Qwen2.5-VL

Qwen2.5-VL 모델은 시각적 질의응답(VQA), 이미지 캡셔닝, 비디오 이해와 같은 작업을 위해 설계된 비전-언어 모델입니다. 이미지와 비디오 입력을 텍스트와 함께 처리할 수 있어 멀티모달 애플리케이션에 매우 유연하게 사용됩니다. RBLN NPU는 Optimum RBLN을 사용하여 Qwen2.5-VL 모델 추론을 가속화할 수 있습니다.

API Reference

Classes

RBLNQwen2_5_VisionTransformerPretrainedModel

Bases: RBLNModel

RBLN optimized Qwen2.5-VL vision transformer model.

This class provides hardware-accelerated inference for Qwen2.5-VL vision transformers on RBLN devices, supporting image and video encoding for multimodal vision-language tasks with window-based attention mechanisms.

Functions

from_model(model, config=None, rbln_config=None, model_save_dir=None, subfolder='', **kwargs) classmethod

Converts and compiles a pre-trained HuggingFace library model into a RBLN model. This method performs the actual model conversion and compilation process.

Parameters:

Name	Type	Description	Default
model	PreTrainedModel	The PyTorch model to be compiled. The object must be an instance of the HuggingFace transformers PreTrainedModel class.	required
config	Optional[PretrainedConfig]	The configuration object associated with the model.	None
rbln_config	Optional[Union[RBLNModelConfig, Dict]]	Configuration for RBLN model compilation and runtime. This can be provided as a dictionary or an instance of the model's configuration class (e.g., RBLNLlamaForCausalLMConfig for Llama models). For detailed configuration options, see the specific model's configuration class documentation.	None
kwargs	Any	Additional keyword arguments. Arguments with the prefix rbln_ are passed to rbln_config, while the remaining arguments are passed to the HuggingFace library.	{}

The method performs the following steps:

1. Compiles the PyTorch model into an optimized RBLN graph
2. Configures the model for the specified NPU device
3. Creates the necessary runtime objects if requested
4. Saves the compiled model and configurations

Returns:

Type	Description
RBLNModel	A RBLN model instance ready for inference on RBLN NPU devices.

from_pretrained(model_id, export=None, rbln_config=None, **kwargs) classmethod

The from_pretrained() function is utilized in its standard form as in the HuggingFace transformers library. User can use this function to load a pre-trained model from the HuggingFace library and convert it to a RBLN model to be run on RBLN NPUs.

Parameters:

Name	Type	Description	Default
model_id	Union[str, Path]	The model id of the pre-trained model to be loaded. It can be downloaded from the HuggingFace model hub or a local path, or a model id of a compiled model using the RBLN Compiler.	required
export	Optional[bool]	A boolean flag to indicate whether the model should be compiled. If None, it will be determined based on the existence of the compiled model files in the model_id.	None
rbln_config	Optional[Union[Dict, RBLNModelConfig]]	Configuration for RBLN model compilation and runtime. This can be provided as a dictionary or an instance of the model's configuration class (e.g., RBLNLlamaForCausalLMConfig for Llama models). For detailed configuration options, see the specific model's configuration class documentation.	None
kwargs	Any	Additional keyword arguments. Arguments with the prefix rbln_ are passed to rbln_config, while the remaining arguments are passed to the HuggingFace library.	{}

Returns:

Type	Description
RBLNModel	A RBLN model instance ready for inference on RBLN NPU devices.

save_pretrained(save_directory, push_to_hub=False, **kwargs)

Saves a model and its configuration file to a directory, so that it can be re-loaded using the [~optimum.rbln.modeling_base.RBLNBaseModel.from_pretrained] class method.

Parameters:

Name	Type	Description	Default
save_directory	Union[str, Path]	Directory where to save the model file.	required
push_to_hub	bool	Whether or not to push your model to the HuggingFace model hub after saving it.	False

RBLNQwen2_5_VLForConditionalGeneration

Bases: RBLNDecoderOnlyModelForCausalLM

RBLNQwen2_5_VLForConditionalGeneration is a multi-modal model that integrates vision and language processing capabilities, optimized for RBLN NPUs. It is designed for conditional generation tasks that involve both image and text inputs.

This model inherits from [RBLNDecoderOnlyModelForCausalLM]. Check the superclass documentation for the generic methods the library implements for all its models.

Important Note

This model includes a Large Language Model (LLM). For optimal performance, it is highly recommended to use tensor parallelism for the language model. This can be achieved by using the rbln_config parameter in the from_pretrained method. Refer to the from_pretrained documentation and the RBLNQwen2_5_VLForConditionalGenerationConfig class for details.

Examples:

from optimum.rbln import RBLNQwen2_5_VLForConditionalGeneration

model = RBLNQwen2_5_VLForConditionalGeneration.from_pretrained(
    "Qwen/Qwen2.5-VL-7B-Instruct",
    export=True,
    rbln_config={
        "visual": {
            "max_seq_lens": 6400,
            "device": 0,
        },
        "tensor_parallel_size": 8,
        "kvcache_partition_len": 16_384,
        "max_seq_len": 114_688,
        "device": [0, 1, 2, 3, 4, 5, 6, 7],
    },
)

model.save_pretrained("compiled-qwen2.5-vl-7b-instruct")

Functions

generate(input_ids, attention_mask=None, max_length=None, **kwargs)

The generate function is utilized in its standard form as in the HuggingFace transformers library. User can use this function to generate text from the model.

Parameters:

Name	Type	Description	Default
input_ids	LongTensor	The input ids to the model.	required
attention_mask	Optional[LongTensor]	The attention mask to the model.	None
max_length	Optional[int]	The maximum length of the generated text.	None
kwargs		Additional arguments passed to the generate function. See the HuggingFace transformers documentation for more details.	{}

from_pretrained(model_id, export=None, rbln_config=None, **kwargs) classmethod

The from_pretrained() function is utilized in its standard form as in the HuggingFace transformers library. User can use this function to load a pre-trained model from the HuggingFace library and convert it to a RBLN model to be run on RBLN NPUs.

Parameters:

Name	Type	Description	Default
model_id	Union[str, Path]	The model id of the pre-trained model to be loaded. It can be downloaded from the HuggingFace model hub or a local path, or a model id of a compiled model using the RBLN Compiler.	required
export	Optional[bool]	A boolean flag to indicate whether the model should be compiled. If None, it will be determined based on the existence of the compiled model files in the model_id.	None
rbln_config	Optional[Union[Dict, RBLNModelConfig]]	Configuration for RBLN model compilation and runtime. This can be provided as a dictionary or an instance of the model's configuration class (e.g., RBLNLlamaForCausalLMConfig for Llama models). For detailed configuration options, see the specific model's configuration class documentation.	None
kwargs	Any	Additional keyword arguments. Arguments with the prefix rbln_ are passed to rbln_config, while the remaining arguments are passed to the HuggingFace library.	{}

Returns:

Type	Description
RBLNModel	A RBLN model instance ready for inference on RBLN NPU devices.

save_pretrained(save_directory, push_to_hub=False, **kwargs)

Saves a model and its configuration file to a directory, so that it can be re-loaded using the [~optimum.rbln.modeling_base.RBLNBaseModel.from_pretrained] class method.

Parameters:

Name	Type	Description	Default
save_directory	Union[str, Path]	Directory where to save the model file.	required
push_to_hub	bool	Whether or not to push your model to the HuggingFace model hub after saving it.	False

from_model(model, config=None, rbln_config=None, model_save_dir=None, subfolder='', **kwargs) classmethod

Converts and compiles a pre-trained HuggingFace library model into a RBLN model. This method performs the actual model conversion and compilation process.

Parameters:

Name	Type	Description	Default
model	PreTrainedModel	The PyTorch model to be compiled. The object must be an instance of the HuggingFace transformers PreTrainedModel class.	required
config	Optional[PretrainedConfig]	The configuration object associated with the model.	None
rbln_config	Optional[Union[RBLNModelConfig, Dict]]	Configuration for RBLN model compilation and runtime. This can be provided as a dictionary or an instance of the model's configuration class (e.g., RBLNLlamaForCausalLMConfig for Llama models). For detailed configuration options, see the specific model's configuration class documentation.	None
kwargs	Any	Additional keyword arguments. Arguments with the prefix rbln_ are passed to rbln_config, while the remaining arguments are passed to the HuggingFace library.	{}

The method performs the following steps:

1. Compiles the PyTorch model into an optimized RBLN graph
2. Configures the model for the specified NPU device
3. Creates the necessary runtime objects if requested
4. Saves the compiled model and configurations

Returns:

Type	Description
RBLNModel	A RBLN model instance ready for inference on RBLN NPU devices.

Functions

Classes

RBLNQwen2_5_VLForConditionalGenerationConfig

Bases: RBLNDecoderOnlyModelForCausalLMConfig

Configuration class for RBLNQwen2_5_VLForConditionalGeneration.

This configuration class stores the configuration parameters specific to RBLN-optimized Qwen2.5-VL models for multimodal conditional generation tasks that combine vision and language processing capabilities.

Functions

__init__(use_inputs_embeds=True, visual=None, **kwargs)

Parameters:

Name	Type	Description	Default
use_inputs_embeds	bool	Whether or not to use inputs_embeds as input. Defaults to True.	True
visual	Optional[RBLNModelConfig]	Configuration for the vision encoder component.	None
kwargs	Any	Additional arguments passed to the parent RBLNDecoderOnlyModelForCausalLMConfig.	{}

Raises:

Type	Description
ValueError	If use_inputs_embeds is False.
ValueError	If the visual configuration is provided but contains invalid settings, such as an invalid max_seq_lens (e.g., not a positive integer, not a multiple of the window-based attention unit, or insufficient for the expected resolution).
ValueError	If visual is None and no default vision configuration can be inferred for the model architecture.
ValueError	If any inherited parameters violate constraints defined in the parent class, such as batch_size not being a positive integer, prefill_chunk_size not being divisible by 64, or max_seq_len not meeting requirements for Flash Attention.

load(path, **kwargs) classmethod

Load a RBLNModelConfig from a path.

Parameters:

Name	Type	Description	Default
path	str	Path to the RBLNModelConfig file or directory containing the config file.	required
kwargs	Any	Additional keyword arguments to override configuration values. Keys starting with 'rbln_' will have the prefix removed and be used to update the configuration.	{}

Returns:

Name	Type	Description
RBLNModelConfig	RBLNModelConfig	The loaded configuration instance.

Note

This method loads the configuration from the specified path and applies any provided overrides. If the loaded configuration class doesn't match the expected class, a warning will be logged.

RBLNQwen2_5_VisionTransformerPretrainedModelConfig

Bases: RBLNModelConfig

Configuration class for RBLNQwen2_5_VisionTransformerPretrainedModel.

This configuration class stores the configuration parameters specific to RBLN-optimized Qwen2.5-VL vision transformer models with window-based attention mechanisms for processing images and videos.

Functions

__init__(max_seq_lens=None, **kwargs)

Parameters:

Name	Type	Description	Default
max_seq_lens	Optional[Union[int, List[int]]]	Maximum sequence lengths for Vision Transformer attention. Can be an integer or list of integers, each indicating the number of patches in a sequence for an image or video. For example, an image of 224x196 pixels with patch size 14 and window size 112 has its width padded to 224, forming a 224x224 image. This yields 256 patches [(224/14) * (224/14)], so max_seq_len must be at least 256. For window-based attention, max_seq_len must be a multiple of (window_size / patch_size)^2, e.g., (112/14)^2 = 64, making 256 (64 * 4) valid. RBLN optimization runs inference per image or video frame, so set max_seq_len to match the maximum expected resolution to reduce computation. If not provided, a ValueError is raised.	None
kwargs	Any	Additional arguments passed to the parent RBLNModelConfig.	{}

Raises:

Type	Description
ValueError	If max_seq_lens is None or not provided.
ValueError	If max_seq_lens (or any value in the list) is not a positive integer.
ValueError	If max_seq_lens is not a multiple of (window_size / patch_size)^2 for window-based attention, or is insufficient for the expected image/video resolution.
ValueError	If batch_size (inherited from RBLNModelConfig) is not a positive integer.

Max Seq Lens

Since Qwen2_5_VLForConditionalGeneration performs inference on a per-image or per-frame basis, max_seq_lens should be set based on the maximum expected resolution of the input images or video frames, according to the following guidelines:

1. Minimum Value: max_seq_lens must be greater than or equal to the number of patches generated from the input image. For example, a 224x224 image with a patch size of 14 results in (224 / 14) * (224 / 14) = 256 patches. Therefore, max_seq_lens must be at least 256.
2. Alignment Requirement: max_seq_lens must be a multiple of (window_size / patch_size)^2 due to the requirements of the window-based attention mechanism. For instance, if window_size is 112 and patch_size is 14, then (112 / 14)^2 = 64, meaning valid values for max_seq_lens include 64, 128, 192, 256, etc.

load(path, **kwargs) classmethod

Load a RBLNModelConfig from a path.

Parameters:

Name	Type	Description	Default
path	str	Path to the RBLNModelConfig file or directory containing the config file.	required
kwargs	Any	Additional keyword arguments to override configuration values. Keys starting with 'rbln_' will have the prefix removed and be used to update the configuration.	{}

Returns:

Name	Type	Description
RBLNModelConfig	RBLNModelConfig	The loaded configuration instance.

Note

This method loads the configuration from the specified path and applies any provided overrides. If the loaded configuration class doesn't match the expected class, a warning will be logged.