microsoft/qlib · error · ValueError

unknown loss `%s`

Error message

unknown loss `%s`

What it means

Thrown by TransformerTSModel.loss_fn. The time-series Transformer implements exactly one loss, 'mse', evaluated over the non-NaN label mask; any other `loss` hyperparameter value reaches the terminal ValueError on the first batch.

Source

Thrown at qlib/contrib/model/pytorch_transformer_ts.py:92

        self.fitted = False
        self.model.to(self.device)

    @property
    def use_gpu(self):
        return self.device != torch.device("cpu")

    def mse(self, pred, label):
        loss = (pred.float() - label.float()) ** 2
        return torch.mean(loss)

    def loss_fn(self, pred, label):
        mask = ~torch.isnan(label)

        if self.loss == "mse":
            return self.mse(pred[mask], label[mask])

        raise ValueError("unknown loss `%s`" % self.loss)

    def metric_fn(self, pred, label):
        mask = torch.isfinite(label)

        if self.metric in ("", "loss"):
            return -self.loss_fn(pred[mask], label[mask])

        raise ValueError("unknown metric `%s`" % self.metric)

    def train_epoch(self, data_loader):
        self.model.train()

        for data in data_loader:
            feature = data[:, :, 0:-1].to(self.device)
            label = data[:, -1, -1].to(self.device)

            pred = self.model(feature.float())  # .float()
            loss = self.loss_fn(pred, label)

View on GitHub (pinned to 79633dd950)

Solutions

  1. Set loss='mse' (exact lowercase) — the only supported value.
  2. Subclass TransformerTSModel and override loss_fn to add other losses while keeping the NaN mask.

Example fix

# before
model = TransformerTSModel(..., loss="smoothl1")

# after
model = TransformerTSModel(..., loss="mse")
Defensive patterns

Strategy: validation

Validate before calling

assert model_kwargs.get("loss", "mse") == "mse", "TransformerTSModel supports only loss='mse'"

Try / catch

try:
    model.fit(ds, valid)
except ValueError as e:
    if "unknown loss" in str(e):
        model_kwargs["loss"] = "mse"
        model = TransformerTSModel(**model_kwargs)
        model.fit(ds, valid)
    else:
        raise

Prevention

When it happens

Trigger: TransformerTSModel(..., loss='mae'|'smoothl1'|'MSE') followed by fit(); train_epoch's loss_fn call raises immediately.

Common situations: Case-sensitive typo 'MSE'; hyperparameter dicts shared across models where another loss name was valid; assuming the loss vocabulary of the wider qlib/TRA stack applies here.

Related errors

AI-assisted analysis of microsoft/qlib@79633dd950 (2026-08-15). Data as JSON: /api/errors/db8843eabbdb3b38. Report an issue: GitHub.