Upload policy weights, train config and readme

Browse files

Files changed (4) hide show

README.md +62 -0
config.json +119 -0
model.safetensors +3 -0
train_config.json +257 -0

README.md ADDED Viewed

	@@ -0,0 +1,62 @@

+---
+datasets: DAVIAN-Robotics/robocasa-MG_300
+library_name: lerobot
+license: apache-2.0
+model_name: vla_jepa
+pipeline_tag: robotics
+tags:
+- robotics
+- vla_jepa
+- lerobot
+---
+# Model Card for vla_jepa
+<!-- Provide a quick summary of what the model is/does. -->
+_Model type not recognized — please update this template._
+This policy has been trained and pushed to the Hub using [LeRobot](https://github.com/huggingface/lerobot).
+See the full documentation at [LeRobot Docs](https://huggingface.co/docs/lerobot/index).
+---
+## How to Get Started with the Model
+For a complete walkthrough, see the [training guide](https://huggingface.co/docs/lerobot/il_robots#train-a-policy).
+Below is the short version on how to train and run inference/eval:
+### Train from scratch
+```bash
+lerobot-train \
+  --dataset.repo_id=${HF_USER}/<dataset> \
+  --policy.type=act \
+  --output_dir=outputs/train/<desired_policy_repo_id> \
+  --job_name=lerobot_training \
+  --policy.device=cuda \
+  --policy.repo_id=${HF_USER}/<desired_policy_repo_id>
+  --wandb.enable=true
+```
+_Writes checkpoints to `outputs/train/<desired_policy_repo_id>/checkpoints/`._
+### Evaluate the policy/run inference
+```bash
+lerobot-record \
+  --robot.type=so100_follower \
+  --dataset.repo_id=<hf_user>/eval_<dataset> \
+  --policy.path=<hf_user>/<desired_policy_repo_id> \
+  --episodes=10
+```
+Prefix the dataset repo with **eval\_** and supply `--policy.path` pointing to a local or hub checkpoint.
+---
+## Model Details
+- **License:** apache-2.0

config.json ADDED Viewed

	@@ -0,0 +1,119 @@

+{
+    "type": "vla_jepa",
+    "n_obs_steps": 1,
+    "input_features": {
+        "observation.state": {
+            "type": "STATE",
+            "shape": [
+                16
+            ]
+        },
+        "observation.images.robot0_agentview_left_image": {
+            "type": "VISUAL",
+            "shape": [
+                3,
+                128,
+                128
+            ]
+        },
+        "observation.images.robot0_agentview_right_image": {
+            "type": "VISUAL",
+            "shape": [
+                3,
+                128,
+                128
+            ]
+        },
+        "observation.images.robot0_eye_in_hand_image": {
+            "type": "VISUAL",
+            "shape": [
+                3,
+                128,
+                128
+            ]
+        }
+    },
+    "output_features": {
+        "action": {
+            "type": "ACTION",
+            "shape": [
+                12
+            ]
+        }
+    },
+    "device": "cuda",
+    "use_amp": false,
+    "use_peft": false,
+    "push_to_hub": true,
+    "repo_id": "deksprime/vla-jepa-robocasa-ri",
+    "private": null,
+    "tags": null,
+    "license": null,
+    "pretrained_path": "/__modal/volumes/vo-MzoUtjIoo7jtxnPqAorQDI/base/lerobot__VLA-JEPA-Pretrain",
+    "chunk_size": 7,
+    "n_action_steps": 7,
+    "normalization_mapping": {
+        "VISUAL": "IDENTITY",
+        "STATE": "MEAN_STD",
+        "ACTION": "MIN_MAX"
+    },
+    "qwen_model_name": "Qwen/Qwen3-VL-2B-Instruct",
+    "jepa_encoder_name": "facebook/vjepa2-vitl-fpc64-256",
+    "freeze_qwen": true,
+    "enable_world_model": false,
+    "reinit_modules": [
+        "model.action_model.action_encoder",
+        "model.action_model.action_decoder",
+        "model.action_model.state_encoder"
+    ],
+    "tokenizer_padding_side": "left",
+    "prompt_template": "Your task is {instruction}. Infer the temporal dynamics from frames {actions} and produce the corresponding policy actions {e_actions}.",
+    "special_action_token": "<|action_{}|>",
+    "embodied_action_token": "<|embodied_action|>",
+    "action_dim": 12,
+    "state_dim": 16,
+    "num_action_tokens_per_timestep": 8,
+    "num_embodied_action_tokens_per_instruction": 32,
+    "num_inference_timesteps": 4,
+    "action_hidden_size": 1024,
+    "action_model_type": "DiT-B",
+    "action_num_layers": 16,
+    "action_num_heads": 12,
+    "action_attention_head_dim": 64,
+    "action_dropout": 0.2,
+    "action_num_timestep_buckets": 1000,
+    "action_noise_beta_alpha": 1.5,
+    "action_noise_beta_beta": 1.0,
+    "action_noise_s": 0.999,
+    "num_target_vision_tokens": 32,
+    "action_max_seq_len": 1024,
+    "num_video_frames": 8,
+    "predictor_depth": 12,
+    "predictor_num_heads": 8,
+    "predictor_mlp_ratio": 4.0,
+    "predictor_dropout": 0.0,
+    "world_model_loss_weight": 0.1,
+    "jepa_tubelet_size": 2,
+    "repeated_diffusion_steps": 8,
+    "resize_images_to": [
+        224,
+        224
+    ],
+    "binarize_gripper_action": true,
+    "pre_snap_gripper_action": true,
+    "clip_normalized_actions": true,
+    "gripper_dim": 6,
+    "gripper_threshold": 0.5,
+    "torch_dtype": "bfloat16",
+    "optimizer_lr": 0.0001,
+    "optimizer_betas": [
+        0.9,
+        0.95
+    ],
+    "optimizer_eps": 1e-08,
+    "optimizer_weight_decay": 1e-08,
+    "optimizer_grad_clip_norm": 1.0,
+    "scheduler_warmup_steps": 5000,
+    "scheduler_decay_steps": 30000,
+    "scheduler_decay_lr": 1e-06
+}

model.safetensors ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:f9be0a41eda387e92065a7a70ad08a72ac4b9342fd60b105136d363324b29c3b
+size 4875982288

train_config.json ADDED Viewed

	@@ -0,0 +1,257 @@

+{
+    "dataset": {
+        "repo_id": "DAVIAN-Robotics/robocasa-MG_300",
+        "root": null,
+        "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": "v3.0",
+        "use_imagenet_stats": true,
+        "video_backend": "pyav",
+        "return_uint8": false,
+        "streaming": false
+    },
+    "env": null,
+    "policy": {
+        "type": "vla_jepa",
+        "n_obs_steps": 1,
+        "input_features": {
+            "observation.state": {
+                "type": "STATE",
+                "shape": [
+                    16
+                ]
+            },
+            "observation.images.robot0_agentview_left_image": {
+                "type": "VISUAL",
+                "shape": [
+                    3,
+                    128,
+                    128
+                ]
+            },
+            "observation.images.robot0_agentview_right_image": {
+                "type": "VISUAL",
+                "shape": [
+                    3,
+                    128,
+                    128
+                ]
+            },
+            "observation.images.robot0_eye_in_hand_image": {
+                "type": "VISUAL",
+                "shape": [
+                    3,
+                    128,
+                    128
+                ]
+            }
+        },
+        "output_features": {
+            "action": {
+                "type": "ACTION",
+                "shape": [
+                    12
+                ]
+            }
+        },
+        "device": "cuda",
+        "use_amp": false,
+        "use_peft": false,
+        "push_to_hub": true,
+        "repo_id": "deksprime/vla-jepa-robocasa-ri",
+        "private": null,
+        "tags": null,
+        "license": null,
+        "pretrained_path": "/__modal/volumes/vo-MzoUtjIoo7jtxnPqAorQDI/base/lerobot__VLA-JEPA-Pretrain",
+        "chunk_size": 7,
+        "n_action_steps": 7,
+        "normalization_mapping": {
+            "VISUAL": "IDENTITY",
+            "STATE": "MEAN_STD",
+            "ACTION": "MIN_MAX"
+        },
+        "qwen_model_name": "Qwen/Qwen3-VL-2B-Instruct",
+        "jepa_encoder_name": "facebook/vjepa2-vitl-fpc64-256",
+        "freeze_qwen": true,
+        "enable_world_model": false,
+        "reinit_modules": [
+            "model.action_model.action_encoder",
+            "model.action_model.action_decoder",
+            "model.action_model.state_encoder"
+        ],
+        "tokenizer_padding_side": "left",
+        "prompt_template": "Your task is {instruction}. Infer the temporal dynamics from frames {actions} and produce the corresponding policy actions {e_actions}.",
+        "special_action_token": "<|action_{}|>",
+        "embodied_action_token": "<|embodied_action|>",
+        "action_dim": 12,
+        "state_dim": 16,
+        "num_action_tokens_per_timestep": 8,
+        "num_embodied_action_tokens_per_instruction": 32,
+        "num_inference_timesteps": 4,
+        "action_hidden_size": 1024,
+        "action_model_type": "DiT-B",
+        "action_num_layers": 16,
+        "action_num_heads": 12,
+        "action_attention_head_dim": 64,
+        "action_dropout": 0.2,
+        "action_num_timestep_buckets": 1000,
+        "action_noise_beta_alpha": 1.5,
+        "action_noise_beta_beta": 1.0,
+        "action_noise_s": 0.999,
+        "num_target_vision_tokens": 32,
+        "action_max_seq_len": 1024,
+        "num_video_frames": 8,
+        "predictor_depth": 12,
+        "predictor_num_heads": 8,
+        "predictor_mlp_ratio": 4.0,
+        "predictor_dropout": 0.0,
+        "world_model_loss_weight": 0.1,
+        "jepa_tubelet_size": 2,
+        "repeated_diffusion_steps": 8,
+        "resize_images_to": [
+            224,
+            224
+        ],
+        "binarize_gripper_action": true,
+        "pre_snap_gripper_action": true,
+        "clip_normalized_actions": true,
+        "gripper_dim": 6,
+        "gripper_threshold": 0.5,
+        "torch_dtype": "bfloat16",
+        "optimizer_lr": 0.0001,
+        "optimizer_betas": [
+            0.9,
+            0.95
+        ],
+        "optimizer_eps": 1e-08,
+        "optimizer_weight_decay": 1e-08,
+        "optimizer_grad_clip_norm": 1.0,
+        "scheduler_warmup_steps": 5000,
+        "scheduler_decay_steps": 30000,
+        "scheduler_decay_lr": 1e-06
+    },
+    "reward_model": null,
+    "output_dir": "/vol/outputs/vla_jepa_robocasa",
+    "job_name": "vla_jepa_robocasa",
+    "resume": false,
+    "seed": 1000,
+    "cudnn_deterministic": false,
+    "num_workers": 16,
+    "batch_size": 32,
+    "prefetch_factor": 4,
+    "persistent_workers": true,
+    "steps": 30000,
+    "eval_freq": 20000,
+    "log_freq": 200,
+    "tolerance_s": 0.0001,
+    "save_checkpoint": true,
+    "save_freq": 5000,
+    "use_policy_training_preset": true,
+    "optimizer": {
+        "type": "adamw",
+        "lr": 0.0001,
+        "weight_decay": 1e-08,
+        "grad_clip_norm": 1.0,
+        "betas": [
+            0.9,
+            0.95
+        ],
+        "eps": 1e-08
+    },
+    "scheduler": {
+        "type": "cosine_decay_with_warmup",
+        "num_warmup_steps": 5000,
+        "num_decay_steps": 30000,
+        "peak_lr": 0.0001,
+        "decay_lr": 1e-06
+    },
+    "eval": {
+        "n_episodes": 50,
+        "batch_size": 11,
+        "use_async_envs": true
+    },
+    "wandb": {
+        "enable": false,
+        "disable_artifact": false,
+        "project": "lerobot",
+        "entity": null,
+        "notes": null,
+        "run_id": null,
+        "mode": null,
+        "add_tags": true
+    },
+    "peft": null,
+    "sample_weighting": null,
+    "rename_map": {},
+    "checkpoint_path": null
+}