Spaces:
Running
Running
File size: 2,926 Bytes
b4f3263 67bae95 b4f3263 |
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 33 34 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 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 |
from device_manager import DeviceManager
from transformers import AlbertModel, AlbertTokenizerFast
import torch.nn as nn
import torch
import numpy as np
class AlbertCustomClassificationHead(nn.Module):
def __init__(self, albert_model, num_additional_features=25, dropout_rate=0.1):
super(AlbertCustomClassificationHead, self).__init__()
self.albert_model = albert_model
self.dropout = nn.Dropout(dropout_rate)
self.classifier = nn.Linear(1024 + num_additional_features, 1)
def forward(self, input_ids, attention_mask, additional_features, labels=None):
albert_output = self.albert_model(
input_ids=input_ids, attention_mask=attention_mask).pooler_output
combined_features = torch.cat(
[albert_output, additional_features], dim=1)
dropout_output = self.dropout(combined_features)
logits = self.classifier(dropout_output)
if labels is not None:
loss_fn = nn.BCEWithLogitsLoss()
labels = labels.unsqueeze(1)
loss = loss_fn(logits, labels.float())
return logits, loss
else:
return logits
class PredictMainModel:
_instance = None
def __new__(cls):
if cls._instance is None:
cls._instance = super(PredictMainModel, cls).__new__()
cls._instance.initialize()
return cls._instance
def initialize(self):
self.model_name = "albert-large-v2"
self.tokenizer = AlbertTokenizerFast.from_pretrained(self.model_name)
self.albert_model = AlbertModel.from_pretrained(self.model_name)
self.device = DeviceManager()
self.model = AlbertCustomClassificationHead(
self.albert_model).to(self.device)
# TODO : CHANGE MODEL STATE DICT PATH
self.model.load_state_dict(torch.load("models/albert_model.pth"))
def preprocess_input(self, text: str, additional_features: np.ndarray):
encoding = self.tokenizer.encode_plus(
text,
add_special_tokens=True,
max_length=512,
return_token_type_ids=False,
padding="max_length",
truncation=True,
return_attention_mask=True,
return_tensors="pt"
)
additional_features_tensor = torch.tensor(
additional_features, dtype=torch.float)
return {
"input_ids": encoding["input_ids"].to(self.device),
"attention_mask": encoding["attention_mask"].to(self.device),
"additional_features": additional_features_tensor.to(self.device)
}
def predict(self, text: str, additional_features: np.ndarray) -> float:
self.model.eval()
with torch.no_grad():
data = self.preprocess_input(text, additional_features)
logits = self.model(**data)
return torch.sigmoid(logits).cpu().numpy()[0][0]
|