Add files using upload-large-folder tool

results/simvla_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_25_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/added_tokens.json

@@ -0,0 +1,3 @@
+{
+  "<PAD>": 32000
+}

results/simvla_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_25_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/dataset_statistics.json

@@ -0,0 +1,218 @@
+{
+  "grab_roller_aloha_agilex_50": {
+    "action": {
+      "mean": [
+        -0.7234118580818176,
+        1.632694959640503,
+        1.139991283416748,
+        -0.7696658968925476,
+        -0.004497084766626358,
+        -1.915460228919983,
+        0.6765856742858887,
+        0.4867003262042999,
+        1.567619800567627,
+        1.0547168254852295,
+        -0.740928053855896,
+        0.0007320955046452582,
+        1.6736525297164917,
+        0.6765856742858887
+      ],
+      "std": [
+        0.9490994811058044,
+        0.7159450650215149,
+        0.5567411184310913,
+        0.3428436517715454,
+        0.023813901469111443,
+        0.9220959544181824,
+        0.44203680753707886,
+        0.2887645363807678,
+        0.6954315304756165,
+        0.5172013640403748,
+        0.32894495129585266,
+        0.02310887910425663,
+        0.8557336330413818,
+        0.44203680753707886
+      ],
+      "max": [
+        0.0,
+        2.482893466949463,
+        2.353947639465332,
+        0.0,
+        0.05954868718981743,
+        0.0,
+        1.0,
+        1.1409132480621338,
+        2.443716287612915,
+        2.2799878120422363,
+        0.0,
+        0.15795069932937622,
+        3.4252119064331055,
+        1.0
+      ],
+      "min": [
+        -7.338869571685791,
+        0.0,
+        0.0,
+        -1.5001972913742065,
+        -0.12379012256860733,
+        -3.4617013931274414,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.385493516921997,
+        -0.048340000212192535,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -7.335215625762939,
+        0.0,
+        0.0,
+        -1.3311991214752197,
+        -0.11849832087755204,
+        -3.424123525619507,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.2851634454727172,
+        -0.0435758301615715,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        0.0,
+        2.478337526321411,
+        2.1660122013092042,
+        0.0,
+        0.03330111399292946,
+        0.0,
+        1.0,
+        1.1318105459213257,
+        2.3855650424957275,
+        1.9841104364395143,
+        0.0,
+        0.11385425746440887,
+        3.409965982437134,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true
+      ]
+    },
+    "proprio": {
+      "mean": [
+        -0.7144084572792053,
+        1.6121407747268677,
+        1.1227450370788574,
+        -0.7565815448760986,
+        -0.0046524289064109325,
+        -1.8903743028640747,
+        0.687203586101532,
+        0.4801884591579437,
+        1.5479238033294678,
+        1.0385884046554565,
+        -0.7280691862106323,
+        0.0008704152423888445,
+        1.651688575744629,
+        0.687203586101532
+      ],
+      "std": [
+        0.9468903541564941,
+        0.7339856028556824,
+        0.5653066039085388,
+        0.3482204079627991,
+        0.023658812046051025,
+        0.9399275779724121,
+        0.4376404583454132,
+        0.29170694947242737,
+        0.7125725150108337,
+        0.5250204801559448,
+        0.33385151624679565,
+        0.022993318736553192,
+        0.8697940111160278,
+        0.4376404583454132
+      ],
+      "max": [
+        0.0,
+        2.482893466949463,
+        2.353947639465332,
+        0.0,
+        0.05954868718981743,
+        0.0,
+        1.0,
+        1.1409132480621338,
+        2.443716287612915,
+        2.2799878120422363,
+        0.0,
+        0.15795069932937622,
+        3.4252119064331055,
+        1.0
+      ],
+      "min": [
+        -7.338869571685791,
+        0.0,
+        0.0,
+        -1.5001972913742065,
+        -0.12379012256860733,
+        -3.4617013931274414,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.385493516921997,
+        -0.048340000212192535,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -7.335215625762939,
+        0.0,
+        0.0,
+        -1.3265374326705932,
+        -0.11849832087755204,
+        -3.424123525619507,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.2640411233901978,
+        -0.0435758301615715,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        0.0,
+        2.478337526321411,
+        2.151273918151856,
+        0.0,
+        0.03329093188047409,
+        0.0,
+        1.0,
+        1.1318105459213257,
+        2.3855650424957275,
+        1.9724502563476562,
+        0.0,
+        0.11385425746440887,
+        3.409965982437134,
+        1.0
+      ]
+    },
+    "num_transitions": 4709,
+    "num_trajectories": 50
+  }
+}

results/simvla_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_25_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/lora_adapter/README.md

@@ -0,0 +1,202 @@
+---
+base_model: /inspire/hdd/ws-f4d69b29-e0a5-44e6-bd92-acf4de9990f0/public-project/chengdongzhou-240108390137/ai_models/openvla/openvla-7b
+library_name: peft
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.11.1

results/simvla_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_25_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/lora_adapter/adapter_config.json

@@ -0,0 +1,45 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": {
+    "base_model_class": "OpenVLAForActionPrediction",
+    "parent_library": "transformers_modules.openvla-7b.modeling_prismatic"
+  },
+  "base_model_name_or_path": "/inspire/hdd/ws-f4d69b29-e0a5-44e6-bd92-acf4de9990f0/public-project/chengdongzhou-240108390137/ai_models/openvla/openvla-7b",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": "gaussian",
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_dropout": 0.0,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "v_proj",
+    "o_proj",
+    "kv",
+    "up_proj",
+    "fc2",
+    "down_proj",
+    "q_proj",
+    "k_proj",
+    "q",
+    "gate_proj",
+    "qkv",
+    "fc3",
+    "lm_head",
+    "fc1",
+    "proj"
+  ],
+  "task_type": null,
+  "use_dora": false,
+  "use_rslora": false
+}

results/simvla_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_25_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/preprocessor_config.json

@@ -0,0 +1,114 @@
+{
+  "auto_map": {
+    "AutoImageProcessor": "processing_prismatic.PrismaticImageProcessor",
+    "AutoProcessor": "processing_prismatic.PrismaticProcessor"
+  },
+  "image_processor_type": "PrismaticImageProcessor",
+  "image_resize_strategy": "resize-naive",
+  "input_sizes": [
+    [
+      3,
+      224,
+      224
+    ],
+    [
+      3,
+      224,
+      224
+    ]
+  ],
+  "interpolations": [
+    "bicubic",
+    "bicubic"
+  ],
+  "means": [
+    [
+      0.485,
+      0.456,
+      0.406
+    ],
+    [
+      0.5,
+      0.5,
+      0.5
+    ]
+  ],
+  "processor_class": "PrismaticProcessor",
+  "stds": [
+    [
+      0.229,
+      0.224,
+      0.225
+    ],
+    [
+      0.5,
+      0.5,
+      0.5
+    ]
+  ],
+  "tvf_crop_params": [
+    {
+      "output_size": [
+        224,
+        224
+      ]
+    },
+    {
+      "output_size": [
+        224,
+        224
+      ]
+    }
+  ],
+  "tvf_do_letterbox": false,
+  "tvf_letterbox_fill": null,
+  "tvf_normalize_params": [
+    {
+      "inplace": false,
+      "mean": [
+        0.484375,
+        0.455078125,
+        0.40625
+      ],
+      "std": [
+        0.228515625,
+        0.2236328125,
+        0.224609375
+      ]
+    },
+    {
+      "inplace": false,
+      "mean": [
+        0.5,
+        0.5,
+        0.5
+      ],
+      "std": [
+        0.5,
+        0.5,
+        0.5
+      ]
+    }
+  ],
+  "tvf_resize_params": [
+    {
+      "antialias": true,
+      "interpolation": 3,
+      "max_size": null,
+      "size": [
+        224,
+        224
+      ]
+    },
+    {
+      "antialias": true,
+      "interpolation": 3,
+      "max_size": null,
+      "size": [
+        224,
+        224
+      ]
+    }
+  ],
+  "use_fused_vision_backbone": true
+}

results/simvla_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_25_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/processing_prismatic.py

@@ -0,0 +1,257 @@
+"""
+processing_prismatic.py
+
+HuggingFace-style preprocessor definitions for Prismatic VLMs, inheriting from `ProcessorMixin`. Default configuration
+specifies `siglip-224px+7b`.
+"""
+
+from typing import Any, ClassVar, List, Optional, Tuple, Union
+
+import timm.data
+import torch
+import torchvision.transforms.functional as TVF
+from PIL import Image
+from torchvision.transforms import CenterCrop, Compose, Normalize, Resize, ToTensor
+from transformers import PreTrainedTokenizerBase
+from transformers.image_processing_utils import BatchFeature, ImageProcessingMixin
+from transformers.processing_utils import ProcessorMixin
+from transformers.tokenization_utils import PaddingStrategy, PreTokenizedInput, TextInput, TruncationStrategy
+from transformers.utils import TensorType
+
+
+# === Image Processing ===
+def letterbox_pad_transform(image: Image.Image, padding_fill_value: Tuple[int, int, int]) -> Image.Image:
+    """Given a PIL.Image, pad to square by adding a symmetric border around the height/width."""
+    (w, h), max_wh = image.size, max(image.size)
+    horizontal_pad, vertical_pad = int((max_wh - w) / 2), int((max_wh - h) / 2)
+    padding = (horizontal_pad, vertical_pad, horizontal_pad, vertical_pad)
+
+    return TVF.pad(image, padding, fill=padding_fill_value, padding_mode="constant")
+
+
+class PrismaticImageProcessor(ImageProcessingMixin):
+    model_input_names: ClassVar[List[str]] = ["pixel_values"]
+
+    def __init__(
+        self,
+        use_fused_vision_backbone: bool = False,
+        image_resize_strategy: str = "letterbox",
+        input_sizes: Optional[List[Tuple[int, int, int]]] = None,
+        interpolations: Optional[List[str]] = None,
+        means: Optional[List[Tuple[float, float, float]]] = None,
+        stds: Optional[List[Tuple[float, float, float]]] = None,
+        **kwargs: str,
+    ) -> None:
+        """
+        Initialize a PrismaticImageProcessor as a wrapper around a torchvision transform; this transform will be
+        created by TIMM, and edited to follow our custom `image_resize_strategy` logic.
+
+        @param use_fused_vision_backbone: Boolean indicating single or fused (dual) vision backbone
+        @param image_resize_strategy: Prismatic image resize strategy in < resize-naive | resize-crop | letterbox >
+        @param input_size: [TIMM :: `data_cfg`] Input image size as tuple (channels, width, height)
+        @param interpolation: [TIMM :: `data_cfg`] Interpolation as string (default: "bicubic")
+        @param mean: [TIMM :: `data_cfg`] Normalization mean as float tuple (or two-tuple if `fused_backbone`)
+        @param std: [TIMM :: `data_cfg`] Normalization std as float tuple (or two-tuple if `fused_backbone`)
+        """
+        self.use_fused_vision_backbone = use_fused_vision_backbone
+        self.image_resize_strategy = image_resize_strategy
+
+        # Handle `None` default values
+        input_sizes = [(3, 224, 224)] if input_sizes is None else input_sizes
+        means = [(0.5, 0.5, 0.5)] if means is None else means
+        stds = [(0.5, 0.5, 0.5)] if stds is None else stds
+
+        # TIMM `data_cfg` Parameters
+        self.input_sizes, self.interpolations, self.means, self.stds = input_sizes, interpolations, means, stds
+
+        # Grab torchvision transforms via TIMM =>> need to parse for specific "functional" transform values!
+        self.tvf_resize_params, self.tvf_crop_params, self.tvf_normalize_params = [], [], []
+        self.tvf_do_letterbox, self.tvf_letterbox_fill = False, None
+
+        for idx in range(len(input_sizes)):
+            transform = timm.data.create_transform(
+                input_size=self.input_sizes[idx],
+                interpolation=self.interpolations[idx],
+                mean=self.means[idx],
+                std=self.stds[idx],
+                crop_pct=1.0,  # Set to 1.0 to ignore cropping (initial Resize sets `input_size`)
+                crop_mode="center",  # Default crop mode -- no-op when `crop_pct == 1.0`
+                is_training=False,  # No image augmentations when loading the transform!
+            )
+
+            # [Validation] Ensure appropriate transform structure, expected sizes
+            if not (
+                isinstance(transform, Compose)
+                and (len(transform.transforms) == 4)
+                and isinstance(transform.transforms[0], Resize)
+                and isinstance(transform.transforms[1], CenterCrop)
+                and isinstance(transform.transforms[2], ToTensor)
+                and isinstance(transform.transforms[3], Normalize)
+                and (transform.transforms[0].size == self.input_sizes[idx][-1])
+                and (transform.transforms[1].size == self.input_sizes[idx][-2:])
+            ):
+                raise ValueError(f"Unexpected TIMM image transformation structure/sizes: `{transform}`")
+
+            # HF Image Processors *must* be JSON-serializable; as such, cannot have torchvision. as an attribute.
+            #   => Instead, we're going to parse the transform and call "torchvision.transforms.functional" (`tvf`)
+            resize_t, crop_t, norm_t = transform.transforms[0], transform.transforms[1], transform.transforms[3]
+            self.tvf_resize_params.append(
+                {
+                    "size": resize_t.size,
+                    "interpolation": TVF.pil_modes_mapping[resize_t.interpolation],
+                    "max_size": None,
+                    "antialias": True,
+                }
+            )
+            self.tvf_crop_params.append({"output_size": crop_t.size})
+            self.tvf_normalize_params.append(
+                {
+                    "mean": norm_t.mean.float().numpy().tolist(),
+                    "std": norm_t.std.float().numpy().tolist(),
+                    "inplace": False,
+                }
+            )
+            self.tvf_do_letterbox, self.tvf_letterbox_fill = False, None
+
+            # Handle Prismatic `image_resize_strategy`
+            if self.image_resize_strategy == "resize-naive":
+                self.tvf_resize_params[idx]["size"] = (resize_t.size, resize_t.size)
+            elif self.image_resize_strategy == "letterbox":
+                self.tvf_do_letterbox, self.tvf_letterbox_fill = True, tuple([int(x * 255) for x in self.means[idx]])
+            elif self.image_resize_strategy == "resize-crop":
+                pass
+            else:
+                raise ValueError(f"Image resize strategy `{self.image_resize_strategy}` is not supported!")
+
+        # Dispatch **kwargs to super()
+        super().__init__(**kwargs)
+
+    def apply_transform(self, img: Image.Image) -> torch.Tensor:
+        """Apply `functional` variant of TIMM's Transform = Compose([Resize -> CenterCrop -> ToTensor -> Normalize])"""
+        if self.tvf_do_letterbox:
+            img = letterbox_pad_transform(img, self.tvf_letterbox_fill)
+
+        # [Contract] Fused Backbones expect "channel-stacked" inputs; we'll unpack on the model side!
+        imgs_t = []
+        for idx in range(len(self.input_sizes)):
+            img_idx = TVF.resize(img, **self.tvf_resize_params[idx])
+            img_idx = TVF.center_crop(img_idx, **self.tvf_crop_params[idx])
+            img_idx_t = TVF.to_tensor(img_idx)
+            img_idx_t = TVF.normalize(img_idx_t, **self.tvf_normalize_params[idx])
+            imgs_t.append(img_idx_t)
+
+        # [Contract] `imgs_t` is a list of Tensors of shape [3, input_size, input_size]; stack along dim = 0
+        img_t = torch.vstack(imgs_t)
+
+        return img_t
+
+    def preprocess(
+        self,
+        images: Union[Image.Image, List[Image.Image]],
+        return_tensors: Optional[Union[str, TensorType]] = None,
+        **_: str,
+    ) -> BatchFeature:
+        """
+        Preprocess an image (or batch of images); note that unlike the `transformers :: BaseImageProcessor` we
+        explicitly only handle PIL.Image.Image instances for simplicity.
+
+        @param images: A (batch of) PIL.Image.Image instance(s) to preprocess.
+        @param return_tensors: BatchFeature default Tensor format (e.g., "pt" for torch); if None, returns np.ndarray
+
+        @return: Instance of `transformers :: BatchFeature` with a single key "pixel_values"
+        """
+        if not isinstance(images, list):
+            images = [images]
+
+        # Apply `self.img_transform` to each image (will return list of torch.Tensors); stack into "batched" Tensor
+        pixel_values = torch.stack([self.apply_transform(img.convert("RGB")) for img in images])
+
+        # Return BatchFeature =>> note that for compatibility, constructor expects Dict[str, np.ndarray], so we convert
+        return BatchFeature(data={"pixel_values": pixel_values.float().numpy()}, tensor_type=return_tensors)
+
+    def __call__(self, images: Union[Image.Image, List[Image.Image]], **kwargs) -> BatchFeature:
+        return self.preprocess(images, **kwargs)
+
+
+# === PrismaticProcessor =>> Wraps both ImageProcessor and Tokenizer ===
+#   =>> https://github.com/huggingface/transformers/blob/main/src/transformers/models/llava/processing_llava.py
+class PrismaticProcessor(ProcessorMixin):
+    attributes: ClassVar[List[str]] = ["image_processor", "tokenizer"]
+    image_processor_class: str = "AutoImageProcessor"
+    tokenizer_class: str = "AutoTokenizer"
+
+    def __init__(
+        self,
+        image_processor: Optional[ImageProcessingMixin] = None,
+        tokenizer: Optional[PreTrainedTokenizerBase] = None,
+    ) -> None:
+        super().__init__(image_processor, tokenizer)
+
+    def __call__(
+        self,
+        text: Union[TextInput, PreTokenizedInput, List[TextInput], List[PreTokenizedInput]],
+        images: Union[Image.Image, List[Image.Image]],
+        padding: Union[bool, str, PaddingStrategy] = False,
+        truncation: Optional[Union[bool, str, TruncationStrategy]] = None,
+        max_length: Optional[int] = None,
+        return_tensors: Optional[Union[str, TensorType]] = TensorType.PYTORCH,
+    ) -> BatchFeature:
+        """
+        Preprocess a given (batch) of text/images for a Prismatic VLM; forwards text to the underlying LLM's tokenizer,
+        forwards images to PrismaticImageProcessor.
+
+        @param text: The (batch) of text to encode; must be a string or list of strings.
+        @param images: A (batch of) PIL.Image.Image instance(s) to preprocess.
+        @param padding: Sequence padding strategy (if multiple specified) in < True = "longest" | "max_length" | False >
+        @param truncation: Truncation strategy for the output sequences; requires `max_length` to be specified
+        @param max_length: Maximum length (in tokens) to truncate
+        @param return_tensors: Type of return tensors (usually "pt" or TensorType.PYTORCH)
+
+        @return: BatchFeature with keys for `input_ids`, `attention_mask` and `pixel_values`.
+        """
+        pixel_values = self.image_processor(images, return_tensors=return_tensors)["pixel_values"]
+        text_inputs = self.tokenizer(
+            text, return_tensors=return_tensors, padding=padding, truncation=truncation, max_length=max_length
+        )
+
+        # [Validate] Need same number of images and text inputs!
+        if pixel_values.shape[0] != text_inputs.input_ids.shape[0]:
+            raise ValueError("Batch is malformed; expected same number of images and text inputs!")
+
+        return BatchFeature(data={**text_inputs, "pixel_values": pixel_values})
+
+    # === Tokenizer Dispatch Utilities =>> check `PreTrainedTokenizerBase` for documentation ===
+    def batch_decode(
+        self,
+        sequences: Union[List[int], List[List[int]], torch.Tensor, Any],  # `Any` = np.ndarray | tf.Tensor
+        skip_special_tokens: bool = False,
+        clean_up_tokenization_spaces: Optional[bool] = None,
+        **kwargs: str,
+    ) -> List[str]:
+        return self.tokenizer.batch_decode(
+            sequences=sequences,
+            skip_special_tokens=skip_special_tokens,
+            clean_up_tokenization_spaces=clean_up_tokenization_spaces,
+            **kwargs,
+        )
+
+    def decode(
+        self,
+        token_ids: Union[int, List[int], torch.Tensor, Any],  # `Any` = np.ndarray | tf.Tensor
+        skip_special_tokens: bool = False,
+        clean_up_tokenization_spaces: Optional[bool] = None,
+        **kwargs: str,
+    ) -> str:
+        return self.tokenizer.decode(
+            token_ids=token_ids,
+            skip_special_tokens=skip_special_tokens,
+            clean_up_tokenization_spaces=clean_up_tokenization_spaces,
+            **kwargs,
+        )
+
+    @property
+    def model_input_names(self) -> List[str]:
+        tokenizer_input_names = self.tokenizer.model_input_names
+        image_processor_input_names = self.image_processor.model_input_names
+
+        return list(dict.fromkeys(tokenizer_input_names + image_processor_input_names))

results/simvla_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_25_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/processor_config.json

@@ -0,0 +1,6 @@
+{
+  "auto_map": {
+    "AutoProcessor": "processing_prismatic.PrismaticProcessor"
+  },
+  "processor_class": "PrismaticProcessor"
+}

results/simvla_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_25_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/special_tokens_map.json

@@ -0,0 +1,30 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<PAD>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

results/simvla_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_25_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/tokenizer_config.json

@@ -0,0 +1,53 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "32000": {
+      "content": "<PAD>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "auto_map": {
+    "AutoProcessor": "processing_prismatic.PrismaticProcessor"
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "</s>",
+  "legacy": false,
+  "model_max_length": 2048,
+  "pad_token": "<PAD>",
+  "padding_side": "right",
+  "processor_class": "PrismaticProcessor",
+  "sp_model_kwargs": {},
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}

results/simvla_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_25_inner1_proj_type_onlynorm_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/added_tokens.json

@@ -0,0 +1,3 @@
+{
+  "<PAD>": 32000
+}

results/simvla_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_25_inner1_proj_type_onlynorm_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/dataset_statistics.json

@@ -0,0 +1,218 @@
+{
+  "grab_roller_aloha_agilex_50": {
+    "action": {
+      "mean": [
+        -0.7234118580818176,
+        1.632694959640503,
+        1.139991283416748,
+        -0.7696658968925476,
+        -0.004497084766626358,
+        -1.915460228919983,
+        0.6765856742858887,
+        0.4867003262042999,
+        1.567619800567627,
+        1.0547168254852295,
+        -0.740928053855896,
+        0.0007320955046452582,
+        1.6736525297164917,
+        0.6765856742858887
+      ],
+      "std": [
+        0.9490994811058044,
+        0.7159450650215149,
+        0.5567411184310913,
+        0.3428436517715454,
+        0.023813901469111443,
+        0.9220959544181824,
+        0.44203680753707886,
+        0.2887645363807678,
+        0.6954315304756165,
+        0.5172013640403748,
+        0.32894495129585266,
+        0.02310887910425663,
+        0.8557336330413818,
+        0.44203680753707886
+      ],
+      "max": [
+        0.0,
+        2.482893466949463,
+        2.353947639465332,
+        0.0,
+        0.05954868718981743,
+        0.0,
+        1.0,
+        1.1409132480621338,
+        2.443716287612915,
+        2.2799878120422363,
+        0.0,
+        0.15795069932937622,
+        3.4252119064331055,
+        1.0
+      ],
+      "min": [
+        -7.338869571685791,
+        0.0,
+        0.0,
+        -1.5001972913742065,
+        -0.12379012256860733,
+        -3.4617013931274414,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.385493516921997,
+        -0.048340000212192535,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -7.335215625762939,
+        0.0,
+        0.0,
+        -1.3311991214752197,
+        -0.11849832087755204,
+        -3.424123525619507,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.2851634454727172,
+        -0.0435758301615715,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        0.0,
+        2.478337526321411,
+        2.1660122013092042,
+        0.0,
+        0.03330111399292946,
+        0.0,
+        1.0,
+        1.1318105459213257,
+        2.3855650424957275,
+        1.9841104364395143,
+        0.0,
+        0.11385425746440887,
+        3.409965982437134,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true
+      ]
+    },
+    "proprio": {
+      "mean": [
+        -0.7144084572792053,
+        1.6121407747268677,
+        1.1227450370788574,
+        -0.7565815448760986,
+        -0.0046524289064109325,
+        -1.8903743028640747,
+        0.687203586101532,
+        0.4801884591579437,
+        1.5479238033294678,
+        1.0385884046554565,
+        -0.7280691862106323,
+        0.0008704152423888445,
+        1.651688575744629,
+        0.687203586101532
+      ],
+      "std": [
+        0.9468903541564941,
+        0.7339856028556824,
+        0.5653066039085388,
+        0.3482204079627991,
+        0.023658812046051025,
+        0.9399275779724121,
+        0.4376404583454132,
+        0.29170694947242737,
+        0.7125725150108337,
+        0.5250204801559448,
+        0.33385151624679565,
+        0.022993318736553192,
+        0.8697940111160278,
+        0.4376404583454132
+      ],
+      "max": [
+        0.0,
+        2.482893466949463,
+        2.353947639465332,
+        0.0,
+        0.05954868718981743,
+        0.0,
+        1.0,
+        1.1409132480621338,
+        2.443716287612915,
+        2.2799878120422363,
+        0.0,
+        0.15795069932937622,
+        3.4252119064331055,
+        1.0
+      ],
+      "min": [
+        -7.338869571685791,
+        0.0,
+        0.0,
+        -1.5001972913742065,
+        -0.12379012256860733,
+        -3.4617013931274414,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.385493516921997,
+        -0.048340000212192535,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -7.335215625762939,
+        0.0,
+        0.0,
+        -1.3265374326705932,
+        -0.11849832087755204,
+        -3.424123525619507,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.2640411233901978,
+        -0.0435758301615715,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        0.0,
+        2.478337526321411,
+        2.151273918151856,
+        0.0,
+        0.03329093188047409,
+        0.0,
+        1.0,
+        1.1318105459213257,
+        2.3855650424957275,
+        1.9724502563476562,
+        0.0,
+        0.11385425746440887,
+        3.409965982437134,
+        1.0
+      ]
+    },
+    "num_transitions": 4709,
+    "num_trajectories": 50
+  }
+}

results/simvla_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_25_inner1_proj_type_onlynorm_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/lora_adapter/adapter_config.json

@@ -0,0 +1,45 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": {
+    "base_model_class": "OpenVLAForActionPrediction",
+    "parent_library": "transformers_modules.openvla-7b.modeling_prismatic"
+  },
+  "base_model_name_or_path": "/inspire/hdd/ws-f4d69b29-e0a5-44e6-bd92-acf4de9990f0/public-project/chengdongzhou-240108390137/ai_models/openvla/openvla-7b",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": "gaussian",
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_dropout": 0.0,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q",
+    "down_proj",
+    "k_proj",
+    "fc1",
+    "qkv",
+    "fc2",
+    "v_proj",
+    "gate_proj",
+    "up_proj",
+    "fc3",
+    "lm_head",
+    "proj",
+    "q_proj",
+    "kv",
+    "o_proj"
+  ],
+  "task_type": null,
+  "use_dora": false,
+  "use_rslora": false
+}

results/simvla_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_25_inner1_proj_type_onlynorm_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/preprocessor_config.json

@@ -0,0 +1,114 @@
+{
+  "auto_map": {
+    "AutoImageProcessor": "processing_prismatic.PrismaticImageProcessor",
+    "AutoProcessor": "processing_prismatic.PrismaticProcessor"
+  },
+  "image_processor_type": "PrismaticImageProcessor",
+  "image_resize_strategy": "resize-naive",
+  "input_sizes": [
+    [
+      3,
+      224,
+      224
+    ],
+    [
+      3,
+      224,
+      224
+    ]
+  ],
+  "interpolations": [
+    "bicubic",
+    "bicubic"
+  ],
+  "means": [
+    [
+      0.485,
+      0.456,
+      0.406
+    ],
+    [
+      0.5,
+      0.5,
+      0.5
+    ]
+  ],
+  "processor_class": "PrismaticProcessor",
+  "stds": [
+    [
+      0.229,
+      0.224,
+      0.225
+    ],
+    [
+      0.5,
+      0.5,
+      0.5
+    ]
+  ],
+  "tvf_crop_params": [
+    {
+      "output_size": [
+        224,
+        224
+      ]
+    },
+    {
+      "output_size": [
+        224,
+        224
+      ]
+    }
+  ],
+  "tvf_do_letterbox": false,
+  "tvf_letterbox_fill": null,
+  "tvf_normalize_params": [
+    {
+      "inplace": false,
+      "mean": [
+        0.484375,
+        0.455078125,
+        0.40625
+      ],
+      "std": [
+        0.228515625,
+        0.2236328125,
+        0.224609375
+      ]
+    },
+    {
+      "inplace": false,
+      "mean": [
+        0.5,
+        0.5,
+        0.5
+      ],
+      "std": [
+        0.5,
+        0.5,
+        0.5
+      ]
+    }
+  ],
+  "tvf_resize_params": [
+    {
+      "antialias": true,
+      "interpolation": 3,
+      "max_size": null,
+      "size": [
+        224,
+        224
+      ]
+    },
+    {
+      "antialias": true,
+      "interpolation": 3,
+      "max_size": null,
+      "size": [
+        224,
+        224
+      ]
+    }
+  ],
+  "use_fused_vision_backbone": true
+}

results/simvla_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_25_inner1_proj_type_onlynorm_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/processing_prismatic.py

@@ -0,0 +1,257 @@
+"""
+processing_prismatic.py
+
+HuggingFace-style preprocessor definitions for Prismatic VLMs, inheriting from `ProcessorMixin`. Default configuration
+specifies `siglip-224px+7b`.
+"""
+
+from typing import Any, ClassVar, List, Optional, Tuple, Union
+
+import timm.data
+import torch
+import torchvision.transforms.functional as TVF
+from PIL import Image
+from torchvision.transforms import CenterCrop, Compose, Normalize, Resize, ToTensor
+from transformers import PreTrainedTokenizerBase
+from transformers.image_processing_utils import BatchFeature, ImageProcessingMixin
+from transformers.processing_utils import ProcessorMixin
+from transformers.tokenization_utils import PaddingStrategy, PreTokenizedInput, TextInput, TruncationStrategy
+from transformers.utils import TensorType
+
+
+# === Image Processing ===
+def letterbox_pad_transform(image: Image.Image, padding_fill_value: Tuple[int, int, int]) -> Image.Image:
+    """Given a PIL.Image, pad to square by adding a symmetric border around the height/width."""
+    (w, h), max_wh = image.size, max(image.size)
+    horizontal_pad, vertical_pad = int((max_wh - w) / 2), int((max_wh - h) / 2)
+    padding = (horizontal_pad, vertical_pad, horizontal_pad, vertical_pad)
+
+    return TVF.pad(image, padding, fill=padding_fill_value, padding_mode="constant")
+
+
+class PrismaticImageProcessor(ImageProcessingMixin):
+    model_input_names: ClassVar[List[str]] = ["pixel_values"]
+
+    def __init__(
+        self,
+        use_fused_vision_backbone: bool = False,
+        image_resize_strategy: str = "letterbox",
+        input_sizes: Optional[List[Tuple[int, int, int]]] = None,
+        interpolations: Optional[List[str]] = None,
+        means: Optional[List[Tuple[float, float, float]]] = None,
+        stds: Optional[List[Tuple[float, float, float]]] = None,
+        **kwargs: str,
+    ) -> None:
+        """
+        Initialize a PrismaticImageProcessor as a wrapper around a torchvision transform; this transform will be
+        created by TIMM, and edited to follow our custom `image_resize_strategy` logic.
+
+        @param use_fused_vision_backbone: Boolean indicating single or fused (dual) vision backbone
+        @param image_resize_strategy: Prismatic image resize strategy in < resize-naive | resize-crop | letterbox >
+        @param input_size: [TIMM :: `data_cfg`] Input image size as tuple (channels, width, height)
+        @param interpolation: [TIMM :: `data_cfg`] Interpolation as string (default: "bicubic")
+        @param mean: [TIMM :: `data_cfg`] Normalization mean as float tuple (or two-tuple if `fused_backbone`)
+        @param std: [TIMM :: `data_cfg`] Normalization std as float tuple (or two-tuple if `fused_backbone`)
+        """
+        self.use_fused_vision_backbone = use_fused_vision_backbone
+        self.image_resize_strategy = image_resize_strategy
+
+        # Handle `None` default values
+        input_sizes = [(3, 224, 224)] if input_sizes is None else input_sizes
+        means = [(0.5, 0.5, 0.5)] if means is None else means
+        stds = [(0.5, 0.5, 0.5)] if stds is None else stds
+
+        # TIMM `data_cfg` Parameters
+        self.input_sizes, self.interpolations, self.means, self.stds = input_sizes, interpolations, means, stds
+
+        # Grab torchvision transforms via TIMM =>> need to parse for specific "functional" transform values!
+        self.tvf_resize_params, self.tvf_crop_params, self.tvf_normalize_params = [], [], []
+        self.tvf_do_letterbox, self.tvf_letterbox_fill = False, None
+
+        for idx in range(len(input_sizes)):
+            transform = timm.data.create_transform(
+                input_size=self.input_sizes[idx],
+                interpolation=self.interpolations[idx],
+                mean=self.means[idx],
+                std=self.stds[idx],
+                crop_pct=1.0,  # Set to 1.0 to ignore cropping (initial Resize sets `input_size`)
+                crop_mode="center",  # Default crop mode -- no-op when `crop_pct == 1.0`
+                is_training=False,  # No image augmentations when loading the transform!
+            )
+
+            # [Validation] Ensure appropriate transform structure, expected sizes
+            if not (
+                isinstance(transform, Compose)
+                and (len(transform.transforms) == 4)
+                and isinstance(transform.transforms[0], Resize)
+                and isinstance(transform.transforms[1], CenterCrop)
+                and isinstance(transform.transforms[2], ToTensor)
+                and isinstance(transform.transforms[3], Normalize)
+                and (transform.transforms[0].size == self.input_sizes[idx][-1])
+                and (transform.transforms[1].size == self.input_sizes[idx][-2:])
+            ):
+                raise ValueError(f"Unexpected TIMM image transformation structure/sizes: `{transform}`")
+
+            # HF Image Processors *must* be JSON-serializable; as such, cannot have torchvision. as an attribute.
+            #   => Instead, we're going to parse the transform and call "torchvision.transforms.functional" (`tvf`)
+            resize_t, crop_t, norm_t = transform.transforms[0], transform.transforms[1], transform.transforms[3]
+            self.tvf_resize_params.append(
+                {
+                    "size": resize_t.size,
+                    "interpolation": TVF.pil_modes_mapping[resize_t.interpolation],
+                    "max_size": None,
+                    "antialias": True,
+                }
+            )
+            self.tvf_crop_params.append({"output_size": crop_t.size})
+            self.tvf_normalize_params.append(
+                {
+                    "mean": norm_t.mean.float().numpy().tolist(),
+                    "std": norm_t.std.float().numpy().tolist(),
+                    "inplace": False,
+                }
+            )
+            self.tvf_do_letterbox, self.tvf_letterbox_fill = False, None
+
+            # Handle Prismatic `image_resize_strategy`
+            if self.image_resize_strategy == "resize-naive":
+                self.tvf_resize_params[idx]["size"] = (resize_t.size, resize_t.size)
+            elif self.image_resize_strategy == "letterbox":
+                self.tvf_do_letterbox, self.tvf_letterbox_fill = True, tuple([int(x * 255) for x in self.means[idx]])
+            elif self.image_resize_strategy == "resize-crop":
+                pass
+            else:
+                raise ValueError(f"Image resize strategy `{self.image_resize_strategy}` is not supported!")
+
+        # Dispatch **kwargs to super()
+        super().__init__(**kwargs)
+
+    def apply_transform(self, img: Image.Image) -> torch.Tensor:
+        """Apply `functional` variant of TIMM's Transform = Compose([Resize -> CenterCrop -> ToTensor -> Normalize])"""
+        if self.tvf_do_letterbox:
+            img = letterbox_pad_transform(img, self.tvf_letterbox_fill)
+
+        # [Contract] Fused Backbones expect "channel-stacked" inputs; we'll unpack on the model side!
+        imgs_t = []
+        for idx in range(len(self.input_sizes)):
+            img_idx = TVF.resize(img, **self.tvf_resize_params[idx])
+            img_idx = TVF.center_crop(img_idx, **self.tvf_crop_params[idx])
+            img_idx_t = TVF.to_tensor(img_idx)
+            img_idx_t = TVF.normalize(img_idx_t, **self.tvf_normalize_params[idx])
+            imgs_t.append(img_idx_t)
+
+        # [Contract] `imgs_t` is a list of Tensors of shape [3, input_size, input_size]; stack along dim = 0
+        img_t = torch.vstack(imgs_t)
+
+        return img_t
+
+    def preprocess(
+        self,
+        images: Union[Image.Image, List[Image.Image]],
+        return_tensors: Optional[Union[str, TensorType]] = None,
+        **_: str,
+    ) -> BatchFeature:
+        """
+        Preprocess an image (or batch of images); note that unlike the `transformers :: BaseImageProcessor` we
+        explicitly only handle PIL.Image.Image instances for simplicity.
+
+        @param images: A (batch of) PIL.Image.Image instance(s) to preprocess.
+        @param return_tensors: BatchFeature default Tensor format (e.g., "pt" for torch); if None, returns np.ndarray
+
+        @return: Instance of `transformers :: BatchFeature` with a single key "pixel_values"
+        """
+        if not isinstance(images, list):
+            images = [images]
+
+        # Apply `self.img_transform` to each image (will return list of torch.Tensors); stack into "batched" Tensor
+        pixel_values = torch.stack([self.apply_transform(img.convert("RGB")) for img in images])
+
+        # Return BatchFeature =>> note that for compatibility, constructor expects Dict[str, np.ndarray], so we convert
+        return BatchFeature(data={"pixel_values": pixel_values.float().numpy()}, tensor_type=return_tensors)
+
+    def __call__(self, images: Union[Image.Image, List[Image.Image]], **kwargs) -> BatchFeature:
+        return self.preprocess(images, **kwargs)
+
+
+# === PrismaticProcessor =>> Wraps both ImageProcessor and Tokenizer ===
+#   =>> https://github.com/huggingface/transformers/blob/main/src/transformers/models/llava/processing_llava.py
+class PrismaticProcessor(ProcessorMixin):
+    attributes: ClassVar[List[str]] = ["image_processor", "tokenizer"]
+    image_processor_class: str = "AutoImageProcessor"
+    tokenizer_class: str = "AutoTokenizer"
+
+    def __init__(
+        self,
+        image_processor: Optional[ImageProcessingMixin] = None,
+        tokenizer: Optional[PreTrainedTokenizerBase] = None,
+    ) -> None:
+        super().__init__(image_processor, tokenizer)
+
+    def __call__(
+        self,
+        text: Union[TextInput, PreTokenizedInput, List[TextInput], List[PreTokenizedInput]],
+        images: Union[Image.Image, List[Image.Image]],
+        padding: Union[bool, str, PaddingStrategy] = False,
+        truncation: Optional[Union[bool, str, TruncationStrategy]] = None,
+        max_length: Optional[int] = None,
+        return_tensors: Optional[Union[str, TensorType]] = TensorType.PYTORCH,
+    ) -> BatchFeature:
+        """
+        Preprocess a given (batch) of text/images for a Prismatic VLM; forwards text to the underlying LLM's tokenizer,
+        forwards images to PrismaticImageProcessor.
+
+        @param text: The (batch) of text to encode; must be a string or list of strings.
+        @param images: A (batch of) PIL.Image.Image instance(s) to preprocess.
+        @param padding: Sequence padding strategy (if multiple specified) in < True = "longest" | "max_length" | False >
+        @param truncation: Truncation strategy for the output sequences; requires `max_length` to be specified
+        @param max_length: Maximum length (in tokens) to truncate
+        @param return_tensors: Type of return tensors (usually "pt" or TensorType.PYTORCH)
+
+        @return: BatchFeature with keys for `input_ids`, `attention_mask` and `pixel_values`.
+        """
+        pixel_values = self.image_processor(images, return_tensors=return_tensors)["pixel_values"]
+        text_inputs = self.tokenizer(
+            text, return_tensors=return_tensors, padding=padding, truncation=truncation, max_length=max_length
+        )
+
+        # [Validate] Need same number of images and text inputs!
+        if pixel_values.shape[0] != text_inputs.input_ids.shape[0]:
+            raise ValueError("Batch is malformed; expected same number of images and text inputs!")
+
+        return BatchFeature(data={**text_inputs, "pixel_values": pixel_values})
+
+    # === Tokenizer Dispatch Utilities =>> check `PreTrainedTokenizerBase` for documentation ===
+    def batch_decode(
+        self,
+        sequences: Union[List[int], List[List[int]], torch.Tensor, Any],  # `Any` = np.ndarray | tf.Tensor
+        skip_special_tokens: bool = False,
+        clean_up_tokenization_spaces: Optional[bool] = None,
+        **kwargs: str,
+    ) -> List[str]:
+        return self.tokenizer.batch_decode(
+            sequences=sequences,
+            skip_special_tokens=skip_special_tokens,
+            clean_up_tokenization_spaces=clean_up_tokenization_spaces,
+            **kwargs,
+        )
+
+    def decode(
+        self,
+        token_ids: Union[int, List[int], torch.Tensor, Any],  # `Any` = np.ndarray | tf.Tensor
+        skip_special_tokens: bool = False,
+        clean_up_tokenization_spaces: Optional[bool] = None,
+        **kwargs: str,
+    ) -> str:
+        return self.tokenizer.decode(
+            token_ids=token_ids,
+            skip_special_tokens=skip_special_tokens,
+            clean_up_tokenization_spaces=clean_up_tokenization_spaces,
+            **kwargs,
+        )
+
+    @property
+    def model_input_names(self) -> List[str]:
+        tokenizer_input_names = self.tokenizer.model_input_names
+        image_processor_input_names = self.image_processor.model_input_names
+
+        return list(dict.fromkeys(tokenizer_input_names + image_processor_input_names))

results/simvla_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_25_inner1_proj_type_onlynorm_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/processor_config.json

@@ -0,0 +1,6 @@
+{
+  "auto_map": {
+    "AutoProcessor": "processing_prismatic.PrismaticProcessor"
+  },
+  "processor_class": "PrismaticProcessor"
+}

results/simvla_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_25_inner1_proj_type_onlynorm_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/special_tokens_map.json

@@ -0,0 +1,30 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<PAD>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

results/simvla_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_25_inner1_proj_type_onlynorm_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/tokenizer_config.json

@@ -0,0 +1,53 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "32000": {
+      "content": "<PAD>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "auto_map": {
+    "AutoProcessor": "processing_prismatic.PrismaticProcessor"
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "</s>",
+  "legacy": false,
+  "model_max_length": 2048,
+  "pad_token": "<PAD>",
+  "padding_side": "right",
+  "processor_class": "PrismaticProcessor",
+  "sp_model_kwargs": {},
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}

results/simvla_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_25_inner1_proj_type_relu_linear_ffn_type_relu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/added_tokens.json

@@ -0,0 +1,3 @@
+{
+  "<PAD>": 32000
+}

results/simvla_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_25_inner1_proj_type_relu_linear_ffn_type_relu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/lora_adapter/adapter_config.json

@@ -0,0 +1,45 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": {
+    "base_model_class": "OpenVLAForActionPrediction",
+    "parent_library": "transformers_modules.openvla-7b.modeling_prismatic"
+  },
+  "base_model_name_or_path": "/inspire/hdd/ws-f4d69b29-e0a5-44e6-bd92-acf4de9990f0/public-project/chengdongzhou-240108390137/ai_models/openvla/openvla-7b",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": "gaussian",
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_dropout": 0.0,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "k_proj",
+    "up_proj",
+    "v_proj",
+    "fc3",
+    "fc1",
+    "o_proj",
+    "qkv",
+    "q_proj",
+    "proj",
+    "gate_proj",
+    "lm_head",
+    "fc2",
+    "q",
+    "down_proj",
+    "kv"
+  ],
+  "task_type": null,
+  "use_dora": false,
+  "use_rslora": false
+}

results/simvla_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_25_inner1_proj_type_relu_linear_ffn_type_relu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/preprocessor_config.json

@@ -0,0 +1,114 @@
+{
+  "auto_map": {
+    "AutoImageProcessor": "processing_prismatic.PrismaticImageProcessor",
+    "AutoProcessor": "processing_prismatic.PrismaticProcessor"
+  },
+  "image_processor_type": "PrismaticImageProcessor",
+  "image_resize_strategy": "resize-naive",
+  "input_sizes": [
+    [
+      3,
+      224,
+      224
+    ],
+    [
+      3,
+      224,
+      224
+    ]
+  ],
+  "interpolations": [
+    "bicubic",
+    "bicubic"
+  ],
+  "means": [
+    [
+      0.485,
+      0.456,
+      0.406
+    ],
+    [
+      0.5,
+      0.5,
+      0.5
+    ]
+  ],
+  "processor_class": "PrismaticProcessor",
+  "stds": [
+    [
+      0.229,
+      0.224,
+      0.225
+    ],
+    [
+      0.5,
+      0.5,
+      0.5
+    ]
+  ],
+  "tvf_crop_params": [
+    {
+      "output_size": [
+        224,
+        224
+      ]
+    },
+    {
+      "output_size": [
+        224,
+        224
+      ]
+    }
+  ],
+  "tvf_do_letterbox": false,
+  "tvf_letterbox_fill": null,
+  "tvf_normalize_params": [
+    {
+      "inplace": false,
+      "mean": [
+        0.484375,
+        0.455078125,
+        0.40625
+      ],
+      "std": [
+        0.228515625,
+        0.2236328125,
+        0.224609375
+      ]
+    },
+    {
+      "inplace": false,
+      "mean": [
+        0.5,
+        0.5,
+        0.5
+      ],
+      "std": [
+        0.5,
+        0.5,
+        0.5
+      ]
+    }
+  ],
+  "tvf_resize_params": [
+    {
+      "antialias": true,
+      "interpolation": 3,
+      "max_size": null,
+      "size": [
+        224,
+        224
+      ]
+    },
+    {
+      "antialias": true,
+      "interpolation": 3,
+      "max_size": null,
+      "size": [
+        224,
+        224
+      ]
+    }
+  ],
+  "use_fused_vision_backbone": true
+}

results/simvla_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_25_inner1_proj_type_relu_linear_ffn_type_relu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/special_tokens_map.json

@@ -0,0 +1,30 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<PAD>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

results/simvla_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_25_inner1_proj_type_relu_linear_ffn_type_relu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/tokenizer_config.json

@@ -0,0 +1,53 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "32000": {
+      "content": "<PAD>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "auto_map": {
+    "AutoProcessor": "processing_prismatic.PrismaticProcessor"
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "</s>",
+  "legacy": false,
+  "model_max_length": 2048,
+  "pad_token": "<PAD>",
+  "padding_side": "right",
+  "processor_class": "PrismaticProcessor",
+  "sp_model_kwargs": {},
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}

results/simvla_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.001+lora-r32+dropout-0.0--image_aug--simvla_25_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000/dataset_statistics.json

@@ -0,0 +1,218 @@
+{
+  "grab_roller_aloha_agilex_50": {
+    "action": {
+      "mean": [
+        -0.7234118580818176,
+        1.632694959640503,
+        1.139991283416748,
+        -0.7696658968925476,
+        -0.004497084766626358,
+        -1.915460228919983,
+        0.6765856742858887,
+        0.4867003262042999,
+        1.567619800567627,
+        1.0547168254852295,
+        -0.740928053855896,
+        0.0007320955046452582,
+        1.6736525297164917,
+        0.6765856742858887
+      ],
+      "std": [
+        0.9490994811058044,
+        0.7159450650215149,
+        0.5567411184310913,
+        0.3428436517715454,
+        0.023813901469111443,
+        0.9220959544181824,
+        0.44203680753707886,
+        0.2887645363807678,
+        0.6954315304756165,
+        0.5172013640403748,
+        0.32894495129585266,
+        0.02310887910425663,
+        0.8557336330413818,
+        0.44203680753707886
+      ],
+      "max": [
+        0.0,
+        2.482893466949463,
+        2.353947639465332,
+        0.0,
+        0.05954868718981743,
+        0.0,
+        1.0,
+        1.1409132480621338,
+        2.443716287612915,
+        2.2799878120422363,
+        0.0,
+        0.15795069932937622,
+        3.4252119064331055,
+        1.0
+      ],
+      "min": [
+        -7.338869571685791,
+        0.0,
+        0.0,
+        -1.5001972913742065,
+        -0.12379012256860733,
+        -3.4617013931274414,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.385493516921997,
+        -0.048340000212192535,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -7.335215625762939,
+        0.0,
+        0.0,
+        -1.3311991214752197,
+        -0.11849832087755204,
+        -3.424123525619507,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.2851634454727172,
+        -0.0435758301615715,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        0.0,
+        2.478337526321411,
+        2.1660122013092042,
+        0.0,
+        0.03330111399292946,
+        0.0,
+        1.0,
+        1.1318105459213257,
+        2.3855650424957275,
+        1.9841104364395143,
+        0.0,
+        0.11385425746440887,
+        3.409965982437134,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true
+      ]
+    },
+    "proprio": {
+      "mean": [
+        -0.7144084572792053,
+        1.6121407747268677,
+        1.1227450370788574,
+        -0.7565815448760986,
+        -0.0046524289064109325,
+        -1.8903743028640747,
+        0.687203586101532,
+        0.4801884591579437,
+        1.5479238033294678,
+        1.0385884046554565,
+        -0.7280691862106323,
+        0.0008704152423888445,
+        1.651688575744629,
+        0.687203586101532
+      ],
+      "std": [
+        0.9468903541564941,
+        0.7339856028556824,
+        0.5653066039085388,
+        0.3482204079627991,
+        0.023658812046051025,
+        0.9399275779724121,
+        0.4376404583454132,
+        0.29170694947242737,
+        0.7125725150108337,
+        0.5250204801559448,
+        0.33385151624679565,
+        0.022993318736553192,
+        0.8697940111160278,
+        0.4376404583454132
+      ],
+      "max": [
+        0.0,
+        2.482893466949463,
+        2.353947639465332,
+        0.0,
+        0.05954868718981743,
+        0.0,
+        1.0,
+        1.1409132480621338,
+        2.443716287612915,
+        2.2799878120422363,
+        0.0,
+        0.15795069932937622,
+        3.4252119064331055,
+        1.0
+      ],
+      "min": [
+        -7.338869571685791,
+        0.0,
+        0.0,
+        -1.5001972913742065,
+        -0.12379012256860733,
+        -3.4617013931274414,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.385493516921997,
+        -0.048340000212192535,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -7.335215625762939,
+        0.0,
+        0.0,
+        -1.3265374326705932,
+        -0.11849832087755204,
+        -3.424123525619507,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.2640411233901978,
+        -0.0435758301615715,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        0.0,
+        2.478337526321411,
+        2.151273918151856,
+        0.0,
+        0.03329093188047409,
+        0.0,
+        1.0,
+        1.1318105459213257,
+        2.3855650424957275,
+        1.9724502563476562,
+        0.0,
+        0.11385425746440887,
+        3.409965982437134,
+        1.0
+      ]
+    },
+    "num_transitions": 4709,
+    "num_trajectories": 50
+  }
+}

results/simvla_newbase_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_newbase_25_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/added_tokens.json

@@ -0,0 +1,3 @@
+{
+  "<PAD>": 32000
+}

results/simvla_newbase_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_newbase_25_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/dataset_statistics.json

@@ -0,0 +1,218 @@
+{
+  "grab_roller_aloha_agilex_50": {
+    "action": {
+      "mean": [
+        -0.7234118580818176,
+        1.632694959640503,
+        1.139991283416748,
+        -0.7696658968925476,
+        -0.004497084766626358,
+        -1.915460228919983,
+        0.6765856742858887,
+        0.4867003262042999,
+        1.567619800567627,
+        1.0547168254852295,
+        -0.740928053855896,
+        0.0007320955046452582,
+        1.6736525297164917,
+        0.6765856742858887
+      ],
+      "std": [
+        0.9490994811058044,
+        0.7159450650215149,
+        0.5567411184310913,
+        0.3428436517715454,
+        0.023813901469111443,
+        0.9220959544181824,
+        0.44203680753707886,
+        0.2887645363807678,
+        0.6954315304756165,
+        0.5172013640403748,
+        0.32894495129585266,
+        0.02310887910425663,
+        0.8557336330413818,
+        0.44203680753707886
+      ],
+      "max": [
+        0.0,
+        2.482893466949463,
+        2.353947639465332,
+        0.0,
+        0.05954868718981743,
+        0.0,
+        1.0,
+        1.1409132480621338,
+        2.443716287612915,
+        2.2799878120422363,
+        0.0,
+        0.15795069932937622,
+        3.4252119064331055,
+        1.0
+      ],
+      "min": [
+        -7.338869571685791,
+        0.0,
+        0.0,
+        -1.5001972913742065,
+        -0.12379012256860733,
+        -3.4617013931274414,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.385493516921997,
+        -0.048340000212192535,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -7.335215625762939,
+        0.0,
+        0.0,
+        -1.3311991214752197,
+        -0.11849832087755204,
+        -3.424123525619507,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.2851634454727172,
+        -0.0435758301615715,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        0.0,
+        2.478337526321411,
+        2.1660122013092042,
+        0.0,
+        0.03330111399292946,
+        0.0,
+        1.0,
+        1.1318105459213257,
+        2.3855650424957275,
+        1.9841104364395143,
+        0.0,
+        0.11385425746440887,
+        3.409965982437134,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true
+      ]
+    },
+    "proprio": {
+      "mean": [
+        -0.7144084572792053,
+        1.6121407747268677,
+        1.1227450370788574,
+        -0.7565815448760986,
+        -0.0046524289064109325,
+        -1.8903743028640747,
+        0.687203586101532,
+        0.4801884591579437,
+        1.5479238033294678,
+        1.0385884046554565,
+        -0.7280691862106323,
+        0.0008704152423888445,
+        1.651688575744629,
+        0.687203586101532
+      ],
+      "std": [
+        0.9468903541564941,
+        0.7339856028556824,
+        0.5653066039085388,
+        0.3482204079627991,
+        0.023658812046051025,
+        0.9399275779724121,
+        0.4376404583454132,
+        0.29170694947242737,
+        0.7125725150108337,
+        0.5250204801559448,
+        0.33385151624679565,
+        0.022993318736553192,
+        0.8697940111160278,
+        0.4376404583454132
+      ],
+      "max": [
+        0.0,
+        2.482893466949463,
+        2.353947639465332,
+        0.0,
+        0.05954868718981743,
+        0.0,
+        1.0,
+        1.1409132480621338,
+        2.443716287612915,
+        2.2799878120422363,
+        0.0,
+        0.15795069932937622,
+        3.4252119064331055,
+        1.0
+      ],
+      "min": [
+        -7.338869571685791,
+        0.0,
+        0.0,
+        -1.5001972913742065,
+        -0.12379012256860733,
+        -3.4617013931274414,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.385493516921997,
+        -0.048340000212192535,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -7.335215625762939,
+        0.0,
+        0.0,
+        -1.3265374326705932,
+        -0.11849832087755204,
+        -3.424123525619507,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.2640411233901978,
+        -0.0435758301615715,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        0.0,
+        2.478337526321411,
+        2.151273918151856,
+        0.0,
+        0.03329093188047409,
+        0.0,
+        1.0,
+        1.1318105459213257,
+        2.3855650424957275,
+        1.9724502563476562,
+        0.0,
+        0.11385425746440887,
+        3.409965982437134,
+        1.0
+      ]
+    },
+    "num_transitions": 4709,
+    "num_trajectories": 50
+  }
+}

results/simvla_newbase_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_newbase_25_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/lora_adapter/README.md

@@ -0,0 +1,202 @@
+---
+base_model: /inspire/hdd/ws-f4d69b29-e0a5-44e6-bd92-acf4de9990f0/public-project/chengdongzhou-240108390137/ai_models/openvla/openvla-7b
+library_name: peft
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.11.1

results/simvla_newbase_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_newbase_25_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/lora_adapter/adapter_config.json

@@ -0,0 +1,45 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": {
+    "base_model_class": "OpenVLAForActionPrediction",
+    "parent_library": "transformers_modules.openvla-7b.modeling_prismatic"
+  },
+  "base_model_name_or_path": "/inspire/hdd/ws-f4d69b29-e0a5-44e6-bd92-acf4de9990f0/public-project/chengdongzhou-240108390137/ai_models/openvla/openvla-7b",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": "gaussian",
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_dropout": 0.0,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "fc2",
+    "q_proj",
+    "up_proj",
+    "qkv",
+    "proj",
+    "o_proj",
+    "gate_proj",
+    "q",
+    "kv",
+    "fc3",
+    "v_proj",
+    "k_proj",
+    "fc1",
+    "down_proj",
+    "lm_head"
+  ],
+  "task_type": null,
+  "use_dora": false,
+  "use_rslora": false
+}

results/simvla_newbase_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_newbase_25_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/preprocessor_config.json

@@ -0,0 +1,114 @@
+{
+  "auto_map": {
+    "AutoImageProcessor": "processing_prismatic.PrismaticImageProcessor",
+    "AutoProcessor": "processing_prismatic.PrismaticProcessor"
+  },
+  "image_processor_type": "PrismaticImageProcessor",
+  "image_resize_strategy": "resize-naive",
+  "input_sizes": [
+    [
+      3,
+      224,
+      224
+    ],
+    [
+      3,
+      224,
+      224
+    ]
+  ],
+  "interpolations": [
+    "bicubic",
+    "bicubic"
+  ],
+  "means": [
+    [
+      0.485,
+      0.456,
+      0.406
+    ],
+    [
+      0.5,
+      0.5,
+      0.5
+    ]
+  ],
+  "processor_class": "PrismaticProcessor",
+  "stds": [
+    [
+      0.229,
+      0.224,
+      0.225
+    ],
+    [
+      0.5,
+      0.5,
+      0.5
+    ]
+  ],
+  "tvf_crop_params": [
+    {
+      "output_size": [
+        224,
+        224
+      ]
+    },
+    {
+      "output_size": [
+        224,
+        224
+      ]
+    }
+  ],
+  "tvf_do_letterbox": false,
+  "tvf_letterbox_fill": null,
+  "tvf_normalize_params": [
+    {
+      "inplace": false,
+      "mean": [
+        0.484375,
+        0.455078125,
+        0.40625
+      ],
+      "std": [
+        0.228515625,
+        0.2236328125,
+        0.224609375
+      ]
+    },
+    {
+      "inplace": false,
+      "mean": [
+        0.5,
+        0.5,
+        0.5
+      ],
+      "std": [
+        0.5,
+        0.5,
+        0.5
+      ]
+    }
+  ],
+  "tvf_resize_params": [
+    {
+      "antialias": true,
+      "interpolation": 3,
+      "max_size": null,
+      "size": [
+        224,
+        224
+      ]
+    },
+    {
+      "antialias": true,
+      "interpolation": 3,
+      "max_size": null,
+      "size": [
+        224,
+        224
+      ]
+    }
+  ],
+  "use_fused_vision_backbone": true
+}

results/simvla_newbase_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_newbase_25_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/processing_prismatic.py

@@ -0,0 +1,257 @@
+"""
+processing_prismatic.py
+
+HuggingFace-style preprocessor definitions for Prismatic VLMs, inheriting from `ProcessorMixin`. Default configuration
+specifies `siglip-224px+7b`.
+"""
+
+from typing import Any, ClassVar, List, Optional, Tuple, Union
+
+import timm.data
+import torch
+import torchvision.transforms.functional as TVF
+from PIL import Image
+from torchvision.transforms import CenterCrop, Compose, Normalize, Resize, ToTensor
+from transformers import PreTrainedTokenizerBase
+from transformers.image_processing_utils import BatchFeature, ImageProcessingMixin
+from transformers.processing_utils import ProcessorMixin
+from transformers.tokenization_utils import PaddingStrategy, PreTokenizedInput, TextInput, TruncationStrategy
+from transformers.utils import TensorType
+
+
+# === Image Processing ===
+def letterbox_pad_transform(image: Image.Image, padding_fill_value: Tuple[int, int, int]) -> Image.Image:
+    """Given a PIL.Image, pad to square by adding a symmetric border around the height/width."""
+    (w, h), max_wh = image.size, max(image.size)
+    horizontal_pad, vertical_pad = int((max_wh - w) / 2), int((max_wh - h) / 2)
+    padding = (horizontal_pad, vertical_pad, horizontal_pad, vertical_pad)
+
+    return TVF.pad(image, padding, fill=padding_fill_value, padding_mode="constant")
+
+
+class PrismaticImageProcessor(ImageProcessingMixin):
+    model_input_names: ClassVar[List[str]] = ["pixel_values"]
+
+    def __init__(
+        self,
+        use_fused_vision_backbone: bool = False,
+        image_resize_strategy: str = "letterbox",
+        input_sizes: Optional[List[Tuple[int, int, int]]] = None,
+        interpolations: Optional[List[str]] = None,
+        means: Optional[List[Tuple[float, float, float]]] = None,
+        stds: Optional[List[Tuple[float, float, float]]] = None,
+        **kwargs: str,
+    ) -> None:
+        """
+        Initialize a PrismaticImageProcessor as a wrapper around a torchvision transform; this transform will be
+        created by TIMM, and edited to follow our custom `image_resize_strategy` logic.
+
+        @param use_fused_vision_backbone: Boolean indicating single or fused (dual) vision backbone
+        @param image_resize_strategy: Prismatic image resize strategy in < resize-naive | resize-crop | letterbox >
+        @param input_size: [TIMM :: `data_cfg`] Input image size as tuple (channels, width, height)
+        @param interpolation: [TIMM :: `data_cfg`] Interpolation as string (default: "bicubic")
+        @param mean: [TIMM :: `data_cfg`] Normalization mean as float tuple (or two-tuple if `fused_backbone`)
+        @param std: [TIMM :: `data_cfg`] Normalization std as float tuple (or two-tuple if `fused_backbone`)
+        """
+        self.use_fused_vision_backbone = use_fused_vision_backbone
+        self.image_resize_strategy = image_resize_strategy
+
+        # Handle `None` default values
+        input_sizes = [(3, 224, 224)] if input_sizes is None else input_sizes
+        means = [(0.5, 0.5, 0.5)] if means is None else means
+        stds = [(0.5, 0.5, 0.5)] if stds is None else stds
+
+        # TIMM `data_cfg` Parameters
+        self.input_sizes, self.interpolations, self.means, self.stds = input_sizes, interpolations, means, stds
+
+        # Grab torchvision transforms via TIMM =>> need to parse for specific "functional" transform values!
+        self.tvf_resize_params, self.tvf_crop_params, self.tvf_normalize_params = [], [], []
+        self.tvf_do_letterbox, self.tvf_letterbox_fill = False, None
+
+        for idx in range(len(input_sizes)):
+            transform = timm.data.create_transform(
+                input_size=self.input_sizes[idx],
+                interpolation=self.interpolations[idx],
+                mean=self.means[idx],
+                std=self.stds[idx],
+                crop_pct=1.0,  # Set to 1.0 to ignore cropping (initial Resize sets `input_size`)
+                crop_mode="center",  # Default crop mode -- no-op when `crop_pct == 1.0`
+                is_training=False,  # No image augmentations when loading the transform!
+            )
+
+            # [Validation] Ensure appropriate transform structure, expected sizes
+            if not (
+                isinstance(transform, Compose)
+                and (len(transform.transforms) == 4)
+                and isinstance(transform.transforms[0], Resize)
+                and isinstance(transform.transforms[1], CenterCrop)
+                and isinstance(transform.transforms[2], ToTensor)
+                and isinstance(transform.transforms[3], Normalize)
+                and (transform.transforms[0].size == self.input_sizes[idx][-1])
+                and (transform.transforms[1].size == self.input_sizes[idx][-2:])
+            ):
+                raise ValueError(f"Unexpected TIMM image transformation structure/sizes: `{transform}`")
+
+            # HF Image Processors *must* be JSON-serializable; as such, cannot have torchvision. as an attribute.
+            #   => Instead, we're going to parse the transform and call "torchvision.transforms.functional" (`tvf`)
+            resize_t, crop_t, norm_t = transform.transforms[0], transform.transforms[1], transform.transforms[3]
+            self.tvf_resize_params.append(
+                {
+                    "size": resize_t.size,
+                    "interpolation": TVF.pil_modes_mapping[resize_t.interpolation],
+                    "max_size": None,
+                    "antialias": True,
+                }
+            )
+            self.tvf_crop_params.append({"output_size": crop_t.size})
+            self.tvf_normalize_params.append(
+                {
+                    "mean": norm_t.mean.float().numpy().tolist(),
+                    "std": norm_t.std.float().numpy().tolist(),
+                    "inplace": False,
+                }
+            )
+            self.tvf_do_letterbox, self.tvf_letterbox_fill = False, None
+
+            # Handle Prismatic `image_resize_strategy`
+            if self.image_resize_strategy == "resize-naive":
+                self.tvf_resize_params[idx]["size"] = (resize_t.size, resize_t.size)
+            elif self.image_resize_strategy == "letterbox":
+                self.tvf_do_letterbox, self.tvf_letterbox_fill = True, tuple([int(x * 255) for x in self.means[idx]])
+            elif self.image_resize_strategy == "resize-crop":
+                pass
+            else:
+                raise ValueError(f"Image resize strategy `{self.image_resize_strategy}` is not supported!")
+
+        # Dispatch **kwargs to super()
+        super().__init__(**kwargs)
+
+    def apply_transform(self, img: Image.Image) -> torch.Tensor:
+        """Apply `functional` variant of TIMM's Transform = Compose([Resize -> CenterCrop -> ToTensor -> Normalize])"""
+        if self.tvf_do_letterbox:
+            img = letterbox_pad_transform(img, self.tvf_letterbox_fill)
+
+        # [Contract] Fused Backbones expect "channel-stacked" inputs; we'll unpack on the model side!
+        imgs_t = []
+        for idx in range(len(self.input_sizes)):
+            img_idx = TVF.resize(img, **self.tvf_resize_params[idx])
+            img_idx = TVF.center_crop(img_idx, **self.tvf_crop_params[idx])
+            img_idx_t = TVF.to_tensor(img_idx)
+            img_idx_t = TVF.normalize(img_idx_t, **self.tvf_normalize_params[idx])
+            imgs_t.append(img_idx_t)
+
+        # [Contract] `imgs_t` is a list of Tensors of shape [3, input_size, input_size]; stack along dim = 0
+        img_t = torch.vstack(imgs_t)
+
+        return img_t
+
+    def preprocess(
+        self,
+        images: Union[Image.Image, List[Image.Image]],
+        return_tensors: Optional[Union[str, TensorType]] = None,
+        **_: str,
+    ) -> BatchFeature:
+        """
+        Preprocess an image (or batch of images); note that unlike the `transformers :: BaseImageProcessor` we
+        explicitly only handle PIL.Image.Image instances for simplicity.
+
+        @param images: A (batch of) PIL.Image.Image instance(s) to preprocess.
+        @param return_tensors: BatchFeature default Tensor format (e.g., "pt" for torch); if None, returns np.ndarray
+
+        @return: Instance of `transformers :: BatchFeature` with a single key "pixel_values"
+        """
+        if not isinstance(images, list):
+            images = [images]
+
+        # Apply `self.img_transform` to each image (will return list of torch.Tensors); stack into "batched" Tensor
+        pixel_values = torch.stack([self.apply_transform(img.convert("RGB")) for img in images])
+
+        # Return BatchFeature =>> note that for compatibility, constructor expects Dict[str, np.ndarray], so we convert
+        return BatchFeature(data={"pixel_values": pixel_values.float().numpy()}, tensor_type=return_tensors)
+
+    def __call__(self, images: Union[Image.Image, List[Image.Image]], **kwargs) -> BatchFeature:
+        return self.preprocess(images, **kwargs)
+
+
+# === PrismaticProcessor =>> Wraps both ImageProcessor and Tokenizer ===
+#   =>> https://github.com/huggingface/transformers/blob/main/src/transformers/models/llava/processing_llava.py
+class PrismaticProcessor(ProcessorMixin):
+    attributes: ClassVar[List[str]] = ["image_processor", "tokenizer"]
+    image_processor_class: str = "AutoImageProcessor"
+    tokenizer_class: str = "AutoTokenizer"
+
+    def __init__(
+        self,
+        image_processor: Optional[ImageProcessingMixin] = None,
+        tokenizer: Optional[PreTrainedTokenizerBase] = None,
+    ) -> None:
+        super().__init__(image_processor, tokenizer)
+
+    def __call__(
+        self,
+        text: Union[TextInput, PreTokenizedInput, List[TextInput], List[PreTokenizedInput]],
+        images: Union[Image.Image, List[Image.Image]],
+        padding: Union[bool, str, PaddingStrategy] = False,
+        truncation: Optional[Union[bool, str, TruncationStrategy]] = None,
+        max_length: Optional[int] = None,
+        return_tensors: Optional[Union[str, TensorType]] = TensorType.PYTORCH,
+    ) -> BatchFeature:
+        """
+        Preprocess a given (batch) of text/images for a Prismatic VLM; forwards text to the underlying LLM's tokenizer,
+        forwards images to PrismaticImageProcessor.
+
+        @param text: The (batch) of text to encode; must be a string or list of strings.
+        @param images: A (batch of) PIL.Image.Image instance(s) to preprocess.
+        @param padding: Sequence padding strategy (if multiple specified) in < True = "longest" | "max_length" | False >
+        @param truncation: Truncation strategy for the output sequences; requires `max_length` to be specified
+        @param max_length: Maximum length (in tokens) to truncate
+        @param return_tensors: Type of return tensors (usually "pt" or TensorType.PYTORCH)
+
+        @return: BatchFeature with keys for `input_ids`, `attention_mask` and `pixel_values`.
+        """
+        pixel_values = self.image_processor(images, return_tensors=return_tensors)["pixel_values"]
+        text_inputs = self.tokenizer(
+            text, return_tensors=return_tensors, padding=padding, truncation=truncation, max_length=max_length
+        )
+
+        # [Validate] Need same number of images and text inputs!
+        if pixel_values.shape[0] != text_inputs.input_ids.shape[0]:
+            raise ValueError("Batch is malformed; expected same number of images and text inputs!")
+
+        return BatchFeature(data={**text_inputs, "pixel_values": pixel_values})
+
+    # === Tokenizer Dispatch Utilities =>> check `PreTrainedTokenizerBase` for documentation ===
+    def batch_decode(
+        self,
+        sequences: Union[List[int], List[List[int]], torch.Tensor, Any],  # `Any` = np.ndarray | tf.Tensor
+        skip_special_tokens: bool = False,
+        clean_up_tokenization_spaces: Optional[bool] = None,
+        **kwargs: str,
+    ) -> List[str]:
+        return self.tokenizer.batch_decode(
+            sequences=sequences,
+            skip_special_tokens=skip_special_tokens,
+            clean_up_tokenization_spaces=clean_up_tokenization_spaces,
+            **kwargs,
+        )
+
+    def decode(
+        self,
+        token_ids: Union[int, List[int], torch.Tensor, Any],  # `Any` = np.ndarray | tf.Tensor
+        skip_special_tokens: bool = False,
+        clean_up_tokenization_spaces: Optional[bool] = None,
+        **kwargs: str,
+    ) -> str:
+        return self.tokenizer.decode(
+            token_ids=token_ids,
+            skip_special_tokens=skip_special_tokens,
+            clean_up_tokenization_spaces=clean_up_tokenization_spaces,
+            **kwargs,
+        )
+
+    @property
+    def model_input_names(self) -> List[str]:
+        tokenizer_input_names = self.tokenizer.model_input_names
+        image_processor_input_names = self.image_processor.model_input_names
+
+        return list(dict.fromkeys(tokenizer_input_names + image_processor_input_names))

results/simvla_newbase_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_newbase_25_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/processor_config.json

@@ -0,0 +1,6 @@
+{
+  "auto_map": {
+    "AutoProcessor": "processing_prismatic.PrismaticProcessor"
+  },
+  "processor_class": "PrismaticProcessor"
+}

results/simvla_newbase_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_newbase_25_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/special_tokens_map.json

@@ -0,0 +1,30 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<PAD>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

results/simvla_newbase_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_newbase_25_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000--2000_chkpt/tokenizer_config.json

@@ -0,0 +1,53 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "32000": {
+      "content": "<PAD>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "auto_map": {
+    "AutoProcessor": "processing_prismatic.PrismaticProcessor"
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "</s>",
+  "legacy": false,
+  "model_max_length": 2048,
+  "pad_token": "<PAD>",
+  "padding_side": "right",
+  "processor_class": "PrismaticProcessor",
+  "sp_model_kwargs": {},
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}

results/simvla_newbase_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_newbase_25_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000/dataset_statistics.json

@@ -0,0 +1,218 @@
+{
+  "grab_roller_aloha_agilex_50": {
+    "action": {
+      "mean": [
+        -0.7234118580818176,
+        1.632694959640503,
+        1.139991283416748,
+        -0.7696658968925476,
+        -0.004497084766626358,
+        -1.915460228919983,
+        0.6765856742858887,
+        0.4867003262042999,
+        1.567619800567627,
+        1.0547168254852295,
+        -0.740928053855896,
+        0.0007320955046452582,
+        1.6736525297164917,
+        0.6765856742858887
+      ],
+      "std": [
+        0.9490994811058044,
+        0.7159450650215149,
+        0.5567411184310913,
+        0.3428436517715454,
+        0.023813901469111443,
+        0.9220959544181824,
+        0.44203680753707886,
+        0.2887645363807678,
+        0.6954315304756165,
+        0.5172013640403748,
+        0.32894495129585266,
+        0.02310887910425663,
+        0.8557336330413818,
+        0.44203680753707886
+      ],
+      "max": [
+        0.0,
+        2.482893466949463,
+        2.353947639465332,
+        0.0,
+        0.05954868718981743,
+        0.0,
+        1.0,
+        1.1409132480621338,
+        2.443716287612915,
+        2.2799878120422363,
+        0.0,
+        0.15795069932937622,
+        3.4252119064331055,
+        1.0
+      ],
+      "min": [
+        -7.338869571685791,
+        0.0,
+        0.0,
+        -1.5001972913742065,
+        -0.12379012256860733,
+        -3.4617013931274414,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.385493516921997,
+        -0.048340000212192535,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -7.335215625762939,
+        0.0,
+        0.0,
+        -1.3311991214752197,
+        -0.11849832087755204,
+        -3.424123525619507,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.2851634454727172,
+        -0.0435758301615715,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        0.0,
+        2.478337526321411,
+        2.1660122013092042,
+        0.0,
+        0.03330111399292946,
+        0.0,
+        1.0,
+        1.1318105459213257,
+        2.3855650424957275,
+        1.9841104364395143,
+        0.0,
+        0.11385425746440887,
+        3.409965982437134,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true
+      ]
+    },
+    "proprio": {
+      "mean": [
+        -0.7144084572792053,
+        1.6121407747268677,
+        1.1227450370788574,
+        -0.7565815448760986,
+        -0.0046524289064109325,
+        -1.8903743028640747,
+        0.687203586101532,
+        0.4801884591579437,
+        1.5479238033294678,
+        1.0385884046554565,
+        -0.7280691862106323,
+        0.0008704152423888445,
+        1.651688575744629,
+        0.687203586101532
+      ],
+      "std": [
+        0.9468903541564941,
+        0.7339856028556824,
+        0.5653066039085388,
+        0.3482204079627991,
+        0.023658812046051025,
+        0.9399275779724121,
+        0.4376404583454132,
+        0.29170694947242737,
+        0.7125725150108337,
+        0.5250204801559448,
+        0.33385151624679565,
+        0.022993318736553192,
+        0.8697940111160278,
+        0.4376404583454132
+      ],
+      "max": [
+        0.0,
+        2.482893466949463,
+        2.353947639465332,
+        0.0,
+        0.05954868718981743,
+        0.0,
+        1.0,
+        1.1409132480621338,
+        2.443716287612915,
+        2.2799878120422363,
+        0.0,
+        0.15795069932937622,
+        3.4252119064331055,
+        1.0
+      ],
+      "min": [
+        -7.338869571685791,
+        0.0,
+        0.0,
+        -1.5001972913742065,
+        -0.12379012256860733,
+        -3.4617013931274414,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.385493516921997,
+        -0.048340000212192535,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -7.335215625762939,
+        0.0,
+        0.0,
+        -1.3265374326705932,
+        -0.11849832087755204,
+        -3.424123525619507,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.2640411233901978,
+        -0.0435758301615715,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        0.0,
+        2.478337526321411,
+        2.151273918151856,
+        0.0,
+        0.03329093188047409,
+        0.0,
+        1.0,
+        1.1318105459213257,
+        2.3855650424957275,
+        1.9724502563476562,
+        0.0,
+        0.11385425746440887,
+        3.409965982437134,
+        1.0
+      ]
+    },
+    "num_transitions": 4709,
+    "num_trajectories": 50
+  }
+}

results/simvla_prop_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_prop_25_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_decoder_num_blocks_2-M2000-F2000-D1000/dataset_statistics.json

@@ -0,0 +1,218 @@
+{
+  "grab_roller_aloha_agilex_50": {
+    "action": {
+      "mean": [
+        -0.7234118580818176,
+        1.632694959640503,
+        1.139991283416748,
+        -0.7696658968925476,
+        -0.004497084766626358,
+        -1.915460228919983,
+        0.6765856742858887,
+        0.4867003262042999,
+        1.567619800567627,
+        1.0547168254852295,
+        -0.740928053855896,
+        0.0007320955046452582,
+        1.6736525297164917,
+        0.6765856742858887
+      ],
+      "std": [
+        0.9490994811058044,
+        0.7159450650215149,
+        0.5567411184310913,
+        0.3428436517715454,
+        0.023813901469111443,
+        0.9220959544181824,
+        0.44203680753707886,
+        0.2887645363807678,
+        0.6954315304756165,
+        0.5172013640403748,
+        0.32894495129585266,
+        0.02310887910425663,
+        0.8557336330413818,
+        0.44203680753707886
+      ],
+      "max": [
+        0.0,
+        2.482893466949463,
+        2.353947639465332,
+        0.0,
+        0.05954868718981743,
+        0.0,
+        1.0,
+        1.1409132480621338,
+        2.443716287612915,
+        2.2799878120422363,
+        0.0,
+        0.15795069932937622,
+        3.4252119064331055,
+        1.0
+      ],
+      "min": [
+        -7.338869571685791,
+        0.0,
+        0.0,
+        -1.5001972913742065,
+        -0.12379012256860733,
+        -3.4617013931274414,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.385493516921997,
+        -0.048340000212192535,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -7.335215625762939,
+        0.0,
+        0.0,
+        -1.3311991214752197,
+        -0.11849832087755204,
+        -3.424123525619507,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.2851634454727172,
+        -0.0435758301615715,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        0.0,
+        2.478337526321411,
+        2.1660122013092042,
+        0.0,
+        0.03330111399292946,
+        0.0,
+        1.0,
+        1.1318105459213257,
+        2.3855650424957275,
+        1.9841104364395143,
+        0.0,
+        0.11385425746440887,
+        3.409965982437134,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true
+      ]
+    },
+    "proprio": {
+      "mean": [
+        -0.7144084572792053,
+        1.6121407747268677,
+        1.1227450370788574,
+        -0.7565815448760986,
+        -0.0046524289064109325,
+        -1.8903743028640747,
+        0.687203586101532,
+        0.4801884591579437,
+        1.5479238033294678,
+        1.0385884046554565,
+        -0.7280691862106323,
+        0.0008704152423888445,
+        1.651688575744629,
+        0.687203586101532
+      ],
+      "std": [
+        0.9468903541564941,
+        0.7339856028556824,
+        0.5653066039085388,
+        0.3482204079627991,
+        0.023658812046051025,
+        0.9399275779724121,
+        0.4376404583454132,
+        0.29170694947242737,
+        0.7125725150108337,
+        0.5250204801559448,
+        0.33385151624679565,
+        0.022993318736553192,
+        0.8697940111160278,
+        0.4376404583454132
+      ],
+      "max": [
+        0.0,
+        2.482893466949463,
+        2.353947639465332,
+        0.0,
+        0.05954868718981743,
+        0.0,
+        1.0,
+        1.1409132480621338,
+        2.443716287612915,
+        2.2799878120422363,
+        0.0,
+        0.15795069932937622,
+        3.4252119064331055,
+        1.0
+      ],
+      "min": [
+        -7.338869571685791,
+        0.0,
+        0.0,
+        -1.5001972913742065,
+        -0.12379012256860733,
+        -3.4617013931274414,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.385493516921997,
+        -0.048340000212192535,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -7.335215625762939,
+        0.0,
+        0.0,
+        -1.3265374326705932,
+        -0.11849832087755204,
+        -3.424123525619507,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.2640411233901978,
+        -0.0435758301615715,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        0.0,
+        2.478337526321411,
+        2.151273918151856,
+        0.0,
+        0.03329093188047409,
+        0.0,
+        1.0,
+        1.1318105459213257,
+        2.3855650424957275,
+        1.9724502563476562,
+        0.0,
+        0.11385425746440887,
+        3.409965982437134,
+        1.0
+      ]
+    },
+    "num_transitions": 4709,
+    "num_trajectories": 50
+  }
+}

results/simvla_reg16_lib_16/openvla-7b+libero_4_task_suites_no_noops+b16+lr-0.0005+lora-r32+dropout-0.0--image_aug--simvla_reg16_lib_16_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_decoder_num_blocks_4-M40000-F40000-D20000--40000_chkpt/dataset_statistics.json

@@ -0,0 +1,526 @@
+{
+  "libero_spatial_no_noops": {
+    "action": {
+      "mean": [
+        0.15312479436397552,
+        0.13707277178764343,
+        -0.15526802837848663,
+        -0.005176450591534376,
+        -0.01120874285697937,
+        -0.020194264128804207,
+        0.4578818082809448
+      ],
+      "std": [
+        0.41272708773612976,
+        0.34724321961402893,
+        0.50869220495224,
+        0.037266165018081665,
+        0.07244449853897095,
+        0.05762382969260216,
+        0.49827873706817627
+      ],
+      "max": [
+        0.9375,
+        0.9375,
+        0.9375,
+        0.1971428543329239,
+        0.33642858266830444,
+        0.375,
+        1.0
+      ],
+      "min": [
+        -0.9375,
+        -0.9375,
+        -0.9375,
+        -0.1875,
+        -0.3675000071525574,
+        -0.36000001430511475,
+        0.0
+      ],
+      "q01": [
+        -0.7454732114076613,
+        -0.6616071462631226,
+        -0.9375,
+        -0.1071428582072258,
+        -0.20678570866584778,
+        -0.1842857152223587,
+        0.0
+      ],
+      "q99": [
+        0.9375,
+        0.8758928775787354,
+        0.9321428537368774,
+        0.1039285734295845,
+        0.17678570747375488,
+        0.14571428298950195,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        false
+      ]
+    },
+    "proprio": {
+      "mean": [
+        -0.024462558329105377,
+        0.106529600918293,
+        1.0580483675003052,
+        3.0628468990325928,
+        -0.10464039444923401,
+        0.08307311683893204,
+        0.01995457336306572,
+        -0.020162804052233696
+      ],
+      "std": [
+        0.1101478561758995,
+        0.13784688711166382,
+        0.1044282391667366,
+        0.10451053828001022,
+        0.4112098217010498,
+        0.2176690548658371,
+        0.017260896041989326,
+        0.0171116404235363
+      ],
+      "max": [
+        0.1759040206670761,
+        0.3904820382595062,
+        1.3290715217590332,
+        3.4566118717193604,
+        1.2268599271774292,
+        1.0429412126541138,
+        0.041053611785173416,
+        0.000775813648942858
+      ],
+      "min": [
+        -0.3095473051071167,
+        -0.29250794649124146,
+        0.9095591306686401,
+        2.497488260269165,
+        -1.8006486892700195,
+        -0.7207611203193665,
+        -0.0004703797458205372,
+        -0.041536275297403336
+      ],
+      "q01": [
+        -0.2727657300233841,
+        -0.23721413239836692,
+        0.9160063165426254,
+        2.77949666261673,
+        -1.3187511622905732,
+        -0.41989982962608335,
+        0.001503719249740243,
+        -0.03989770736545324
+      ],
+      "q99": [
+        0.13529365032911292,
+        0.3629165390133857,
+        1.2862326657772063,
+        3.2829698753356933,
+        0.9332760351896285,
+        0.6325724506378171,
+        0.039933966137468815,
+        -0.001671919699292631
+      ]
+    },
+    "num_transitions": 52970,
+    "num_trajectories": 432
+  },
+  "libero_object_no_noops": {
+    "action": {
+      "mean": [
+        0.07096529006958008,
+        0.13498851656913757,
+        -0.04601382836699486,
+        0.00123520044144243,
+        0.006998839322477579,
+        -0.015027612447738647,
+        0.46428999304771423
+      ],
+      "std": [
+        0.2681235373020172,
+        0.43846824765205383,
+        0.4474974274635315,
+        0.024446550756692886,
+        0.049355510622262955,
+        0.042107198387384415,
+        0.49879148602485657
+      ],
+      "max": [
+        0.9375,
+        0.8919642567634583,
+        0.9375,
+        0.17678570747375488,
+        0.35035714507102966,
+        0.1810714304447174,
+        1.0
+      ],
+      "min": [
+        -0.8839285969734192,
+        -0.9375,
+        -0.9375,
+        -0.15000000596046448,
+        -0.29035714268684387,
+        -0.32892856001853943,
+        0.0
+      ],
+      "q01": [
+        -0.5383928418159485,
+        -0.8758928775787354,
+        -0.9375,
+        -0.06964285671710968,
+        -0.11678571254014969,
+        -0.15964286029338837,
+        0.0
+      ],
+      "q99": [
+        0.8464285731315613,
+        0.84375,
+        0.9375,
+        0.08142857253551483,
+        0.14892856776714325,
+        0.0867857113480568,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        false
+      ]
+    },
+    "proprio": {
+      "mean": [
+        -0.02999030612409115,
+        -0.007947085425257683,
+        0.20293472707271576,
+        3.1086409091949463,
+        -0.21404768526554108,
+        -0.11307074874639511,
+        0.029380427673459053,
+        -0.030556727200746536
+      ],
+      "std": [
+        0.06694897264242172,
+        0.17608462274074554,
+        0.07807064801454544,
+        0.0868484303355217,
+        0.33540457487106323,
+        0.20728276669979095,
+        0.00956575945019722,
+        0.009197483770549297
+      ],
+      "max": [
+        0.14580604434013367,
+        0.33216384053230286,
+        0.3857804834842682,
+        3.4003844261169434,
+        0.7954911589622498,
+        0.6642207503318787,
+        0.04104341194033623,
+        -0.00018117300351150334
+      ],
+      "min": [
+        -0.1765444278717041,
+        -0.29457300901412964,
+        0.008128180168569088,
+        2.2890501022338867,
+        -1.883241891860962,
+        -1.0600427389144897,
+        0.0006495157140307128,
+        -0.041782498359680176
+      ],
+      "q01": [
+        -0.14911890715360643,
+        -0.25978428691625594,
+        0.009925739830359817,
+        2.7545341420173646,
+        -1.3996034812927245,
+        -0.6867720144987106,
+        0.008197814421728254,
+        -0.04015838988125324
+      ],
+      "q99": [
+        0.09063626825809479,
+        0.29066365867853167,
+        0.3370887073874472,
+        3.2611824750900267,
+        0.32092821151018125,
+        0.4037663781642913,
+        0.039891827926039694,
+        -0.009106044843792932
+      ]
+    },
+    "num_transitions": 66984,
+    "num_trajectories": 454
+  },
+  "libero_goal_no_noops": {
+    "action": {
+      "mean": [
+        0.04721052572131157,
+        0.028835246339440346,
+        -0.1485840231180191,
+        -0.0025010062381625175,
+        0.026408178731799126,
+        0.027379808947443962,
+        0.6299911737442017
+      ],
+      "std": [
+        0.3968801498413086,
+        0.3473387360572815,
+        0.49239858984947205,
+        0.055331431329250336,
+        0.07844757288694382,
+        0.10008802264928818,
+        0.48270025849342346
+      ],
+      "max": [
+        0.9375,
+        0.9375,
+        0.9375,
+        0.3557142913341522,
+        0.375,
+        0.375,
+        1.0
+      ],
+      "min": [
+        -0.9375,
+        -0.9375,
+        -0.9375,
+        -0.2582142949104309,
+        -0.375,
+        -0.2871428430080414,
+        0.0
+      ],
+      "q01": [
+        -0.8785714507102966,
+        -0.7553571462631226,
+        -0.9375,
+        -0.1510714292526245,
+        -0.1639285683631897,
+        -0.13777500048279764,
+        0.0
+      ],
+      "q99": [
+        0.9375,
+        0.9107142686843872,
+        0.9375,
+        0.20357142388820648,
+        0.26357144117355347,
+        0.375,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        false
+      ]
+    },
+    "proprio": {
+      "mean": [
+        -0.09923473745584488,
+        0.013597904704511166,
+        1.0694637298583984,
+        2.82898211479187,
+        0.30799180269241333,
+        -0.274286687374115,
+        0.028092455118894577,
+        -0.027339335530996323
+      ],
+      "std": [
+        0.11653962731361389,
+        0.11478105187416077,
+        0.10487838834524155,
+        0.5570293664932251,
+        0.7221656441688538,
+        0.36479514837265015,
+        0.01507475133985281,
+        0.014990941621363163
+      ],
+      "max": [
+        0.13579000532627106,
+        0.33316105604171753,
+        1.3660105466842651,
+        3.473310708999634,
+        2.6688623428344727,
+        0.8255361318588257,
+        0.04233968257904053,
+        0.0010111660230904818
+      ],
+      "min": [
+        -0.46141114830970764,
+        -0.30129560828208923,
+        0.9083037972450256,
+        0.35277295112609863,
+        -1.4858465194702148,
+        -1.5227035284042358,
+        -0.0013586411951109767,
+        -0.042040832340717316
+      ],
+      "q01": [
+        -0.42401049643754957,
+        -0.27338370531797407,
+        0.911226047873497,
+        1.3085840785503386,
+        -0.691297555565834,
+        -1.130668159723282,
+        0.0016738151130266487,
+        -0.040336399003863335
+      ],
+      "q99": [
+        0.08990443304181095,
+        0.26473945528268716,
+        1.2910678112506866,
+        3.2425890421867365,
+        2.3376442337036116,
+        0.4659483411908149,
+        0.040610933862626555,
+        -0.0015016929572448147
+      ]
+    },
+    "num_transitions": 52042,
+    "num_trajectories": 428
+  },
+  "libero_10_no_noops": {
+    "action": {
+      "mean": [
+        0.01820324920117855,
+        0.05858374014496803,
+        -0.05592384561896324,
+        0.004626928828656673,
+        0.00289608770981431,
+        -0.007673131301999092,
+        0.5457824468612671
+      ],
+      "std": [
+        0.2825464606285095,
+        0.35904666781425476,
+        0.3673802614212036,
+        0.03770702704787254,
+        0.05429719388484955,
+        0.08725254982709885,
+        0.49815231561660767
+      ],
+      "max": [
+        0.9375,
+        0.9375,
+        0.9375,
+        0.30000001192092896,
+        0.29357144236564636,
+        0.375,
+        1.0
+      ],
+      "min": [
+        -0.9375,
+        -0.9375,
+        -0.9375,
+        -0.23642857372760773,
+        -0.3053571283817291,
+        -0.3675000071525574,
+        0.0
+      ],
+      "q01": [
+        -0.6348214149475098,
+        -0.7741071581840515,
+        -0.7633928656578064,
+        -0.09749999642372131,
+        -0.14819999992847435,
+        -0.2742857038974762,
+        0.0
+      ],
+      "q99": [
+        0.7714285850524902,
+        0.8464285731315613,
+        0.9375,
+        0.13928571343421936,
+        0.15964286029338837,
+        0.3246428668498993,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        false
+      ]
+    },
+    "proprio": {
+      "mean": [
+        -0.04190658777952194,
+        0.03539430722594261,
+        0.8257141709327698,
+        2.908308267593384,
+        -0.5562185049057007,
+        -0.16649018228054047,
+        0.028316624462604523,
+        -0.028561657294631004
+      ],
+      "std": [
+        0.10743364691734314,
+        0.14424669742584229,
+        0.2572328448295593,
+        0.3441362977027893,
+        1.234421730041504,
+        0.3579835891723633,
+        0.013308707624673843,
+        0.013174631632864475
+      ],
+      "max": [
+        0.21031762659549713,
+        0.39128610491752625,
+        1.3332009315490723,
+        3.6714255809783936,
+        3.560650587081909,
+        1.386339545249939,
+        0.04160946607589722,
+        0.0013633022317662835
+      ],
+      "min": [
+        -0.4828203022480011,
+        -0.3255046010017395,
+        0.445506751537323,
+        1.1321442127227783,
+        -3.641430377960205,
+        -1.842738389968872,
+        -0.0010040868073701859,
+        -0.04111652821302414
+      ],
+      "q01": [
+        -0.3899900782108307,
+        -0.2838300323486328,
+        0.44795057058334353,
+        1.8810229921340942,
+        -2.886677579879761,
+        -1.1599004411697387,
+        0.002066459748893976,
+        -0.04001387819647789
+      ],
+      "q99": [
+        0.1530261474847791,
+        0.32915401458740223,
+        1.2546923208236693,
+        3.303542451858519,
+        2.7496529006957933,
+        0.6893712210655194,
+        0.040048558115959164,
+        -0.0017598449345678235
+      ]
+    },
+    "num_transitions": 101469,
+    "num_trajectories": 379
+  }
+}

results/simvla_reg4_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_reg4_25_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_register4-M2000-F2000-D1000/dataset_statistics.json

@@ -0,0 +1,218 @@
+{
+  "grab_roller_aloha_agilex_50": {
+    "action": {
+      "mean": [
+        -0.7234118580818176,
+        1.632694959640503,
+        1.139991283416748,
+        -0.7696658968925476,
+        -0.004497084766626358,
+        -1.915460228919983,
+        0.6765856742858887,
+        0.4867003262042999,
+        1.567619800567627,
+        1.0547168254852295,
+        -0.740928053855896,
+        0.0007320955046452582,
+        1.6736525297164917,
+        0.6765856742858887
+      ],
+      "std": [
+        0.9490994811058044,
+        0.7159450650215149,
+        0.5567411184310913,
+        0.3428436517715454,
+        0.023813901469111443,
+        0.9220959544181824,
+        0.44203680753707886,
+        0.2887645363807678,
+        0.6954315304756165,
+        0.5172013640403748,
+        0.32894495129585266,
+        0.02310887910425663,
+        0.8557336330413818,
+        0.44203680753707886
+      ],
+      "max": [
+        0.0,
+        2.482893466949463,
+        2.353947639465332,
+        0.0,
+        0.05954868718981743,
+        0.0,
+        1.0,
+        1.1409132480621338,
+        2.443716287612915,
+        2.2799878120422363,
+        0.0,
+        0.15795069932937622,
+        3.4252119064331055,
+        1.0
+      ],
+      "min": [
+        -7.338869571685791,
+        0.0,
+        0.0,
+        -1.5001972913742065,
+        -0.12379012256860733,
+        -3.4617013931274414,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.385493516921997,
+        -0.048340000212192535,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -7.335215625762939,
+        0.0,
+        0.0,
+        -1.3311991214752197,
+        -0.11849832087755204,
+        -3.424123525619507,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.2851634454727172,
+        -0.0435758301615715,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        0.0,
+        2.478337526321411,
+        2.1660122013092042,
+        0.0,
+        0.03330111399292946,
+        0.0,
+        1.0,
+        1.1318105459213257,
+        2.3855650424957275,
+        1.9841104364395143,
+        0.0,
+        0.11385425746440887,
+        3.409965982437134,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true
+      ]
+    },
+    "proprio": {
+      "mean": [
+        -0.7144084572792053,
+        1.6121407747268677,
+        1.1227450370788574,
+        -0.7565815448760986,
+        -0.0046524289064109325,
+        -1.8903743028640747,
+        0.687203586101532,
+        0.4801884591579437,
+        1.5479238033294678,
+        1.0385884046554565,
+        -0.7280691862106323,
+        0.0008704152423888445,
+        1.651688575744629,
+        0.687203586101532
+      ],
+      "std": [
+        0.9468903541564941,
+        0.7339856028556824,
+        0.5653066039085388,
+        0.3482204079627991,
+        0.023658812046051025,
+        0.9399275779724121,
+        0.4376404583454132,
+        0.29170694947242737,
+        0.7125725150108337,
+        0.5250204801559448,
+        0.33385151624679565,
+        0.022993318736553192,
+        0.8697940111160278,
+        0.4376404583454132
+      ],
+      "max": [
+        0.0,
+        2.482893466949463,
+        2.353947639465332,
+        0.0,
+        0.05954868718981743,
+        0.0,
+        1.0,
+        1.1409132480621338,
+        2.443716287612915,
+        2.2799878120422363,
+        0.0,
+        0.15795069932937622,
+        3.4252119064331055,
+        1.0
+      ],
+      "min": [
+        -7.338869571685791,
+        0.0,
+        0.0,
+        -1.5001972913742065,
+        -0.12379012256860733,
+        -3.4617013931274414,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.385493516921997,
+        -0.048340000212192535,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -7.335215625762939,
+        0.0,
+        0.0,
+        -1.3265374326705932,
+        -0.11849832087755204,
+        -3.424123525619507,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.2640411233901978,
+        -0.0435758301615715,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        0.0,
+        2.478337526321411,
+        2.151273918151856,
+        0.0,
+        0.03329093188047409,
+        0.0,
+        1.0,
+        1.1318105459213257,
+        2.3855650424957275,
+        1.9724502563476562,
+        0.0,
+        0.11385425746440887,
+        3.409965982437134,
+        1.0
+      ]
+    },
+    "num_transitions": 4709,
+    "num_trajectories": 50
+  }
+}

results/simvla_reg_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_reg_25_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_register16-M3000-F3000-D2000--3000_chkpt/added_tokens.json

@@ -0,0 +1,3 @@
+{
+  "<PAD>": 32000
+}

results/simvla_reg_25/openvla-7b+grab_roller_aloha_agilex_50+b4+lr-0.0001+lora-r32+dropout-0.0--image_aug--simvla_reg_25_inner1_proj_type_gelu_linear_ffn_type_gelu_mlp_ffn_register16-M3000-F3000-D2000--3000_chkpt/processor_config.json

@@ -0,0 +1,6 @@
+{
+  "auto_map": {
+    "AutoProcessor": "processing_prismatic.PrismaticProcessor"
+  },
+  "processor_class": "PrismaticProcessor"
+}

run_scripts/ffn_q2a/aloha/aloha_robotwin2_ffn_10 copy.sh

@@ -0,0 +1,104 @@
+#========== settings ==========#
+PROJECT_PATH=simvla_twin2
+ROOT_PATH=/inspire/hdd/ws-f4d69b29-e0a5-44e6-bd92-acf4de9990f0/public-project/chengdongzhou-240108390137
+#========== !NOTE! ==========#
+RUN_MODE=simvla_10
+use_predict_future_prop=False
+batch_size=4
+use_action_ts_head=True
+use_one_embed=True
+use_multi_scaling=False
+mlp_type=ffn
+decoder_num_blocks=4
+robot_platform=10_al
+proj_type=gelu_linear
+ffn_type=gelu
+expand_inner_ratio=1
+linear_drop_ratio=0.0
+multi_queries_num=2
+multi_query_norm_type=layernorm
+action_norm=layernorm
+use_fredf=False
+MODE=${RUN_MODE}_inner${expand_inner_ratio}_proj_type_${proj_type}_ffn_type_${ffn_type}_mlp_${mlp_type}_decoder_num_blocks_${decoder_num_blocks}
+#========== !NOTE! ==========#
+use_l1_regression=True
+num_images_in_input=3     
+wandb_entity=chenghaha
+wandb_project=robotwin
+wandb_log_freq=1
+use_proprio=True
+use_diffusion=False
+use_film=True
+num_steps_before_decay=20000
+save_freq=10000
+max_steps=30000
+vla_path=$ROOT_PATH/ai_models/openvla/openvla-7b
+data_root_dir=$ROOT_PATH/datasets/TianxingChen/RoboTwin2.0/tfds
+dataset_name=aloha_agilex_robotwin2_benchmark
+run_root_dir=$ROOT_PATH/vla_projects/$PROJECT_PATH/results/$RUN_MODE
+#========== get run_id ==========#
+note_parts=("${MODE}")
+
+# if [ "$use_l1_regression" = "True" ]; then
+#     note_parts+=("L1_regression")
+# fi
+
+# if [ "$num_images_in_input" == 1 ]; then
+#     note_parts+=("3rd_person_img")
+# else
+#     note_parts+=("3rd_person_img_and_wrist")
+# fi
+
+# if [ "$use_l1_regression" = "True" ]; then
+#     note_parts+=("proprio_state")
+# fi
+
+# if [ "$use_film" = "True" ]; then
+#     note_parts+=("Film")
+# fi
+note_parts+=("M$max_steps-F$save_freq-D$num_steps_before_decay")
+run_id_note_value=$(IFS='--'; echo "${note_parts[*]}")
+
+#========== enter environment ==========#
+conda activate openvla-oft
+cd $ROOT_PATH/vla_projects/$PROJECT_PATH
+export PYTHONPATH=$ROOT_PATH/vla_projects/$PROJECT_PATH
+
+#========== run ==========#
+WANDB_CONSOLE=off WANDB_MODE=offline torchrun --standalone --nnodes 1 --nproc-per-node 4 vla-scripts/finetune.py \
+  --vla_path "$vla_path" \
+  --data_root_dir "$data_root_dir" \
+  --dataset_name "$dataset_name" \
+  --run_root_dir "$run_root_dir" \
+  --use_l1_regression "$use_l1_regression" \
+  --use_diffusion "$use_diffusion" \
+  --use_film "$use_film" \
+  --num_images_in_input "$num_images_in_input" \
+  --use_proprio "$use_proprio" \
+  --batch_size "$batch_size" \
+  --learning_rate 5e-5 \
+  --num_steps_before_decay "$num_steps_before_decay" \
+  --max_steps "$max_steps" \
+  --save_freq "$save_freq" \
+  --save_latest_checkpoint_only False \
+  --image_aug True \
+  --lora_rank 32 \
+  --wandb_entity "$wandb_entity" \
+  --wandb_project "$wandb_project" \
+  --wandb_log_freq "$wandb_log_freq" \
+  --run_id_note "$run_id_note_value" \
+  --use_predict_future_prop "$use_predict_future_prop" \
+  --use_action_ts_head "$use_action_ts_head" \
+  --use_one_embed "$use_one_embed" \
+  --use_multi_scaling "$use_multi_scaling" \
+  --mlp_type "$mlp_type" \
+  --decoder_num_blocks "$decoder_num_blocks" \
+  --robot_platform "$robot_platform" \
+  --proj_type "$proj_type" \
+  --ffn_type "$ffn_type" \
+  --expand_inner_ratio "$expand_inner_ratio" \
+  --linear_drop_ratio "$linear_drop_ratio" \
+  --multi_query_norm_type "$multi_query_norm_type" \
+  --multi_queries_num "$multi_queries_num" \
+  --action_norm "$action_norm" \
+  --use_fredf "$use_fredf"

run_scripts/ffn_q2a/aloha/aloha_robotwin2_ffn_10.sh

@@ -0,0 +1,104 @@
+#========== settings ==========#
+PROJECT_PATH=simvla_twin2
+ROOT_PATH=/inspire/hdd/ws-f4d69b29-e0a5-44e6-bd92-acf4de9990f0/public-project/chengdongzhou-240108390137
+#========== !NOTE! ==========#
+RUN_MODE=simvla_10
+use_predict_future_prop=False
+batch_size=4
+use_action_ts_head=True
+use_one_embed=True
+use_multi_scaling=False
+mlp_type=ffn
+decoder_num_blocks=4
+robot_platform=10_al
+proj_type=gelu_linear
+ffn_type=gelu
+expand_inner_ratio=1
+linear_drop_ratio=0.0
+multi_queries_num=2
+multi_query_norm_type=layernorm
+action_norm=layernorm
+use_fredf=False
+MODE=${RUN_MODE}_inner${expand_inner_ratio}_proj_type_${proj_type}_ffn_type_${ffn_type}_mlp_${mlp_type}_decoder_num_blocks_${decoder_num_blocks}
+#========== !NOTE! ==========#
+use_l1_regression=True
+num_images_in_input=3     
+wandb_entity=chenghaha
+wandb_project=robotwin
+wandb_log_freq=1
+use_proprio=True
+use_diffusion=False
+use_film=True
+num_steps_before_decay=20000
+save_freq=10000
+max_steps=20000
+vla_path=$ROOT_PATH/ai_models/openvla/openvla-7b
+data_root_dir=$ROOT_PATH/datasets/TianxingChen/RoboTwin2.0/tfds
+dataset_name=aloha_agilex_robotwin2_benchmark
+run_root_dir=$ROOT_PATH/vla_projects/$PROJECT_PATH/results/$RUN_MODE
+#========== get run_id ==========#
+note_parts=("${MODE}")
+
+# if [ "$use_l1_regression" = "True" ]; then
+#     note_parts+=("L1_regression")
+# fi
+
+# if [ "$num_images_in_input" == 1 ]; then
+#     note_parts+=("3rd_person_img")
+# else
+#     note_parts+=("3rd_person_img_and_wrist")
+# fi
+
+# if [ "$use_l1_regression" = "True" ]; then
+#     note_parts+=("proprio_state")
+# fi
+
+# if [ "$use_film" = "True" ]; then
+#     note_parts+=("Film")
+# fi
+note_parts+=("M$max_steps-F$save_freq-D$num_steps_before_decay")
+run_id_note_value=$(IFS='--'; echo "${note_parts[*]}")
+
+#========== enter environment ==========#
+conda activate openvla-oft
+cd $ROOT_PATH/vla_projects/$PROJECT_PATH
+export PYTHONPATH=$ROOT_PATH/vla_projects/$PROJECT_PATH
+
+#========== run ==========#
+WANDB_CONSOLE=off WANDB_MODE=offline torchrun --standalone --nnodes 1 --nproc-per-node 4 vla-scripts/finetune.py \
+  --vla_path "$vla_path" \
+  --data_root_dir "$data_root_dir" \
+  --dataset_name "$dataset_name" \
+  --run_root_dir "$run_root_dir" \
+  --use_l1_regression "$use_l1_regression" \
+  --use_diffusion "$use_diffusion" \
+  --use_film "$use_film" \
+  --num_images_in_input "$num_images_in_input" \
+  --use_proprio "$use_proprio" \
+  --batch_size "$batch_size" \
+  --learning_rate 5e-5 \
+  --num_steps_before_decay "$num_steps_before_decay" \
+  --max_steps "$max_steps" \
+  --save_freq "$save_freq" \
+  --save_latest_checkpoint_only False \
+  --image_aug True \
+  --lora_rank 32 \
+  --wandb_entity "$wandb_entity" \
+  --wandb_project "$wandb_project" \
+  --wandb_log_freq "$wandb_log_freq" \
+  --run_id_note "$run_id_note_value" \
+  --use_predict_future_prop "$use_predict_future_prop" \
+  --use_action_ts_head "$use_action_ts_head" \
+  --use_one_embed "$use_one_embed" \
+  --use_multi_scaling "$use_multi_scaling" \
+  --mlp_type "$mlp_type" \
+  --decoder_num_blocks "$decoder_num_blocks" \
+  --robot_platform "$robot_platform" \
+  --proj_type "$proj_type" \
+  --ffn_type "$ffn_type" \
+  --expand_inner_ratio "$expand_inner_ratio" \
+  --linear_drop_ratio "$linear_drop_ratio" \
+  --multi_query_norm_type "$multi_query_norm_type" \
+  --multi_queries_num "$multi_queries_num" \
+  --action_norm "$action_norm" \
+  --use_fredf "$use_fredf"

run_scripts/ffn_q2a/aloha/aloha_robotwin2_ffn_25_registers16.sh

@@ -0,0 +1,106 @@
+#========== settings ==========#
+PROJECT_PATH=simvla_twin2
+ROOT_PATH=/inspire/hdd/ws-f4d69b29-e0a5-44e6-bd92-acf4de9990f0/public-project/chengdongzhou-240108390137
+#========== !NOTE! ==========#
+RUN_MODE=simvla_allreg16_25
+use_predict_future_prop=False
+batch_size=4
+use_action_ts_head=True
+use_one_embed=True
+use_multi_scaling=False
+mlp_type=ffn
+decoder_num_blocks=4
+robot_platform=aloha
+proj_type=gelu_linear
+ffn_type=gelu
+expand_inner_ratio=1
+linear_drop_ratio=0.0
+multi_queries_num=25
+registers_num=16
+use_registers=True
+multi_query_norm_type=layernorm
+action_norm=layernorm
+MODE=${RUN_MODE}_inner${expand_inner_ratio}_proj_type_${proj_type}_ffn_type_${ffn_type}_mlp_${mlp_type}_register${registers_num}
+#========== !NOTE! ==========#
+use_l1_regression=True
+num_images_in_input=3     
+wandb_entity=chenghaha
+wandb_project=robotwin
+wandb_log_freq=1
+use_proprio=True
+use_diffusion=False
+use_film=True
+num_steps_before_decay=10000
+save_freq=10000
+max_steps=20000
+vla_path=$ROOT_PATH/ai_models/openvla/openvla-7b
+data_root_dir=$ROOT_PATH/datasets/TianxingChen/RoboTwin2.0/tfds
+dataset_name=aloha_agilex_robotwin2_benchmark
+run_root_dir=$ROOT_PATH/vla_projects/$PROJECT_PATH/results/$RUN_MODE
+#========== get run_id ==========#
+note_parts=("${MODE}")
+
+# if [ "$use_l1_regression" = "True" ]; then
+#     note_parts+=("L1_regression")
+# fi
+
+# if [ "$num_images_in_input" == 1 ]; then
+#     note_parts+=("3rd_person_img")
+# else
+#     note_parts+=("3rd_person_img_and_wrist")
+# fi
+
+# if [ "$use_l1_regression" = "True" ]; then
+#     note_parts+=("proprio_state")
+# fi
+
+# if [ "$use_film" = "True" ]; then
+#     note_parts+=("Film")
+# fi
+note_parts+=("M$max_steps-F$save_freq-D$num_steps_before_decay")
+run_id_note_value=$(IFS='--'; echo "${note_parts[*]}")
+
+#========== enter environment ==========#
+conda activate openvla-oft
+cd $ROOT_PATH/vla_projects/$PROJECT_PATH
+export PYTHONPATH=$ROOT_PATH/vla_projects/$PROJECT_PATH
+
+#========== run ==========#
+WANDB_CONSOLE=off WANDB_MODE=offline torchrun --standalone --nnodes 1 --nproc-per-node 4 vla-scripts/finetune_registers.py \
+  --vla_path "$vla_path" \
+  --data_root_dir "$data_root_dir" \
+  --dataset_name "$dataset_name" \
+  --run_root_dir "$run_root_dir" \
+  --use_l1_regression "$use_l1_regression" \
+  --use_diffusion "$use_diffusion" \
+  --use_film "$use_film" \
+  --num_images_in_input "$num_images_in_input" \
+  --use_proprio "$use_proprio" \
+  --batch_size "$batch_size" \
+  --learning_rate 1e-4 \
+  --num_steps_before_decay "$num_steps_before_decay" \
+  --max_steps "$max_steps" \
+  --save_freq "$save_freq" \
+  --save_latest_checkpoint_only False \
+  --image_aug True \
+  --lora_rank 32 \
+  --wandb_entity "$wandb_entity" \
+  --wandb_project "$wandb_project" \
+  --wandb_log_freq "$wandb_log_freq" \
+  --run_id_note "$run_id_note_value" \
+  --use_predict_future_prop "$use_predict_future_prop" \
+  --use_action_ts_head "$use_action_ts_head" \
+  --use_one_embed "$use_one_embed" \
+  --use_multi_scaling "$use_multi_scaling" \
+  --mlp_type "$mlp_type" \
+  --decoder_num_blocks "$decoder_num_blocks" \
+  --robot_platform "$robot_platform" \
+  --proj_type "$proj_type" \
+  --ffn_type "$ffn_type" \
+  --expand_inner_ratio "$expand_inner_ratio" \
+  --linear_drop_ratio "$linear_drop_ratio" \
+  --multi_query_norm_type "$multi_query_norm_type" \
+  --multi_queries_num "$multi_queries_num" \
+  --action_norm "$action_norm" \
+  --registers_num "$registers_num" \
+  --use_registers "$use_registers"

run_scripts/ffn_q2a/aloha/run_test.sh

@@ -1,2 +1,2 @@
-bash run_scripts/ffn_q2a/aloha/test_action_aloha_robotwin2_ffn_50_cosine.sh
-bash run_scripts/ffn_q2a/aloha/test_aloha_robotwin2_ffn_50.sh
+bash run_scripts/ffn_q2a/libero/libero_register.sh
+bash run_scripts/ffn_q2a/aloha/aloha_robotwin2_ffn_25_registers16.sh

run_scripts/ffn_q2a/aloha/test_aloha_robotwin2_ffn_10.sh

@@ -15,7 +15,7 @@ proj_type=onlynorm
 ffn_type=relu
 expand_inner_ratio=1
 linear_drop_ratio=0.0
-multi_queries_num=2
+multi_queries_num=1
 multi_query_norm_type=layernorm
 action_norm=layernorm
 use_patch_wise_loss=True
@@ -29,9 +29,9 @@ wandb_log_freq=1
 use_proprio=True
 use_diffusion=False
 use_film=True
-num_steps_before_decay=1000
-save_freq=2000
-max_steps=2000
+num_steps_before_decay=2000
+save_freq=3000
+max_steps=3000
 vla_path=$ROOT_PATH/ai_models/openvla/openvla-7b
 data_root_dir=$ROOT_PATH/datasets/TianxingChen/RoboTwin2.0/tfds
 dataset_name=grab_roller_aloha_agilex_50
@@ -101,4 +101,4 @@ WANDB_CONSOLE=off WANDB_MODE=offline torchrun --standalone --nnodes 1 --nproc-pe
   --multi_query_norm_type "$multi_query_norm_type" \
   --multi_queries_num "$multi_queries_num" \
   --action_norm "$action_norm" \
-  --use_patch_wise_loss "$use_patch_wise_loss" 
+  --use_patch_wise_loss "$use_patch_wise_loss"