microsoft/qlib · error · NotImplementedError

optimizer {} is not supported!

Error message

optimizer {} is not supported!

What it means

LSTMModel's constructor wires up exactly two optimizers: 'adam' (optim.Adam, lr only) and 'gd' (optim.SGD, lr only) — note neither applies weight decay in this model. Any other optimizer string raises NotImplementedError('optimizer {} is not supported!') during __init__, so the model is unusable until fixed.

Source

Thrown at qlib/contrib/model/pytorch_lstm.py:123

            )
        )

        if self.seed is not None:
            np.random.seed(self.seed)
            torch.manual_seed(self.seed)

        self.lstm_model = LSTMModel(
            d_feat=self.d_feat,
            hidden_size=self.hidden_size,
            num_layers=self.num_layers,
            dropout=self.dropout,
        )
        if optimizer.lower() == "adam":
            self.train_optimizer = optim.Adam(self.lstm_model.parameters(), lr=self.lr)
        elif optimizer.lower() == "gd":
            self.train_optimizer = optim.SGD(self.lstm_model.parameters(), lr=self.lr)
        else:
            raise NotImplementedError("optimizer {} is not supported!".format(optimizer))

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

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

    def mse(self, pred, label):
        loss = (pred - label) ** 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])

View on GitHub (pinned to 79633dd950)

Solutions

  1. Change the model kwarg to optimizer='adam' or optimizer='gd'.
  2. For other optimizers, subclass, call super().__init__(), then set self.train_optimizer = optim.<Opt>(self.lstm_model.parameters(), lr=self.lr, ...).
  3. Double-check spelling; matching is lowercased but exact.

Example fix

# before
model = LSTMModel(..., optimizer="sgd")  # NotImplementedError

# after
model = LSTMModel(..., optimizer="adam")  # or "gd" for SGD
Defensive patterns

Strategy: validation

Validate before calling

assert optimizer.lower() in {"adam", "gd"}, "LSTMModel supports only 'adam' and 'gd'"
model = LSTMModel(..., optimizer=optimizer)

Type guard

def is_supported_optimizer(name: str) -> bool:
    return isinstance(name, str) and name.lower() in {"adam", "gd"}

Try / catch

try:
    model = LSTMModel(..., optimizer=opt)
except NotImplementedError as e:
    raise ValueError(f"{e} — use 'adam' or 'gd'") from e

Prevention

When it happens

Trigger: Constructing LSTMModel(..., optimizer=X) in qlib/contrib/model/pytorch_lstm.py with X.lower() not in {'adam','gd'} — e.g. 'sgd', 'adamw', 'rmsprop', or a typo.

Common situations: Reusing YAML from another contrib model; expecting 'sgd' as the SGD keyword when this codebase uses 'gd'; copy-paste of optimizer names from raw PyTorch examples.

Related errors

AI-assisted analysis of microsoft/qlib@79633dd950 (2026-08-15). Data as JSON: /api/errors/573848b44cf32bd9. Report an issue: GitHub.