unslothai/unsloth · error · HTTPException

Export is not supported on this platform.

Error message

Export is not supported on this platform.

What it means

HTTP 400 raised when utils.hardware.export_capability reports export_supported=False for this host. The capability probe (run off the event loop because a cold import can be slow) decides whether the platform/GPU/driver stack can run export at all; the message comes from export_unsupported_message or falls back to this generic string. Read-only endpoints are not gated so the UI can still show the reason.

Source

Thrown at studio/backend/routes/export.py:70

    Keeps the backend authoritative even if a client bypasses the UI gate. Read-only endpoints
    (scan/status/logs) are intentionally NOT gated so the Export page can still render the reason.
    Also refuses (409) while a latest-transformers install is swapping .venv_t5_latest: an
    export worker spawned mid-swap could activate a half-replaced sidecar.
    """
    from utils.transformers_latest import is_install_in_progress

    if is_install_in_progress():
        raise HTTPException(
            status_code = 409,
            detail = "A transformers installation is in progress. Retry when it completes.",
        )

    from utils.hardware import export_capability

    # Off-loop: detection is deferred past bind, so the first call can wait on a cold import.
    cap = await asyncio.to_thread(export_capability)
    if not cap.get("export_supported", True):
        raise HTTPException(
            status_code = 400,
            detail = cap.get("export_unsupported_message")
            or "Export is not supported on this platform.",
        )


@router.post("/load-checkpoint", response_model = ExportOperationResponse)
async def load_checkpoint(
    request: LoadCheckpointRequest, current_subject: str = Depends(get_current_subject)
):
    """Load a checkpoint into the export backend (ExportBackend.load_checkpoint).

    Export runs in its own subprocess and is allowed to run in parallel with
    training and inference. We deliberately do NOT stop training or unload the
    chat model here -- if the GPU runs out of memory the load/export fails with
    a clear error instead of tearing down the user's other running workloads.
    """
    try:

View on GitHub (pinned to 203007d190)

Solutions

1. Call GET /export/status or the capability probe first and read export_unsupported_message for the precise reason.
2. Fix the underlying hardware/software gap (install CUDA-capable drivers, enable GPU passthrough, or install the missing export toolchain).
3. If export is not needed on this host, disable the Export page / hide mutating controls based on the read-only status response.

const cap = await api.get('/export/capability'); // or read from /export/status
if (!cap.export_supported) {
  renderUnsupported(cap.export_unsupported_message);
  return;
}

When it happens

Trigger: POST to any mutating export endpoint on a machine whose hardware probe fails — e.g. no CUDA device, an unsupported GPU/driver combo, or a missing export-only dependency detected by export_capability().

Common situations: Running the studio backend on a CPU-only box or in a container without GPU passthrough; driver/toolkit versions that the capability check rejects; a fresh environment where the export stack was never installed.

Related errors

• {message}
• Unsupported format. Use webm or gif.
• GGUF conversion produced a symlink, refusing to relocate it:
• GGUF conversion produced no files: no .gguf outputs for {abs
• GGUF conversion produced a symlinked Modelfile, refusing to

AI-assisted analysis of unslothai/unsloth@203007d190 (2026-08-15). Data as JSON: /api/errors/f1dae23aa8e52aab. Report an issue: GitHub.