Skip to content

unitorch.models.bart¤

BartProcessor¤

Bases: HfTextGenerationProcessor

Processor for BART model. Inherits from HfTextGenerationProcessor.

Initializes the BartProcessor.

Parameters:

Name Type Description Default
vocab_path str

Path to the vocabulary file.

 required
merge_path str

Path to the BPE merge file.

 required
special_input_ids Optional[Dict]

Optional dictionary of special input IDs.

 dict()
max_seq_length Optional[int]

Maximum sequence length.

 128
max_gen_seq_length Optional[int]

Maximum generation sequence length.

 48
Source code in src/unitorch/models/bart/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
58
59
60
61
62
def __init__(
    self,
    vocab_path: str,
    merge_path: str,
    special_input_ids: Optional[Dict] = dict(),
    max_seq_length: Optional[int] = 128,
    max_gen_seq_length: Optional[int] = 48,
):
    """
    Initializes the BartProcessor.

    Args:
        vocab_path (str): Path to the vocabulary file.
        merge_path (str): Path to the BPE merge file.
        special_input_ids (Optional[Dict]): Optional dictionary of special input IDs.
        max_seq_length (Optional[int]): Maximum sequence length.
        max_gen_seq_length (Optional[int]): Maximum generation sequence length.
    """
    tokenizer = get_bart_tokenizer(
        vocab_path,
        merge_path,
        special_input_ids=special_input_ids,
    )
    super().__init__(
        tokenizer=tokenizer,
        max_seq_length=max_seq_length,
        max_gen_seq_length=max_gen_seq_length,
    )

BartForGeneration¤

Bases: GenericModel

BART model for text generation.

Initializes a BartForGeneration model.

Parameters:

Name Type Description Default
config_path str

Path to the BART model configuration file.

 required
gradient_checkpointing bool

Whether to use gradient checkpointing during training. Defaults to False.

 False
Source code in src/unitorch/models/bart/modeling.py 
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
def __init__(
    self,
    config_path: str,
    gradient_checkpointing: Optional[bool] = False,
):
    """
    Initializes a BartForGeneration model.

    Args:
        config_path (str): Path to the BART model configuration file.
        gradient_checkpointing (bool, optional): Whether to use gradient checkpointing during training. Defaults to False.
    """
    super().__init__()
    self.config = BartConfig.from_json_file(config_path)
    self.config.gradient_checkpointing = gradient_checkpointing
    self.config.forced_bos_token_id = None
    self.model = BartForConditionalGeneration(self.config)
    self.init_weights()

forward ¤

forward(
    input_ids: Tensor,
    attention_mask: Tensor,
    decoder_input_ids: Tensor,
    decoder_attention_mask: Tensor,
)

Performs forward pass of the BartForGeneration model.

Parameters:

Name Type Description Default
input_ids Tensor

Tensor of input token IDs.

 required
attention_mask Tensor

Tensor indicating which tokens should be attended to.

 required
decoder_input_ids Tensor

Tensor of decoder input token IDs.

 required
decoder_attention_mask Tensor

Tensor indicating which decoder tokens should be attended to.

 required

Returns:

Type Description
Tensor

Output logits of the model.
Source code in src/unitorch/models/bart/modeling.py 
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
def forward(
    self,
    input_ids: torch.Tensor,
    attention_mask: torch.Tensor,
    decoder_input_ids: torch.Tensor,
    decoder_attention_mask: torch.Tensor,
):
    """
    Performs forward pass of the BartForGeneration model.

    Args:
        input_ids (torch.Tensor): Tensor of input token IDs.
        attention_mask (torch.Tensor): Tensor indicating which tokens should be attended to.
        decoder_input_ids (torch.Tensor): Tensor of decoder input token IDs.
        decoder_attention_mask (torch.Tensor): Tensor indicating which decoder tokens should be attended to.

    Returns:
        (torch.Tensor): Output logits of the model.
    """
    outputs = self.model(
        input_ids=input_ids,
        attention_mask=attention_mask,
        decoder_input_ids=decoder_input_ids,
        decoder_attention_mask=decoder_attention_mask,
        return_dict=True,
    )
    logits = outputs.logits
    return logits

generate ¤

generate(
    input_ids: Tensor,
    num_beams: Optional[int] = 5,
    decoder_start_token_id: Optional[int] = 2,
    decoder_end_token_id: Optional[
        Union[int, List[int]]
    ] = 2,
    num_return_sequences: Optional[int] = 1,
    min_gen_seq_length: Optional[int] = 0,
    max_gen_seq_length: Optional[int] = 48,
    repetition_penalty: Optional[float] = 1.0,
    no_repeat_ngram_size: Optional[int] = 0,
    early_stopping: Optional[bool] = True,
    length_penalty: Optional[float] = 1.0,
    num_beam_groups: Optional[int] = 1,
    diversity_penalty: Optional[float] = 0.0,
    do_sample: Optional[bool] = False,
    temperature: Optional[float] = 1.0,
    top_k: Optional[int] = 50,
    top_p: Optional[float] = 1.0,
)

Generates text using the BartForGeneration model.

Parameters:

Name Type Description Default
input_ids Tensor

Tensor of input token IDs.

 required
num_beams int

Number of beams for beam search. Defaults to 5.

 5
decoder_start_token_id int

The ID of the decoder start token. Defaults to 2.

 2
decoder_end_token_id int or List[int]

The ID(s) of the decoder end token(s). Defaults to 2.

 2
num_return_sequences int

Number of sequences to return. Defaults to 1.

 1
min_gen_seq_length int

Minimum length of generated sequences. Defaults to 0.

 0
max_gen_seq_length int

Maximum length of generated sequences. Defaults to 48.

 48
repetition_penalty float

Repetition penalty for generated sequences. Defaults to 1.0.

 1.0
no_repeat_ngram_size int

Size of n-grams to avoid repetition. Defaults to 0.

 0
early_stopping bool

Whether to stop generation early. Defaults to True.

 True
length_penalty float

Length penalty for generated sequences. Defaults to 1.0.

 1.0
num_beam_groups int

Number of beam groups for diverse beam search. Defaults to 1.

 1
diversity_penalty float

Diversity penalty for diverse beam search. Defaults to 0.0.

 0.0
do_sample bool

Whether to use sampling during generation. Defaults to False.

 False
temperature float

Temperature for sampling. Defaults to 1.0.

 1.0
top_k int

Value for top-k sampling. Defaults to 50.

 50
top_p float

Value for top-p sampling. Defaults to 1.0.

 1.0

Returns:

Name Type Description
GenericOutputs

Generated sequences and their scores.
Source code in src/unitorch/models/bart/modeling.py 
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
@torch.no_grad()
def generate(
    self,
    input_ids: torch.Tensor,
    num_beams: Optional[int] = 5,
    decoder_start_token_id: Optional[int] = 2,
    decoder_end_token_id: Optional[Union[int, List[int]]] = 2,
    num_return_sequences: Optional[int] = 1,
    min_gen_seq_length: Optional[int] = 0,
    max_gen_seq_length: Optional[int] = 48,
    repetition_penalty: Optional[float] = 1.0,
    no_repeat_ngram_size: Optional[int] = 0,
    early_stopping: Optional[bool] = True,
    length_penalty: Optional[float] = 1.0,
    num_beam_groups: Optional[int] = 1,
    diversity_penalty: Optional[float] = 0.0,
    do_sample: Optional[bool] = False,
    temperature: Optional[float] = 1.0,
    top_k: Optional[int] = 50,
    top_p: Optional[float] = 1.0,
):
    """
    Generates text using the BartForGeneration model.

    Args:
        input_ids (torch.Tensor): Tensor of input token IDs.
        num_beams (int, optional): Number of beams for beam search. Defaults to 5.
        decoder_start_token_id (int, optional): The ID of the decoder start token. Defaults to 2.
        decoder_end_token_id (int or List[int], optional): The ID(s) of the decoder end token(s). Defaults to 2.
        num_return_sequences (int, optional): Number of sequences to return. Defaults to 1.
        min_gen_seq_length (int, optional): Minimum length of generated sequences. Defaults to 0.
        max_gen_seq_length (int, optional): Maximum length of generated sequences. Defaults to 48.
        repetition_penalty (float, optional): Repetition penalty for generated sequences. Defaults to 1.0.
        no_repeat_ngram_size (int, optional): Size of n-grams to avoid repetition. Defaults to 0.
        early_stopping (bool, optional): Whether to stop generation early. Defaults to True.
        length_penalty (float, optional): Length penalty for generated sequences. Defaults to 1.0.
        num_beam_groups (int, optional): Number of beam groups for diverse beam search. Defaults to 1.
        diversity_penalty (float, optional): Diversity penalty for diverse beam search. Defaults to 0.0.
        do_sample (bool, optional): Whether to use sampling during generation. Defaults to False.
        temperature (float, optional): Temperature for sampling. Defaults to 1.0.
        top_k (int, optional): Value for top-k sampling. Defaults to 50.
        top_p (float, optional): Value for top-p sampling. Defaults to 1.0.

    Returns:
        GenericOutputs: Generated sequences and their scores.
    """
    outputs = self.model.generate(
        input_ids,
        max_length=max_gen_seq_length,
        min_length=min_gen_seq_length,
        num_beams=num_beams,
        do_sample=do_sample,
        decoder_start_token_id=decoder_start_token_id,
        no_repeat_ngram_size=no_repeat_ngram_size,
        early_stopping=early_stopping,
        length_penalty=length_penalty,
        repetition_penalty=repetition_penalty,
        num_return_sequences=num_return_sequences,
        bos_token_id=decoder_start_token_id,
        eos_token_id=decoder_end_token_id,
        num_beam_groups=num_beam_groups,
        diversity_penalty=diversity_penalty,
        temperature=temperature,
        top_k=top_k,
        top_p=top_p,
        return_dict_in_generate=True,
        output_scores=True,
    )

    sequences = outputs.sequences.reshape(
        -1, num_return_sequences, outputs.sequences.size(-1)
    )
    outputs.sequences = torch.zeros(
        sequences.size(0), num_return_sequences, max_gen_seq_length
    ).to(device=sequences.device)
    outputs.sequences[:, :, : sequences.size(-1)].copy_(sequences)

    if num_return_sequences == 1:
        outputs.sequences = outputs.sequences.reshape(-1, max_gen_seq_length)

    return GenericOutputs(
        sequences=outputs.sequences, sequences_scores=outputs.sequences_scores
    )