콘텐츠로 이동

ColQwen2

ColQwen2 모델은 새로운 아키텍처와 학습 전략을 사용해 문서를 시각적 특징으로 효율적으로 색인하는 시각 언어 모델(Vision Language Model, VLM)입니다. RBLN Optimum을 사용하면 RBLN NPU에서 ColQwen2 모델 추론을 가속할 수 있습니다.

API 참조

RBLNColQwen2ForRetrieval

Bases: RBLNModel

RBLNColQwen2ForRetrieval is a model for document retrieval using vision-language models. This model inherits from [RBLNModel]. Check the superclass documentation for the generic methods the library implements for all its models.

A class to convert and run pre-trained transformers based ColQwen2ForRetrieval model on RBLN devices. It implements the methods to convert a pre-trained transformers ColQwen2ForRetrieval model into a RBLN transformer model by:

  • transferring the checkpoint weights of the original into an optimized RBLN graph,
  • compiling the resulting graph using the RBLN compiler.

Examples:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
import torch
from PIL import Image
from transformers import ColQwen2Processor

from optimum.rbln import RBLNColQwen2ForRetrieval

rbln_config = {
    "vlm": {
        "visual": {
            "max_seq_len": 6400,
        },
        "num_devices": 4,
        "attn_impl": "flash_attn",
        "max_seq_len": 32_768,
    },
}
model = RBLNColQwen2ForRetrieval.from_pretrained("vidore/colqwen2-v1.0-hf", export=True, rbln_config=rbln_config)
model.save_pretrained("compiled-colqwen2-v1.0-hf")

# The document page screenshots from your corpus. Below are dummy images.
images = [
    Image.new("RGB", (128, 128), color="white"),
    Image.new("RGB", (64, 32), color="black"),
]
processor = ColQwen2Processor.from_pretrained("vidore/colqwen2-v1.0-hf")

queries = [
    "When was the United States Declaration of Independence proclaimed?",
    "Who printed the edition of Romeo and Juliet?",
]
inputs_images = processor(images=images)
inputs_text = processor(text=queries)

# Forward pass
with torch.no_grad():
    image_embeddings = model(**inputs_images).embeddings
    query_embeddings = model(**inputs_text).embeddings

scores = processor.score_retrieval(query_embeddings, image_embeddings)
print("Retrieval scores (query x image):")
print(scores)

forward(input_ids=None, inputs_embeds=None, attention_mask=None, pixel_values=None, image_grid_thw=None, output_hidden_states=None, return_dict=None, **kwargs)

Runs a ColQwen2 retrieval forward pass on text tokens and optional visual inputs.

Parameters:

Name Type Description Default
input_ids LongTensor

Indices of the textual tokens. Mutually exclusive with inputs_embeds.

 None
inputs_embeds FloatTensor

Pre-computed embeddings fed directly into the language model.

 None
attention_mask Tensor

Mask that selects which token positions contribute to the loss/embeddings.

 None
pixel_values Tensor

Flattened image patches produced by ColQwen2Processor for document pages.

 None
image_grid_thw LongTensor

Per-image (t, h, w) grid metadata that allows unpadding of pixel_values.

 None
output_hidden_states bool

If True, expose intermediate decoder hidden states.

 None
return_dict bool

If True, return a ColQwen2ForRetrievalOutput; otherwise return a tuple.

 None
**kwargs dict[str, Any]

Extra multimodal args forwarded to the wrapped VLM (e.g. pixel_values_videos, video_grid_thw, second_per_grid_ts).

 {}

Returns:

Type Description
tuple | ColQwen2ForRetrievalOutput

Dataclass containing the embeddings and hidden states of the VLM model.

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 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 bool | None

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 dict | RBLNModelConfig | None

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.

get_hf_class() classmethod

Lazily load and cache the HuggingFace class this model wraps.

The class is looked up in hf_library_name under the name without the RBLN prefix (e.g. RBLNLlamaForCausalLM -> LlamaForCausalLM).

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 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 | ModelMixin

The PyTorch model to be compiled. The object must be a HuggingFace transformers PreTrainedModel or a Diffusers component based on diffusers.ModelMixin, such as a VAE, UNet or ControlNet.

 required
config PretrainedConfig | None

The configuration object associated with the model.

 None
rbln_config RBLNModelConfig | dict | None

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.

RBLNColQwen2ForRetrievalConfig

Bases: RBLNDecoderOnlyModelConfig

Configuration class for RBLN ColQwen2 models for document retrieval.

This class extends RBLNModelConfig with specific configurations for ColQwen2 models, including vision tower settings and multi-sequence length support.

Example usage 
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
from optimum.rbln import RBLNColQwen2ForRetrieval, RBLNColQwen2ForRetrievalConfig

# Create a configuration object
config = RBLNColQwen2ForRetrievalConfig(
    vlm = {
        "visual": {
            "max_seq_len": 6400,
            "device": 0,
        },
        "max_seq_len": 32_768,
        "attn_impl": "flash_attn",
        "num_devices": 4,
        "device": [0, 1, 2, 3],
        "output_hidden_states": False,
        }
)

# Use the configuration with from_pretrained
model = RBLNColQwen2ForRetrieval.from_pretrained(
    "vidore/colqwen2-v1.0-hf",
    export=True,
    rbln_config=config
)

is_auto_num_blocks property

Returns True if kvcache_num_blocks will be automatically determined during compilation to fit within the available DRAM on the NPU.

__init__(batch_size=None, output_hidden_states=None, vlm=None, **kwargs)

Parameters:

Name Type Description Default
batch_size int | None

The batch size for the model.

 None
output_hidden_states bool | None

Whether to output the hidden states of the VLM model.

 None
vlm RBLNModelConfig | None

Configuration for the VLM component.

 None
kwargs Any

Additional arguments passed to the parent RBLNModelConfig.

 {}

Raises: ValueError: If batch_size is not a positive integer.

initialize_submodule_config(submodule_config=None, force_kwargs=False, **kwargs)

Build the config of a submodule from the parent's runtime options and kwargs.

Returns the submodule config instance when its class is known (an instance was passed, or cls_name was given), otherwise the merged kwargs dict for the model class to resolve later.

get_load_overrides()

Extract the options that must cross the load boundary, for passing to from_pretrained.

Returns the explicitly-set runtime options (device, device_map, create_runtimes, activate_profiler, timeout) and the subclass-specific non-save attributes (load-behavior flags such as _load_visual_runtime, which are never serialized so they can only travel with the caller), recursively including submodules. Compile-time attributes are not included: when loading a compiled model they come from the artifact's rbln_config.json, and this dict is merged on top of it.

Returns:

Type Description
dict[str, Any]

Dictionary of load overrides, keyed like the rbln_config dict accepted by from_pretrained.

from_pretrained(path, rbln_config=None, return_unused_kwargs=False, **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
rbln_config dict[str, Any] | None

Additional configuration to override.

 None
return_unused_kwargs bool

Whether to return unused kwargs.

 False
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 Self | tuple[Self, dict[str, Any]]

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.

Examples:

1
config = RBLNResNetForImageClassificationConfig.from_pretrained("/path/to/model")