Hugging Face's logo

Transformers documentation

Molmo2

You are viewing main version, which requires installation from source. If you'd like regular pip install, checkout the latest stable version (v5.19.0).
Hugging Face's logo
Join the Hugging Face community

and get access to the augmented documentation experience

to get started

This model was published in HF papers on 2026-01-15 and contributed to Hugging Face Transformers on 2026-10-08.

FlashAttention SDPA

Molmo2

Molmo2 is a family of open-weight vision-language models by AllenAI that are state-of-the-art among open-source models, with exceptional capabilities in point-driven grounding for single image, multi-image, and video tasks. The architecture combines a Vision Transformer (ViT) for image processing with an adapter layer connecting vision and text modalities, and a text decoder based on transformer architecture with rotary position embeddings.

The abstract from the paper is the following:

Today’s strongest video-language models (VLMs) remain proprietary. The strongest open-weight models either rely on synthetic data from proprietary VLMs, effectively distilling from them, or do not disclose their training data or recipe. As a result, the open-source community lacks the foundations needed to improve on the state-of-the-art video (and image) language models. Crucially, many downstream applications require more than just high-level video understanding; they require grounding — either by pointing or by tracking in pixels. Even proprietary models lack this capability. We present Molmo2, a new family of VLMs that are state-of-the-art among open-source models and demonstrate exceptional new capabilities in point-driven grounding in single image, multi-image, and video tasks. Our key contribution is a collection of 7 new video datasets and 2 multi-image datasets, including a dataset of highly detailed video captions for pre-training, a free-form video Q&A dataset for fine-tuning, a new object tracking dataset with complex queries, and an innovative new video pointing dataset, all collected without the use of closed VLMs. We also present a training recipe for this data utilizing an efficient packing and message-tree encoding scheme, and show bi-directional attention on vision tokens and a novel token-weight strategy improves performance. Our best-in-class 8B model outperforms others in the class of open weight and data models on short videos, counting, and captioning, and is competitive on long-videos. On video-grounding Molmo2 significantly outperforms existing open-weight models like Qwen3-VL (35.5 vs 29.6 accuracy on video counting) and surpasses proprietary models like Gemini 3 Pro on some tasks (38.4 vs 20.0 F1 on video pointing and 56.2 vs 41.1 J&F on video tracking).

You can find all the original Molmo2 checkpoints under the Molmo2 collection.

Usage example

Image-text-to-text generation

The example below demonstrates how to generate text based on an image with Pipeline or the AutoModel class.

Pipeline
AutoModel
from transformers import pipeline

pipeline = pipeline(
    task="image-text-to-text",
    model="allenai/Molmo2-8B",
    device_map="auto",
)

messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/pipeline-cat-chonk.jpeg"},
            {"type": "text", "text": "Describe this image."},
        ],
    }
]
pipeline(text=messages, max_new_tokens=128, return_full_text=False)

Video pointing

Molmo2 can also generate grounded points for video inputs. The points are returned in the generated text as a <points coords="..."> payload, where each coordinate group contains a timestamp followed by point ids and normalized image coordinates.

import torch

from transformers import AutoModelForImageTextToText, AutoProcessor


model_id = "allenai/Molmo2-8B"
video_url = "https://storage.googleapis.com/oe-training-public/demo_videos/many_penguins.mp4"

processor = AutoProcessor.from_pretrained(model_id)
model = AutoModelForImageTextToText.from_pretrained(model_id, device_map="auto")

messages = [
    {
        "role": "user",
        "content": [
            {"type": "text", "text": "Point to the penguins."},
            {"type": "video", "video": video_url},
        ],
    }
]
inputs = processor.apply_chat_template(
    messages,
    tokenize=True,
    add_generation_prompt=True,
    return_tensors="pt",
    return_dict=True,
)

generated_ids = model.generate(**inputs, max_new_tokens=128, do_sample=False)

input_len = inputs["input_ids"].shape[1]
generated_text = processor.decode(generated_ids[0][input_len:], skip_special_tokens=True)
print(generated_text)

Molmo2Config

class transformers.Molmo2Config

< >

( transformers_version: str | None = Nonearchitectures: list[str] | None = Noneoutput_hidden_states: bool | None = Falsereturn_dict: bool | None = Truedtype: str | torch.dtype | None = Nonechunk_size_feed_forward: int = 0is_encoder_decoder: bool = Falseid2label: dict[int, str] | dict[str, str] | None = Nonelabel2id: dict[str, int] | dict[str, str] | None = Noneproblem_type: Literal['regression', 'single_label_classification', 'multi_label_classification'] | None = Nonevision_config: dict | transformers.configuration_utils.PreTrainedConfig | None = Noneadapter_config: dict | transformers.configuration_utils.PreTrainedConfig | None = Nonetext_config: dict | transformers.configuration_utils.PreTrainedConfig | None = Noneimage_start_token_id: int = 151936low_res_image_start_token_id: int = 151940image_end_token_id: int = 151937image_patch_id: int = 151938frame_start_token_id: int = 151943frame_end_token_id: int = 151944initializer_range: float = 0.02tie_word_embeddings: bool = False )

Parameters

  • vision_config (Molmo2VisionConfig, optional) — Configuration for the vision transformer backbone.
  • adapter_config (Molmo2AdapterConfig, optional) — Configuration for the vision-to-language adapter.
  • text_config (Union[dict, ~configuration_utils.PreTrainedConfig], optional) — The config object or dictionary of the text backbone.
  • image_start_token_id (int, optional, defaults to 151936) — Token ID marking the start of an image region.
  • low_res_image_start_token_id (int, optional, defaults to 151940) — Token ID marking the start of a low-resolution image crop.
  • image_end_token_id (int, optional, defaults to 151937) — Token ID marking the end of an image region.
  • image_patch_id (int, optional, defaults to 151938) — Token ID for image patches.
  • frame_start_token_id (int, optional, defaults to 151943) — Token ID marking the start of a video frame.
  • frame_end_token_id (int, optional, defaults to 151944) — Token ID marking the end of a video frame.
  • initializer_range (float, optional, defaults to 0.02) — The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
  • tie_word_embeddings (bool, optional, defaults to False) — Whether the model’s input and output word embeddings should be tied.

This is the configuration class to store the configuration of a Molmo2Model. It is used to instantiate a Molmo2 model according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of the allenai/Molmo2-8B

Configuration objects inherit from PreTrainedConfig and can be used to control the model outputs. Read the documentation from PreTrainedConfig for more information.

Molmo2VisionConfig

class transformers.Molmo2VisionConfig

< >

( transformers_version: str | None = Nonearchitectures: list[str] | None = Noneoutput_hidden_states: bool | None = Falsereturn_dict: bool | None = Truedtype: str | torch.dtype | None = Nonechunk_size_feed_forward: int = 0is_encoder_decoder: bool = Falseid2label: dict[int, str] | dict[str, str] | None = Nonelabel2id: dict[str, int] | dict[str, str] | None = Noneproblem_type: Literal['regression', 'single_label_classification', 'multi_label_classification'] | None = Nonehidden_size: int = 1152intermediate_size: int = 4304num_hidden_layers: int = 25num_attention_heads: int = 16num_key_value_heads: int = 16head_dim: int = 72hidden_act: str = 'gelu_pytorch_tanh'layer_norm_eps: float = 1e-06image_size: list[int] | None = Nonepatch_size: int = 14num_position_embeddings: int | None = Nonenum_channels: int = 3attention_dropout: float = 0.0attention_bias: bool = Trueinitializer_range: float = 0.02 )

Parameters

  • hidden_size (int, optional, defaults to 1152) — Dimension of the hidden representations.
  • intermediate_size (int, optional, defaults to 4304) — Dimension of the MLP representations.
  • num_hidden_layers (int, optional, defaults to 25) — Number of hidden layers in the Transformer decoder.
  • num_attention_heads (int, optional, defaults to 16) — Number of attention heads for each attention layer in the Transformer decoder.
  • num_key_value_heads (int, optional, defaults to 16) — This is the number of key_value heads that should be used to implement Grouped Query Attention. If num_key_value_heads=num_attention_heads, the model will use Multi Head Attention (MHA), if num_key_value_heads=1 the model will use Multi Query Attention (MQA) otherwise GQA is used. When converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed by meanpooling all the original heads within that group. For more details, check out this paper. If it is not specified, will default to num_attention_heads.
  • head_dim (int, optional, defaults to 72) — The attention head dimension. If None, it will default to hidden_size // num_attention_heads
  • hidden_act (str, optional, defaults to gelu_pytorch_tanh) — The non-linear activation function (function or string) in the decoder. For example, "gelu", "relu", "silu", etc.
  • layer_norm_eps (float, optional, defaults to 1e-06) — The epsilon used by the layer normalization layers.
  • image_size (list[int], optional) — Input image size as (height, width), [378, 378] when not provided.
  • patch_size (int, optional, defaults to 14) — The size (resolution) of each patch.
  • num_position_embeddings (int, optional) — Number of positional embeddings, one per image patch. Derived from image_size and patch_size when not provided.
  • num_channels (int, optional, defaults to 3) — The number of input channels.
  • attention_dropout (float, optional, defaults to 0.0) — The dropout ratio for the attention probabilities.
  • attention_bias (bool, optional, defaults to True) — Whether to use a bias in the query, key, value and output projection layers during self-attention.
  • initializer_range (float, optional, defaults to 0.02) — The standard deviation of the truncated_normal_initializer for initializing all weight matrices.

This is the configuration class to store the configuration of a Molmo2Model. It is used to instantiate a Molmo2 model according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of the allenai/Molmo2-8B

Configuration objects inherit from PreTrainedConfig and can be used to control the model outputs. Read the documentation from PreTrainedConfig for more information.

Molmo2AdapterConfig

class transformers.Molmo2AdapterConfig

< >

( transformers_version: str | None = Nonearchitectures: list[str] | None = Noneoutput_hidden_states: bool | None = Falsereturn_dict: bool | None = Truedtype: str | torch.dtype | None = Nonechunk_size_feed_forward: int = 0is_encoder_decoder: bool = Falseid2label: dict[int, str] | dict[str, str] | None = Nonelabel2id: dict[str, int] | dict[str, str] | None = Noneproblem_type: Literal['regression', 'single_label_classification', 'multi_label_classification'] | None = Nonevision_feature_layer: list[int] | None = Nonehidden_size: int = 1152num_attention_heads: int = 16num_key_value_heads: int = 16head_dim: int = 72attention_dropout: float = 0.0attention_bias: bool = Truehidden_act: str = 'silu'intermediate_size: int = 18944text_hidden_size: int = 3584mlp_bias: bool = Falseinitializer_range: float = 0.02 )

Parameters

  • vision_feature_layer (list[int], optional) — Indices of the ViT layers whose outputs are concatenated and pooled, [24, 18] when not provided. Negative indices count from the last layer.
  • hidden_size (int, optional, defaults to 1152) — Dimension of the hidden representations.
  • num_attention_heads (int, optional, defaults to 16) — Number of attention heads for each attention layer in the Transformer decoder.
  • num_key_value_heads (int, optional, defaults to 16) — This is the number of key_value heads that should be used to implement Grouped Query Attention. If num_key_value_heads=num_attention_heads, the model will use Multi Head Attention (MHA), if num_key_value_heads=1 the model will use Multi Query Attention (MQA) otherwise GQA is used. When converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed by meanpooling all the original heads within that group. For more details, check out this paper. If it is not specified, will default to num_attention_heads.
  • head_dim (int, optional, defaults to 72) — The attention head dimension. If None, it will default to hidden_size // num_attention_heads
  • attention_dropout (float, optional, defaults to 0.0) — The dropout ratio for the attention probabilities.
  • attention_bias (bool, optional, defaults to True) — Whether to use a bias in the query, key, value and output projection layers during self-attention.
  • hidden_act (str, optional, defaults to silu) — The non-linear activation function (function or string) in the decoder. For example, "gelu", "relu", "silu", etc.
  • intermediate_size (int, optional, defaults to 18944) — Dimension of the MLP representations.
  • text_hidden_size (int, optional, defaults to 3584) — Hidden size of the text model (used for projection).
  • mlp_bias (bool, optional, defaults to False) — Whether to use a bias in up_proj, down_proj and gate_proj layers in the MLP layers.
  • initializer_range (float, optional, defaults to 0.02) — The standard deviation of the truncated_normal_initializer for initializing all weight matrices.

This is the configuration class to store the configuration of a Molmo2Model. It is used to instantiate a Molmo2 model according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of the allenai/Molmo2-8B

Configuration objects inherit from PreTrainedConfig and can be used to control the model outputs. Read the documentation from PreTrainedConfig for more information.

Molmo2TextConfig

class transformers.Molmo2TextConfig

< >

( transformers_version: str | None = Nonearchitectures: list[str] | None = Noneoutput_hidden_states: bool | None = Falsereturn_dict: bool | None = Truedtype: str | torch.dtype | None = Nonechunk_size_feed_forward: int = 0is_encoder_decoder: bool = Falseid2label: dict[int, str] | dict[str, str] | None = Nonelabel2id: dict[str, int] | dict[str, str] | None = Noneproblem_type: Literal['regression', 'single_label_classification', 'multi_label_classification'] | None = Nonehidden_size: int = 4096num_attention_heads: int = 32num_key_value_heads: int = 8head_dim: int = 128vocab_size: int = 151936additional_vocab_size: int = 128attention_bias: bool = Falseqk_norm_type: str = 'qwen3'num_hidden_layers: int = 36intermediate_size: int = 12288hidden_act: str = 'silu'attention_dropout: float = 0.0max_position_embeddings: int = 36864rope_parameters: transformers.modeling_rope_utils.RopeParameters | dict | None = Nonerope_scaling_layers: list[int] | None = Nonerms_norm_eps: float = 1e-06norm_after: bool = Falseinitializer_range: float = 0.02use_cache: bool = Truetie_word_embeddings: bool = False )

Parameters

  • hidden_size (int, optional, defaults to 4096) — Dimension of the hidden representations.
  • num_attention_heads (int, optional, defaults to 32) — Number of attention heads for each attention layer in the Transformer decoder.
  • num_key_value_heads (int, optional, defaults to 8) — This is the number of key_value heads that should be used to implement Grouped Query Attention. If num_key_value_heads=num_attention_heads, the model will use Multi Head Attention (MHA), if num_key_value_heads=1 the model will use Multi Query Attention (MQA) otherwise GQA is used. When converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed by meanpooling all the original heads within that group. For more details, check out this paper. If it is not specified, will default to num_attention_heads.
  • head_dim (int, optional, defaults to 128) — The attention head dimension. If None, it will default to hidden_size // num_attention_heads
  • vocab_size (int, optional, defaults to 151936) — Vocabulary size of the model. Defines the number of different tokens that can be represented by the input_ids.
  • additional_vocab_size (int, optional, defaults to 128) — Number of additional vocabulary tokens beyond the base vocabulary.
  • attention_bias (bool, optional, defaults to False) — Whether to use a bias in the query, key, value and output projection layers during self-attention.
  • qk_norm_type (str, optional, defaults to "qwen3") — Query/key normalization layout used by the checkpoint. "qwen3" normalizes per head; "olmo" normalizes the full projected query/key tensors.
  • num_hidden_layers (int, optional, defaults to 36) — Number of hidden layers in the Transformer decoder.
  • intermediate_size (int, optional, defaults to 12288) — Dimension of the MLP representations.
  • hidden_act (str, optional, defaults to silu) — The non-linear activation function (function or string) in the decoder. For example, "gelu", "relu", "silu", etc.
  • attention_dropout (float, optional, defaults to 0.0) — The dropout ratio for the attention probabilities.
  • max_position_embeddings (int, optional, defaults to 36864) — The maximum sequence length that this model might ever be used with.
  • rope_parameters (Union[~modeling_rope_utils.RopeParameters, dict], optional) — Dictionary containing the configuration parameters for the RoPE embeddings. The dictionary should contain a value for rope_theta and optionally parameters used for scaling in case you want to use RoPE with longer max_position_embeddings.
  • rope_scaling_layers (list[int], optional) — Indices of the layers that apply the scaled RoPE described by rope_parameters. The remaining layers use an unscaled RoPE with the same theta. All layers are scaled when not provided.
  • rms_norm_eps (float, optional, defaults to 1e-06) — The epsilon used by the rms normalization layers.
  • norm_after (bool, optional, defaults to False) — Whether to apply layer normalization after the attention/FFN blocks instead of before.
  • initializer_range (float, optional, defaults to 0.02) — The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
  • use_cache (bool, optional, defaults to True) — Whether or not the model should return the last key/values attentions (not used by all models). Only relevant if config.is_decoder=True or when the model is a decoder-only generative model.
  • tie_word_embeddings (bool, optional, defaults to False) — Whether to tie weight embeddings according to model’s tied_weights_keys mapping.

This is the configuration class to store the configuration of a Molmo2Model. It is used to instantiate a Molmo2 model according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of the allenai/Molmo2-8B

Configuration objects inherit from PreTrainedConfig and can be used to control the model outputs. Read the documentation from PreTrainedConfig for more information.

Molmo2Processor

class transformers.Molmo2Processor

< >

( image_processor = Nonevideo_processor = Nonetokenizer = Nonechat_template: str | None = Noneimage_use_col_tokens: bool | None = Trueuse_single_crop_col_tokens: bool | None = Noneuse_single_crop_start_token: bool | None = Truevideo_use_col_tokens: bool | None = Falseuse_frame_special_tokens: bool | None = True**kwargs )

Parameters

  • image_processor (Molmo2ImageProcessor) — The image processor is a required input.
  • video_processor (Molmo2VideoProcessor) — The video processor is a required input.
  • tokenizer (TokenizersBackend) — The tokenizer is a required input.
  • chat_template (str, optional) — A Jinja template to convert lists of messages in a chat into a tokenizable string.
  • image_use_col_tokens (bool, optional, defaults to True) — Whether to append column-separator tokens (<im_col>) after each patch row of the high-resolution image view.
  • use_single_crop_col_tokens (bool, optional) — Whether to append column-separator tokens after each patch row of the low-resolution (single-crop) image view. If None, falls back to image_use_col_tokens.
  • use_single_crop_start_token (bool, optional, defaults to True) — Whether to start the low-resolution image view with <low_res_im_start> instead of the regular <im_start>.
  • video_use_col_tokens (bool, optional, defaults to False) — Whether to append column-separator tokens after each patch row of video frames.
  • use_frame_special_tokens (bool, optional, defaults to True) — Whether to wrap each video frame with <frame_start> / <frame_end> tokens. If False, falls back to <im_start> / <im_end>.

Constructs a Molmo2Processor which wraps a image processor, a video processor, and a tokenizer into a single processor.

Molmo2Processor offers all the functionalities of Molmo2ImageProcessor, Molmo2VideoProcessor, and TokenizersBackend. See the ~Molmo2ImageProcessor, ~Molmo2VideoProcessor, and ~TokenizersBackend for more information.

__call__

< >

( images: typing.Union[ForwardRef('PIL.Image.Image'), numpy.ndarray, ForwardRef('torch.Tensor'), list['PIL.Image.Image'], list[numpy.ndarray], list['torch.Tensor'], NoneType] = Nonetext: str | list[str] | list[list[str]] | None = Nonevideos: typing.Union[list['PIL.Image.Image'], numpy.ndarray, ForwardRef('torch.Tensor'), list[numpy.ndarray], list['torch.Tensor'], list[list['PIL.Image.Image']], list[list[numpy.ndarray]], list[list['torch.Tensor']], transformers.video_utils.URL, list[transformers.video_utils.URL], list[list[transformers.video_utils.URL]], transformers.video_utils.Path, list[transformers.video_utils.Path], list[list[transformers.video_utils.Path]], NoneType] = None**kwargs: Unpack )

Parameters

  • images (Union[PIL.Image.Image, numpy.ndarray, torch.Tensor, list[PIL.Image.Image], list[numpy.ndarray], list[torch.Tensor]], optional) — Image to preprocess. Expects a single or batch of images with pixel values ranging from 0 to 255. If passing in images with pixel values between 0 and 1, set do_rescale=False.
  • text (Union[str, list[str], list[list[str]]], optional) — The sequence or batch of sequences to be encoded. Each sequence can be a string or a list of strings (pretokenized string). If you pass a pretokenized input, set is_split_into_words=True to avoid ambiguity with batched inputs.
  • videos (Union[list[PIL.Image.Image], numpy.ndarray, torch.Tensor, list[numpy.ndarray], list[torch.Tensor], list[list[PIL.Image.Image]], list[list[numpy.ndarray]], list[list[torch.Tensor]], ~video_utils.URL, list[~video_utils.URL], list[list[~video_utils.URL]], ~video_utils.Path, list[~video_utils.Path], list[list[~video_utils.Path]]], optional) — Video to preprocess. Expects a single or batch of videos with pixel values ranging from 0 to 255. If passing in videos with pixel values between 0 and 1, set do_rescale=False.
  • max_crops (int, kwargs, optional, defaults to 8) — Maximum number of crops to use per image.
  • overlap_margins (list[int], kwargs, optional, defaults to [4, 4]) — Overlap margins (in patches) for overlapping crop extraction.
  • patch_size (int, kwargs, optional, defaults to 14) — The spatial patch size of the vision encoder.
  • pooling_size (list[int], kwargs, optional, defaults to [2, 2]) — The pooling size of the vision adapter.
  • patch_size (int, kwargs, optional) — Side length in pixels of each ViT patch for video frames.
  • pooling_size (list[int], kwargs, optional) — [pool_h, pool_w] pooling window applied to video patch features.
  • max_fps (int, kwargs, optional) — Maximum sampling rate in frames per second for short videos.
  • return_tensors (str or TensorType, optional) — If set, will return tensors of a particular framework. Acceptable values are:

    • 'pt': Return PyTorch torch.Tensor objects.
    • 'np': Return NumPy np.ndarray objects.
  • **kwargs (ProcessingKwargs, optional) — Additional processing options for each modality (text, images, videos, audio). Model-specific parameters are listed above; see the TypedDict class for the complete list of supported arguments.

Molmo2ImageProcessor

class transformers.Molmo2ImageProcessor

< >

( **kwargs: Unpack )

Parameters

  • do_convert_rgb (bool, kwargs, optional, defaults to True) — Whether to convert the image to RGB.
  • do_resize (bool, kwargs, optional, defaults to True) — Whether to resize the image.
  • size (Annotated[int | list[int] | tuple[int, ...] | dict[str, int] | None, None], kwargs, defaults to {'height' -- 378, 'width': 378}): Describes the maximum input dimensions to the model.
  • default_to_square (bool, kwargs, optional, defaults to True) — Whether to default to a square image when resizing, if size is an int.
  • crop_size (Annotated[int | list[int] | tuple[int, ...] | dict[str, int] | None, None], kwargs) — Size of the output image after applying center_crop.
  • resample (Annotated[Union[int, PILImageResampling, NoneType], None], kwargs, defaults to Resampling.BILINEAR) — Resampling filter to use if resizing the image. This can be one of the enum PILImageResampling. Only has an effect if do_resize is set to True.
  • do_rescale (bool, kwargs, optional, defaults to True) — Whether to rescale the image.
  • rescale_factor (float, kwargs, optional, defaults to 0.00392156862745098) — Rescale factor to rescale the image by if do_rescale is set to True.
  • do_normalize (bool, kwargs, optional, defaults to True) — Whether to normalize the image.
  • image_mean (Union[float, list[float], tuple[float, ...]], kwargs, optional, defaults to [0.5, 0.5, 0.5]) — Image mean to use for normalization. Only has an effect if do_normalize is set to True.
  • image_std (Union[float, list[float], tuple[float, ...]], kwargs, optional, defaults to [0.5, 0.5, 0.5]) — Image standard deviation to use for normalization. Only has an effect if do_normalize is set to True.
  • do_pad (bool, kwargs, optional) — Whether to pad the image. Padding is done either to the largest size in the batch or to a fixed square size per image. The exact padding strategy depends on the model.
  • pad_size (Annotated[int | list[int] | tuple[int, ...] | dict[str, int] | None, None], kwargs) — The size in {"height": int, "width" int} to pad the images to. Must be larger than any image size provided for preprocessing. If pad_size is not provided, images will be padded to the largest height and width in the batch. Applied only when do_pad=True.
  • do_center_crop (bool, kwargs, optional) — Whether to center crop the image.
  • data_format (Union[str, ~image_utils.ChannelDimension], kwargs, optional) — Only ChannelDimension.FIRST is supported. Added for compatibility with slow processors.
  • input_data_format (Union[str, ~image_utils.ChannelDimension], kwargs, optional) — The channel dimension format for the input image. If unset, the channel dimension format is inferred from the input image. Can be one of:
    • "channels_first" or ChannelDimension.FIRST: image in (num_channels, height, width) format.
    • "channels_last" or ChannelDimension.LAST: image in (height, width, num_channels) format.
    • "none" or ChannelDimension.NONE: image in (height, width) format.
  • device (Annotated[Union[str, torch.device, NoneType], None], kwargs) — The device to process the videos on. If unset, the device is inferred from the input videos.
  • return_tensors (Annotated[str | ~utils.generic.TensorType | None, None], kwargs) — Returns stacked tensors if set to 'pt', otherwise returns a list of tensors.
  • disable_grouping (bool, kwargs, optional) — Whether to disable grouping of images by size to process them individually and not in batches. If None, will be set to True if the images are on CPU, and False otherwise. This choice is based on empirical observations, as detailed here: https://github.com/huggingface/transformers/pull/38157
  • image_seq_length (int, kwargs, optional) — The number of image tokens to be used for each image in the input. Added for backward compatibility but this should be set as a processor attribute in future models.
  • max_crops (int, kwargs, optional, defaults to 8) — Maximum number of crops to use per image.
  • overlap_margins (list[int], kwargs, optional, defaults to [4, 4]) — Overlap margins (in patches) for overlapping crop extraction.
  • patch_size (int, kwargs, optional, defaults to 14) — The spatial patch size of the vision encoder.
  • pooling_size (list[int], kwargs, optional, defaults to [2, 2]) — The pooling size of the vision adapter.

Constructs a Molmo2ImageProcessor image processor.

__call__

< >

( images: typing.Union[ForwardRef('PIL.Image.Image'), numpy.ndarray, ForwardRef('torch.Tensor'), list['PIL.Image.Image'], list[numpy.ndarray], list['torch.Tensor']]*args**kwargs: Unpack )

Preprocess an image or a batch of images.

preprocess

< >

( images: typing.Union[ForwardRef('PIL.Image.Image'), numpy.ndarray, ForwardRef('torch.Tensor'), list['PIL.Image.Image'], list[numpy.ndarray], list['torch.Tensor']]**kwargs: Unpack ) → ~image_processing_base.BatchFeature

Parameters

  • images (Union[PIL.Image.Image, numpy.ndarray, torch.Tensor, list[PIL.Image.Image], list[numpy.ndarray], list[torch.Tensor]]) — Image to preprocess. Expects a single or batch of images with pixel values ranging from 0 to 255. If passing in images with pixel values between 0 and 1, set do_rescale=False.
  • do_convert_rgb (bool, kwargs, optional) — Whether to convert the image to RGB.
  • do_resize (bool, kwargs, optional) — Whether to resize the image.
  • size (Annotated[int | list[int] | tuple[int, ...] | dict[str, int] | None, None], kwargs) — Describes the maximum input dimensions to the model.
  • default_to_square (bool, kwargs, optional) — Whether to default to a square image when resizing, if size is an int.
  • crop_size (Annotated[int | list[int] | tuple[int, ...] | dict[str, int] | None, None], kwargs) — Size of the output image after applying center_crop.
  • resample (Annotated[Union[int, PILImageResampling, NoneType], None], kwargs) — Resampling filter to use if resizing the image. This can be one of the enum PILImageResampling. Only has an effect if do_resize is set to True.
  • do_rescale (bool, kwargs, optional) — Whether to rescale the image.
  • rescale_factor (float, kwargs, optional) — Rescale factor to rescale the image by if do_rescale is set to True.
  • do_normalize (bool, kwargs, optional) — Whether to normalize the image.
  • image_mean (Union[float, list[float], tuple[float, ...]], kwargs, optional) — Image mean to use for normalization. Only has an effect if do_normalize is set to True.
  • image_std (Union[float, list[float], tuple[float, ...]], kwargs, optional) — Image standard deviation to use for normalization. Only has an effect if do_normalize is set to True.
  • do_pad (bool, kwargs, optional) — Whether to pad the image. Padding is done either to the largest size in the batch or to a fixed square size per image. The exact padding strategy depends on the model.
  • pad_size (Annotated[int | list[int] | tuple[int, ...] | dict[str, int] | None, None], kwargs) — The size in {"height": int, "width" int} to pad the images to. Must be larger than any image size provided for preprocessing. If pad_size is not provided, images will be padded to the largest height and width in the batch. Applied only when do_pad=True.
  • do_center_crop (bool, kwargs, optional) — Whether to center crop the image.
  • data_format (Union[str, ~image_utils.ChannelDimension], kwargs, optional) — Only ChannelDimension.FIRST is supported. Added for compatibility with slow processors.
  • input_data_format (Union[str, ~image_utils.ChannelDimension], kwargs, optional) — The channel dimension format for the input image. If unset, the channel dimension format is inferred from the input image. Can be one of:
    • "channels_first" or ChannelDimension.FIRST: image in (num_channels, height, width) format.
    • "channels_last" or ChannelDimension.LAST: image in (height, width, num_channels) format.
    • "none" or ChannelDimension.NONE: image in (height, width) format.
  • device (Annotated[Union[str, torch.device, NoneType], None], kwargs) — The device to process the videos on. If unset, the device is inferred from the input videos.
  • return_tensors (Annotated[str | ~utils.generic.TensorType | None, None], kwargs) — Returns stacked tensors if set to 'pt', otherwise returns a list of tensors.
  • disable_grouping (bool, kwargs, optional) — Whether to disable grouping of images by size to process them individually and not in batches. If None, will be set to True if the images are on CPU, and False otherwise. This choice is based on empirical observations, as detailed here: https://github.com/huggingface/transformers/pull/38157
  • image_seq_length (int, kwargs, optional) — The number of image tokens to be used for each image in the input. Added for backward compatibility but this should be set as a processor attribute in future models.
  • max_crops (int, kwargs, optional, defaults to 8) — Maximum number of crops to use per image.
  • overlap_margins (list[int], kwargs, optional, defaults to [4, 4]) — Overlap margins (in patches) for overlapping crop extraction.
  • patch_size (int, kwargs, optional, defaults to 14) — The spatial patch size of the vision encoder.
  • pooling_size (list[int], kwargs, optional, defaults to [2, 2]) — The pooling size of the vision adapter.

Returns

~image_processing_base.BatchFeature

  • data (dict) — Dictionary of lists/arrays/tensors returned by the call method (‘pixel_values’, etc.).
  • tensor_type (Union[None, str, TensorType], optional) — You can give a tensor_type here to convert the lists of integers in PyTorch/Numpy Tensors at initialization.

Molmo2VideoProcessor

class transformers.Molmo2VideoProcessor

< >

( **kwargs: Unpack )

Parameters

  • do_convert_rgb (bool, kwargs, optional) — Whether to convert the image to RGB.
  • do_resize (bool, kwargs, optional) — Whether to resize the image.
  • size (Annotated[int | list[int] | tuple[int, ...] | dict[str, int] | None, None], kwargs, defaults to {'height' -- 378, 'width': 378}): Describes the maximum input dimensions to the model.
  • default_to_square (bool, kwargs, optional, defaults to True) — Whether to default to a square image when resizing, if size is an int.
  • resample (Annotated[Union[int, PILImageResampling, NoneType], None], kwargs, defaults to Resampling.BILINEAR) — Resampling filter to use if resizing the image. This can be one of the enum PILImageResampling. Only has an effect if do_resize is set to True.
  • do_rescale (bool, kwargs, optional) — Whether to rescale the image.
  • rescale_factor (float, kwargs, optional, defaults to 0.00392156862745098) — Rescale factor to rescale the image by if do_rescale is set to True.
  • do_normalize (bool, kwargs, optional) — Whether to normalize the image.
  • image_mean (Union[float, list[float], tuple[float, ...]], kwargs, optional, defaults to [0.5, 0.5, 0.5]) — Image mean to use for normalization. Only has an effect if do_normalize is set to True.
  • image_std (Union[float, list[float], tuple[float, ...]], kwargs, optional, defaults to [0.5, 0.5, 0.5]) — Image standard deviation to use for normalization. Only has an effect if do_normalize is set to True.
  • do_center_crop (bool, kwargs, optional) — Whether to center crop the image.
  • do_pad (bool, kwargs, optional) — Whether to pad the image. Padding is done either to the largest size in the batch or to a fixed square size per image. The exact padding strategy depends on the model.
  • crop_size (Annotated[int | list[int] | tuple[int, ...] | dict[str, int] | None, None], kwargs) — Size of the output image after applying center_crop.
  • data_format (Union[str, ~image_utils.ChannelDimension], kwargs, optional) — Only ChannelDimension.FIRST is supported. Added for compatibility with slow processors.
  • input_data_format (Union[str, ~image_utils.ChannelDimension], kwargs, optional) — The channel dimension format for the input image. If unset, the channel dimension format is inferred from the input image. Can be one of:

    • "channels_first" or ChannelDimension.FIRST: image in (num_channels, height, width) format.
    • "channels_last" or ChannelDimension.LAST: image in (height, width, num_channels) format.
    • "none" or ChannelDimension.NONE: image in (height, width) format.
  • device (Annotated[Union[str, torch.device, NoneType], None], kwargs) — The device to process the videos on. If unset, the device is inferred from the input videos.
  • do_sample_frames (bool, kwargs, optional, defaults to True) — Whether to sample frames from the video before processing or to process the whole video.
  • video_metadata (Annotated[~video_utils.VideoMetadata | dict | list[dict | ~video_utils.VideoMetadata] | list[list[dict | ~video_utils.VideoMetadata]] | None, None], kwargs) — Metadata of the video containing information about total duration, fps and total number of frames. It will be used to sample frames from video or compute timestamps. Don’t pass any metadata unless you are trying to decode the video manually before processing
  • fps (Annotated[int | float | None, None], kwargs) — Target frames to sample per second when do_sample_frames=True.
  • num_frames (Annotated[int | None, None], kwargs, defaults to 64) — Maximum number of frames to sample when do_sample_frames=True.
  • return_metadata (bool, kwargs, optional, defaults to False) — Whether to return video metadata or not. Video metadats is an object containing info about video duration, fps, decoding backend, etc.
  • return_tensors (Annotated[str | ~utils.generic.TensorType | None, None], kwargs) — If set, will return tensors of a particular framework. Acceptable values are:

    • 'pt': Return PyTorch torch.Tensor objects.
    • 'np': Return NumPy np.ndarray objects.
  • patch_size (int, kwargs, optional) — Side length in pixels of each ViT patch for video frames.
  • pooling_size (list[int], kwargs, optional) — [pool_h, pool_w] pooling window applied to video patch features.
  • max_fps (int, kwargs, optional) — Maximum sampling rate in frames per second for short videos.
  • **kwargs (VideosKwargs, optional) — Additional processing options for each modality (text, images, videos, audio). Model-specific parameters are listed above; see the TypedDict class for the complete list of supported arguments.

Constructs a Molmo2VideoProcessor video processor.

__call__

< >

( videos**kwargs )

Molmo2Model

class transformers.Molmo2Model

< >

( config: Molmo2Config )

Parameters

  • config (Molmo2Config) — Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the from_pretrained() method to load the model weights.

The Molmo2 model which consists of a vision backbone, a pooling adapter and a language model, without a language modeling head.

This model inherits from PreTrainedModel. Check the superclass documentation for the generic methods the library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads etc.)

This model is also a PyTorch torch.nn.Module subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.

forward

< >

( input_ids: typing.Optional[torch.LongTensor] = Nonepixel_values: typing.Optional[torch.FloatTensor] = Noneimage_token_pooling: typing.Optional[torch.Tensor] = Noneimage_grids: typing.Optional[torch.Tensor] = Noneimage_num_crops: typing.Optional[torch.Tensor] = Nonepixel_values_videos: typing.Optional[torch.Tensor] = Nonevideo_token_pooling: typing.Optional[torch.Tensor] = Nonevideo_grids: typing.Optional[torch.Tensor] = Noneattention_mask: typing.Optional[torch.Tensor] = Noneposition_ids: typing.Optional[torch.Tensor] = Nonepast_key_values: transformers.cache_utils.Cache | None = Nonemm_token_type_ids: typing.Optional[torch.LongTensor] = Noneinputs_embeds: typing.Optional[torch.FloatTensor] = Noneuse_cache: bool | None = Nonemm_encoder_outputs: dict[str, transformers.modeling_outputs.BaseModelOutputWithPooling] | None = None**kwargs: Unpack ) → Molmo2ModelOutputWithPast or tuple(torch.FloatTensor)

Parameters

  • input_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) — Indices of input sequence tokens in the vocabulary. Padding will be ignored by default.

    Indices can be obtained using AutoTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.

    What are input IDs?

  • pixel_values (torch.FloatTensor of shape (batch_size, num_channels, image_size, image_size), optional) — The tensors corresponding to the input images. Pixel values can be obtained using Molmo2ImageProcessor. See Molmo2ImageProcessor.call() for details (Molmo2Processor uses Molmo2ImageProcessor for processing images).
  • image_token_pooling (torch.Tensor of shape (num_image_tokens, pool_h * pool_w), optional) — Indices into the flattened image patch sequence gathered by each pooled image token, with -1 marking padding slots. Can be obtained using AutoProcessor. See ProcessorMixin.call() for details.
  • image_grids (torch.Tensor of shape (num_images, 4), optional) — Per-image [low_res_h, low_res_w, high_res_h, high_res_w] pooled token grid. Can be obtained using AutoProcessor. See ProcessorMixin.call() for details.
  • image_num_crops (torch.Tensor of shape (num_images,), optional) — Number of crops per image, low-resolution view included. Can be obtained using AutoProcessor. See ProcessorMixin.call() for details.
  • pixel_values_videos (torch.Tensor of shape (batch_size, num_frames, num_channels, frame_size, frame_size), optional) — The tensors corresponding to the input video. Pixel values for videos can be obtained using Molmo2VideoProcessor. See Molmo2VideoProcessor.call() for details (Molmo2Processor uses Molmo2VideoProcessor for processing videos).
  • video_token_pooling (torch.Tensor of shape (num_video_tokens, pool_h * pool_w), optional) — Indices into the flattened video frame-patch sequence gathered by each pooled video token, with -1 marking padding slots. Can be obtained using AutoProcessor. See ProcessorMixin.call() for details.
  • video_grids (torch.Tensor of shape (num_videos, 3), optional) — Per-video [num_frames, pooled_h, pooled_w] token grid. Can be obtained using AutoProcessor. See ProcessorMixin.call() for details.
  • attention_mask (torch.Tensor of shape (batch_size, sequence_length), optional) — Mask to avoid performing attention on padding token indices. Mask values selected in [0, 1]:

    • 1 for tokens that are not masked,
    • 0 for tokens that are masked.

    What are attention masks?

  • position_ids (torch.Tensor of shape (batch_size, sequence_length), optional) — Indices of positions of each input sequence tokens in the position embeddings. Selected in the range [0, config.n_positions - 1].

    What are position IDs?

  • past_key_values (~cache_utils.Cache, optional) — Pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention blocks) that can be used to speed up sequential decoding. This typically consists in the past_key_values returned by the model at a previous stage of decoding, when use_cache=True or config.use_cache=True.

    Only Cache instance is allowed as input, see our kv cache guide. If no past_key_values are passed, DynamicCache will be initialized by default.

    The model will output the same cache format that is fed as input.

    If past_key_values are used, the user is expected to input only unprocessed input_ids (those that don’t have their past key value states given to this model) of shape (batch_size, unprocessed_length) instead of all input_ids of shape (batch_size, sequence_length).

  • mm_token_type_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) — Indices of input sequence tokens matching each modality. For example text (0), image (1), video (2). Multimodal token type ids can be obtained using AutoProcessor. See ProcessorMixin.call() for details.
  • inputs_embeds (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size), optional) — Optionally, instead of passing input_ids you can choose to directly pass an embedded representation. This is useful if you want more control over how to convert input_ids indices into associated vectors than the model’s internal embedding lookup matrix.
  • use_cache (bool, optional) — If set to True, past_key_values key value states are returned and can be used to speed up decoding (see past_key_values).
  • mm_encoder_outputs (dict[str, ~modeling_outputs.BaseModelOutputWithPooling], optional) — Dict where keys are supported modalities and values are encoded outputs for that modality. Each encoded output is a tuple that consists of (pooler_output, optional: last_hidden_states, optional: hidden_states, optional: attentions) pooler_output of shape (batch_size, sequence_length, hidden_size), optional) is a sequence of multimmodal features of the encoder merged into text embeddings.

Returns

Molmo2ModelOutputWithPast or tuple(torch.FloatTensor)

A Molmo2ModelOutputWithPast or a tuple of torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various elements depending on the configuration (Molmo2Config) and inputs.

The Molmo2Model forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

  • last_hidden_state (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size)) — Sequence of hidden-states at the output of the last layer of the model.

    If past_key_values is used only the last hidden-state of the sequences of shape (batch_size, 1, hidden_size) is output.

  • past_key_values (Cache, optional, returned when use_cache=True is passed or when config.use_cache=True) — It is a Cache instance. For more details, see our kv cache guide.

    Contains pre-computed hidden-states (key and values in the self-attention blocks) that can be used (see past_key_values input) to speed up sequential decoding.

  • hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings, if the model has an embedding layer, + one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

    Hidden-states of the model at the output of each layer plus the optional initial embedding outputs.

  • attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True is passed or when config.output_attentions=True) — Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length).

    Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.

  • image_hidden_states (torch.FloatTensor, optional) — A torch.FloatTensor of size (batch_size, num_images, sequence_length, hidden_size). image_hidden_states of the model produced by the vision encoder and after projecting the last hidden state.

Molmo2TextModel

class transformers.Molmo2TextModel

< >

( config: Molmo2TextConfig )

Parameters

  • config (Molmo2TextConfig) — Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the from_pretrained() method to load the model weights.

The bare Molmo2 Text Model outputting raw hidden-states without any specific head on top.

This model inherits from PreTrainedModel. Check the superclass documentation for the generic methods the library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads etc.)

This model is also a PyTorch torch.nn.Module subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.

forward

< >

( input_ids: typing.Optional[torch.LongTensor] = Noneattention_mask: typing.Optional[torch.Tensor] = Noneposition_ids: typing.Optional[torch.LongTensor] = Nonepast_key_values: transformers.cache_utils.Cache | None = Noneinputs_embeds: typing.Optional[torch.FloatTensor] = Noneuse_cache: bool | None = None**kwargs: Unpack ) → BaseModelOutputWithPast or tuple(torch.FloatTensor)

Parameters

  • input_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) — Indices of input sequence tokens in the vocabulary. Padding will be ignored by default.

    Indices can be obtained using AutoTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.

    What are input IDs?

  • attention_mask (torch.Tensor of shape (batch_size, sequence_length), optional) — Mask to avoid performing attention on padding token indices. Mask values selected in [0, 1]:

    • 1 for tokens that are not masked,
    • 0 for tokens that are masked.

    What are attention masks?

  • position_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) — Indices of positions of each input sequence tokens in the position embeddings. Selected in the range [0, config.n_positions - 1].

    What are position IDs?

  • past_key_values (~cache_utils.Cache, optional) — Pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention blocks) that can be used to speed up sequential decoding. This typically consists in the past_key_values returned by the model at a previous stage of decoding, when use_cache=True or config.use_cache=True.

    Only Cache instance is allowed as input, see our kv cache guide. If no past_key_values are passed, DynamicCache will be initialized by default.

    The model will output the same cache format that is fed as input.

    If past_key_values are used, the user is expected to input only unprocessed input_ids (those that don’t have their past key value states given to this model) of shape (batch_size, unprocessed_length) instead of all input_ids of shape (batch_size, sequence_length).

  • inputs_embeds (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size), optional) — Optionally, instead of passing input_ids you can choose to directly pass an embedded representation. This is useful if you want more control over how to convert input_ids indices into associated vectors than the model’s internal embedding lookup matrix.
  • use_cache (bool, optional) — If set to True, past_key_values key value states are returned and can be used to speed up decoding (see past_key_values).

Returns

BaseModelOutputWithPast or tuple(torch.FloatTensor)

A BaseModelOutputWithPast or a tuple of torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various elements depending on the configuration (Molmo2Config) and inputs.

The Molmo2TextModel forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

  • last_hidden_state (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size)) — Sequence of hidden-states at the output of the last layer of the model.

    If past_key_values is used only the last hidden-state of the sequences of shape (batch_size, 1, hidden_size) is output.

  • past_key_values (Cache, optional, returned when use_cache=True is passed or when config.use_cache=True) — It is a Cache instance. For more details, see our kv cache guide.

    Contains pre-computed hidden-states (key and values in the self-attention blocks and optionally if config.is_encoder_decoder=True in the cross-attention blocks) that can be used (see past_key_values input) to speed up sequential decoding.

  • hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings, if the model has an embedding layer, + one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

    Hidden-states of the model at the output of each layer plus the optional initial embedding outputs.

  • attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True is passed or when config.output_attentions=True) — Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length).

    Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.

Molmo2Adapter

class transformers.Molmo2Adapter

< >

( config: Molmo2AdapterConfig )

Parameters

  • config (Molmo2AdapterConfig) — Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the from_pretrained() method to load the model weights.

The Molmo2 vision adapter: pools ViT patch features with a cross-attention layer and projects them into the language model’s embedding space.

This model inherits from PreTrainedModel. Check the superclass documentation for the generic methods the library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads etc.)

This model is also a PyTorch torch.nn.Module subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.

forward

< >

( image_features: Tensorpooled_patches_idx: Tensor**kwargs: Unpack ) → BaseModelOutput or tuple(torch.FloatTensor)

Parameters

  • image_features (torch.Tensor of shape (num_crops, num_patches, hidden_size * len(vision_feature_layer))) — Concatenated intermediate ViT features of every crop.
  • pooled_patches_idx (torch.Tensor of shape (num_tokens, pool_h * pool_w)) — Indices into the flattened patch sequence pooled by each output token; -1 marks padding slots.

Returns

BaseModelOutput or tuple(torch.FloatTensor)

A BaseModelOutput or a tuple of torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various elements depending on the configuration (Molmo2Config) and inputs.

The Molmo2Adapter forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

  • last_hidden_state (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size)) — Sequence of hidden-states at the output of the last layer of the model.

  • hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings, if the model has an embedding layer, + one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

    Hidden-states of the model at the output of each layer plus the optional initial embedding outputs.

  • attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True is passed or when config.output_attentions=True) — Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length).

    Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.

Molmo2VisionModel

class transformers.Molmo2VisionModel

< >

( config: Molmo2VisionConfig )

Parameters

  • config (Molmo2VisionConfig) — Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the from_pretrained() method to load the model weights.

The bare Molmo2 Model outputting raw hidden-states without any specific head on top.

This model inherits from PreTrainedModel. Check the superclass documentation for the generic methods the library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads etc.)

This model is also a PyTorch torch.nn.Module subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.

forward

< >

( pixel_values: Tensor**kwargs: Unpack ) → BaseModelOutputWithPooling or tuple(torch.FloatTensor)

Parameters

Returns

BaseModelOutputWithPooling or tuple(torch.FloatTensor)

A BaseModelOutputWithPooling or a tuple of torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various elements depending on the configuration (Molmo2Config) and inputs.

The Molmo2VisionModel forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

  • last_hidden_state (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size)) — Sequence of hidden-states at the output of the last layer of the model.

  • pooler_output (torch.FloatTensor of shape (batch_size, hidden_size)) — Last layer hidden-state of the first token of the sequence (classification token) after further processing through the layers used for the auxiliary pretraining task. E.g. for BERT-family of models, this returns the classification token after processing through a linear layer and a tanh activation function. The linear layer weights are trained from the next sentence prediction (classification) objective during pretraining.

  • hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings, if the model has an embedding layer, + one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

    Hidden-states of the model at the output of each layer plus the optional initial embedding outputs.

  • attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True is passed or when config.output_attentions=True) — Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length).

    Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.

Molmo2ForConditionalGeneration

class transformers.Molmo2ForConditionalGeneration

< >

( config: Molmo2Config )

Parameters

  • config (Molmo2Config) — Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the from_pretrained() method to load the model weights.

The Molmo2 model which consists of a vision backbone, a pooling adapter and a language model.

This model inherits from PreTrainedModel. Check the superclass documentation for the generic methods the library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads etc.)

This model is also a PyTorch torch.nn.Module subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.

forward

< >

( input_ids: LongTensor = Nonepixel_values: typing.Optional[torch.Tensor] = Noneimage_token_pooling: typing.Optional[torch.Tensor] = Noneimage_grids: typing.Optional[torch.Tensor] = Noneimage_num_crops: typing.Optional[torch.Tensor] = Nonepixel_values_videos: typing.Optional[torch.Tensor] = Nonevideo_token_pooling: typing.Optional[torch.Tensor] = Nonevideo_grids: typing.Optional[torch.Tensor] = Noneattention_mask: typing.Optional[torch.Tensor] = Noneposition_ids: typing.Optional[torch.LongTensor] = Nonepast_key_values: list[torch.FloatTensor] | None = Nonemm_token_type_ids: typing.Optional[torch.LongTensor] = Noneinputs_embeds: typing.Optional[torch.FloatTensor] = Nonelabels: typing.Optional[torch.LongTensor] = Noneuse_cache: bool | None = Nonelogits_to_keep: typing.Union[int, torch.Tensor] = 0mm_encoder_outputs: dict[str, transformers.modeling_outputs.BaseModelOutputWithPooling] | None = None**kwargs: Unpack ) → Molmo2CausalLMOutputWithPast or tuple(torch.FloatTensor)

Parameters

  • input_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) — Indices of input sequence tokens in the vocabulary. Padding will be ignored by default.

    Indices can be obtained using AutoTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.

    What are input IDs?

  • pixel_values (torch.Tensor of shape (batch_size, num_channels, image_size, image_size), optional) — The tensors corresponding to the input images. Pixel values can be obtained using Molmo2ImageProcessor. See Molmo2ImageProcessor.call() for details (Molmo2Processor uses Molmo2ImageProcessor for processing images).
  • image_token_pooling (torch.Tensor of shape (num_image_tokens, pool_h * pool_w), optional) — Indices into the flattened image patch sequence gathered by each pooled image token, with -1 marking padding slots. Can be obtained using AutoProcessor. See ProcessorMixin.call() for details.
  • image_grids (torch.Tensor of shape (num_images, 4), optional) — Per-image [low_res_h, low_res_w, high_res_h, high_res_w] pooled token grid. Can be obtained using AutoProcessor. See ProcessorMixin.call() for details.
  • image_num_crops (torch.Tensor of shape (num_images,), optional) — Number of crops per image, low-resolution view included. Can be obtained using AutoProcessor. See ProcessorMixin.call() for details.
  • pixel_values_videos (torch.Tensor of shape (batch_size, num_frames, num_channels, frame_size, frame_size), optional) — The tensors corresponding to the input video. Pixel values for videos can be obtained using Molmo2VideoProcessor. See Molmo2VideoProcessor.call() for details (Molmo2Processor uses Molmo2VideoProcessor for processing videos).
  • video_token_pooling (torch.Tensor of shape (num_video_tokens, pool_h * pool_w), optional) — Indices into the flattened video frame-patch sequence gathered by each pooled video token, with -1 marking padding slots. Can be obtained using AutoProcessor. See ProcessorMixin.call() for details.
  • video_grids (torch.Tensor of shape (num_videos, 3), optional) — Per-video [num_frames, pooled_h, pooled_w] token grid. Can be obtained using AutoProcessor. See ProcessorMixin.call() for details.
  • attention_mask (torch.Tensor of shape (batch_size, sequence_length), optional) — Mask to avoid performing attention on padding token indices. Mask values selected in [0, 1]:

    • 1 for tokens that are not masked,
    • 0 for tokens that are masked.

    What are attention masks?

  • position_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) — Indices of positions of each input sequence tokens in the position embeddings. Selected in the range [0, config.n_positions - 1].

    What are position IDs?

  • past_key_values (list[torch.FloatTensor], optional) — Pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention blocks) that can be used to speed up sequential decoding. This typically consists in the past_key_values returned by the model at a previous stage of decoding, when use_cache=True or config.use_cache=True.

    Only Cache instance is allowed as input, see our kv cache guide. If no past_key_values are passed, DynamicCache will be initialized by default.

    The model will output the same cache format that is fed as input.

    If past_key_values are used, the user is expected to input only unprocessed input_ids (those that don’t have their past key value states given to this model) of shape (batch_size, unprocessed_length) instead of all input_ids of shape (batch_size, sequence_length).

  • mm_token_type_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) — Indices of input sequence tokens matching each modality. For example text (0), image (1), video (2). Multimodal token type ids can be obtained using AutoProcessor. See ProcessorMixin.call() for details.
  • inputs_embeds (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size), optional) — Optionally, instead of passing input_ids you can choose to directly pass an embedded representation. This is useful if you want more control over how to convert input_ids indices into associated vectors than the model’s internal embedding lookup matrix.
  • labels (torch.LongTensor of shape (batch_size, sequence_length), optional) — Labels for computing the masked language modeling loss. Indices should either be in [0, ..., config.vocab_size] or -100 (see input_ids docstring). Tokens with indices set to -100 are ignored (masked), the loss is only computed for the tokens with labels in [0, ..., config.vocab_size].
  • use_cache (bool, optional) — If set to True, past_key_values key value states are returned and can be used to speed up decoding (see past_key_values).
  • logits_to_keep (Union[int, torch.Tensor], optional, defaults to 0) — If an int, compute logits for the last logits_to_keep tokens. If 0, calculate logits for all input_ids (special case). Only last token logits are needed for generation, and calculating them only for that token can save memory, which becomes pretty significant for long sequences or large vocabulary size. If a torch.Tensor, must be 1D corresponding to the indices to keep in the sequence length dimension. This is useful when using packed tensor format (single dimension for batch and sequence length).
  • mm_encoder_outputs (dict[str, ~modeling_outputs.BaseModelOutputWithPooling], optional) — Dict where keys are supported modalities and values are encoded outputs for that modality. Each encoded output is a tuple that consists of (pooler_output, optional: last_hidden_states, optional: hidden_states, optional: attentions) pooler_output of shape (batch_size, sequence_length, hidden_size), optional) is a sequence of multimmodal features of the encoder merged into text embeddings.

Returns

Molmo2CausalLMOutputWithPast or tuple(torch.FloatTensor)

A Molmo2CausalLMOutputWithPast or a tuple of torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various elements depending on the configuration (Molmo2Config) and inputs.

The Molmo2ForConditionalGeneration forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

  • loss (torch.FloatTensor of shape (1,), optional, returned when labels is provided) — Language modeling loss (for next-token prediction).

  • logits (torch.FloatTensor of shape (batch_size, sequence_length, config.vocab_size)) — Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).

  • past_key_values (Cache, optional, returned when use_cache=True is passed or when config.use_cache=True) — It is a Cache instance. For more details, see our kv cache guide.

    Contains pre-computed hidden-states (key and values in the self-attention blocks) that can be used (see past_key_values input) to speed up sequential decoding.

  • hidden_states (tuple[torch.FloatTensor], optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings, if the model has an embedding layer, + one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

    Hidden-states of the model at the output of each layer plus the optional initial embedding outputs.

  • attentions (tuple[torch.FloatTensor], optional, returned when output_attentions=True is passed or when config.output_attentions=True) — Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length).

    Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.

  • image_hidden_states (torch.FloatTensor, optional) — A torch.FloatTensor of size (batch_size, num_images, sequence_length, hidden_size). image_hidden_states of the model produced by the vision encoder and after projecting the last hidden state.

Example:

>>> from PIL import Image
>>> import requests
>>> from transformers import AutoProcessor, Molmo2ForConditionalGeneration

>>> model = Molmo2ForConditionalGeneration.from_pretrained("allenai/Molmo2-8B")
>>> processor = AutoProcessor.from_pretrained("allenai/Molmo2-8B")

>>> prompt = "What's the content of the image?"
>>> url = "https://www.ilankelman.org/stopsigns/australia.jpg"
>>> image = Image.open(requests.get(url, stream=True).raw)

>>> messages = [{"role": "user", "content": [{"type": "text", "text": prompt}, {"type": "image", "image": image}]}]

>>> inputs = processor.apply_chat_template(messages, tokenize=True, add_generation_prompt=True, return_tensors="pt", return_dict=True)

>>> # Generate
>>> generated_ids = model.generate(**inputs, max_new_tokens=15)
>>> generated_tokens = generated_ids[:, inputs['input_ids'].size(1):]
>>> processor.post_process_image_text_to_text(generated_tokens, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
"The image shows a bustling street scene in what appears to be a Chinatown area. There's ..."
Update on GitHub

←MM Grounding DINO MusicFlamingo→