0,
+                    0.0,
+                    0.0
+                ],
+                "normalize_state": true,
+                "pad_action_dim": 36,
+                "pad_state_dim": 36
+            }
+        },
+        "root_dir": "/root/workspace/Haoxuan/data/Psi0/data/simple-depth",
+        "train_repo_ids": [
+            "G1WholebodyXMovePickTeleop-v0"
+        ],
+        "val_repo_ids": [
+            "G1WholebodyXMovePickTeleop-v0"
+        ]
+    },
+    "model": {
+        "resnet_store_path": null,
+        "pretrained_action_header_path": "/root/workspace/Haoxuan/data/Psi0/cache/checkpoints/psi0/postpre.1by1.pad36.2601131206.ckpt.he30k",
+        "rtc": true,
+        "max_delay": 8,
+        "action_dim": 36,
+        "action_chunk_size": 30,
+        "action_exec_horizon": 30,
+        "observation_horizon": 1,
+        "img_chunk": 1,
+        "n_cams": 1,
+        "use_obs": "add_token",
+        "dropout": 0.1,
+        "noise_scheduler": "flow",
+        "train_diffusion_steps": 1000,
+        "eval_diffusion_steps": 10,
+        "share_cam_features": false,
+        "early_fusion": false,
+        "odim": 36,
+        "n_conditions": 0,
+        "token_fusion": "concat",
+        "loss_w": [
+            0.1,
+            0.2,
+            0.1
+        ],
+        "time_dim": 256,
+        "hidden_dim": 1536,
+        "num_blocks": 6,
+        "dim_feedforward": 2048,
+        "nhead": 24,
+        "activation": "gelu",
+        "view_feature_dim": 2048,
+        "use_film": false,
+        "combined_temb": false,
+        "use_dit": false,
+        "use_depth": true,
+        "depth_encoder_name": "defm_vit_s14",
+        "depth_encoder_path": "/root/workspace/Haoxuan/data/Psi0/cache/checkpoints/defm/defm_vit_s14.pth",
+        "depth_encoder_dim": 384,
+        "depth_encoder_input_size": [
+            196,
+            350
+        ],
+        "depth_encoder_patch_size": 14,
+        "freeze_depth_encoder": true,
+        "depth_cross_attention_layers": [
+            0,
+            1,
+            2,
+            3,
+            4,
+            5
+        ],
+        "depth_fusion_mode": "query_moe",
+        "depth_query_count": 8,
+        "depth_query_moe_expert_count": 8,
+        "depth_query_moe_top_k": 2,
+        "depth_query_moe_expert_dim": 384,
+        "depth_query_load_balance_weight": 0.1,
+        "depth_query_orthogonality_weight": 0.1,
+        "depth_gate_init": 0.001,
+        "depth_module_lr": 0.0001,
+        "action_expert_lr": 0.0001,
+        "weight_decay": 0.01,
+        "model_name_or_path": "/root/workspace/Haoxuan/data/Psi0/cache/checkpoints/psi0/pre.fast.1by1.2601091803.ckpt.ego200k.he30k",
+        "vlm_ckpt_step": null,
+        "tune_vlm": false,
+        "tune_mm_llm": false,
+        "tune_mm_vision": false,
+        "tune_mm_mlp": false,
+        "gradient_checkpointing": true,
+        "lang_backbone_lr": 0.00001,
+        "mm_projector_lr": 0.00001,
+        "vision_tower_lr": 1e-6,
+        "optim": "adamw_torch",
+        "model_max_length": 4096,
+        "data_flatten": true,
+        "data_packing": true,
+        "max_pixels": 451584,
+        "min_pixels": 12544
+    }
+}