Upload audiodit/__init__.py with huggingface_hub

audiodit/__init__.py

@@ -0,0 +1,37 @@
+"""AudioDiT — Conditional Flow Matching TTS with DiT backbone.
+
+Auto-registers AudioDiTConfig and AudioDiTModel with HuggingFace Transformers
+so that ``AutoConfig.from_pretrained`` and ``AutoModel.from_pretrained`` work
+out of the box.
+
+Usage:
+    import audiodit  # triggers auto-registration
+    from audiodit import AudioDiTConfig, AudioDiTModel
+
+    model = AudioDiTModel.from_pretrained("path/to/hf_audiodit_1b")
+"""
+
+from .configuration_audiodit import AudioDiTConfig, AudioDiTVaeConfig
+from .modeling_audiodit import (
+    AudioDiTModel,
+    AudioDiTPreTrainedModel,
+    AudioDiTTransformer,
+    AudioDiTVae,
+    AudioDiTOutput,
+)
+
+# Auto-register with transformers so AutoConfig/AutoModel work
+from transformers import AutoConfig, AutoModel
+
+AutoConfig.register("audiodit", AudioDiTConfig, exist_ok=True)
+AutoModel.register(AudioDiTConfig, AudioDiTModel, exist_ok=True)
+
+__all__ = [
+    "AudioDiTConfig",
+    "AudioDiTVaeConfig",
+    "AudioDiTModel",
+    "AudioDiTPreTrainedModel",
+    "AudioDiTTransformer",
+    "AudioDiTVae",
+    "AudioDiTOutput",
+]