The Gemma 3 model was proposed in the Gemma 3 Techncial Report by Google. It is a vision-language model composed by a SigLIP vision encoder and a Gemma 2 language decoder, linked by a multimodal linear projection. It cuts an image into a fixed number of tokens, in the same way as SigLIP, as long as the image does not exceed certain aspect ratio. For images that exceed the given aspect ratio, it crops the image into multiple smaller patches and concatenates them with the base image embedding. One particularity is that the model uses bidirectional attention on all the image tokens. In addition, the model interleaves sliding window local attention with full causal attention in the language backbone, where each sixth layer is a full causal attention layer.
This model was contributed by Ryan Mullins, Raushan Turganbay Arthur Zucker, and Pedro Cuenca.
Gemma3ForConditionalGeneration
.Gemma3ForCausalLM
for generation to avoid loading the vision tower.<start_of_image>
token wherever an image should be inserted.apply_chat_template
method to convert chat messages to model inputs. See the examples below for more details on how to use it.The model supports cropping images into smaller patches when the image aspect ratio exceeds a certain value. By default the images are not cropped and only the base image is forwarded to the model. Users can set do_pan_and_scan=True
to obtain several crops per image along with the base image to improve the quality in DocVQA or similar tasks requiring higher resolution images.
Pan and scan is an inference time optimization to handle images with skewed aspect ratios. When enabled, it improves performance on tasks related to document understanding, infographics, OCR, etc.
processor = AutoProcessor.from_pretrained("google/gemma-3-4b-it", padding_side="left")
url = "https://media.istockphoto.com/id/1192867753/photo/cow-in-berchida-beach-siniscola.jpg?s=612x612&w=0&k=20&c=v0hjjniwsMNfJSuKWZuIn8pssmD5h5bSN1peBd1CmH4="
messages = [
{
"role": "system",
"content": [
{"type": "text", "text": "You are a helpful assistant."}
]
},
{
"role": "user", "content": [
{"type": "image", "url": url},
{"type": "text", "text": "What is shown in this image?"},
]
},
]
inputs = processor.apply_chat_template(
messages,
tokenize=True,
return_dict=True,
return_tensors="pt",
add_generation_prompt=True,
do_pan_and_scan=True,
).to(model.device)
from transformers import AutoProcessor, Gemma3ForConditionalGeneration
model_id = "google/gemma-3-4b-it"
model = Gemma3ForConditionalGeneration.from_pretrained(model_id, device_map="auto")
processor = AutoProcessor.from_pretrained(model_id, padding_side="left")
url = "https://media.istockphoto.com/id/1192867753/photo/cow-in-berchida-beach-siniscola.jpg?s=612x612&w=0&k=20&c=v0hjjniwsMNfJSuKWZuIn8pssmD5h5bSN1peBd1CmH4="
messages = [
{
"role": "system",
"content": [
{"type": "text", "text": "You are a helpful assistant."}
]
},
{
"role": "user", "content": [
{"type": "image", "url": url},
{"type": "text", "text": "What is shown in this image?"},
]
},
]
inputs = processor.apply_chat_template(
messages,
tokenize=True,
return_dict=True,
return_tensors="pt",
add_generation_prompt=True,
).to(model.device)
output = model.generate(**inputs, max_new_tokens=50)
print(processor.decode(output[0], skip_special_tokens=True)[inputs.input_ids.shape[1]: ])
model_id = "google/gemma-3-4b-it"
model = Gemma3ForConditionalGeneration.from_pretrained(model_id, device_map="auto")
processor = AutoProcessor.from_pretrained(model_id, padding_side="left")
url_cow = "https://media.istockphoto.com/id/1192867753/photo/cow-in-berchida-beach-siniscola.jpg?s=612x612&w=0&k=20&c=v0hjjniwsMNfJSuKWZuIn8pssmD5h5bSN1peBd1CmH4="
url_stop = "https://www.ilankelman.org/stopsigns/australia.jpg"
messages = [
{
"role": "system",
"content": [
{"type": "text", "text": "You are a helpful assistant."}
]
},
{
"role": "user", "content": [
{"type": "image", "url": url_cow},
{"type": "image", "url": url_stop},
{"type": "text", "text": "Are these two images identical?"},
]
},
]
inputs = processor.apply_chat_template(
messages,
tokenize=True,
return_dict=True,
return_tensors="pt",
add_generation_prompt=True,
).to(model.device)
output = model.generate(**inputs, max_new_tokens=50)
print(processor.decode(output[0], skip_special_tokens=True)[inputs.input_ids.shape[1]: ])
You can use the VLMs for text-only generation by omitting images in your input. However, you can also load the models in text-only mode as shown below. This will skip loading the vision tower and will save resources when you just need the LLM capabilities.
from transformers import AutoTokenizer, Gemma3ForCausalLM
model_id = "google/gemma-3-1b-it"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = Gemma3ForCausalLM.from_pretrained(model_id, device_map="auto")
input_ids = tokenizer("Write me a poem about Machine Learning.", return_tensors="pt").to(model.device)
outputs = model.generate(**input_ids, max_new_tokens=100)
text = tokenizer.batch_decode(outputs, skip_special_tokens=True)
print(text)
( do_resize: bool = True size: typing.Dict[str, int] = None resample: Resampling = <Resampling.BILINEAR: 2> do_rescale: bool = True rescale_factor: typing.Union[int, float] = 0.00392156862745098 do_normalize: bool = True image_mean: typing.Union[float, typing.List[float], NoneType] = None image_std: typing.Union[float, typing.List[float], NoneType] = None do_convert_rgb: bool = None do_pan_and_scan: bool = None pan_and_scan_min_crop_size: int = None pan_and_scan_max_num_crops: int = None pan_and_scan_min_ratio_to_activate: float = None **kwargs )
Parameters
bool
, optional, defaults to True
) —
Whether to resize the image’s (height, width) dimensions to the specified size
. Can be overridden by
do_resize
in the preprocess
method. Dict[str, int]
optional, defaults to {"height" -- 224, "width": 224}
):
Size of the image after resizing. Can be overridden by size
in the preprocess
method. PILImageResampling
, optional, defaults to Resampling.BILINEAR
) —
Resampling filter to use if resizing the image. Can be overridden by resample
in the preprocess
method. bool
, optional, defaults to True
) —
Whether to rescale the image by the specified scale rescale_factor
. Can be overridden by do_rescale
in
the preprocess
method. int
or float
, optional, defaults to 1/255
) —
Scale factor to use if rescaling the image. Can be overridden by rescale_factor
in the preprocess
method. bool
, optional, defaults to True
) —
Whether to normalize the image by the specified mean and standard deviation. Can be overridden by
do_normalize
in the preprocess
method. float
or List[float]
, optional, defaults to [0.5, 0.5, 0.5]
) —
Mean to use if normalizing the image. This is a float or list of floats the length of the number of
channels in the image. Can be overridden by the image_mean
parameter in the preprocess
method. float
or List[float]
, optional, defaults to [0.5, 0.5, 0.5]
) —
Standard deviation to use if normalizing the image. This is a float or list of floats the length of the
number of channels in the image. Can be overridden by the image_std
parameter in the preprocess
method.
Can be overridden by the image_std
parameter in the preprocess
method. bool
, optional, defaults to True
) —
Whether to convert the image to RGB. bool
, optional) —
Whether to apply pan_and_scan
to images. int
, optional) —
Minimum size of each crop in pan and scan. int
, optional) —
Maximum number of crops per image in pan and scan. float
, optional) —
Minimum aspect ratio to activate pan and scan. Constructs a SigLIP image processor.
( image: ndarray pan_and_scan_min_crop_size: int pan_and_scan_max_num_crops: int pan_and_scan_min_ratio_to_activate: float data_format: typing.Union[str, transformers.image_utils.ChannelDimension, NoneType] = None input_data_format: typing.Union[str, transformers.image_utils.ChannelDimension, NoneType] = None )
Parameters
np.ndarray
) —
Image to resize. int
, optional) —
Minimum size of each crop in pan and scan. int
, optional) —
Maximum number of crops per image in pan and scan. float
, optional) —
Minimum aspect ratio to activate pan and scan. str
or ChannelDimension
, optional) —
The channel dimension format of the image. If not provided, it will be the same as the input image. ChannelDimension
or str
, optional) —
The channel dimension format of the input image. If not provided, it will be inferred. Pan and Scan and image, by cropping into smaller images when the aspect ratio exceeds minumum allowed ratio.
( images: typing.Union[ForwardRef('PIL.Image.Image'), numpy.ndarray, ForwardRef('torch.Tensor'), typing.List[ForwardRef('PIL.Image.Image')], typing.List[numpy.ndarray], typing.List[ForwardRef('torch.Tensor')]] do_resize: bool = None size: typing.Dict[str, int] = None resample: Resampling = None do_rescale: bool = None rescale_factor: float = None do_normalize: bool = None image_mean: typing.Union[float, typing.List[float], NoneType] = None image_std: typing.Union[float, typing.List[float], NoneType] = None return_tensors: typing.Union[str, transformers.utils.generic.TensorType, NoneType] = None data_format: typing.Optional[transformers.image_utils.ChannelDimension] = <ChannelDimension.FIRST: 'channels_first'> input_data_format: typing.Union[str, transformers.image_utils.ChannelDimension, NoneType] = None do_convert_rgb: bool = None do_pan_and_scan: bool = None pan_and_scan_min_crop_size: int = None pan_and_scan_max_num_crops: int = None pan_and_scan_min_ratio_to_activate: float = None )
Parameters
ImageInput
) —
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
. bool
, optional, defaults to self.do_resize
) —
Whether to resize the image. Dict[str, int]
, optional, defaults to self.size
) —
Size of the image after resizing. int
, optional, defaults to self.resample
) —
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
. bool
, optional, defaults to self.do_rescale
) —
Whether to rescale the image. float
, optional, defaults to self.rescale_factor
) —
Rescale factor to rescale the image by if do_rescale
is set to True
. bool
, optional, defaults to self.do_normalize
) —
Whether to normalize the image. float
or List[float]
, optional, defaults to self.image_mean
) —
Image mean to use for normalization. Only has an effect if do_normalize
is set to True
. float
or List[float]
, optional, defaults to self.image_std
) —
Image standard deviation to use for normalization. Only has an effect if do_normalize
is set to
True
. str
or TensorType
, optional) —
The type of tensors to return. Can be one of:np.ndarray
.TensorType.TENSORFLOW
or 'tf'
: Return a batch of type tf.Tensor
.TensorType.PYTORCH
or 'pt'
: Return a batch of type torch.Tensor
.TensorType.NUMPY
or 'np'
: Return a batch of type np.ndarray
.TensorType.JAX
or 'jax'
: Return a batch of type jax.numpy.ndarray
.ChannelDimension
or str
, optional, defaults to ChannelDimension.FIRST
) —
The channel dimension format for the output 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.ChannelDimension
or str
, 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.bool
, optional, defaults to self.do_convert_rgb
) —
Whether to convert the image to RGB. bool
, optional, defaults to self.do_convert_rgb
) —
Whether to apply pan_and_scan
to images. int
, optional, defaults to self.pan_and_scan_min_crop_size
) —
Minimum size of each crop in pan and scan. int
, optional, defaults to self.pan_and_scan_max_num_crops
) —
Maximum number of crops per image in pan and scan. float
, optional, defaults to self.pan_and_scan_min_ratio_to_activate
) —
Minimum aspect ratio to activate pan and scan. Preprocess an image or batch of images.
( **kwargs: typing_extensions.Unpack[transformers.models.gemma3.image_processing_gemma3_fast.Gemma3FastImageProcessorKwargs] )
Parameters
bool
, optional, defaults to self.do_resize
) —
Whether to resize the image’s (height, width) dimensions to the specified size
. Can be overridden by the
do_resize
parameter in the preprocess
method. dict
, optional, defaults to self.size
) —
Size of the output image after resizing. Can be overridden by the size
parameter in the preprocess
method. bool
, optional, defaults to self.default_to_square
) —
Whether to default to a square image when resizing, if size is an int. PILImageResampling
, optional, defaults to self.resample
) —
Resampling filter to use if resizing the image. Only has an effect if do_resize
is set to True
. Can be
overridden by the resample
parameter in the preprocess
method. bool
, optional, defaults to self.do_center_crop
) —
Whether to center crop the image to the specified crop_size
. Can be overridden by do_center_crop
in the
preprocess
method. Dict[str, int]
optional, defaults to self.crop_size
) —
Size of the output image after applying center_crop
. Can be overridden by crop_size
in the preprocess
method. bool
, optional, defaults to self.do_rescale
) —
Whether to rescale the image by the specified scale rescale_factor
. Can be overridden by the
do_rescale
parameter in the preprocess
method. int
or float
, optional, defaults to self.rescale_factor
) —
Scale factor to use if rescaling the image. Only has an effect if do_rescale
is set to True
. Can be
overridden by the rescale_factor
parameter in the preprocess
method. bool
, optional, defaults to self.do_normalize
) —
Whether to normalize the image. Can be overridden by the do_normalize
parameter in the preprocess
method. Can be overridden by the do_normalize
parameter in the preprocess
method. float
or List[float]
, optional, defaults to self.image_mean
) —
Mean to use if normalizing the image. This is a float or list of floats the length of the number of
channels in the image. Can be overridden by the image_mean
parameter in the preprocess
method. Can be
overridden by the image_mean
parameter in the preprocess
method. float
or List[float]
, optional, defaults to self.image_std
) —
Standard deviation to use if normalizing the image. This is a float or list of floats the length of the
number of channels in the image. Can be overridden by the image_std
parameter in the preprocess
method.
Can be overridden by the image_std
parameter in the preprocess
method. bool
, optional, defaults to self.do_convert_rgb
) —
Whether to convert the image to RGB. str
or TensorType
, optional, defaults to self.return_tensors
) —
Returns stacked tensors if set to `pt, otherwise returns a list of tensors. ChannelDimension
or str
, optional, defaults to self.data_format
) —
Only ChannelDimension.FIRST
is supported. Added for compatibility with slow processors. ChannelDimension
or str
, optional, defaults to self.input_data_format
) —
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.torch.device
, optional, defaults to self.device
) —
The device to process the images on. If unset, the device is inferred from the input images. bool
, optional) —
Whether to apply pan_and_scan
to images. int
, optional) —
Minimum size of each crop in pan and scan. int
, optional) —
Maximum number of crops per image in pan and scan. float
, optional) —
Minimum aspect ratio to activate pan and scan. Constructs a fast ConvNeXT image processor. Based on SiglipImageProcessor with incorporation of Pan adn Scan cropping method.
( image: torch.Tensor pan_and_scan_min_crop_size: int pan_and_scan_max_num_crops: int pan_and_scan_min_ratio_to_activate: float )
Parameters
torch.Tensor
) —
Image to resize. int
, optional) —
Minimum size of each crop in pan and scan. int
, optional) —
Maximum number of crops per image in pan and scan. float
, optional) —
Minimum aspect ratio to activate pan and scan. Pan and Scan an image, by cropping into smaller images when the aspect ratio exceeds minumum allowed ratio.
( images: typing.Union[ForwardRef('PIL.Image.Image'), numpy.ndarray, ForwardRef('torch.Tensor'), typing.List[ForwardRef('PIL.Image.Image')], typing.List[numpy.ndarray], typing.List[ForwardRef('torch.Tensor')]] **kwargs: typing_extensions.Unpack[transformers.models.gemma3.image_processing_gemma3_fast.Gemma3FastImageProcessorKwargs] )
Parameters
ImageInput
) —
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
. bool
, optional, defaults to self.do_resize
) —
Whether to resize the image. Dict[str, int]
, optional, defaults to self.size
) —
Describes the maximum input dimensions to the model. PILImageResampling
or InterpolationMode
, optional, defaults to self.resample
) —
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
. bool
, optional, defaults to self.do_center_crop
) —
Whether to center crop the image. Dict[str, int]
, optional, defaults to self.crop_size
) —
Size of the output image after applying center_crop
. bool
, optional, defaults to self.do_rescale
) —
Whether to rescale the image. float
, optional, defaults to self.rescale_factor
) —
Rescale factor to rescale the image by if do_rescale
is set to True
. bool
, optional, defaults to self.do_normalize
) —
Whether to normalize the image. float
or List[float]
, optional, defaults to self.image_mean
) —
Image mean to use for normalization. Only has an effect if do_normalize
is set to True
. float
or List[float]
, optional, defaults to self.image_std
) —
Image standard deviation to use for normalization. Only has an effect if do_normalize
is set to
True
. bool
, optional, defaults to self.do_convert_rgb
) —
Whether to convert the image to RGB. str
or TensorType
, optional, defaults to self.return_tensors
) —
Returns stacked tensors if set to `pt, otherwise returns a list of tensors. ChannelDimension
or str
, optional, defaults to self.data_format
) —
Only ChannelDimension.FIRST
is supported. Added for compatibility with slow processors. ChannelDimension
or str
, optional, defaults to self.input_data_format
) —
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.torch.device
, optional, defaults to self.device
) —
The device to process the images on. If unset, the device is inferred from the input images.
do_pan_and_scan (bool
, optional):
Whether to apply pan_and_scan
to images.
pan_and_scan_min_crop_size (int
, optional):
Minimum size of each crop in pan and scan.
pan_and_scan_max_num_crops (int
, optional):
Maximum number of crops per image in pan and scan.
pan_and_scan_min_ratio_to_activate (float
, optional):
Minimum aspect ratio to activate pan and scan. Preprocess an image or batch of images.
( image_processor tokenizer chat_template = None image_seq_length: int = 256 **kwargs )
This method forwards all its arguments to GemmaTokenizerFast’s batch_decode(). Please refer to the docstring of this method for more information.
This method forwards all its arguments to GemmaTokenizerFast’s decode(). Please refer to the docstring of this method for more information.
( vocab_size = 262208 hidden_size = 2304 intermediate_size = 9216 num_hidden_layers = 26 num_attention_heads = 8 num_key_value_heads = 4 head_dim = 256 hidden_activation = 'gelu_pytorch_tanh' max_position_embeddings = 131072 initializer_range = 0.02 rms_norm_eps = 1e-06 use_cache = True pad_token_id = 0 eos_token_id = 1 bos_token_id = 2 tie_word_embeddings = True rope_theta = 1000000.0 attention_bias = False attention_dropout = 0.0 query_pre_attn_scalar = 256 sliding_window = 4096 final_logit_softcapping = None attn_logit_softcapping = None cache_implementation = 'hybrid' rope_scaling = None rope_local_base_freq = 10000.0 sliding_window_pattern = 6 **kwargs )
Parameters
int
, optional, defaults to 262208) —
Vocabulary size of the Gemma3Text model. Defines the number of different tokens that can be represented by the
inputs_ids
passed when calling Gemma3TextModel int
, optional, defaults to 2304) —
Dimension of the hidden representations. int
, optional, defaults to 9216) —
Dimension of the MLP representations. int
, optional, defaults to 26) —
Number of hidden layers in the Transformer decoder. int
, optional, defaults to 8) —
Number of attention heads for each attention layer in the Transformer decoder. int
, optional, defaults to 4) —
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 checkout this
paper. If it is not specified, will default to
num_attention_heads
. int
, optional, defaults to 256) —
The attention head dimension. str
or function
, optional, defaults to "gelu_pytorch_tanh"
) —
The non-linear activation function (function or string) in the decoder. Will default to "gelu_pytorch_tanh"
if not specified. "gelu_pytorch_tanh"
uses an approximation of the "gelu"
activation function. int
, optional, defaults to 131072) —
The maximum sequence length that this model might ever be used with. float
, optional, defaults to 0.02) —
The standard deviation of the truncated_normal_initializer for initializing all weight matrices. float
, optional, defaults to 1e-06) —
The epsilon used by the rms normalization layers. 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
. int
, optional, defaults to 0) —
Padding token id. int
, optional, defaults to 1) —
End of stream token id. int
, optional, defaults to 2) —
Beginning of stream token id. bool
, optional, defaults to True
) —
Whether to tie weight embeddings float
, optional, defaults to 1000000.0) —
The base period of the RoPE embeddings. bool
, defaults to False
, optional, defaults to False
) —
Whether to use a bias in the query, key, value and output projection layers during self-attention. float
, optional, defaults to 0.0) —
The dropout ratio for the attention probabilities. float
, optional, defaults to 256) —
Scaling factor used on the attention scores int
, optional, defaults to 4096) — in Gemma3Text, every other layer uses sliding window attention. This is the
size of the sliding window. float
, optional) —
Scaling factor when applying tanh softcapping on the logits. float
, optional) —
Scaling factor when applying tanh softcapping on the attention scores. str
, optional, defaults to "hybrid"
) — the cache type to be used with generate
. Dict
, optional) —
Dictionary containing the scaling configuration for the RoPE embeddings used in gloabl attention. NOTE: if you apply new rope type
and you expect the model to work on longer max_position_embeddings
, we recommend you to update this value
accordingly.
Expected contents:
rope_type
(str
):
The sub-variant of RoPE to use. Can be one of [‘default’, ‘linear’, ‘dynamic’, ‘yarn’, ‘longrope’,
‘llama3’], with ‘default’ being the original RoPE implementation.
factor
(float
, optional):
Used with all rope types except ‘default’. The scaling factor to apply to the RoPE embeddings. In
most scaling types, a factor
of x will enable the model to handle sequences of length x
original maximum pre-trained length.
original_max_position_embeddings
(int
, optional):
Used with ‘dynamic’, ‘longrope’ and ‘llama3’. The original max position embeddings used during
pretraining.
attention_factor
(float
, optional):
Used with ‘yarn’ and ‘longrope’. The scaling factor to be applied on the attention
computation. If unspecified, it defaults to value recommended by the implementation, using the
factor
field to infer the suggested value.
beta_fast
(float
, optional):
Only used with ‘yarn’. Parameter to set the boundary for extrapolation (only) in the linear
ramp function. If unspecified, it defaults to 32.
beta_slow
(float
, optional):
Only used with ‘yarn’. Parameter to set the boundary for interpolation (only) in the linear
ramp function. If unspecified, it defaults to 1.
short_factor
(List[float]
, optional):
Only used with ‘longrope’. The scaling factor to be applied to short contexts (<
original_max_position_embeddings
). Must be a list of numbers with the same length as the hidden
size divided by the number of attention heads divided by 2
long_factor
(List[float]
, optional):
Only used with ‘longrope’. The scaling factor to be applied to long contexts (<
original_max_position_embeddings
). Must be a list of numbers with the same length as the hidden
size divided by the number of attention heads divided by 2
low_freq_factor
(float
, optional):
Only used with ‘llama3’. Scaling factor applied to low frequency components of the RoPE
high_freq_factor
(float
, optional*):
Only used with ‘llama3’. Scaling factor applied to high frequency components of the RoPE int
, optional, defaults to 6) —
Pattern for the sliding window attention. This is the configuration class to store the configuration of a Gemma3TextModel. It is used to instantiate an Gemma3Text 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 Gemma3Text-7B. e.g. google/gemma3_text-7b Configuration objects inherit from PretrainedConfig and can be used to control the model outputs. Read the documentation from PretrainedConfig for more information.
>>> from transformers import Gemma3TextModel, Gemma3TextConfig
>>> # Initializing a Gemma3Text gemma3_text-7b style configuration
>>> configuration = Gemma3TextConfig()
>>> # Initializing a model from the gemma3_text-7b style configuration
>>> model = Gemma3TextModel(configuration)
>>> # Accessing the model configuration
>>> configuration = model.config
( text_config: typing.Optional[transformers.models.gemma3.configuration_gemma3.Gemma3TextConfig] = None vision_config: typing.Optional[transformers.models.siglip.configuration_siglip.SiglipVisionConfig] = None mm_tokens_per_image: int = 256 boi_token_index: int = 255999 eoi_token_index: int = 256000 image_token_index: int = 262144 initializer_range: float = 0.02 **kwargs )
Parameters
Union[Gemma3TextConfig, dict]
, optional) —
The config object of the text backbone. Union[AutoConfig, dict]
, optional) —
Custom vision config or dict. int
, optional, defaults to 256) —
The number of tokens per image embedding. int
, optional, defaults to 255999) —
The begin-of-image token index to wrap the image prompt. int
, optional, defaults to 256000) —
The end-of-image token index to wrap the image prompt. int
, optional, defaults to 262144) —
The image token index to encode the image prompt. 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 Gemma3ForConditionalGeneration. It is used to instantiate an Gemma3ForConditionalGeneration according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of the PaliGemma-2B.
e.g. google/gemma-3-4b
Configuration objects inherit from PretrainedConfig and can be used to control the model outputs. Read the documentation from PretrainedConfig for more information.
Example:
>>> from transformers import Gemma3ForConditionalGeneration, Gemma3Config, SiglipVisionConfig, Gemma3TextConfig
>>> # Initializing a Siglip-like vision config
>>> vision_config = SiglipVisionConfig()
>>> # Initializing a Gemma3 Text config
>>> text_config = Gemma3TextConfig()
>>> # Initializing a Gemma3 gemma-3-4b style configuration
>>> configuration = Gemma3Config(vision_config, text_config)
>>> # Initializing a model from the gemma-3-4b style configuration
>>> model = Gemma3TextConfig(configuration)
>>> # Accessing the model configuration
>>> configuration = model.config
( config: Gemma3TextConfig )
Parameters
The bare Gemma3Text 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.
Transformer decoder consisting of config.num_hidden_layers layers. Each layer is a Gemma3TextDecoderLayer
( input_ids: typing.Optional[torch.LongTensor] = None attention_mask: typing.Optional[torch.Tensor] = None position_ids: typing.Optional[torch.LongTensor] = None past_key_values: typing.Optional[transformers.cache_utils.HybridCache] = None inputs_embeds: typing.Optional[torch.FloatTensor] = None use_cache: typing.Optional[bool] = None output_attentions: typing.Optional[bool] = None output_hidden_states: typing.Optional[bool] = None return_dict: typing.Optional[bool] = None cache_position: typing.Optional[torch.LongTensor] = None last_cache_position: typing.Optional[int] = None **flash_attn_kwargs: typing_extensions.Unpack[transformers.modeling_flash_attention_utils.FlashAttentionKwargs] )
Parameters
torch.LongTensor
of shape (batch_size, sequence_length)
) —
Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
it.
Indices can be obtained using AutoTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.
torch.Tensor
of shape (batch_size, sequence_length)
, optional) —
Mask to avoid performing attention on padding token indices. Mask values selected in [0, 1]
:
Indices can be obtained using AutoTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.
If past_key_values
is used, optionally only the last input_ids
have to be input (see
past_key_values
).
If you want to change padding behavior, you should read modeling_opt._prepare_decoder_attention_mask
and modify to your needs. See diagram 1 in the paper for more
information on the default strategy.
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]
.
Cache
or tuple(tuple(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
.
Two formats are allowed:
tuple(torch.FloatTensor)
of length config.n_layers
, with each tuple having 2 tensors of
shape (batch_size, num_heads, sequence_length, embed_size_per_head)
). This is also known as the legacy
cache format.The model will output the same cache format that is fed as input. If no past_key_values
are passed, the
legacy cache format will be returned.
If past_key_values
are used, the user can optionally input only the last input_ids
(those that don’t
have their past key value states given to this model) of shape (batch_size, 1)
instead of all input_ids
of shape (batch_size, sequence_length)
.
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. 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
). bool
, optional) —
Whether or not to return the attentions tensors of all attention layers. See attentions
under returned
tensors for more detail. bool
, optional) —
Whether or not to return the hidden states of all layers. See hidden_states
under returned tensors for
more detail. bool
, optional) —
Whether or not to return a ModelOutput instead of a plain tuple. torch.LongTensor
of shape (sequence_length)
, optional) —
Indices depicting the position of the input sequence tokens in the sequence. Contrarily to position_ids
,
this tensor is not affected by padding. It is used to update the cache in the correct position and to infer
the complete sequence length. The Gemma3TextModel 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.
( input_ids: LongTensor = None attention_mask: typing.Optional[torch.Tensor] = None position_ids: typing.Optional[torch.LongTensor] = None past_key_values: typing.Optional[transformers.cache_utils.HybridCache] = None inputs_embeds: typing.Optional[torch.FloatTensor] = None labels: typing.Optional[torch.LongTensor] = None use_cache: typing.Optional[bool] = None output_attentions: typing.Optional[bool] = None output_hidden_states: typing.Optional[bool] = None return_dict: typing.Optional[bool] = None cache_position: typing.Optional[torch.LongTensor] = None logits_to_keep: typing.Union[int, torch.Tensor] = 0 **loss_kwargs ) → transformers.modeling_outputs.CausalLMOutputWithPast or tuple(torch.FloatTensor)
Parameters
torch.LongTensor
of shape (batch_size, sequence_length)
) —
Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
it.
Indices can be obtained using AutoTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.
torch.Tensor
of shape (batch_size, sequence_length)
, optional) —
Mask to avoid performing attention on padding token indices. Mask values selected in [0, 1]
:
Indices can be obtained using AutoTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.
If past_key_values
is used, optionally only the last input_ids
have to be input (see
past_key_values
).
If you want to change padding behavior, you should read modeling_opt._prepare_decoder_attention_mask
and modify to your needs. See diagram 1 in the paper for more
information on the default strategy.
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]
.
Cache
or tuple(tuple(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
.
Two formats are allowed:
tuple(torch.FloatTensor)
of length config.n_layers
, with each tuple having 2 tensors of
shape (batch_size, num_heads, sequence_length, embed_size_per_head)
). This is also known as the legacy
cache format.The model will output the same cache format that is fed as input. If no past_key_values
are passed, the
legacy cache format will be returned.
If past_key_values
are used, the user can optionally input only the last input_ids
(those that don’t
have their past key value states given to this model) of shape (batch_size, 1)
instead of all input_ids
of shape (batch_size, sequence_length)
.
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. 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
). bool
, optional) —
Whether or not to return the attentions tensors of all attention layers. See attentions
under returned
tensors for more detail. bool
, optional) —
Whether or not to return the hidden states of all layers. See hidden_states
under returned tensors for
more detail. bool
, optional) —
Whether or not to return a ModelOutput instead of a plain tuple. torch.LongTensor
of shape (sequence_length)
, optional) —
Indices depicting the position of the input sequence tokens in the sequence. Contrarily to position_ids
,
this tensor is not affected by padding. It is used to update the cache in the correct position and to infer
the complete sequence length. 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]
. int
or torch.Tensor
, optional) —
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). Returns
transformers.modeling_outputs.CausalLMOutputWithPast or tuple(torch.FloatTensor)
A transformers.modeling_outputs.CausalLMOutputWithPast 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 (Gemma3Config) and inputs.
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 (tuple(tuple(torch.FloatTensor))
, optional, returned when use_cache=True
is passed or when config.use_cache=True
) — Tuple of tuple(torch.FloatTensor)
of length config.n_layers
, with each tuple having 2 tensors of shape
(batch_size, num_heads, sequence_length, embed_size_per_head)
)
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.
The Gemma3ForCausalLM 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.
Example:
>>> from transformers import AutoTokenizer, Gemma3ForCausalLM
>>> model = Gemma3ForCausalLM.from_pretrained("google/gemma-2-9b")
>>> tokenizer = AutoTokenizer.from_pretrained("google/gemma-2-9b")
>>> prompt = "What is your favorite condiment?"
>>> inputs = tokenizer(prompt, return_tensors="pt")
>>> # Generate
>>> generate_ids = model.generate(inputs.input_ids, max_length=30)
>>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
"What is your favorite condiment?"
( config: Gemma3Config )
Parameters
The GEMMA3 model which consists of a vision backbone 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.
( input_ids: LongTensor = None pixel_values: FloatTensor = None attention_mask: typing.Optional[torch.Tensor] = None position_ids: typing.Optional[torch.LongTensor] = None past_key_values: typing.Union[typing.List[torch.FloatTensor], transformers.cache_utils.Cache, NoneType] = None token_type_ids: typing.Optional[torch.LongTensor] = None cache_position: typing.Optional[torch.LongTensor] = None inputs_embeds: typing.Optional[torch.FloatTensor] = None labels: typing.Optional[torch.LongTensor] = None use_cache: typing.Optional[bool] = None output_attentions: typing.Optional[bool] = None output_hidden_states: typing.Optional[bool] = None return_dict: typing.Optional[bool] = None logits_to_keep: typing.Union[int, torch.Tensor] = 0 **lm_kwargs ) → transformers.models.gemma3.modeling_gemma3.Gemma3CausalLMOutputWithPast
or tuple(torch.FloatTensor)
Parameters
torch.LongTensor
of shape (batch_size, sequence_length)
) —
Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
it.
Indices can be obtained using AutoTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.
torch.Tensor
of shape (batch_size, sequence_length)
, optional) —
Mask to avoid performing attention on padding token indices. Mask values selected in [0, 1]
:
Indices can be obtained using AutoTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.
If past_key_values
is used, optionally only the last input_ids
have to be input (see
past_key_values
).
If you want to change padding behavior, you should read modeling_opt._prepare_decoder_attention_mask
and modify to your needs. See diagram 1 in the paper for more
information on the default strategy.
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]
.
Cache
or tuple(tuple(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
.
Two formats are allowed:
tuple(torch.FloatTensor)
of length config.n_layers
, with each tuple having 2 tensors of
shape (batch_size, num_heads, sequence_length, embed_size_per_head)
). This is also known as the legacy
cache format.The model will output the same cache format that is fed as input. If no past_key_values
are passed, the
legacy cache format will be returned.
If past_key_values
are used, the user can optionally input only the last input_ids
(those that don’t
have their past key value states given to this model) of shape (batch_size, 1)
instead of all input_ids
of shape (batch_size, sequence_length)
.
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. 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
). bool
, optional) —
Whether or not to return the attentions tensors of all attention layers. See attentions
under returned
tensors for more detail. bool
, optional) —
Whether or not to return the hidden states of all layers. See hidden_states
under returned tensors for
more detail. bool
, optional) —
Whether or not to return a ModelOutput instead of a plain tuple. torch.LongTensor
of shape (sequence_length)
, optional) —
Indices depicting the position of the input sequence tokens in the sequence. Contrarily to position_ids
,
this tensor is not affected by padding. It is used to update the cache in the correct position and to infer
the complete sequence length. torch.LongTensor
of shape (batch_size, sequence_length)
, optional) —
Labels for computing the masked language modeling loss. Indices should either be in [0, ..., config.text_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.text_config.vocab_size]
. int
or torch.Tensor
, optional) —
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). Returns
transformers.models.gemma3.modeling_gemma3.Gemma3CausalLMOutputWithPast
or tuple(torch.FloatTensor)
A transformers.models.gemma3.modeling_gemma3.Gemma3CausalLMOutputWithPast
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 (Gemma3Config) and inputs.
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.text_config.vocab_size)
) — Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).
past_key_values (tuple(tuple(torch.FloatTensor))
, optional, returned when use_cache=True
is passed or when config.use_cache=True
) — Tuple of tuple(torch.FloatTensor)
of length config.n_layers
, with each tuple having 2 tensors of shape
(batch_size, num_heads, sequence_length, embed_size_per_head)
)
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, sequence_length, hidden_size)
.
image_hidden_states of the model produced by the vision encoder after projecting last hidden state.
The Gemma3ForConditionalGeneration 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.
Example:
>>> from PIL import Image
>>> import requests
>>> from transformers import AutoProcessor, Gemma3ForConditionalGeneration
>>> model = Gemma3ForConditionalGeneration.from_pretrained("google/Gemma3-test-224px-hf")
>>> processor = AutoProcessor.from_pretrained("google/Gemma3-test-224px-hf")
>>> prompt = "answer en Where is the cow standing?"
>>> url = "https://huggingface.co/gv-hf/Gemma3-test-224px-hf/resolve/main/cow_beach_1.png"
>>> image = Image.open(requests.get(url, stream=True).raw)
>>> inputs = processor(images=image, text=prompt, return_tensors="pt")
>>> # Generate
>>> generate_ids = model.generate(**inputs, max_length=30)
>>> processor.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
"answer en Where is the cow standing?\nbeach"