Spaces:
Running
Running
File size: 3,648 Bytes
ae4e2a6 |
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 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 |
"""
Model Loader
============
Responsible for loading and initializing models (Single Responsibility)
"""
import torch
from transformers import AutoModel, AutoTokenizer
from typing import Tuple
import logging
from app.models.phobert_model import PhoBERTFineTuned
from app.core.config import settings
from app.core.exceptions import ModelNotLoadedException
logger = logging.getLogger(__name__)
class ModelLoader:
"""
Model loader service
Responsibilities:
- Load tokenizer
- Load base model
- Load fine-tuned weights
- Initialize model on correct device
"""
def __init__(self):
self._model: PhoBERTFineTuned | None = None
self._tokenizer: AutoTokenizer | None = None
self._device: torch.device | None = None
def load(self) -> Tuple[PhoBERTFineTuned, AutoTokenizer, torch.device]:
"""
Load model, tokenizer, and set device
Returns:
model: Loaded model
tokenizer: Loaded tokenizer
device: Device (CPU/CUDA)
Raises:
ModelNotLoadedException: If loading fails
"""
try:
# Set device
self._device = torch.device(settings.DEVICE)
logger.info(f"Using device: {self._device}")
# Load tokenizer
logger.info(f"Loading tokenizer: {settings.MODEL_NAME}")
self._tokenizer = AutoTokenizer.from_pretrained(settings.MODEL_NAME)
# Load base model
logger.info(f"Loading base model: {settings.MODEL_NAME}")
phobert = AutoModel.from_pretrained(settings.MODEL_NAME)
# Initialize fine-tuned model
logger.info("Initializing fine-tuned model")
self._model = PhoBERTFineTuned(
embedding_model=phobert,
hidden_dim=768,
dropout=0.3,
num_classes=2,
num_layers_to_finetune=4,
pooling='mean'
)
# Load weights
logger.info(f"Loading weights from: {settings.MODEL_PATH}")
state_dict = torch.load(
settings.MODEL_PATH,
map_location=self._device
)
self._model.load_state_dict(state_dict)
# Move to device and set eval mode
self._model = self._model.to(self._device)
self._model.eval()
logger.info("Model loaded successfully")
return self._model, self._tokenizer, self._device
except Exception as e:
logger.error(f"Failed to load model: {str(e)}")
raise ModelNotLoadedException()
@property
def model(self) -> PhoBERTFineTuned:
"""Get loaded model"""
if self._model is None:
raise ModelNotLoadedException()
return self._model
@property
def tokenizer(self) -> AutoTokenizer:
"""Get loaded tokenizer"""
if self._tokenizer is None:
raise ModelNotLoadedException()
return self._tokenizer
@property
def device(self) -> torch.device:
"""Get device"""
if self._device is None:
raise ModelNotLoadedException()
return self._device
def is_loaded(self) -> bool:
"""Check if model is loaded"""
return all([
self._model is not None,
self._tokenizer is not None,
self._device is not None
])
# Singleton instance
model_loader = ModelLoader()
|