Skip to content

vllm.transformers_utils.configs.nemotron_vl

Nemotron_Nano_VL_Config

Bases: PretrainedConfig

Source code in vllm/transformers_utils/configs/nemotron_vl.py 
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
class Nemotron_Nano_VL_Config(PretrainedConfig):
    model_type = 'Llama_Nemotron_Nano_VL'
    is_composition = True

    def __init__(
        self,
        vision_config=None,
        llm_config=None,
        force_image_size=None,
        downsample_ratio=0.5,
        template=None,
        ps_version='v1',
        image_tag_type="internvl",
        projector_hidden_size=4096,
        vit_hidden_size=1280,
        **kwargs
    ):
        super().__init__(**kwargs)

        if vision_config is not None:
            assert "auto_map" in vision_config and "AutoConfig" in vision_config["auto_map"]
            vision_auto_config = get_class_from_dynamic_module(*vision_config["auto_map"]["AutoConfig"].split("--")[::-1])
            self.vision_config = vision_auto_config(**vision_config)
        else:
            self.vision_config = PretrainedConfig()

        if llm_config is None:
            self.text_config = LlamaConfig()
        else:
            self.text_config = LlamaConfig(**llm_config)

        # Assign configuration values
        self.force_image_size = force_image_size
        self.downsample_ratio = downsample_ratio
        self.template = template  # TODO move out of here and into the tokenizer
        self.ps_version = ps_version  # Pixel shuffle version
        self.image_tag_type = image_tag_type # TODO: into the tokenizer too?
        self.projector_hidden_size = projector_hidden_size
        self.vit_hidden_size = vit_hidden_size

downsample_ratio instance-attribute 

downsample_ratio = downsample_ratio

force_image_size instance-attribute 

force_image_size = force_image_size

image_tag_type instance-attribute 

image_tag_type = image_tag_type

is_composition class-attribute instance-attribute 

is_composition = True

model_type class-attribute instance-attribute 

model_type = 'Llama_Nemotron_Nano_VL'

projector_hidden_size instance-attribute 

projector_hidden_size = projector_hidden_size

ps_version instance-attribute 

ps_version = ps_version

template instance-attribute 

template = template

text_config instance-attribute 

text_config = LlamaConfig()

vision_config instance-attribute 

vision_config = vision_auto_config(**vision_config)

vit_hidden_size instance-attribute 

vit_hidden_size = vit_hidden_size

__init__ 

__init__(
    vision_config=None,
    llm_config=None,
    force_image_size=None,
    downsample_ratio=0.5,
    template=None,
    ps_version="v1",
    image_tag_type="internvl",
    projector_hidden_size=4096,
    vit_hidden_size=1280,
    **kwargs,
)
Source code in vllm/transformers_utils/configs/nemotron_vl.py 
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
def __init__(
    self,
    vision_config=None,
    llm_config=None,
    force_image_size=None,
    downsample_ratio=0.5,
    template=None,
    ps_version='v1',
    image_tag_type="internvl",
    projector_hidden_size=4096,
    vit_hidden_size=1280,
    **kwargs
):
    super().__init__(**kwargs)

    if vision_config is not None:
        assert "auto_map" in vision_config and "AutoConfig" in vision_config["auto_map"]
        vision_auto_config = get_class_from_dynamic_module(*vision_config["auto_map"]["AutoConfig"].split("--")[::-1])
        self.vision_config = vision_auto_config(**vision_config)
    else:
        self.vision_config = PretrainedConfig()

    if llm_config is None:
        self.text_config = LlamaConfig()
    else:
        self.text_config = LlamaConfig(**llm_config)

    # Assign configuration values
    self.force_image_size = force_image_size
    self.downsample_ratio = downsample_ratio
    self.template = template  # TODO move out of here and into the tokenizer
    self.ps_version = ps_version  # Pixel shuffle version
    self.image_tag_type = image_tag_type # TODO: into the tokenizer too?
    self.projector_hidden_size = projector_hidden_size
    self.vit_hidden_size = vit_hidden_size