RNABERT
Pre-trained model on non-coding RNA (ncRNA) using masked language modeling (MLM) and structural alignment learning (SAL) objectives.
Disclaimer
This is an UNOFFICIAL implementation of the Informative RNA-base embedding for functional RNA clustering and structural alignment by Manato Akiyama and Yasubumi Sakakibara.
The OFFICIAL repository of RNABERT is at mana438/RNABERT.
Caution
The MultiMolecule team is aware of a potential risk in reproducing the results of RNABERT.
The original implementation of RNABERT does not prepend <bos> (<cls>) and append <eos> tokens to the input sequence.
This should not affect the performance of the model in most cases, but it can lead to unexpected behavior in some cases.
Please set bos_token=None, eos_token=None in the tokenizer and set bos_token_id=None, eos_token_id=None in the model configuration if you want the exact behavior of the original implementation.
Tip
The MultiMolecule team has confirmed that the provided model and checkpoints are producing the same intermediate representations as the original implementation.
The team releasing RNABERT did not write this model card for this model so this model card has been written by the MultiMolecule team.
Model Details
RNABERT is a bert-style model pre-trained on a large corpus of non-coding RNA sequences in a self-supervised fashion. This means that the model was trained on the raw nucleotides of RNA sequences only, with an automatic process to generate inputs and labels from those texts. Please refer to the Training Details section for more information on the training process.
Model Specification
| Num Layers |
Hidden Size |
Num Heads |
Intermediate Size |
Num Parameters (M) |
FLOPs (G) |
MACs (G) |
Max Num Tokens |
| 6 |
120 |
12 |
40 |
0.48 |
0.15 |
0.08 |
440 |
Links
Usage
The model file depends on the multimolecule library. You can install it using pip:
| Bash |
|---|
| pip install multimolecule
|
Direct Use
Masked Language Modeling
You can use this model directly with a pipeline for masked language modeling:
| Python |
|---|
| import multimolecule # you must import multimolecule to register models
from transformers import pipeline
predictor = pipeline("fill-mask", model="multimolecule/rnabert")
output = predictor("gguc<mask>cucugguuagaccagaucugagccu")
|
Downstream Use
Here is how to use this model to get the features of a given sequence in PyTorch:
| Python |
|---|
| from multimolecule import RnaTokenizer, RnaBertModel
tokenizer = RnaTokenizer.from_pretrained("multimolecule/rnabert")
model = RnaBertModel.from_pretrained("multimolecule/rnabert")
text = "UAGCUUAUCAGACUGAUGUUG"
input = tokenizer(text, return_tensors="pt")
output = model(**input)
|
Sequence Classification / Regression
Note
This model is not fine-tuned for any specific task. You will need to fine-tune the model on a downstream task to use it for sequence classification or regression.
Here is how to use this model as backbone to fine-tune for a sequence-level task in PyTorch:
| Python |
|---|
| import torch
from multimolecule import RnaTokenizer, RnaBertForSequencePrediction
tokenizer = RnaTokenizer.from_pretrained("multimolecule/rnabert")
model = RnaBertForSequencePrediction.from_pretrained("multimolecule/rnabert")
text = "UAGCUUAUCAGACUGAUGUUG"
input = tokenizer(text, return_tensors="pt")
label = torch.tensor([1])
output = model(**input, labels=label)
|
Token Classification / Regression
Note
This model is not fine-tuned for any specific task. You will need to fine-tune the model on a downstream task to use it for token classification or regression.
Here is how to use this model as backbone to fine-tune for a nucleotide-level task in PyTorch:
| Python |
|---|
| import torch
from multimolecule import RnaTokenizer, RnaBertForTokenPrediction
tokenizer = RnaTokenizer.from_pretrained("multimolecule/rnabert")
model = RnaBertForTokenPrediction.from_pretrained("multimolecule/rnabert")
text = "UAGCUUAUCAGACUGAUGUUG"
input = tokenizer(text, return_tensors="pt")
label = torch.randint(2, (len(text), ))
output = model(**input, labels=label)
|
Note
This model is not fine-tuned for any specific task. You will need to fine-tune the model on a downstream task to use it for contact classification or regression.
Here is how to use this model as backbone to fine-tune for a contact-level task in PyTorch:
| Python |
|---|
| import torch
from multimolecule import RnaTokenizer, RnaBertForContactPrediction
tokenizer = RnaTokenizer.from_pretrained("multimolecule/rnabert")
model = RnaBertForContactPrediction.from_pretrained("multimolecule/rnabert")
text = "UAGCUUAUCAGACUGAUGUUG"
input = tokenizer(text, return_tensors="pt")
label = torch.randint(2, (len(text), len(text)))
output = model(**input, labels=label)
|
Training Details
RNABERT has two pre-training objectives: masked language modeling (MLM) and structural alignment learning (SAL).
- Masked Language Modeling (MLM): taking a sequence, the model randomly masks 15% of the tokens in the input then runs the entire masked sentence through the model and has to predict the masked tokens. This is comparable to the Cloze task in language modeling.
- Structural Alignment Learning (SAL): the model learns to predict the structural alignment of two RNA sequences. The model is trained to predict the alignment score of two RNA sequences using the Needleman-Wunsch algorithm.
Training Data
The RNABERT model was pre-trained on RNAcentral.
RNAcentral is a free, public resource that offers integrated access to a comprehensive and up-to-date set of non-coding RNA sequences provided by a collaborating group of Expert Databases representing a broad range of organisms and RNA types.
RNABERT used a subset of 76, 237 human ncRNA sequences from RNAcentral for pre-training.
RNABERT preprocessed all tokens by replacing “U”s with “T”s.
Note that during model conversions, “T” is replaced with “U”. RnaTokenizer will convert “T”s to “U”s for you, you may disable this behaviour by passing replace_T_with_U=False.
Training Procedure
Preprocessing
RNABERT preprocess the dataset by applying 10 different mask patterns to the 72, 237 human ncRNA sequences. The final dataset contains 722, 370 sequences. The masking procedure is similar to the one used in BERT:
- 15% of the tokens are masked.
- In 80% of the cases, the masked tokens are replaced by
<mask>.
- In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace.
- In the 10% remaining cases, the masked tokens are left as is.
Pre-training
The model was trained on 1 NVIDIA V100 GPU.
Citation
| BibTeX |
|---|
| @article{akiyama2022informative,
author = {Akiyama, Manato and Sakakibara, Yasubumi},
title = "{Informative RNA base embedding for RNA structural alignment and clustering by deep representation learning}",
journal = {NAR Genomics and Bioinformatics},
volume = {4},
number = {1},
pages = {lqac012},
year = {2022},
month = {02},
abstract = "{Effective embedding is actively conducted by applying deep learning to biomolecular information. Obtaining better embeddings enhances the quality of downstream analyses, such as DNA sequence motif detection and protein function prediction. In this study, we adopt a pre-training algorithm for the effective embedding of RNA bases to acquire semantically rich representations and apply this algorithm to two fundamental RNA sequence problems: structural alignment and clustering. By using the pre-training algorithm to embed the four bases of RNA in a position-dependent manner using a large number of RNA sequences from various RNA families, a context-sensitive embedding representation is obtained. As a result, not only base information but also secondary structure and context information of RNA sequences are embedded for each base. We call this ‘informative base embedding’ and use it to achieve accuracies superior to those of existing state-of-the-art methods on RNA structural alignment and RNA family clustering tasks. Furthermore, upon performing RNA sequence alignment by combining this informative base embedding with a simple Needleman–Wunsch alignment algorithm, we succeed in calculating structural alignments with a time complexity of O(n2) instead of the O(n6) time complexity of the naive implementation of Sankoff-style algorithm for input RNA sequence of length n.}",
issn = {2631-9268},
doi = {10.1093/nargab/lqac012},
url = {https://doi.org/10.1093/nargab/lqac012},
eprint = {https://academic.oup.com/nargab/article-pdf/4/1/lqac012/42577168/lqac012.pdf},
}
|
Note
The artifacts distributed in this repository are part of the MultiMolecule project.
If you use MultiMolecule in your research, you must cite the MultiMolecule project as follows:
| BibTeX |
|---|
| @software{chen_2024_12638419,
author = {Chen, Zhiyuan and Zhu, Sophia Y.},
title = {MultiMolecule},
doi = {10.5281/zenodo.12638419},
publisher = {Zenodo},
url = {https://doi.org/10.5281/zenodo.12638419},
year = 2024,
month = may,
day = 4
}
|
Please use GitHub issues of MultiMolecule for any questions or comments on the model card.
Please contact the authors of the RNABERT paper for questions or comments on the paper/model.
License
This model is licensed under the GNU Affero General Public License.
For additional terms and clarifications, please refer to our License FAQ.
| Text Only |
|---|
| SPDX-License-Identifier: AGPL-3.0-or-later
|
Bases: Tokenizer
Tokenizer for RNA sequences.
Parameters:
| Name |
Type |
Description |
Default |
alphabet
|
Alphabet | str | List[str] | None
|
alphabet to use for tokenization.
- If is
None, the standard RNA alphabet will be used.
- If is a
string, it should correspond to the name of a predefined alphabet. The options include
standardextendedstreamlinenucleobase
- If is an alphabet or a list of characters, that specific alphabet will be used.
|
None
|
nmers
|
int
|
Size of kmer to tokenize.
|
1
|
codon
|
bool
|
Whether to tokenize into codons.
|
False
|
replace_T_with_U
|
bool
|
Whether to replace T with U.
|
True
|
do_upper_case
|
bool
|
Whether to convert input to uppercase.
|
True
|
Examples:
| Python Console Session |
|---|
| >>> from multimolecule import RnaTokenizer
>>> tokenizer = RnaTokenizer()
>>> tokenizer('<pad><cls><eos><unk><mask><null>ACGUNRYSWKMBDHV.X*-I')["input_ids"]
[1, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 2]
>>> tokenizer('acgu')["input_ids"]
[1, 6, 7, 8, 9, 2]
>>> tokenizer('acgt')["input_ids"]
[1, 6, 7, 8, 9, 2]
>>> tokenizer = RnaTokenizer(replace_T_with_U=False)
>>> tokenizer('acgt')["input_ids"]
[1, 6, 7, 8, 3, 2]
>>> tokenizer = RnaTokenizer(nmers=3)
>>> tokenizer('uagcuuauc')["input_ids"]
[1, 83, 17, 64, 49, 96, 84, 22, 2]
>>> tokenizer = RnaTokenizer(codon=True)
>>> tokenizer('uagcuuauc')["input_ids"]
[1, 83, 49, 22, 2]
>>> tokenizer('uagcuuauca')["input_ids"]
Traceback (most recent call last):
ValueError: length of input sequence must be a multiple of 3 for codon tokenization, but got 10
|
Source code in multimolecule/tokenisers/rna/tokenization_rna.py
| Python |
|---|
| class RnaTokenizer(Tokenizer):
"""
Tokenizer for RNA sequences.
Args:
alphabet: alphabet to use for tokenization.
- If is `None`, the standard RNA alphabet will be used.
- If is a `string`, it should correspond to the name of a predefined alphabet. The options include
+ `standard`
+ `extended`
+ `streamline`
+ `nucleobase`
- If is an alphabet or a list of characters, that specific alphabet will be used.
nmers: Size of kmer to tokenize.
codon: Whether to tokenize into codons.
replace_T_with_U: Whether to replace T with U.
do_upper_case: Whether to convert input to uppercase.
Examples:
>>> from multimolecule import RnaTokenizer
>>> tokenizer = RnaTokenizer()
>>> tokenizer('<pad><cls><eos><unk><mask><null>ACGUNRYSWKMBDHV.X*-I')["input_ids"]
[1, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 2]
>>> tokenizer('acgu')["input_ids"]
[1, 6, 7, 8, 9, 2]
>>> tokenizer('acgt')["input_ids"]
[1, 6, 7, 8, 9, 2]
>>> tokenizer = RnaTokenizer(replace_T_with_U=False)
>>> tokenizer('acgt')["input_ids"]
[1, 6, 7, 8, 3, 2]
>>> tokenizer = RnaTokenizer(nmers=3)
>>> tokenizer('uagcuuauc')["input_ids"]
[1, 83, 17, 64, 49, 96, 84, 22, 2]
>>> tokenizer = RnaTokenizer(codon=True)
>>> tokenizer('uagcuuauc')["input_ids"]
[1, 83, 49, 22, 2]
>>> tokenizer('uagcuuauca')["input_ids"]
Traceback (most recent call last):
ValueError: length of input sequence must be a multiple of 3 for codon tokenization, but got 10
"""
model_input_names = ["input_ids", "attention_mask"]
def __init__(
self,
alphabet: Alphabet | str | List[str] | None = None,
nmers: int = 1,
codon: bool = False,
replace_T_with_U: bool = True,
do_upper_case: bool = True,
additional_special_tokens: List | Tuple | None = None,
**kwargs,
):
if codon and (nmers > 1 and nmers != 3):
raise ValueError("Codon and nmers cannot be used together.")
if codon:
nmers = 3 # set to 3 to get correct vocab
if not isinstance(alphabet, Alphabet):
alphabet = get_alphabet(alphabet, nmers=nmers)
super().__init__(
alphabet=alphabet,
nmers=nmers,
codon=codon,
replace_T_with_U=replace_T_with_U,
do_upper_case=do_upper_case,
additional_special_tokens=additional_special_tokens,
**kwargs,
)
self.replace_T_with_U = replace_T_with_U
self.nmers = nmers
self.codon = codon
def _tokenize(self, text: str, **kwargs):
if self.do_upper_case:
text = text.upper()
if self.replace_T_with_U:
text = text.replace("T", "U")
if self.codon:
if len(text) % 3 != 0:
raise ValueError(
f"length of input sequence must be a multiple of 3 for codon tokenization, but got {len(text)}"
)
return [text[i : i + 3] for i in range(0, len(text), 3)]
if self.nmers > 1:
return [text[i : i + self.nmers] for i in range(len(text) - self.nmers + 1)] # noqa: E203
return list(text)
|
Bases: PreTrainedConfig
This is the configuration class to store the configuration of a RnaBertModel.
It is used to instantiate a RNABERT model according to the specified arguments, defining the model architecture.
Instantiating a configuration with the defaults will yield a similar configuration to that of the RNABERT
mana438/RNABERT architecture.
Configuration objects inherit from PreTrainedConfig and can be used to
control the model outputs. Read the documentation from PreTrainedConfig
for more information.
Parameters:
| Name |
Type |
Description |
Default |
vocab_size
|
int
|
Vocabulary size of the RNABERT model. Defines the number of different tokens that can be represented by the
input_ids passed when calling [RnaBertModel].
|
26
|
ss_vocab_size
|
int
|
Vocabulary size for secondary-structure tokens.
|
8
|
hidden_size
|
int | None
|
Dimensionality of the encoder layers and the pooler layer.
|
None
|
multiple
|
int | None
|
Optional multiplier used to derive hidden_size from num_attention_heads when hidden_size is not set.
|
None
|
num_hidden_layers
|
int
|
Number of hidden layers in the Transformer encoder.
|
6
|
num_attention_heads
|
int
|
Number of attention heads for each attention layer in the Transformer encoder.
|
12
|
|
|
int
|
Dimensionality of the “intermediate” (often named feed-forward) layer in the Transformer encoder.
|
40
|
hidden_act
|
str
|
The non-linear activation function (function or string) in the encoder and pooler. If string, "gelu",
"relu", "silu" and "gelu_new" are supported.
|
'gelu'
|
hidden_dropout
|
float
|
The dropout probability for all fully connected layers in the embeddings, encoder, and pooler.
|
0.0
|
attention_dropout
|
float
|
The dropout ratio for the attention probabilities.
|
0.0
|
max_position_embeddings
|
int
|
The maximum sequence length that this model might ever be used with. Typically set this to something large
just in case (e.g., 512 or 1024 or 2048).
|
440
|
initializer_range
|
float
|
The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
|
0.02
|
layer_norm_eps
|
float
|
The epsilon used by the layer normalization layers.
|
1e-12
|
position_embedding_type
|
str
|
|
'absolute'
|
is_decoder
|
bool
|
Whether the model is used as a decoder or not. If False, the model is used as an encoder.
|
False
|
use_cache
|
bool
|
Whether or not the model should return the last key/values attentions (not used by all models). Only
relevant if config.is_decoder=True.
|
True
|
head
|
HeadConfig | None
|
The configuration of the head.
|
None
|
lm_head
|
MaskedLMHeadConfig | None
|
The configuration of the masked language model head.
|
None
|
add_cross_attention
|
bool
|
Whether to add cross-attention layers when the model is used as a decoder.
|
False
|
Examples:
| Python Console Session |
|---|
| >>> from multimolecule import RnaBertConfig, RnaBertModel
>>> # Initializing a RNABERT multimolecule/rnabert style configuration
>>> configuration = RnaBertConfig()
>>> # Initializing a model (with random weights) from the multimolecule/rnabert style configuration
>>> model = RnaBertModel(configuration)
>>> # Accessing the model configuration
>>> configuration = model.config
|
Source code in multimolecule/models/rnabert/configuration_rnabert.py
| Python |
|---|
| class RnaBertConfig(PreTrainedConfig):
r"""
This is the configuration class to store the configuration of a [`RnaBertModel`][multimolecule.models.RnaBertModel].
It is used to instantiate a RNABERT model according to the specified arguments, defining the model architecture.
Instantiating a configuration with the defaults will yield a similar configuration to that of the RNABERT
[mana438/RNABERT](https://github.com/mana438/RNABERT) architecture.
Configuration objects inherit from [`PreTrainedConfig`][multimolecule.models.PreTrainedConfig] and can be used to
control the model outputs. Read the documentation from [`PreTrainedConfig`][multimolecule.models.PreTrainedConfig]
for more information.
Args:
vocab_size:
Vocabulary size of the RNABERT model. Defines the number of different tokens that can be represented by the
`input_ids` passed when calling [`RnaBertModel`].
ss_vocab_size:
Vocabulary size for secondary-structure tokens.
hidden_size:
Dimensionality of the encoder layers and the pooler layer.
multiple:
Optional multiplier used to derive `hidden_size` from `num_attention_heads` when `hidden_size` is not set.
num_hidden_layers:
Number of hidden layers in the Transformer encoder.
num_attention_heads:
Number of attention heads for each attention layer in the Transformer encoder.
intermediate_size:
Dimensionality of the "intermediate" (often named feed-forward) layer in the Transformer encoder.
hidden_act:
The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`,
`"relu"`, `"silu"` and `"gelu_new"` are supported.
hidden_dropout:
The dropout probability for all fully connected layers in the embeddings, encoder, and pooler.
attention_dropout:
The dropout ratio for the attention probabilities.
max_position_embeddings:
The maximum sequence length that this model might ever be used with. Typically set this to something large
just in case (e.g., 512 or 1024 or 2048).
initializer_range:
The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
layer_norm_eps:
The epsilon used by the layer normalization layers.
position_embedding_type:
Type of position embedding. Choose one of `"absolute"`, `"relative_key"`, `"relative_key_query"`. For
positional embeddings use `"absolute"`. For more information on `"relative_key"`, please refer to
[Self-Attention with Relative Position Representations (Shaw et al.)](https://arxiv.org/abs/1803.02155).
For more information on `"relative_key_query"`, please refer to *Method 4* in [Improve Transformer Models
with Better Relative Position Embeddings (Huang et al.)](https://arxiv.org/abs/2009.13658).
is_decoder:
Whether the model is used as a decoder or not. If `False`, the model is used as an encoder.
use_cache:
Whether or not the model should return the last key/values attentions (not used by all models). Only
relevant if `config.is_decoder=True`.
head:
The configuration of the head.
lm_head:
The configuration of the masked language model head.
add_cross_attention:
Whether to add cross-attention layers when the model is used as a decoder.
Examples:
>>> from multimolecule import RnaBertConfig, RnaBertModel
>>> # Initializing a RNABERT multimolecule/rnabert style configuration
>>> configuration = RnaBertConfig()
>>> # Initializing a model (with random weights) from the multimolecule/rnabert style configuration
>>> model = RnaBertModel(configuration)
>>> # Accessing the model configuration
>>> configuration = model.config
"""
model_type = "rnabert"
def __init__(
self,
vocab_size: int = 26,
ss_vocab_size: int = 8,
hidden_size: int | None = None,
multiple: int | None = None,
num_hidden_layers: int = 6,
num_attention_heads: int = 12,
intermediate_size: int = 40,
hidden_act: str = "gelu",
hidden_dropout: float = 0.0,
attention_dropout: float = 0.0,
max_position_embeddings: int = 440,
initializer_range: float = 0.02,
layer_norm_eps: float = 1e-12,
position_embedding_type: str = "absolute",
is_decoder: bool = False,
use_cache: bool = True,
head: HeadConfig | None = None,
lm_head: MaskedLMHeadConfig | None = None,
add_cross_attention: bool = False,
**kwargs,
):
super().__init__(**kwargs)
if hidden_size is None:
hidden_size = num_attention_heads * multiple if multiple is not None else 120
self.vocab_size = vocab_size
self.ss_vocab_size = ss_vocab_size
self.type_vocab_size = 2
self.hidden_size = hidden_size
self.num_hidden_layers = num_hidden_layers
self.num_attention_heads = num_attention_heads
self.intermediate_size = intermediate_size
self.hidden_act = hidden_act
self.hidden_dropout = hidden_dropout
self.attention_dropout = attention_dropout
self.max_position_embeddings = max_position_embeddings
self.initializer_range = initializer_range
self.layer_norm_eps = layer_norm_eps
self.position_embedding_type = position_embedding_type
self.is_decoder = is_decoder
self.use_cache = use_cache
self.head = HeadConfig(**head) if head is not None else None
self.lm_head = MaskedLMHeadConfig(**lm_head) if lm_head is not None else None
self.add_cross_attention = add_cross_attention
|
Bases: RnaBertPreTrainedModel
Examples:
| Python Console Session |
|---|
| >>> import torch
>>> from multimolecule import RnaBertConfig, RnaBertForContactPrediction, RnaTokenizer
>>> config = RnaBertConfig()
>>> model = RnaBertForContactPrediction(config)
>>> tokenizer = RnaTokenizer.from_pretrained("multimolecule/rna")
>>> input = tokenizer("ACGUN", return_tensors="pt")
>>> output = model(**input, labels=torch.randint(2, (1, 5, 5)))
>>> output["logits"].shape
torch.Size([1, 5, 5, 1])
>>> output["loss"]
tensor(..., grad_fn=<BinaryCrossEntropyWithLogitsBackward0>)
|
Source code in multimolecule/models/rnabert/modeling_rnabert.py
| Python |
|---|
| class RnaBertForContactPrediction(RnaBertPreTrainedModel):
"""
Examples:
>>> import torch
>>> from multimolecule import RnaBertConfig, RnaBertForContactPrediction, RnaTokenizer
>>> config = RnaBertConfig()
>>> model = RnaBertForContactPrediction(config)
>>> tokenizer = RnaTokenizer.from_pretrained("multimolecule/rna")
>>> input = tokenizer("ACGUN", return_tensors="pt")
>>> output = model(**input, labels=torch.randint(2, (1, 5, 5)))
>>> output["logits"].shape
torch.Size([1, 5, 5, 1])
>>> output["loss"] # doctest:+ELLIPSIS
tensor(..., grad_fn=<BinaryCrossEntropyWithLogitsBackward0>)
"""
def __init__(self, config: RnaBertConfig):
super().__init__(config)
self.model = RnaBertModel(config, add_pooling_layer=False)
self.contact_head = ContactPredictionHead(config)
self.head_config = self.contact_head.config
self.require_attentions = self.contact_head.require_attentions
# Initialize weights and apply final processing
self.post_init()
@can_return_tuple
def forward(
self,
input_ids: Tensor | NestedTensor | None = None,
attention_mask: Tensor | None = None,
position_ids: Tensor | None = None,
inputs_embeds: Tensor | NestedTensor | None = None,
labels: Tensor | None = None,
**kwargs: Unpack[TransformersKwargs],
) -> Tuple[Tensor, ...] | ContactPredictorOutput:
if self.require_attentions:
output_attentions = kwargs.get("output_attentions", self.config.output_attentions)
if output_attentions is False:
warn("output_attentions must be True since prediction head requires attentions.")
kwargs["output_attentions"] = True
outputs = self.model(
input_ids,
attention_mask=attention_mask,
position_ids=position_ids,
inputs_embeds=inputs_embeds,
return_dict=True,
**kwargs,
)
output = self.contact_head(outputs, attention_mask, input_ids, labels)
logits, loss = output.logits, output.loss
return ContactPredictorOutput(
loss=loss,
logits=logits,
hidden_states=outputs.hidden_states,
attentions=outputs.attentions,
)
|
Bases: RnaBertPreTrainedModel
Examples:
| Python Console Session |
|---|
| >>> import torch
>>> from multimolecule import RnaBertConfig, RnaBertForMaskedLM, RnaTokenizer
>>> config = RnaBertConfig()
>>> model = RnaBertForMaskedLM(config)
>>> tokenizer = RnaTokenizer.from_pretrained("multimolecule/rna")
>>> input = tokenizer("ACGUN", return_tensors="pt")
>>> output = model(**input, labels=input["input_ids"])
>>> output["logits"].shape
torch.Size([1, 7, 26])
>>> output["loss"]
tensor(..., grad_fn=<NllLossBackward0>)
|
Source code in multimolecule/models/rnabert/modeling_rnabert.py
| Python |
|---|
| class RnaBertForMaskedLM(RnaBertPreTrainedModel):
"""
Examples:
>>> import torch
>>> from multimolecule import RnaBertConfig, RnaBertForMaskedLM, RnaTokenizer
>>> config = RnaBertConfig()
>>> model = RnaBertForMaskedLM(config)
>>> tokenizer = RnaTokenizer.from_pretrained("multimolecule/rna")
>>> input = tokenizer("ACGUN", return_tensors="pt")
>>> output = model(**input, labels=input["input_ids"])
>>> output["logits"].shape
torch.Size([1, 7, 26])
>>> output["loss"] # doctest:+ELLIPSIS
tensor(..., grad_fn=<NllLossBackward0>)
"""
_tied_weights_keys = {
"lm_head.decoder.weight": "model.embeddings.word_embeddings.weight",
"lm_head.decoder.bias": "lm_head.bias",
}
def __init__(self, config: RnaBertConfig):
super().__init__(config)
if config.is_decoder:
warn(
"If you want to use `RnaBertForMaskedLM` make sure `config.is_decoder=False` for "
"bi-directional self-attention."
)
self.model = RnaBertModel(config, add_pooling_layer=False)
self.lm_head = MaskedLMHead(config)
# Initialize weights and apply final processing
self.post_init()
def get_output_embeddings(self):
return self.lm_head.decoder
def set_output_embeddings(self, embeddings):
self.lm_head.decoder = embeddings
if hasattr(self.lm_head, "bias"):
self.lm_head.bias = embeddings.bias
@can_return_tuple
def forward(
self,
input_ids: Tensor | NestedTensor | None = None,
attention_mask: Tensor | None = None,
labels: Tensor | None = None,
**kwargs: Unpack[TransformersKwargs],
) -> Tuple[Tensor, ...] | MaskedLMOutput:
outputs = self.model(
input_ids,
attention_mask=attention_mask,
return_dict=True,
**kwargs,
)
output = self.lm_head(outputs, labels)
logits, loss = output.logits, output.loss
return MaskedLMOutput(
loss=loss,
logits=logits,
hidden_states=outputs.hidden_states,
attentions=outputs.attentions,
)
|
Bases: RnaBertPreTrainedModel
Examples:
| Python Console Session |
|---|
| >>> import torch
>>> from multimolecule import RnaBertConfig, RnaBertForPreTraining, RnaTokenizer
>>> config = RnaBertConfig()
>>> model = RnaBertForPreTraining(config)
>>> tokenizer = RnaTokenizer.from_pretrained("multimolecule/rna")
>>> input = tokenizer("ACGUN", return_tensors="pt")
>>> output = model(**input, labels_lm_seq=input["input_ids"])
>>> output["loss"]
tensor(..., grad_fn=<MeanBackward0>)
>>> output["logits_lm"].shape
torch.Size([1, 7, 26])
>>> output["logits_ss"].shape
torch.Size([1, 7, 8])
>>> output["logits_sa"].shape
torch.Size([1, 2])
|
Source code in multimolecule/models/rnabert/modeling_rnabert.py
| Python |
|---|
| class RnaBertForPreTraining(RnaBertPreTrainedModel):
"""
Examples:
>>> import torch
>>> from multimolecule import RnaBertConfig, RnaBertForPreTraining, RnaTokenizer
>>> config = RnaBertConfig()
>>> model = RnaBertForPreTraining(config)
>>> tokenizer = RnaTokenizer.from_pretrained("multimolecule/rna")
>>> input = tokenizer("ACGUN", return_tensors="pt")
>>> output = model(**input, labels_lm_seq=input["input_ids"])
>>> output["loss"] # doctest:+ELLIPSIS
tensor(..., grad_fn=<MeanBackward0>)
>>> output["logits_lm"].shape
torch.Size([1, 7, 26])
>>> output["logits_ss"].shape
torch.Size([1, 7, 8])
>>> output["logits_sa"].shape
torch.Size([1, 2])
"""
_tied_weights_keys = {
"lm_head.decoder.weight": "model.embeddings.word_embeddings.weight",
"lm_head.decoder.bias": "lm_head.bias",
"ss_head.decoder.bias": "ss_head.bias",
}
def __init__(self, config: RnaBertConfig):
super().__init__(config)
if config.is_decoder:
warn(
"If you want to use `RnaBertForPreTraining` make sure `config.is_decoder=False` for "
"bi-directional self-attention."
)
self.model = RnaBertModel(config)
self.lm_head = MaskedLMHead(config)
vocab_size, config.vocab_size = config.vocab_size, config.ss_vocab_size
self.ss_head = MaskedLMHead(config)
config.vocab_size = vocab_size
self.sa_head = SequencePredictionHead(config, HeadConfig(num_labels=2))
# Initialize weights and apply final processing
self.post_init()
@can_return_tuple
def forward(
self,
input_ids: Tensor | NestedTensor | None = None,
attention_mask: Tensor | None = None,
labels_lm_seq: Tensor | None = None,
labels_lm_ss: Tensor | None = None,
labels_sa: Tensor | None = None,
**kwargs: Unpack[TransformersKwargs],
) -> Tuple[Tensor, ...] | RnaBertForPreTrainingOutput:
outputs = self.model(
input_ids,
attention_mask=attention_mask,
return_dict=True,
**kwargs,
)
output_lm = self.lm_head(outputs, labels=labels_lm_seq)
logits_lm, loss_lm = output_lm.logits, output_lm.loss
output_ss = self.ss_head(outputs, labels=labels_lm_ss)
logits_ss, loss_ss = output_ss.logits, output_ss.loss
output_sa = self.sa_head(outputs, labels=labels_sa)
logits_sa, loss_sa = output_sa.logits, output_sa.loss
losses = tuple(l for l in (loss_lm, loss_ss, loss_sa) if l is not None) # noqa: E741
loss = torch.mean(torch.stack(losses)) if losses else None
return RnaBertForPreTrainingOutput(
loss=loss,
logits_lm=logits_lm,
loss_lm=loss_lm,
logits_ss=logits_ss,
loss_ss=loss_ss,
logits_sa=logits_sa,
loss_sa=loss_sa,
hidden_states=outputs.hidden_states,
attentions=outputs.attentions,
)
|
Bases: RnaBertPreTrainedModel
Examples:
| Python Console Session |
|---|
| >>> import torch
>>> from multimolecule import RnaBertConfig, RnaBertForSequencePrediction, RnaTokenizer
>>> config = RnaBertConfig()
>>> model = RnaBertForSequencePrediction(config)
>>> tokenizer = RnaTokenizer.from_pretrained("multimolecule/rna")
>>> input = tokenizer("ACGUN", return_tensors="pt")
>>> output = model(**input, labels=torch.tensor([[1]]))
>>> output["logits"].shape
torch.Size([1, 1])
>>> output["loss"]
tensor(..., grad_fn=<BinaryCrossEntropyWithLogitsBackward0>)
|
Source code in multimolecule/models/rnabert/modeling_rnabert.py
| Python |
|---|
| class RnaBertForSequencePrediction(RnaBertPreTrainedModel):
"""
Examples:
>>> import torch
>>> from multimolecule import RnaBertConfig, RnaBertForSequencePrediction, RnaTokenizer
>>> config = RnaBertConfig()
>>> model = RnaBertForSequencePrediction(config)
>>> tokenizer = RnaTokenizer.from_pretrained("multimolecule/rna")
>>> input = tokenizer("ACGUN", return_tensors="pt")
>>> output = model(**input, labels=torch.tensor([[1]]))
>>> output["logits"].shape
torch.Size([1, 1])
>>> output["loss"] # doctest:+ELLIPSIS
tensor(..., grad_fn=<BinaryCrossEntropyWithLogitsBackward0>)
"""
def __init__(self, config: RnaBertConfig):
super().__init__(config)
self.model = RnaBertModel(config)
self.sequence_head = SequencePredictionHead(config)
self.head_config = self.sequence_head.config
# Initialize weights and apply final processing
self.post_init()
@can_return_tuple
def forward(
self,
input_ids: Tensor | NestedTensor | None = None,
attention_mask: Tensor | None = None,
position_ids: Tensor | None = None,
inputs_embeds: Tensor | NestedTensor | None = None,
labels: Tensor | None = None,
**kwargs: Unpack[TransformersKwargs],
) -> Tuple[Tensor, ...] | SequencePredictorOutput:
outputs = self.model(
input_ids,
attention_mask=attention_mask,
position_ids=position_ids,
inputs_embeds=inputs_embeds,
return_dict=True,
**kwargs,
)
output = self.sequence_head(outputs, labels)
logits, loss = output.logits, output.loss
return SequencePredictorOutput(
loss=loss,
logits=logits,
hidden_states=outputs.hidden_states,
attentions=outputs.attentions,
)
|
Bases: RnaBertPreTrainedModel
Examples:
| Python Console Session |
|---|
| >>> import torch
>>> from multimolecule import RnaBertConfig, RnaBertForTokenPrediction, RnaTokenizer
>>> config = RnaBertConfig()
>>> model = RnaBertForTokenPrediction(config)
>>> tokenizer = RnaTokenizer.from_pretrained("multimolecule/rna")
>>> input = tokenizer("ACGUN", return_tensors="pt")
>>> output = model(**input, labels=torch.randint(2, (1, 5)))
>>> output["logits"].shape
torch.Size([1, 5, 1])
>>> output["loss"]
tensor(..., grad_fn=<BinaryCrossEntropyWithLogitsBackward0>)
|
Source code in multimolecule/models/rnabert/modeling_rnabert.py
| Python |
|---|
| class RnaBertForTokenPrediction(RnaBertPreTrainedModel):
"""
Examples:
>>> import torch
>>> from multimolecule import RnaBertConfig, RnaBertForTokenPrediction, RnaTokenizer
>>> config = RnaBertConfig()
>>> model = RnaBertForTokenPrediction(config)
>>> tokenizer = RnaTokenizer.from_pretrained("multimolecule/rna")
>>> input = tokenizer("ACGUN", return_tensors="pt")
>>> output = model(**input, labels=torch.randint(2, (1, 5)))
>>> output["logits"].shape
torch.Size([1, 5, 1])
>>> output["loss"] # doctest:+ELLIPSIS
tensor(..., grad_fn=<BinaryCrossEntropyWithLogitsBackward0>)
"""
def __init__(self, config: RnaBertConfig):
super().__init__(config)
self.model = RnaBertModel(config, add_pooling_layer=False)
self.token_head = TokenPredictionHead(config)
self.head_config = self.token_head.config
# Initialize weights and apply final processing
self.post_init()
@can_return_tuple
def forward(
self,
input_ids: Tensor | NestedTensor | None = None,
attention_mask: Tensor | None = None,
position_ids: Tensor | None = None,
inputs_embeds: Tensor | NestedTensor | None = None,
labels: Tensor | None = None,
**kwargs: Unpack[TransformersKwargs],
) -> Tuple[Tensor, ...] | TokenPredictorOutput:
outputs = self.model(
input_ids,
attention_mask=attention_mask,
position_ids=position_ids,
inputs_embeds=inputs_embeds,
return_dict=True,
**kwargs,
)
output = self.token_head(outputs, attention_mask, input_ids, labels)
logits, loss = output.logits, output.loss
return TokenPredictorOutput(
loss=loss,
logits=logits,
hidden_states=outputs.hidden_states,
attentions=outputs.attentions,
)
|
Bases: RnaBertPreTrainedModel
Examples:
| Python Console Session |
|---|
| >>> import torch
>>> from multimolecule import RnaBertConfig, RnaBertModel, RnaTokenizer
>>> config = RnaBertConfig()
>>> model = RnaBertModel(config)
>>> tokenizer = RnaTokenizer.from_pretrained("multimolecule/rna")
>>> input = tokenizer("ACGUN", return_tensors="pt")
>>> output = model(**input)
>>> output["last_hidden_state"].shape
torch.Size([1, 7, 120])
>>> output["pooler_output"].shape
torch.Size([1, 120])
|
Source code in multimolecule/models/rnabert/modeling_rnabert.py
| Python |
|---|
| class RnaBertModel(RnaBertPreTrainedModel):
"""
Examples:
>>> import torch
>>> from multimolecule import RnaBertConfig, RnaBertModel, RnaTokenizer
>>> config = RnaBertConfig()
>>> model = RnaBertModel(config)
>>> tokenizer = RnaTokenizer.from_pretrained("multimolecule/rna")
>>> input = tokenizer("ACGUN", return_tensors="pt")
>>> output = model(**input)
>>> output["last_hidden_state"].shape
torch.Size([1, 7, 120])
>>> output["pooler_output"].shape
torch.Size([1, 120])
"""
def __init__(self, config: RnaBertConfig, add_pooling_layer: bool = True):
super().__init__(config)
self.pad_token_id = config.pad_token_id
self.gradient_checkpointing = False
self.embeddings = RnaBertEmbeddings(config)
self.encoder = RnaBertEncoder(config)
self.pooler = RnaBertPooler(config) if add_pooling_layer else None
# Initialize weights and apply final processing
self.post_init()
def get_input_embeddings(self):
return self.embeddings.word_embeddings
def set_input_embeddings(self, value):
self.embeddings.word_embeddings = value
@check_model_inputs
def forward(
self,
input_ids: Tensor | NestedTensor | None = None,
attention_mask: Tensor | None = None,
position_ids: Tensor | None = None,
inputs_embeds: Tensor | NestedTensor | None = None,
encoder_hidden_states: Tensor | None = None,
encoder_attention_mask: Tensor | None = None,
past_key_values: Cache | None = None,
use_cache: bool | None = None,
cache_position: Tensor | None = None,
**kwargs: Unpack[TransformersKwargs],
) -> Tuple[Tensor, ...] | BaseModelOutputWithPoolingAndCrossAttentions:
r"""
Args:
encoder_hidden_states:
Shape: `(batch_size, sequence_length, hidden_size)`
Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention if
the model is configured as a decoder.
encoder_attention_mask:
Shape: `(batch_size, sequence_length)`
Mask to avoid performing attention on the padding token indices of the encoder input. This mask is used
in the cross-attention if the model is configured as a decoder. Mask values selected in `[0, 1]`:
- 1 for tokens that are **not masked**,
- 0 for tokens that are **masked**.
past_key_values:
Tuple of length `config.n_layers` with each tuple having 4 tensors of shape
`(batch_size, num_heads, sequence_length - 1, embed_size_per_head)
Contains precomputed key and value hidden states of the attention blocks. Can be used to speed up
decoding.
If `past_key_values` are used, the user can optionally input only the last `decoder_input_ids` (those
that don't have their past key value states given to this model) of shape `(batch_size, 1)` instead of
all `decoder_input_ids` of shape `(batch_size, sequence_length)`.
use_cache:
If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding
(see `past_key_values`).
"""
if self.config.is_decoder:
use_cache = use_cache if use_cache is not None else self.config.use_cache
else:
use_cache = False
if use_cache and past_key_values is None:
past_key_values = (
EncoderDecoderCache(DynamicCache(config=self.config), DynamicCache(config=self.config))
if encoder_hidden_states is not None or self.config.is_encoder_decoder
else DynamicCache(config=self.config)
)
if isinstance(input_ids, NestedTensor) and attention_mask is None:
attention_mask = input_ids.mask
if (input_ids is None) ^ (inputs_embeds is not None):
raise ValueError("You must specify exactly one of input_ids or inputs_embeds")
if input_ids is not None:
device = input_ids.device
seq_length = input_ids.shape[1]
else:
device = inputs_embeds.device # type: ignore[union-attr]
seq_length = inputs_embeds.shape[1] # type: ignore[union-attr]
# past_key_values_length
past_key_values_length = past_key_values.get_seq_length() if past_key_values is not None else 0
if cache_position is None:
cache_position = torch.arange(past_key_values_length, past_key_values_length + seq_length, device=device)
if attention_mask is None and input_ids is not None and self.pad_token_id is not None:
attention_mask = input_ids.ne(self.pad_token_id)
embedding_output = self.embeddings(
input_ids=input_ids,
position_ids=position_ids,
inputs_embeds=inputs_embeds,
past_key_values_length=past_key_values_length,
)
attention_mask, encoder_attention_mask = self._create_attention_masks(
attention_mask=attention_mask,
encoder_attention_mask=encoder_attention_mask,
embedding_output=embedding_output,
encoder_hidden_states=encoder_hidden_states,
cache_position=cache_position,
past_key_values=past_key_values,
)
encoder_outputs = self.encoder(
embedding_output,
attention_mask,
encoder_hidden_states,
encoder_attention_mask=encoder_attention_mask,
past_key_values=past_key_values,
use_cache=use_cache,
cache_position=cache_position,
position_ids=position_ids,
**kwargs,
)
sequence_output = encoder_outputs.last_hidden_state
pooled_output = self.pooler(sequence_output) if self.pooler is not None else None
return BaseModelOutputWithPoolingAndCrossAttentions(
last_hidden_state=sequence_output,
pooler_output=pooled_output,
past_key_values=encoder_outputs.past_key_values,
)
def _create_attention_masks(
self,
attention_mask,
encoder_attention_mask,
embedding_output,
encoder_hidden_states,
cache_position,
past_key_values,
):
if self.config.is_decoder:
attention_mask = create_causal_mask(
config=self.config,
input_embeds=embedding_output,
attention_mask=attention_mask,
cache_position=cache_position,
past_key_values=past_key_values,
)
else:
attention_mask = create_bidirectional_mask(
config=self.config, input_embeds=embedding_output, attention_mask=attention_mask
)
if encoder_attention_mask is not None:
encoder_attention_mask = create_bidirectional_mask(
config=self.config,
input_embeds=embedding_output,
attention_mask=encoder_attention_mask,
encoder_hidden_states=encoder_hidden_states,
)
return attention_mask, encoder_attention_mask
|
Pythonforward(
input_ids: Tensor | NestedTensor | None = None,
attention_mask: Tensor | None = None,
position_ids: Tensor | None = None,
inputs_embeds: Tensor | NestedTensor | None = None,
encoder_hidden_states: Tensor | None = None,
encoder_attention_mask: Tensor | None = None,
past_key_values: Cache | None = None,
use_cache: bool | None = None,
cache_position: Tensor | None = None,
**kwargs: Unpack[TransformersKwargs]
) -> (
Tuple[Tensor, ...]
| BaseModelOutputWithPoolingAndCrossAttentions
)
Parameters:
| Name |
Type |
Description |
Default |
encoder_hidden_states
|
Tensor | None
|
Shape: (batch_size, sequence_length, hidden_size)
Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention if
the model is configured as a decoder.
|
None
|
encoder_attention_mask
|
Tensor | None
|
Shape: (batch_size, sequence_length)
Mask to avoid performing attention on the padding token indices of the encoder input. This mask is used
in the cross-attention if the model is configured as a decoder. Mask values selected in [0, 1]:
- 1 for tokens that are not masked,
- 0 for tokens that are masked.
|
None
|
past_key_values
|
Cache | None
|
Tuple of length config.n_layers with each tuple having 4 tensors of shape
`(batch_size, num_heads, sequence_length - 1, embed_size_per_head)
Contains precomputed key and value hidden states of the attention blocks. Can be used to speed up
decoding.
If past_key_values are used, the user can optionally input only the last decoder_input_ids (those
that don’t have their past key value states given to this model) of shape (batch_size, 1) instead of
all decoder_input_ids of shape (batch_size, sequence_length).
|
None
|
use_cache
|
bool | None
|
If set to True, past_key_values key value states are returned and can be used to speed up decoding
(see past_key_values).
|
None
|
Source code in multimolecule/models/rnabert/modeling_rnabert.py
| Python |
|---|
| @check_model_inputs
def forward(
self,
input_ids: Tensor | NestedTensor | None = None,
attention_mask: Tensor | None = None,
position_ids: Tensor | None = None,
inputs_embeds: Tensor | NestedTensor | None = None,
encoder_hidden_states: Tensor | None = None,
encoder_attention_mask: Tensor | None = None,
past_key_values: Cache | None = None,
use_cache: bool | None = None,
cache_position: Tensor | None = None,
**kwargs: Unpack[TransformersKwargs],
) -> Tuple[Tensor, ...] | BaseModelOutputWithPoolingAndCrossAttentions:
r"""
Args:
encoder_hidden_states:
Shape: `(batch_size, sequence_length, hidden_size)`
Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention if
the model is configured as a decoder.
encoder_attention_mask:
Shape: `(batch_size, sequence_length)`
Mask to avoid performing attention on the padding token indices of the encoder input. This mask is used
in the cross-attention if the model is configured as a decoder. Mask values selected in `[0, 1]`:
- 1 for tokens that are **not masked**,
- 0 for tokens that are **masked**.
past_key_values:
Tuple of length `config.n_layers` with each tuple having 4 tensors of shape
`(batch_size, num_heads, sequence_length - 1, embed_size_per_head)
Contains precomputed key and value hidden states of the attention blocks. Can be used to speed up
decoding.
If `past_key_values` are used, the user can optionally input only the last `decoder_input_ids` (those
that don't have their past key value states given to this model) of shape `(batch_size, 1)` instead of
all `decoder_input_ids` of shape `(batch_size, sequence_length)`.
use_cache:
If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding
(see `past_key_values`).
"""
if self.config.is_decoder:
use_cache = use_cache if use_cache is not None else self.config.use_cache
else:
use_cache = False
if use_cache and past_key_values is None:
past_key_values = (
EncoderDecoderCache(DynamicCache(config=self.config), DynamicCache(config=self.config))
if encoder_hidden_states is not None or self.config.is_encoder_decoder
else DynamicCache(config=self.config)
)
if isinstance(input_ids, NestedTensor) and attention_mask is None:
attention_mask = input_ids.mask
if (input_ids is None) ^ (inputs_embeds is not None):
raise ValueError("You must specify exactly one of input_ids or inputs_embeds")
if input_ids is not None:
device = input_ids.device
seq_length = input_ids.shape[1]
else:
device = inputs_embeds.device # type: ignore[union-attr]
seq_length = inputs_embeds.shape[1] # type: ignore[union-attr]
# past_key_values_length
past_key_values_length = past_key_values.get_seq_length() if past_key_values is not None else 0
if cache_position is None:
cache_position = torch.arange(past_key_values_length, past_key_values_length + seq_length, device=device)
if attention_mask is None and input_ids is not None and self.pad_token_id is not None:
attention_mask = input_ids.ne(self.pad_token_id)
embedding_output = self.embeddings(
input_ids=input_ids,
position_ids=position_ids,
inputs_embeds=inputs_embeds,
past_key_values_length=past_key_values_length,
)
attention_mask, encoder_attention_mask = self._create_attention_masks(
attention_mask=attention_mask,
encoder_attention_mask=encoder_attention_mask,
embedding_output=embedding_output,
encoder_hidden_states=encoder_hidden_states,
cache_position=cache_position,
past_key_values=past_key_values,
)
encoder_outputs = self.encoder(
embedding_output,
attention_mask,
encoder_hidden_states,
encoder_attention_mask=encoder_attention_mask,
past_key_values=past_key_values,
use_cache=use_cache,
cache_position=cache_position,
position_ids=position_ids,
**kwargs,
)
sequence_output = encoder_outputs.last_hidden_state
pooled_output = self.pooler(sequence_output) if self.pooler is not None else None
return BaseModelOutputWithPoolingAndCrossAttentions(
last_hidden_state=sequence_output,
pooler_output=pooled_output,
past_key_values=encoder_outputs.past_key_values,
)
|
Bases: PreTrainedModel
An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained
models.
Source code in multimolecule/models/rnabert/modeling_rnabert.py
| Python |
|---|
| class RnaBertPreTrainedModel(PreTrainedModel):
"""
An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained
models.
"""
config_class = RnaBertConfig
base_model_prefix = "model"
supports_gradient_checkpointing = True
_supports_flash_attn = True
_supports_sdpa = True
_supports_flex_attn = True
_supports_attention_backend = True
_can_record_outputs: dict[str, Any] | None = None
_no_split_modules = ["RnaBertLayer", "RnaBertEmbeddings"]
@torch.no_grad()
def _init_weights(self, module: nn.Module):
super()._init_weights(module)
if isinstance(module, RnaBertEmbeddings):
init.copy_(module.position_ids, torch.arange(module.position_ids.shape[-1]).expand((1, -1)))
|