unslothai/unsloth · error · HTTPException

dataset_streaming is not supported for embedding training; t

Error message

dataset_streaming is not supported for embedding training; the embedding loader needs the full dataset.

What it means

HTTP 400 when dataset_streaming=true and is_embedding is set: the embedding training loader requires the full dataset in memory (it needs complete pass structure), which conflicts with streaming's lazy iteration.

Source

Thrown at studio/backend/routes/training.py:1337

        from utils.hardware import hardware as _hw
        from utils.hardware import ensure_hardware_detected

        await asyncio.to_thread(ensure_hardware_detected)
        _validate_training_platform(request)

        if request.dataset_streaming:
            if not request.hf_dataset:
                raise HTTPException(
                    status_code = 400,
                    detail = "dataset_streaming requires hf_dataset; streaming is not supported for local datasets.",
                )
            if request.is_dataset_image or request.is_dataset_audio:
                raise HTTPException(
                    status_code = 400,
                    detail = "dataset_streaming is not supported for vision or audio datasets.",
                )
            if request.is_embedding:
                raise HTTPException(
                    status_code = 400,
                    detail = "dataset_streaming is not supported for embedding training; the embedding loader needs the full dataset.",
                )
            if _hw.DEVICE == _hw.DeviceType.MLX:
                raise HTTPException(
                    status_code = 400,
                    detail = "dataset_streaming is not yet supported on Apple Silicon (MLX); the MLX loader materializes the full dataset.",
                )
            if request.max_steps is None or request.max_steps <= 0:
                raise HTTPException(
                    status_code = 422,
                    detail = "dataset_streaming requires max_steps > 0 because streaming datasets have no known length.",
                )
            if request.train_on_completions:
                raise HTTPException(
                    status_code = 422,
                    detail = "dataset_streaming is not supported with train_on_completions yet.",
                )

View on GitHub (pinned to 203007d190)

Solutions

Disable dataset_streaming for embedding training
Reduce dataset size or use a machine with more RAM for large embedding datasets
Reserve streaming for standard text fine-tuning runs only

Example fix

// before
{"dataset_streaming": true, "is_embedding": true}  // 400

// after
{"dataset_streaming": false, "is_embedding": true}

Defensive patterns

Strategy: validation

Validate before calling

def streaming_config_valid(p: dict) -> bool:
    if not p.get("dataset_streaming"):
        return True
    return not p.get("is_embedding")

Try / catch

resp = client.post("/training/start", payload)
if resp.status_code == 400 and "embedding" in resp.text:
    payload["dataset_streaming"] = False
    resp = client.post("/training/start", payload)

Prevention

Never enable streaming for embedding training runs
Budget RAM for the full dataset when training embeddings
Encode the embedding/streaming exclusion in your config schema validation

When it happens

Trigger: POST /training/start with dataset_streaming: true and is_embedding: true.

Common situations: Applying a streaming config preset to embedding fine-tuning; assuming streaming works uniformly across all training modes.

Related errors

AI-assisted analysis of unslothai/unsloth@203007d190 (2026-08-15). Data as JSON: /api/errors/48184289fb3f5b91. Report an issue: GitHub.