configs refactor part 1/3: split _configs.py into a package

_configs.py grew to 2591 lines (9 sub-configs, 19 per-model configs,
3 mega-extractor switch functions, a 530-line ArchitectureConfig.
from_transformers). This is the first of three mechanical refactors
that carve it into a scalable package layout. No behavior change in
this PR — every public name is still importable from mobius._configs.

src/mobius/_configs/
├── __init__.py        # re-exports everything that was in _configs.py
├── _sub_configs.py    # pure-data dataclasses (RoPE/Vision/Audio/Codec/TTS)
├── _quantization.py   # QuantizationConfig + from_transformers
└── _base.py           # BaseModelConfig, ArchitectureConfig, per-model
                      # subclasses, and the _extract_* helpers

Follow-up PRs in this series:
  Part 2/3 — convert the _extract_audio_config / _extract_vision_config
             model_type switches into a decorator-registered dispatch
             so new models add a file under per_model/ instead of a
             branch in the central function.
  Part 3/3 — move per-model config subclasses (Gemma2Config,
             MllamaConfig, NemotronHConfig, ...) into per_model/ and
             carve up ArchitectureConfig.from_transformers.

Tests: 2769 passed (full src/ + tests/build_graph_test.py + cli_test.py).
Ruff: clean.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Signed-off-by: Justin Chu <justinchu@microsoft.com>

Author: justinchuby

src/mobius/_configs/__init__.py

@@ -0,0 +1,128 @@
+# Copyright (c) Microsoft Corporation.
+# Licensed under the MIT License.
+
+"""Model configuration dataclasses and HuggingFace → mobius conversion.
+
+This package re-exports the public surface that used to live in
+``mobius._configs`` (a single 2.6k-line module). The split keeps that
+import path working while letting each concern grow in its own file:
+
+* :mod:`._sub_configs` — pure-data dataclasses for sub-configs
+  (RoPE, vision, audio, codec, TTS) shared across model families.
+* :mod:`._quantization` — :class:`QuantizationConfig` and its HF parser.
+* :mod:`._base` — :class:`BaseModelConfig`, :class:`ArchitectureConfig`,
+  the per-model config subclasses, and the legacy ``_extract_*``
+  helpers that build sub-configs from HuggingFace fields.
+
+Follow-up PRs in this series convert the model-type switches inside the
+``_extract_*`` helpers into a registry-based dispatch, and move per-model
+config subclasses into their own files under ``per_model/``.
+"""
+
+from __future__ import annotations
+
+from mobius._configs._base import (
+    DEFAULT_INT,
+    ArchitectureConfig,
+    BambaConfig,
+    BaseModelConfig,
+    CausalLMConfig,
+    DepthAnythingConfig,
+    EncoderConfig,
+    Gemma2Config,
+    Gemma3nConfig,
+    Gemma4Config,
+    GraniteMoeHybridConfig,
+    JambaConfig,
+    JetMoeConfig,
+    LongcatFlashConfig,
+    Mamba2Config,
+    MambaConfig,
+    MllamaConfig,
+    NanoChatConfig,
+    NemotronHConfig,
+    Sam2Config,
+    SegformerConfig,
+    VisionLanguageConfig,
+    WhisperConfig,
+    YolosConfig,
+    Zamba2Config,
+    _as_int,
+    _extract_audio_config,
+    _extract_mrope_fields,
+    _extract_rope_config,
+    _extract_vision_config,
+    _first,
+    _first_not_none,
+    _nested_rope_theta,
+    _nested_rope_type,
+    _normalize_rope_scaling,
+    _resolve_dtype,
+    _resolve_hidden_act,
+    _shallow_fields,
+    _shared_expert_size,
+)
+from mobius._configs._quantization import QuantizationConfig
+from mobius._configs._sub_configs import (
+    AudioConfig,
+    CodecDecoderConfig,
+    CodecEncoderConfig,
+    CodePredictorConfig,
+    Gemma4AudioConfig,
+    RoPEConfig,
+    SpeakerEncoderConfig,
+    TTSConfig,
+    VisionConfig,
+)
+
+__all__ = [
+    "DEFAULT_INT",
+    "ArchitectureConfig",
+    "AudioConfig",
+    "BambaConfig",
+    "BaseModelConfig",
+    "CausalLMConfig",
+    "CodePredictorConfig",
+    "CodecDecoderConfig",
+    "CodecEncoderConfig",
+    "DepthAnythingConfig",
+    "EncoderConfig",
+    "Gemma2Config",
+    "Gemma3nConfig",
+    "Gemma4AudioConfig",
+    "Gemma4Config",
+    "GraniteMoeHybridConfig",
+    "JambaConfig",
+    "JetMoeConfig",
+    "LongcatFlashConfig",
+    "Mamba2Config",
+    "MambaConfig",
+    "MllamaConfig",
+    "NanoChatConfig",
+    "NemotronHConfig",
+    "QuantizationConfig",
+    "RoPEConfig",
+    "Sam2Config",
+    "SegformerConfig",
+    "SpeakerEncoderConfig",
+    "TTSConfig",
+    "VisionConfig",
+    "VisionLanguageConfig",
+    "WhisperConfig",
+    "YolosConfig",
+    "Zamba2Config",
+    "_as_int",
+    "_extract_audio_config",
+    "_extract_mrope_fields",
+    "_extract_rope_config",
+    "_extract_vision_config",
+    "_first",
+    "_first_not_none",
+    "_nested_rope_theta",
+    "_nested_rope_type",
+    "_normalize_rope_scaling",
+    "_resolve_dtype",
+    "_resolve_hidden_act",
+    "_shallow_fields",
+    "_shared_expert_size",
+]