Skip to content

RoBERTeye

Roberta dataclass

Bases: RoberteyeArgs

Roberta Model (no eye data)

Source code in src/configs/models/dl/RoBERTeye.py 
59
60
61
62
63
64
65
66
67
@register_model_config
@dataclass
class Roberta(RoberteyeArgs):
    """
    Roberta Model (no eye data)
    """

    prepend_eye_features_to_text: bool = False
    use_fixation_report: bool = False

RoberteyeArgs dataclass

Bases: DLModelArgs

prepend_eye_features_to_text (bool): A flag indicating whether to prepend the eye data to the input. If True, the eye data will be added at the beginning of the input; otherwise, eye data is not used.

Source code in src/configs/models/dl/RoBERTeye.py 
 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
@dataclass
class RoberteyeArgs(DLModelArgs):
    """
    prepend_eye_features_to_text (bool): A flag indicating whether to prepend the eye data to the input.
        If True, the eye data will be added at the beginning of the input; otherwise, eye data is not used.
    """

    base_model_name: DLModelNames = DLModelNames.ROBERTEYE_MODEL

    prepend_eye_features_to_text: bool = True
    batch_size: int = 4
    accumulate_grad_batches: int = 16 // batch_size
    backbone: BackboneNames = BackboneNames.ROBERTA_LARGE
    freeze: bool = False
    eye_projection_dropout: float = 0.3
    max_epochs: int = 10
    early_stopping_patience: int = 3
    warmup_proportion: float = 0.1
    eye_projection_MAGModule: bool = False

    token_type_num: int = 2
    vocab_size: int = -1  # specified in instantiate_config
    #! Don't change the following
    n_tokens: int = 0
    eye_token_id: int = 0
    sep_token_id: int = 0
    is_training: bool = False

RoberteyeFixation dataclass

Bases: RoberteyeArgs

Fixation-level RoBERTeye

Source code in src/configs/models/dl/RoBERTeye.py 
37
38
39
40
41
42
43
44
45
46
@register_model_config
@dataclass
class RoberteyeFixation(RoberteyeArgs):
    """
    Fixation-level RoBERTeye
    """

    use_fixation_report: bool = True
    batch_size: int = 2
    accumulate_grad_batches: int = 16 // batch_size

RoberteyeWord dataclass

Bases: RoberteyeArgs

Word-level RoBERTeye

Source code in src/configs/models/dl/RoBERTeye.py 
49
50
51
52
53
54
55
56
@register_model_config
@dataclass
class RoberteyeWord(RoberteyeArgs):
    """
    Word-level RoBERTeye
    """

    use_fixation_report: bool = False