|
|
"""plm.py. |
|
|
|
|
|
File for providing the Plm model implementation. |
|
|
""" |
|
|
|
|
|
import logging |
|
|
|
|
|
import torch |
|
|
from transformers import AutoModelForImageTextToText, AutoProcessor |
|
|
from transformers.feature_extraction_utils import BatchFeature |
|
|
|
|
|
from src.models.base import ModelBase |
|
|
from src.models.config import Config |
|
|
|
|
|
|
|
|
class PlmModel(ModelBase): |
|
|
"""PLM model implementation.""" |
|
|
|
|
|
def __init__(self, config: Config) -> None: |
|
|
"""Initialization of the PLM model. |
|
|
|
|
|
Args: |
|
|
config (Config): Parsed config |
|
|
""" |
|
|
|
|
|
super().__init__(config) |
|
|
|
|
|
def _load_specific_model(self) -> None: |
|
|
"""Overridden function to populate self.model.""" |
|
|
self.model = AutoModelForImageTextToText.from_pretrained( |
|
|
self.model_path, **self.config.model |
|
|
) if hasattr(self.config, 'model') else ( |
|
|
AutoModelForImageTextToText.from_pretrained( |
|
|
self.model_path |
|
|
) |
|
|
) |
|
|
self.model.to(self.config.device) |
|
|
|
|
|
def _init_processor(self) -> None: |
|
|
"""Initialize the self.processor by loading from the path.""" |
|
|
self.processor = AutoProcessor.from_pretrained(self.model_path, use_fast=True) |
|
|
|
|
|
def _forward(self, data: BatchFeature) -> None: |
|
|
"""Given some input data, performs a single forward pass. |
|
|
|
|
|
This function itself can be overriden, while _hook_and_eval |
|
|
should be left in tact. |
|
|
|
|
|
Args: |
|
|
data (BatchFeature): The given data tensor. |
|
|
""" |
|
|
data.to(self.config.device) |
|
|
with torch.no_grad(): |
|
|
_ = self.model.generate(**data, **self.config.forward) |
|
|
logging.debug('Completed forward pass...') |
|
|
|
