Skip to content

vllm.transformers_utils.configs.moonvit

MoonViTConfig

Bases: PretrainedConfig

Source code in vllm/transformers_utils/configs/moonvit.py 
 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
class MoonViTConfig(PretrainedConfig):
    model_type = "moonvit"

    def __init__(
            self,
            patch_size: int = 14,
            init_pos_emb_height: int = 64,
            init_pos_emb_width: int = 64,
            num_attention_heads: int = 16,
            num_hidden_layers: int = 27,
            hidden_size: int = 1152,
            intermediate_size: int = 4304,
            merge_kernel_size: tuple[int, int] = (2, 2),
            **kwargs,
    ):
        super().__init__(**kwargs)
        self.patch_size = patch_size
        # Positional embedding config
        self.init_pos_emb_height = init_pos_emb_height
        self.init_pos_emb_width = init_pos_emb_width
        # Transformer config
        self.num_hidden_layers = num_hidden_layers
        self.num_attention_heads = num_attention_heads
        self.hidden_size = hidden_size
        self.intermediate_size = intermediate_size
        # Patch merger config
        self.merge_kernel_size = merge_kernel_size

hidden_size instance-attribute 

hidden_size = hidden_size

init_pos_emb_height instance-attribute 

init_pos_emb_height = init_pos_emb_height

init_pos_emb_width instance-attribute 

init_pos_emb_width = init_pos_emb_width

intermediate_size instance-attribute 

intermediate_size = intermediate_size

merge_kernel_size instance-attribute 

merge_kernel_size = merge_kernel_size

model_type class-attribute instance-attribute 

model_type = 'moonvit'

num_attention_heads instance-attribute 

num_attention_heads = num_attention_heads

num_hidden_layers instance-attribute 

num_hidden_layers = num_hidden_layers

patch_size instance-attribute 

patch_size = patch_size

__init__ 

__init__(
    patch_size: int = 14,
    init_pos_emb_height: int = 64,
    init_pos_emb_width: int = 64,
    num_attention_heads: int = 16,
    num_hidden_layers: int = 27,
    hidden_size: int = 1152,
    intermediate_size: int = 4304,
    merge_kernel_size: tuple[int, int] = (2, 2),
    **kwargs,
)
Source code in vllm/transformers_utils/configs/moonvit.py 
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
def __init__(
        self,
        patch_size: int = 14,
        init_pos_emb_height: int = 64,
        init_pos_emb_width: int = 64,
        num_attention_heads: int = 16,
        num_hidden_layers: int = 27,
        hidden_size: int = 1152,
        intermediate_size: int = 4304,
        merge_kernel_size: tuple[int, int] = (2, 2),
        **kwargs,
):
    super().__init__(**kwargs)
    self.patch_size = patch_size
    # Positional embedding config
    self.init_pos_emb_height = init_pos_emb_height
    self.init_pos_emb_width = init_pos_emb_width
    # Transformer config
    self.num_hidden_layers = num_hidden_layers
    self.num_attention_heads = num_attention_heads
    self.hidden_size = hidden_size
    self.intermediate_size = intermediate_size
    # Patch merger config
    self.merge_kernel_size = merge_kernel_size