| from peft import PeftModel | |
| import torch | |
| import torch.nn as nn | |
| class MemGenTrigger(nn.Module): | |
| adapter_name = "trigger" | |
| def __init__( | |
| self, | |
| model: PeftModel, | |
| active: bool, | |
| ): | |
| super().__init__() | |
| self.active = active | |
| self.model = model | |
| self.output_layer = nn.Linear(model.base_model.config.hidden_size, 2) | |
| def forward( | |
| self, | |
| input_ids: torch.LongTensor, | |
| attention_mask: torch.LongTensor, | |
| position_ids: torch.Tensor | |
| ) -> torch.FloatTensor: | |
| if self.active: | |
| outputs = self.model( | |
| input_ids=input_ids, | |
| attention_mask=attention_mask, | |
| position_ids=position_ids, | |
| output_hidden_states=True, | |
| ) | |
| hidden_states = outputs.hidden_states[-1] | |
| logits = self.output_layer(hidden_states) | |
| else: | |
| batch_size, seq_len = input_ids.shape | |
| logits = torch.zeros(batch_size, seq_len, 2, device=input_ids.device) # logits: [batch_size, seq_len, 2] | |
| logits[..., 1] = 1.0 | |
| return logits | |