File size: 1,044 Bytes
9c60f47 | 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 26 27 28 29 30 31 32 | import torch.nn as nn
from transformers import AutoModel
class MultiTaskModel(nn.Module):
def __init__(self, encoder_name, num_category_labels, num_urgency_labels):
super().__init__()
self.encoder = AutoModel.from_pretrained(encoder_name)
hidden_size = self.encoder.config.hidden_size
# Changed from category_head to category_classifier
self.category_classifier = nn.Linear(hidden_size, num_category_labels)
# Changed from urgency_head to urgency_classifier
self.urgency_classifier = nn.Linear(hidden_size, num_urgency_labels)
def forward(self, input_ids, attention_mask):
outputs = self.encoder(
input_ids=input_ids,
attention_mask=attention_mask
)
pooled = outputs.last_hidden_state[:, 0]
return type(
"Output",
(),
{
"category_logits": self.category_classifier(pooled),
"urgency_logits": self.urgency_classifier(pooled),
}
)() |