What causes "optimizer {} is not supported!" in microsoft/qlib?

Use optimizer='adam' or optimizer='gd' (other capitalizations like 'Adam' are accepted because of .lower())

microsoft/qlib · error · NotImplementedError

optimizer {} is not supported!

Error message

optimizer {} is not supported!

What it means

IGMTFModel accepts only two optimizers, case-insensitively: 'adam' (optim.Adam) and 'gd' (optim.SGD). Any other optimizer string raises NotImplementedError at the end of __init__, so the object fails during construction, not during fit.

Source

Thrown at qlib/contrib/model/pytorch_igmtf.py:134

            np.random.seed(self.seed)
            torch.manual_seed(self.seed)

        self.igmtf_model = IGMTFModel(
            d_feat=self.d_feat,
            hidden_size=self.hidden_size,
            num_layers=self.num_layers,
            dropout=self.dropout,
            base_model=self.base_model,
        )
        self.logger.info("model:\n{:}".format(self.igmtf_model))
        self.logger.info("model size: {:.4f} MB".format(count_parameters(self.igmtf_model)))

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

        self.fitted = False
        self.igmtf_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

Use optimizer='adam' or optimizer='gd' (other capitalizations like 'Adam' are accepted because of .lower())
If you need a different optimizer, subclass IGMTFModel and add a branch creating it from torch.optim

Example fix

# before
IGMTFModel(optimizer="adamw")

# after
IGMTFModel(optimizer="adam")

Defensive patterns

Strategy: validation

Validate before calling

assert optimizer.lower() in ("adam", "gd"), f"unsupported optimizer {optimizer!r}; use 'adam' or 'gd'"

Type guard

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

Prevention

Keep a per-model allowlist of optimizer strings validated at config load
Remember 'gd' (not 'sgd') is the accepted spelling in qlib torch models

When it happens

Trigger: Constructing IGMTFModel(optimizer='adamw'), 'rmsprop', 'adagrad', or any string other than adam/gd (any case).

Common situations: Porting a config from a framework where 'adamw' or 'rmsprop' are standard; typos like 'Adamm'; copying hyperparameter blocks between models without checking each model's supported optimizer set.

Related errors

AI-assisted analysis of microsoft/qlib@79633dd950 (2026-08-15). Data as JSON: /api/errors/52bf822945ebe415. Report an issue: GitHub.