Skip to content

Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4.1, MiniMax-H3, Gemma 4, FLUX and more.

Studio: train audio datasets on the columns the dataset check found (#11850)

Nilayed55faabd7da
diff --git a/studio/backend/core/training/trainer.py b/studio/backend/core/training/trainer.py
index 53902fed94c..420ee49d637 100644
--- a/studio/backend/core/training/trainer.py
+++ b/studio/backend/core/training/trainer.py
@@ -859,6 +859,14 @@ class UnslothTrainer:
elif "speaker_id" in cols:
speaker_col = "speaker_id"
+ if audio_col is None or text_col is None or speaker_col is None:
+ from hub.utils.dataset_format import detect_multimodal_dataset
+
+ detected = detect_multimodal_dataset(dataset)
+ audio_col = audio_col or detected["detected_audio_column"]
+ text_col = text_col or detected["detected_text_column"]
+ speaker_col = speaker_col or detected["detected_speaker_column"]
+
return {
"audio_col": audio_col,
"text_col": text_col,
diff --git a/studio/backend/tests/test_audio_detected_columns.py b/studio/backend/tests/test_audio_detected_columns.py
new file mode 100644
index 00000000000..89ae55046d2
--- /dev/null
+++ b/studio/backend/tests/test_audio_detected_columns.py
@@ -0,0 +1,112 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""The trainer trains on the audio and text columns the dataset check accepted."""
+
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+
+_BACKEND = Path(__file__).resolve().parent.parent
+if str(_BACKEND) not in sys.path:
+ sys.path.insert(0, str(_BACKEND))
+
+import importlib # noqa: E402
+import types # noqa: E402
+from unittest.mock import MagicMock # noqa: E402
+
+import pytest # noqa: E402
+from datasets import Dataset # noqa: E402
+
+from hub.utils.dataset_format import check_dataset_format # noqa: E402
+
+
+_STUBBED: list[str] = []
+
+
+def _stub_if_missing(name, attrs):
+ if name in sys.modules:
+ return
+ try:
+ importlib.import_module(name)
+ return
+ except Exception: # noqa: BLE001 - stub unusable imports
+ pass
+ _STUBBED.append(name)
+ mod = types.ModuleType(name)
+ mod.__spec__ = None
+ for attr in attrs:
+ setattr(mod, attr, MagicMock())
+ sys.modules[name] = mod
+ parent, _, child = name.rpartition(".")
+ if parent and parent in sys.modules:
+ setattr(sys.modules[parent], child, mod)
+
+
+_stub_if_missing("unsloth", ("FastLanguageModel", "FastVisionModel", "is_bfloat16_supported"))
+_stub_if_missing("unsloth.chat_templates", ("get_chat_template",))
+_stub_if_missing("trl", ("SFTTrainer", "SFTConfig"))
+
+from core.training import trainer as tmod # noqa: E402
+
+for _name in reversed(_STUBBED):
+ sys.modules.pop(_name, None)
+
+
+_AUDIO = {"array": [0.0] * 160, "sampling_rate": 16000}
+
+
+def audio_trainer(monkeypatch):
+ monkeypatch.setattr(tmod, "should_use_mlx_training_backend", lambda *a, **k: False)
+ return tmod.UnslothTrainer()
+
+
+ "row, expected",
+ [
+ ({"audio": _AUDIO, "normalized_text": "hi"}, ("audio", "normalized_text", None)),
+ ({"audio": _AUDIO, "prompt": "hi"}, ("audio", "prompt", None)),
+ ({"audio": _AUDIO, "label": "hi"}, ("audio", "label", None)),
+ ({"wav": _AUDIO, "text": "hi"}, ("wav", "text", None)),
+ ({"clip": _AUDIO, "text": "hi", "speaker": "a"}, ("clip", "text", "speaker")),
+ ({"audio": _AUDIO, "text": "hi", "speaker": "a"}, ("audio", "text", "speaker")),
+ ],
+)
+def test_resolves_the_columns_the_check_accepted(audio_trainer, row, expected):
+ dataset = Dataset.from_list([row])
+ check = check_dataset_format(dataset, is_vlm = False)
+ assert check["requires_manual_mapping"] is False
+
+ resolved = audio_trainer._resolve_audio_columns(dataset, None)
+
+ assert (resolved["audio_col"], resolved["text_col"], resolved["speaker_col"]) == expected
+ assert resolved["audio_col"] == check["detected_audio_column"]
+
+
+def test_exact_names_and_explicit_mapping_are_unchanged(audio_trainer):
+ dataset = Dataset.from_list(
+ [{"audio": _AUDIO, "text": "hi", "prompt": "p", "speaker": "a", "source": "0"}]
+ )
+
+ assert audio_trainer._resolve_audio_columns(dataset, None) == {
+ "audio_col": "audio",
+ "text_col": "text",
+ "speaker_col": "source",
+ }
+ assert audio_trainer._resolve_audio_columns(dataset, {"audio": "audio", "prompt": "text"}) == {
+ "audio_col": "audio",
+ "text_col": "prompt",
+ "speaker_col": None,
+ }
+
+
+def test_text_only_dataset_still_resolves_nothing(audio_trainer):
+ dataset = Dataset.from_list([{"instruction": "a", "output": "b"}])
+
+ assert audio_trainer._resolve_audio_columns(dataset, None) == {
+ "audio_col": None,
+ "text_col": None,
+ "speaker_col": None,
+ }