Views
No views yet
1from transformers import (
2 T5Config,
3 T5EncoderModel,
4 T5Tokenizer,
5 PreTrainedModel,
6 TrainingArguments,
7 Trainer,
8 DataCollatorWithPadding,
9)
10class T5PairRegressionModel(PreTrainedModel):
11 config_class = T5Config
12
13 def __init__(self, config, d_model=None):
14 super().__init__(config)
15 self.encoder = T5EncoderModel.from_pretrained("Rostlab/prot_t5_xl_uniref50")
16 hidden_dim = d_model if d_model is not None else config.d_model
17 self.regression_head = nn.Linear(hidden_dim, 1)
18
19 def forward(self, input_ids=None, attention_mask=None, labels=None, **kwargs):
20 encoder_outputs = self.encoder(input_ids=input_ids, attention_mask=attention_mask)
21 hidden_states = encoder_outputs.last_hidden_state
22
23 mask = attention_mask.unsqueeze(-1)
24 pooled_output = (hidden_states * mask).sum(dim=1) / mask.sum(dim=1)
25 logits = self.regression_head(pooled_output).squeeze(-1) # [batch_size]
26
27 loss = None
28 if labels is not None:
29 labels = labels.to(torch.bfloat16)
30 loss = nn.MSELoss()(logits, labels)
31
32 return {"loss": loss, "logits": logits}1tokenizer = T5Tokenizer.from_pretrained("jiaxie/DeepProtT5-SAbDab-Chen", do_lower_case=False)
2model = T5PairRegressionModel.from_pretrained("jiaxie/DeepProtT5-SAbDab-Chen", torch_dtype=torch.bfloat16).to("cuda")