MTDiT-Spatial multi-task spatial_mask @ 100,000 steps

README.md

@@ -0,0 +1,107 @@
+---
+library_name: lerobot
+pipeline_tag: robotics
+tags:
+- robotics
+- lerobot
+- diffusion-policy
+- multi-task
+- negative-result
+license: apache-2.0
+base_model:
+- openai/clip-vit-base-patch16
+---
+
+# MTDiT-Spatial — multi-task, `spatial_mask`
+
+A 450M-parameter multi-task diffusion transformer trained on 949 real-robot episodes.
+Vision path `spatial_softmax`, state dropout `0.15`, absolute joint actions.
+
+## Read this before using the weights
+
+**These checkpoints are published as a negative result.** They fit their data well and
+largely **ignore their scene cameras**, predicting joint targets from proprioception instead.
+**Arm camera sensitivity is 0.272** against a null-test floor of 0.000 and a grounded reference of ~1.0.
+
+Do not deploy this as a vision-conditioned policy. It is useful as a reproduction target, a
+baseline for work on visual grounding, or a test case for sensitivity metrics.
+
+## Measurements
+
+| Metric | Value |
+|---|---:|
+| Validation loss @ 100,000 | 0.0019 |
+| **Arm camera sensitivity** | **0.272** |
+| Gripper camera sensitivity | 0.930 |
+| Arm prediction error (held-out) | 0.102 |
+| Null test (own cameras) | 0.000 |
+
+Camera sensitivity holds the proprioceptive state fixed, swaps all three camera streams for
+matched-phase frames from a different held-out episode, and reports the mean change in predicted
+joints over the natural cross-episode spread of the ground truth. A policy that tracks the scene
+scores ~1.0; one that ignores it scores ~0.
+
+Two controls make the number trustworthy. Every variant is sampled from an identical noise
+sequence (common random numbers), without which diffusion sampling noise alone contributes 0.077.
+And the null test — feed the anchor its own cameras — returns 0.000 here, so the reported value is
+signal, not floor.
+
+For reference under the same evaluator: stock `multi_task_dit` on LIBERO scores 0.92–1.10, and
+π₀.₅ on LIBERO scores 1.19–1.31.
+
+## The sweep
+
+Ranking by loss is the inverse of ranking by grounding. The best-fitting arms are the blindest,
+and the two arms that withhold proprioception 15% of the time recover 2.5–3.5× the camera
+sensitivity for a ~12% loss penalty. That inversion is the result, so all five arms are published
+together.
+
+| Arm | Vision path | State dropout | Val loss | Arm sensitivity | Gripper sensitivity |
+|---|---|---:|---:|---:|---:|
+| [`cls_nodrop`](https://huggingface.co/maskjp/mtdit-spatial-multitask-cls-nodrop) | `cls` | 0.00 | **0.0017** | 0.077 | 0.706 |
+| [`spatial_nodrop`](https://huggingface.co/maskjp/mtdit-spatial-multitask-spatial-nodrop) | `spatial_softmax` | 0.00 | **0.0017** | 0.089 | 0.693 |
+| [`patch_tokens`](https://huggingface.co/maskjp/mtdit-spatial-multitask-patch-tokens) | `patch_tokens` | 0.00 | 0.0018* | * | * |
+| [`spatial_mask`](https://huggingface.co/maskjp/mtdit-spatial-multitask-spatial-mask) | `spatial_softmax` | 0.15 | 0.0019 | **0.272** | **0.930** |
+| [`cls_mask`](https://huggingface.co/maskjp/mtdit-spatial-multitask-cls-mask) | `cls` | 0.15 | 0.0020 | 0.198 | 0.922 |
+
+<sub>* `patch_tokens` is ~1.8× slower per step and was still training at publication; its 0.0018
+is at 75K and it tracks `cls_nodrop` to within 0.0001 at every shared checkpoint.</sub>
+
+## Training
+
+| | |
+|---|---|
+| Dataset | `l5vel-peng/base4-multitask-eef-merged-v30` — 7 tasks, 949 episodes, 1,524,905 frames, 50 Hz |
+| Held out | `eval_split=0.05`, applied per task |
+| Action space | absolute joints, 10-dim |
+| Steps | 100,000, batch 64, seed 1000 |
+| Architecture | 4 layers, hidden 512, 8 heads, dropout 0.1, RoPE |
+| Objective | DDPM, 100 train timesteps, horizon 32, 24 executed steps, 2 observation steps |
+| Encoders | `openai/clip-vit-base-patch16` for vision and text, vision LR × 0.1 |
+| Images | 3 cameras, 480×640 → resize 240×320 → random crop 224×224 |
+| Optimizer | 3e-4, weight decay 0, no warmup, cosine decay |
+| Normalization | visual MEAN_STD, state MIN_MAX, action MIN_MAX |
+
+The 30K checkpoint was extended to 100K by resuming, which re-stretched the cosine schedule; loss
+bumps at 35K and recovers by ~65K. The floor was already reached at 30K — 3.3× more compute moved
+it by 0.0001.
+
+## Usage
+
+```python
+from lerobot.policies.factory import make_policy
+
+policy = make_policy("maskjp/mtdit-spatial-multitask-spatial-mask")
+```
+
+Requires the `mtdit_spatial` plugin from
+[`maskjp/lerobot_policy_mtdit_spatial`](https://github.com/maskjp/lerobot_policy_mtdit_spatial),
+which also contains `scripts/vision_sensitivity.py` used for every number above.
+
+## Known limitations
+
+- Ignores the scene cameras for arm motion; the left and right cameras contribute least.
+- Trained on absolute joint targets, which correlate 0.96 with the current state — the shortcut
+  this failure runs on. Datasets with per-step deltas (LIBERO, −0.14) do not show it.
+- Single robot, single embodiment, 3 fixed camera poses. No environment-rollout success rate is
+  reported; only offline metrics.

config.json

@@ -0,0 +1,143 @@
+{
+    "type": "mtdit_spatial",
+    "n_obs_steps": 2,
+    "input_features": {
+        "observation.state": {
+            "type": "STATE",
+            "shape": [
+                10
+            ]
+        },
+        "observation.images.left": {
+            "type": "VISUAL",
+            "shape": [
+                3,
+                480,
+                640
+            ]
+        },
+        "observation.images.right": {
+            "type": "VISUAL",
+            "shape": [
+                3,
+                480,
+                640
+            ]
+        },
+        "observation.images.wrist": {
+            "type": "VISUAL",
+            "shape": [
+                3,
+                480,
+                640
+            ]
+        },
+        "observation.base_pose": {
+            "type": "STATE",
+            "shape": [
+                3
+            ]
+        },
+        "observation.eef_state": {
+            "type": "STATE",
+            "shape": [
+                13
+            ]
+        }
+    },
+    "output_features": {
+        "action": {
+            "type": "ACTION",
+            "shape": [
+                10
+            ]
+        },
+        "action.eef": {
+            "type": "ACTION",
+            "shape": [
+                13
+            ]
+        }
+    },
+    "device": "cuda",
+    "use_amp": false,
+    "use_peft": false,
+    "push_to_hub": false,
+    "repo_id": null,
+    "private": null,
+    "tags": null,
+    "license": null,
+    "pretrained_path": "/home/data/pej3435/Code/dev/lerobot_policy_mtdit_spatial/outputs/multitask/spatial_mask/checkpoints/last/pretrained_model",
+    "pretrained_revision": null,
+    "horizon": 32,
+    "n_action_steps": 24,
+    "objective": "diffusion",
+    "noise_scheduler_type": "DDPM",
+    "num_train_timesteps": 100,
+    "beta_schedule": "squaredcos_cap_v2",
+    "beta_start": 0.0001,
+    "beta_end": 0.02,
+    "prediction_type": "epsilon",
+    "clip_sample": true,
+    "clip_sample_range": 1.0,
+    "num_inference_steps": null,
+    "sigma_min": 0.0,
+    "num_integration_steps": 100,
+    "integration_method": "euler",
+    "timestep_sampling_strategy": "beta",
+    "timestep_sampling_s": 0.999,
+    "timestep_sampling_alpha": 1.5,
+    "timestep_sampling_beta": 1.0,
+    "hidden_dim": 512,
+    "num_layers": 4,
+    "num_heads": 8,
+    "dropout": 0.1,
+    "use_positional_encoding": false,
+    "timestep_embed_dim": 256,
+    "use_rope": true,
+    "rope_base": 10000.0,
+    "vision_encoder_name": "openai/clip-vit-base-patch16",
+    "use_separate_rgb_encoder_per_camera": false,
+    "vision_encoder_lr_multiplier": 0.1,
+    "image_resize_shape": [
+        240,
+        320
+    ],
+    "image_crop_shape": [
+        224,
+        224
+    ],
+    "image_crop_is_random": true,
+    "vision_feature_mode": "spatial_softmax",
+    "num_spatial_keypoints": 32,
+    "spatial_softmax_keep_cls": true,
+    "spatial_softmax_learnable_temperature": true,
+    "cross_attention_heads": null,
+    "state_dropout_prob": 0.15,
+    "use_relative_actions": false,
+    "relative_exclude_joints": [],
+    "action_feature_names": [
+        "motors"
+    ],
+    "text_encoder_name": "openai/clip-vit-base-patch16",
+    "tokenizer_max_length": 77,
+    "tokenizer_padding": "max_length",
+    "tokenizer_padding_side": "right",
+    "tokenizer_truncation": true,
+    "normalization_mapping": {
+        "VISUAL": "MEAN_STD",
+        "STATE": "MIN_MAX",
+        "ACTION": "MIN_MAX"
+    },
+    "optimizer_lr": 0.0003,
+    "optimizer_betas": [
+        0.95,
+        0.999
+    ],
+    "optimizer_eps": 1e-08,
+    "optimizer_weight_decay": 0.0,
+    "scheduler_name": "cosine",
+    "scheduler_warmup_steps": 0,
+    "do_mask_loss_for_padding": false,
+    "drop_n_last_frames": 7
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:915f198e294e5cce73c66c73153710db602c93d9b50eb3a9642ff3f78e107002
+size 957895124

policy_postprocessor.json

@@ -0,0 +1,44 @@
+{
+  "name": "policy_postprocessor",
+  "steps": [
+    {
+      "registry_name": "unnormalizer_processor",
+      "config": {
+        "eps": 1e-08,
+        "features": {
+          "action": {
+            "type": "ACTION",
+            "shape": [
+              10
+            ]
+          },
+          "action.eef": {
+            "type": "ACTION",
+            "shape": [
+              13
+            ]
+          }
+        },
+        "norm_map": {
+          "VISUAL": "MEAN_STD",
+          "STATE": "MIN_MAX",
+          "ACTION": "MIN_MAX"
+        }
+      },
+      "state_file": "policy_postprocessor_step_0_unnormalizer_processor.safetensors"
+    },
+    {
+      "registry_name": "absolute_actions_processor",
+      "config": {
+        "enabled": false
+      }
+    },
+    {
+      "registry_name": "device_processor",
+      "config": {
+        "device": "cpu",
+        "float_dtype": null
+      }
+    }
+  ]
+}

policy_postprocessor_step_0_unnormalizer_processor.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d99c485cf8015380852991f5d951535ee5c5c718882a5140a7aea0d687990d7d
+size 6796

policy_preprocessor.json

@@ -0,0 +1,114 @@
+{
+  "name": "policy_preprocessor",
+  "steps": [
+    {
+      "registry_name": "rename_observations_processor",
+      "config": {
+        "rename_map": {}
+      }
+    },
+    {
+      "registry_name": "to_batch_processor",
+      "config": {}
+    },
+    {
+      "registry_name": "tokenizer_processor",
+      "config": {
+        "max_length": 77,
+        "task_key": "task",
+        "padding_side": "right",
+        "padding": "max_length",
+        "truncation": true,
+        "tokenizer_name": "tokenizer"
+      },
+      "artifacts": {
+        "tokenizer_name": "tokenizer"
+      }
+    },
+    {
+      "registry_name": "device_processor",
+      "config": {
+        "device": "cuda",
+        "float_dtype": null
+      }
+    },
+    {
+      "registry_name": "mtdit_spatial_relative_actions_processor",
+      "config": {
+        "enabled": false,
+        "exclude_joints": [],
+        "action_names": [
+          "motors"
+        ]
+      }
+    },
+    {
+      "registry_name": "normalizer_processor",
+      "config": {
+        "eps": 1e-08,
+        "features": {
+          "observation.state": {
+            "type": "STATE",
+            "shape": [
+              10
+            ]
+          },
+          "observation.images.left": {
+            "type": "VISUAL",
+            "shape": [
+              3,
+              480,
+              640
+            ]
+          },
+          "observation.images.right": {
+            "type": "VISUAL",
+            "shape": [
+              3,
+              480,
+              640
+            ]
+          },
+          "observation.images.wrist": {
+            "type": "VISUAL",
+            "shape": [
+              3,
+              480,
+              640
+            ]
+          },
+          "observation.base_pose": {
+            "type": "STATE",
+            "shape": [
+              3
+            ]
+          },
+          "observation.eef_state": {
+            "type": "STATE",
+            "shape": [
+              13
+            ]
+          },
+          "action": {
+            "type": "ACTION",
+            "shape": [
+              10
+            ]
+          },
+          "action.eef": {
+            "type": "ACTION",
+            "shape": [
+              13
+            ]
+          }
+        },
+        "norm_map": {
+          "VISUAL": "MEAN_STD",
+          "STATE": "MIN_MAX",
+          "ACTION": "MIN_MAX"
+        }
+      },
+      "state_file": "policy_preprocessor_step_5_normalizer_processor.safetensors"
+    }
+  ]
+}

policy_preprocessor_step_5_normalizer_processor.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b5a2f5ec520d64a11bccfbff17d59499ec884abf2a9ae90b8b5f0d9a4293ba9
+size 6756

tokenizer/tokenizer_config.json

@@ -0,0 +1,20 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": "<|startoftext|>",
+  "do_lower_case": true,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "is_local": true,
+  "max_length": 77,
+  "model_max_length": 77,
+  "pad_to_multiple_of": null,
+  "pad_token": "<|endoftext|>",
+  "pad_token_type_id": 0,
+  "padding_side": "right",
+  "stride": 0,
+  "tokenizer_class": "CLIPTokenizer",
+  "truncation_side": "right",
+  "truncation_strategy": "longest_first",
+  "unk_token": "<|endoftext|>"
+}

train_config.json

@@ -0,0 +1,344 @@
+{
+    "dataset": {
+        "repo_id": "l5vel-peng/base4-multitask-eef-merged-v30",
+        "repo_type": "dataset",
+        "root": "/home/data/pej3435/.cache/l5vel_eef_merge/outputs/base4-multitask-eef-merged-v30",
+        "episodes": null,
+        "exclude_episodes": null,
+        "image_transforms": {
+            "enable": false,
+            "max_num_transforms": 3,
+            "random_order": false,
+            "tfs": {
+                "brightness": {
+                    "weight": 1.0,
+                    "type": "ColorJitter",
+                    "kwargs": {
+                        "brightness": [
+                            0.8,
+                            1.2
+                        ]
+                    }
+                },
+                "contrast": {
+                    "weight": 1.0,
+                    "type": "ColorJitter",
+                    "kwargs": {
+                        "contrast": [
+                            0.8,
+                            1.2
+                        ]
+                    }
+                },
+                "saturation": {
+                    "weight": 1.0,
+                    "type": "ColorJitter",
+                    "kwargs": {
+                        "saturation": [
+                            0.5,
+                            1.5
+                        ]
+                    }
+                },
+                "hue": {
+                    "weight": 1.0,
+                    "type": "ColorJitter",
+                    "kwargs": {
+                        "hue": [
+                            -0.05,
+                            0.05
+                        ]
+                    }
+                },
+                "sharpness": {
+                    "weight": 1.0,
+                    "type": "SharpnessJitter",
+                    "kwargs": {
+                        "sharpness": [
+                            0.5,
+                            1.5
+                        ]
+                    }
+                },
+                "affine": {
+                    "weight": 1.0,
+                    "type": "RandomAffine",
+                    "kwargs": {
+                        "degrees": [
+                            -5.0,
+                            5.0
+                        ],
+                        "translate": [
+                            0.05,
+                            0.05
+                        ]
+                    }
+                }
+            }
+        },
+        "revision": null,
+        "use_imagenet_stats": true,
+        "video_backend": "torchcodec",
+        "return_uint8": false,
+        "depth_output_unit": "mm",
+        "streaming": false,
+        "eval_split": 0.05
+    },
+    "env": null,
+    "policy": {
+        "type": "mtdit_spatial",
+        "n_obs_steps": 2,
+        "input_features": {
+            "observation.state": {
+                "type": "STATE",
+                "shape": [
+                    10
+                ]
+            },
+            "observation.images.left": {
+                "type": "VISUAL",
+                "shape": [
+                    3,
+                    480,
+                    640
+                ]
+            },
+            "observation.images.right": {
+                "type": "VISUAL",
+                "shape": [
+                    3,
+                    480,
+                    640
+                ]
+            },
+            "observation.images.wrist": {
+                "type": "VISUAL",
+                "shape": [
+                    3,
+                    480,
+                    640
+                ]
+            },
+            "observation.base_pose": {
+                "type": "STATE",
+                "shape": [
+                    3
+                ]
+            },
+            "observation.eef_state": {
+                "type": "STATE",
+                "shape": [
+                    13
+                ]
+            }
+        },
+        "output_features": {
+            "action": {
+                "type": "ACTION",
+                "shape": [
+                    10
+                ]
+            },
+            "action.eef": {
+                "type": "ACTION",
+                "shape": [
+                    13
+                ]
+            }
+        },
+        "device": "cuda",
+        "use_amp": false,
+        "use_peft": false,
+        "push_to_hub": false,
+        "repo_id": null,
+        "private": null,
+        "tags": null,
+        "license": null,
+        "pretrained_path": "/home/data/pej3435/Code/dev/lerobot_policy_mtdit_spatial/outputs/multitask/spatial_mask/checkpoints/last/pretrained_model",
+        "pretrained_revision": null,
+        "horizon": 32,
+        "n_action_steps": 24,
+        "objective": "diffusion",
+        "noise_scheduler_type": "DDPM",
+        "num_train_timesteps": 100,
+        "beta_schedule": "squaredcos_cap_v2",
+        "beta_start": 0.0001,
+        "beta_end": 0.02,
+        "prediction_type": "epsilon",
+        "clip_sample": true,
+        "clip_sample_range": 1.0,
+        "num_inference_steps": null,
+        "sigma_min": 0.0,
+        "num_integration_steps": 100,
+        "integration_method": "euler",
+        "timestep_sampling_strategy": "beta",
+        "timestep_sampling_s": 0.999,
+        "timestep_sampling_alpha": 1.5,
+        "timestep_sampling_beta": 1.0,
+        "hidden_dim": 512,
+        "num_layers": 4,
+        "num_heads": 8,
+        "dropout": 0.1,
+        "use_positional_encoding": false,
+        "timestep_embed_dim": 256,
+        "use_rope": true,
+        "rope_base": 10000.0,
+        "vision_encoder_name": "openai/clip-vit-base-patch16",
+        "use_separate_rgb_encoder_per_camera": false,
+        "vision_encoder_lr_multiplier": 0.1,
+        "image_resize_shape": [
+            240,
+            320
+        ],
+        "image_crop_shape": [
+            224,
+            224
+        ],
+        "image_crop_is_random": true,
+        "vision_feature_mode": "spatial_softmax",
+        "num_spatial_keypoints": 32,
+        "spatial_softmax_keep_cls": true,
+        "spatial_softmax_learnable_temperature": true,
+        "cross_attention_heads": null,
+        "state_dropout_prob": 0.15,
+        "use_relative_actions": false,
+        "relative_exclude_joints": [],
+        "action_feature_names": [
+            "motors"
+        ],
+        "text_encoder_name": "openai/clip-vit-base-patch16",
+        "tokenizer_max_length": 77,
+        "tokenizer_padding": "max_length",
+        "tokenizer_padding_side": "right",
+        "tokenizer_truncation": true,
+        "normalization_mapping": {
+            "VISUAL": "MEAN_STD",
+            "STATE": "MIN_MAX",
+            "ACTION": "MIN_MAX"
+        },
+        "optimizer_lr": 0.0003,
+        "optimizer_betas": [
+            0.95,
+            0.999
+        ],
+        "optimizer_eps": 1e-08,
+        "optimizer_weight_decay": 0.0,
+        "scheduler_name": "cosine",
+        "scheduler_warmup_steps": 0,
+        "do_mask_loss_for_padding": false,
+        "drop_n_last_frames": 7
+    },
+    "reward_model": null,
+    "output_dir": "/home/data/pej3435/Code/dev/lerobot_policy_mtdit_spatial/outputs/multitask/spatial_mask",
+    "job_name": "mtdit_multitask_spatial_mask",
+    "resume": true,
+    "seed": 1000,
+    "cudnn_deterministic": false,
+    "num_workers": 12,
+    "batch_size": 64,
+    "prefetch_factor": 4,
+    "persistent_workers": true,
+    "dataloader_multiprocessing_context": "spawn",
+    "steps": 100000,
+    "env_eval_freq": 0,
+    "log_freq": 200,
+    "eval_steps": 5000,
+    "max_eval_samples": 0,
+    "tolerance_s": 0.0001,
+    "save_checkpoint": true,
+    "save_freq": 10000,
+    "checkpoint_format": "safetensors",
+    "use_policy_training_preset": true,
+    "optimizer": {
+        "type": "adam",
+        "lr": 0.0003,
+        "weight_decay": 0.0,
+        "grad_clip_norm": 10.0,
+        "betas": [
+            0.95,
+            0.999
+        ],
+        "eps": 1e-08
+    },
+    "scheduler": {
+        "type": "diffuser",
+        "num_warmup_steps": 0,
+        "name": "cosine"
+    },
+    "parallelism": {
+        "dp_replicate": 1,
+        "dp_shard": 1,
+        "context_parallel": {
+            "ring_degree": 1,
+            "ulysses_degree": 1
+        },
+        "cfg_parallel": 1
+    },
+    "accelerator": {
+        "mixed_precision": "bf16",
+        "gradient_accumulation": {
+            "steps": 1
+        },
+        "fsdp": {
+            "reshard_after_forward": true,
+            "wrap_modules": null,
+            "min_num_params": null,
+            "cpu_offload": false,
+            "ignored_modules": null
+        },
+        "ddp": {
+            "find_unused_parameters": true,
+            "gradient_as_bucket_view": false,
+            "static_graph": false
+        },
+        "compile": {
+            "enabled": false,
+            "backend": "inductor",
+            "mode": null,
+            "regional": true
+        },
+        "activation_checkpointing": {
+            "mode": "none"
+        }
+    },
+    "eval": {
+        "n_episodes": 50,
+        "batch_size": 50,
+        "use_async_envs": true,
+        "recording": false,
+        "recording_repo_id": null,
+        "recording_private": false
+    },
+    "ema": {
+        "enable": false,
+        "decay": null,
+        "update_after_step": 0,
+        "inv_gamma": 1.0,
+        "power": 0.75,
+        "min_decay": 0.0,
+        "max_decay": 0.9999,
+        "use_for_eval": true
+    },
+    "wandb": {
+        "enable": true,
+        "disable_artifact": true,
+        "project": "mtdit-croissant-multitask",
+        "entity": null,
+        "notes": null,
+        "run_id": "vrg2t4v1",
+        "mode": null,
+        "add_tags": true
+    },
+    "peft": null,
+    "job": {
+        "target": null,
+        "image": "huggingface/lerobot-gpu:latest",
+        "timeout": "2d",
+        "detach": false,
+        "tags": []
+    },
+    "save_checkpoint_to_hub": false,
+    "sample_weighting": null,
+    "rename_map": {},
+    "checkpoint_path": "/home/data/pej3435/Code/dev/lerobot_policy_mtdit_spatial/outputs/multitask/spatial_mask/checkpoints/last"
+}