Skip to content

unitorch.cli.models.roberta¤

RobertaProcessor¤

Tip

core/process/roberta is the section for configuration of RobertaProcessor.

Bases: RobertaProcessor

Processor for the Roberta model.

Source code in src/unitorch/cli/models/roberta/processing.py 
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
def __init__(
    self,
    vocab_path,
    merge_path,
    max_seq_length: Optional[int] = 128,
    source_type_id: Optional[int] = 0,
    target_type_id: Optional[int] = 0,
):
    super().__init__(
        vocab_path=vocab_path,
        merge_path=merge_path,
        max_seq_length=max_seq_length,
        source_type_id=source_type_id,
        target_type_id=target_type_id,
    )

from_config classmethod ¤

from_config(config, **kwargs)
Source code in src/unitorch/cli/models/roberta/processing.py 
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
@classmethod
@config_defaults_init("core/process/roberta")
def from_config(cls, config, **kwargs):
    config.set_default_section("core/process/roberta")
    pretrained_name = config.getoption("pretrained_name", "roberta-base")
    vocab_path = config.getoption("vocab_path", None)
    vocab_path = pop_value(
        vocab_path,
        nested_dict_value(pretrained_roberta_infos, pretrained_name, "vocab"),
    )
    vocab_path = cached_path(vocab_path)

    merge_path = config.getoption("merge_path", None)
    merge_path = pop_value(
        merge_path,
        nested_dict_value(pretrained_roberta_infos, pretrained_name, "merge"),
    )
    merge_path = cached_path(merge_path)

    return {
        "vocab_path": vocab_path,
        "merge_path": merge_path,
    }

_classification ¤

_classification(
    text: str,
    text_pair: Optional[str] = None,
    max_seq_length: Optional[int] = None,
)
Source code in src/unitorch/cli/models/roberta/processing.py 
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
@register_process("core/process/roberta/classification")
def _classification(
    self,
    text: str,
    text_pair: Optional[str] = None,
    max_seq_length: Optional[int] = None,
):
    outputs = super().classification(
        text=text,
        text_pair=text_pair,
        max_seq_length=max_seq_length,
    )
    return TensorInputs(
        input_ids=outputs.input_ids,
        attention_mask=outputs.attention_mask,
        token_type_ids=outputs.token_type_ids,
        position_ids=outputs.position_ids,
    )

RobertaForClassification¤

Tip

core/model/classification/roberta is the section for configuration of RobertaForClassification.

Bases: RobertaForClassification

Roberta model for classification tasks.

Source code in src/unitorch/cli/models/roberta/modeling.py 
24
25
26
27
28
29
30
31
32
33
34
def __init__(
    self,
    config_path: str,
    num_classes: Optional[int] = 1,
    gradient_checkpointing: Optional[bool] = False,
):
    super().__init__(
        config_path=config_path,
        num_classes=num_classes,
        gradient_checkpointing=gradient_checkpointing,
    )

from_config classmethod ¤

from_config(config, **kwargs)
Source code in src/unitorch/cli/models/roberta/modeling.py 
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
@classmethod
@config_defaults_init("core/model/classification/roberta")
def from_config(cls, config, **kwargs):
    config.set_default_section("core/model/classification/roberta")
    pretrained_name = config.getoption("pretrained_name", "roberta-base")
    config_path = config.getoption("config_path", None)
    config_path = pop_value(
        config_path,
        nested_dict_value(pretrained_roberta_infos, pretrained_name, "config"),
    )

    config_path = cached_path(config_path)
    gradient_checkpointing = config.getoption("gradient_checkpointing", False)
    num_classes = config.getoption("num_classes", 1)

    inst = cls(config_path, num_classes, gradient_checkpointing)
    pretrained_weight_path = config.getoption("pretrained_weight_path", None)
    weight_path = pop_value(
        pretrained_weight_path,
        nested_dict_value(pretrained_roberta_infos, pretrained_name, "weight"),
        check_none=False,
    )
    if weight_path is not None:
        inst.from_pretrained(weight_path)

    return inst

forward ¤

forward(
    input_ids: Tensor,
    attention_mask: Optional[Tensor] = None,
    token_type_ids: Optional[Tensor] = None,
    position_ids: Optional[Tensor] = None,
)
Source code in src/unitorch/cli/models/roberta/modeling.py 
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
@autocast(device_type=("cuda" if torch.cuda.is_available() else "cpu"))
def forward(
    self,
    input_ids: torch.Tensor,
    attention_mask: Optional[torch.Tensor] = None,
    token_type_ids: Optional[torch.Tensor] = None,
    position_ids: Optional[torch.Tensor] = None,
):
    outputs = super().forward(
        input_ids=input_ids,
        attention_mask=attention_mask,
        token_type_ids=token_type_ids,
        position_ids=position_ids,
    )
    return ClassificationOutputs(outputs=outputs)