feat: clip text encode

src/domain/model/__init__.py

@@ -1,9 +1,11 @@
 from ._autoencoder import Autoencoder
 from ._clip import Clip
+from ._conditioning import Conditioning
 from ._pipeline import Pipeline
 
 __all__ = [
     "Pipeline",
     "Autoencoder",
     "Clip",
+    "Conditioning",
 ]

src/domain/model/_conditioning.py

@@ -0,0 +1,9 @@
+from dataclasses import dataclass
+
+import torch
+
+
+@dataclass
+class Conditioning:
+    conditioning: torch.Tensor
+    path: str

src/domain/repositories/__init__.py

@@ -1,4 +1,5 @@
 from ._autoencoder_repository import AutoencoderRepository
 from ._pipeline_repository import PipelineRepository
+from ._text_encoder_repository import TextEncoderRepository
 
-__all__ = ["PipelineRepository", "AutoencoderRepository"]
+__all__ = ["PipelineRepository", "AutoencoderRepository", "TextEncoderRepository"]

src/domain/repositories/_text_encoder_repository.py

@@ -0,0 +1,14 @@
+from __future__ import annotations
+
+from abc import ABC, abstractmethod
+
+import torch
+from transformers import CLIPTextModel, CLIPTokenizer
+
+
+class TextEncoderRepository(ABC):
+    @abstractmethod
+    def encode(
+        self, tokenizer: CLIPTokenizer, text_encoder: CLIPTextModel, text: str
+    ) -> torch.Tensor:
+        pass