Upload OpenVLA object_force 100k checkpoint

added_tokens.json

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

dataset_statistics.json

@@ -0,0 +1,277 @@
+{
+  "libero_object_force": {
+    "action": {
+      "mean": [
+        0.071312315762043,
+        0.13514883816242218,
+        -0.0461067333817482,
+        0.0011658848961815238,
+        0.007094137370586395,
+        -0.015087507665157318,
+        0.46519312262535095
+      ],
+      "std": [
+        0.2685853838920593,
+        0.43864676356315613,
+        0.4474373459815979,
+        0.02429334633052349,
+        0.0493454784154892,
+        0.042218245565891266,
+        0.49891483783721924
+      ],
+      "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": [
+        -0.5410714149475098,
+        -0.8758928775787354,
+        -0.9375,
+        -0.06964285671710968,
+        -0.11678571254014969,
+        -0.1607142835855484,
+        0.0
+      ],
+      "q99": [
+        0.8464285731315613,
+        0.84375,
+        0.9375,
+        0.08142857253551483,
+        0.14892856776714325,
+        0.08571428805589676,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        false
+      ]
+    },
+    "proprio": {
+      "mean": [
+        -0.029641686007380486,
+        -0.007346511352807283,
+        0.20262189209461212,
+        3.10734486579895,
+        -0.21544547379016876,
+        -0.11419320851564407,
+        0.029497746378183365,
+        -0.03067079186439514,
+        1.5807781219482422,
+        -0.022773683071136475,
+        0.28887230157852173,
+        -0.0038040464278310537,
+        0.0,
+        0.0,
+        8.588276863098145,
+        0.33169829845428467,
+        0.4518001079559326,
+        0.0,
+        0.0,
+        0.0,
+        1.4875096082687378,
+        0.1563856452703476,
+        0.1289268434047699,
+        -0.0013712065992876887,
+        0.0,
+        0.0,
+        8.779743194580078,
+        0.4979320764541626,
+        0.5377444624900818,
+        0.0,
+        0.0,
+        0.0
+      ],
+      "std": [
+        0.06716274470090866,
+        0.17588521540164948,
+        0.07807677984237671,
+        0.08845069259405136,
+        0.3387415111064911,
+        0.2068338394165039,
+        0.0094814607873559,
+        0.009063540026545525,
+        8.53734016418457,
+        1.449267029762268,
+        2.7356209754943848,
+        0.10999714583158493,
+        0.0,
+        0.0,
+        8.949143409729004,
+        3.068483591079712,
+        2.674851655960083,
+        0.0,
+        0.0,
+        0.0,
+        7.695075511932373,
+        3.0036182403564453,
+        2.915856122970581,
+        0.10801699012517929,
+        0.0,
+        0.0,
+        9.02546215057373,
+        3.711784601211548,
+        2.8412725925445557,
+        0.0,
+        0.0,
+        0.0
+      ],
+      "max": [
+        0.14580604434013367,
+        0.3322603404521942,
+        0.3849222660064697,
+        3.4003379344940186,
+        0.7954724431037903,
+        0.6642153263092041,
+        0.04104502126574516,
+        -0.0005040675168856978,
+        307.7990417480469,
+        87.96394348144531,
+        114.16512298583984,
+        2.443922519683838,
+        0.0,
+        0.0,
+        89.5918960571289,
+        38.36042022705078,
+        58.46033477783203,
+        0.0,
+        0.0,
+        0.0,
+        337.6532287597656,
+        98.4030990600586,
+        148.4830322265625,
+        4.023257732391357,
+        0.0,
+        0.0,
+        66.75836944580078,
+        52.273109436035156,
+        41.93406677246094,
+        0.0,
+        0.0,
+        0.0
+      ],
+      "min": [
+        -0.17652566730976105,
+        -0.28169912099838257,
+        0.008128181099891663,
+        2.2889645099639893,
+        -1.8830392360687256,
+        -1.0422919988632202,
+        0.000678583572153002,
+        -0.041782498359680176,
+        0.0,
+        -31.12724494934082,
+        -37.026729583740234,
+        -4.006844997406006,
+        0.0,
+        0.0,
+        0.0,
+        -89.24981689453125,
+        -69.69068908691406,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -19.259239196777344,
+        -56.92011642456055,
+        -3.8666090965270996,
+        0.0,
+        0.0,
+        0.0,
+        -133.1490478515625,
+        -38.06122589111328,
+        0.0,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -0.1492678302526474,
+        -0.2592164146900177,
+        0.009929686188697814,
+        2.7465669918060303,
+        -1.4019958305358886,
+        -0.6869005513191223,
+        0.008685965463519097,
+        -0.04015249326825142,
+        0.0,
+        -2.4958832359313963,
+        -2.3188658714294434,
+        -0.2632309830188751,
+        0.0,
+        0.0,
+        0.0,
+        -5.7196044349670405,
+        -7.037253799438477,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -1.9764903831481933,
+        -3.1475293350219724,
+        -0.27225555896759035,
+        0.0,
+        0.0,
+        0.0,
+        -7.429137268066406,
+        -7.075016765594483,
+        0.0,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        0.091574721634388,
+        0.29128185868263246,
+        0.33622013688087493,
+        3.260224313735962,
+        0.3196746730804444,
+        0.39607855677604714,
+        0.03989203363656998,
+        -0.010264341309666632,
+        23.248353347778476,
+        1.8256341409683936,
+        4.611621818542482,
+        0.21757932603359245,
+        0.0,
+        0.0,
+        20.265312805175785,
+        10.87531608581547,
+        8.604345703125004,
+        0.0,
+        0.0,
+        0.0,
+        21.16050384521489,
+        5.389982223510743,
+        4.114923648834249,
+        0.225317343473436,
+        0.0,
+        0.0,
+        20.62276679992676,
+        13.933656463623054,
+        9.604595146179202,
+        0.0,
+        0.0,
+        0.0
+      ]
+    },
+    "num_transitions": 67185,
+    "num_trajectories": 456
+  }
+}

lora_adapter/README.md

@@ -0,0 +1,202 @@
+---
+base_model: /gs/fs/tga-aklab/miki/hf_cache/hub/models--openvla--openvla-7b/snapshots/31f090d05236101ebfc381b61c674dd4746d4ce0
+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

lora_adapter/adapter_config.json

@@ -0,0 +1,45 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": {
+    "base_model_class": "OpenVLAForActionPrediction",
+    "parent_library": "transformers_modules.31f090d05236101ebfc381b61c674dd4746d4ce0.modeling_prismatic"
+  },
+  "base_model_name_or_path": "/gs/fs/tga-aklab/miki/hf_cache/hub/models--openvla--openvla-7b/snapshots/31f090d05236101ebfc381b61c674dd4746d4ce0",
+  "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",
+    "fc3",
+    "q_proj",
+    "up_proj",
+    "down_proj",
+    "k_proj",
+    "qkv",
+    "lm_head",
+    "kv",
+    "o_proj",
+    "fc1",
+    "proj",
+    "gate_proj",
+    "v_proj",
+    "fc2"
+  ],
+  "task_type": null,
+  "use_dora": false,
+  "use_rslora": false
+}

lora_adapter/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab979695514b77fbb5c97f2e45e2fbf18e5a2cce89e101d100325dc5b6e25181
+size 484458600

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
+}

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))

processor_config.json

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

proprio_projector--100000_checkpoint.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:167f1a53383019025b4e97b9f7093def10c067a55438847f4f958244d01f5514
+size 67668472

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
+  }
+}

tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
+size 499723

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
+}