microsoft/qlib · error · ValueError

unknown loss `%s`

Error message

unknown loss `%s`

What it means

Thrown by TransformerModel.loss_fn. Only loss='mse' is implemented (computed on the non-NaN label mask); every other value of the `loss` hyperparameter reaches the terminal ValueError on the first training or validation batch.

Source

Thrown at qlib/contrib/model/pytorch_transformer.py:94

        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, x_train, y_train):
        x_train_values = x_train.values
        y_train_values = np.squeeze(y_train.values)

        self.model.train()

        indices = np.arange(len(x_train_values))
        np.random.shuffle(indices)

View on GitHub (pinned to 79633dd950)

Solutions

  1. Set loss='mse' exactly (lowercase) in the TransformerModel config.
  2. For custom losses, subclass TransformerModel and extend loss_fn, preserving the ~torch.isnan(label) mask.

Example fix

# before
model = TransformerModel(..., loss="MSE")

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

Strategy: validation

Validate before calling

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

Try / catch

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

Prevention

When it happens

Trigger: TransformerModel(..., loss='mae'|'huber'|'MSE') then fit(); train_epoch immediately calls loss_fn and raises.

Common situations: Assuming uppercase 'MSE' matches (it does not — comparison is exact lowercase); porting loss names from other frameworks or qlib models.

Related errors

AI-assisted analysis of microsoft/qlib@79633dd950 (2026-08-15). Data as JSON: /api/errors/39c61f0a6b748a09. Report an issue: GitHub.