microsoft/qlib · error · ValueError

model is not fitted yet!

Error message

model is not fitted yet!

What it means

Raised by SANDWICH model predict() in qlib/contrib/model/pytorch_sandwich.py:362 when self.fitted is False. fitted is set True only at the end of a successful fit(); predict() refuses to run inference on an untrained model. This mirrors the guard used across all qlib contrib pytorch models.

Source

Thrown at qlib/contrib/model/pytorch_sandwich.py:362

                stop_steps = 0
                best_epoch = step
                best_param = copy.deepcopy(self.sandwich_model.state_dict())
            else:
                stop_steps += 1
                if stop_steps >= self.early_stop:
                    self.logger.info("early stop")
                    break

        self.logger.info("best score: %.6lf @ %d" % (best_score, best_epoch))
        self.sandwich_model.load_state_dict(best_param)
        torch.save(best_param, save_path)

        if self.use_gpu:
            torch.cuda.empty_cache()

    def predict(self, dataset: DatasetH, segment: Union[Text, slice] = "test"):
        if not self.fitted:
            raise ValueError("model is not fitted yet!")

        x_test = dataset.prepare(segment, col_set="feature", data_key=DataHandlerLP.DK_I)
        index = x_test.index
        self.sandwich_model.eval()
        x_values = x_test.values
        sample_num = x_values.shape[0]
        preds = []

        for begin in range(sample_num)[:: self.batch_size]:
            if sample_num - begin < self.batch_size:
                end = sample_num
            else:
                end = begin + self.batch_size
            x_batch = torch.from_numpy(x_values[begin:end]).float().to(self.device)
            with torch.no_grad():
                pred = self.sandwich_model(x_batch).detach().cpu().numpy()
            preds.append(pred)

View on GitHub (pinned to 79633dd950)

Solutions

  1. Ensure model.fit(dataset) ran to completion (check for 'best score: ... @ ...' in logs) before predict().
  2. Fix any earlier fit()-time exception first — this error is only a symptom.
  3. Use the model's save/load API for checkpoint reuse instead of relying on half-fit objects.

Example fix

model.fit(dataset)                 # must log 'best score: ...' before this flag is set
preds = model.predict(dataset, segment="test")
Defensive patterns

Strategy: validation

Validate before calling

if not model.fitted:
    raise RuntimeError("fit() must complete before predict(); check training logs for failures")
preds = model.predict(dataset, segment="test")

Type guard

def is_fitted(model) -> bool:
    return bool(getattr(model, "fitted", False))

Try / catch

try:
    preds = model.predict(dataset, segment)
except ValueError as e:
    if "not fitted" in str(e):
        raise RuntimeError("Training did not complete; inspect earlier fit() errors") from e
    raise

Prevention

When it happens

Trigger: Calling predict() before fit(); calling predict() after fit() aborted early (e.g. the empty-data ValueError above, NaN loss, CUDA OOM) since fitted is never set on the failure path.

Common situations: Workflow 'record' task run with a model that failed silently in a prior step; interactive sessions where fit raised and the user retries predict; pickle round-trips of partially trained models.

Related errors

AI-assisted analysis of microsoft/qlib@79633dd950 (2026-08-15). Data as JSON: /api/errors/c850e2eff110ce2a. Report an issue: GitHub.