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

Use 'adam' or 'gd' (the model's name for plain SGD), matched case-insensitively.

microsoft/qlib · error · NotImplementedError

optimizer {} is not supported!

Error message

optimizer {} is not supported!

What it means

GATsModel.fit() constructs the optimizer from the optimizer hyperparameter with only two branches: 'adam' (case-insensitive) maps to torch.optim.Adam and 'gd' maps to torch.optim.SGD. Any other string raises NotImplementedError. This happens during fit() before any training begins, so the model is cheap to recover from: fix the string and call fit() again.

Source

Thrown at qlib/contrib/model/pytorch_gats.py:135

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

        self.GAT_model = GATModel(
            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.GAT_model))
        self.logger.info("model size: {:.4f} MB".format(count_parameters(self.GAT_model)))

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

        self.fitted = False
        self.GAT_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 'adam' or 'gd' (the model's name for plain SGD), matched case-insensitively.
If you wrote 'sgd', change it to 'gd'.
For a different optimizer, subclass GATsModel and add a branch constructing the torch.optim class you need.

Example fix

# before
model = GATsModel(optimizer='sgd')

# after
model = GATsModel(optimizer='gd')

Defensive patterns

Strategy: validation

Validate before calling

assert optimizer.lower() in ('adam', 'gd'), f"optimizer must be 'adam' or 'gd', got {optimizer!r}"

Type guard

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

Try / catch

try:
    model.fit(dataset)
except NotImplementedError as e:
    if 'optimizer' in str(e):
        # fall back to the default Adam and retry
        model = GATsModel(optimizer='adam')
        model.fit(dataset)
    else:
        raise

Prevention

Remember 'gd' (not 'sgd') is this family's name for SGD.
Validate optimizer names against the fixed allowlist before training.
Encode the allowlist in your workflow config schema to catch typos early.

When it happens

Trigger: Calling GATsModel(...).fit(dataset) with optimizer set to 'sgd' (note: the code expects 'gd' for SGD), 'adagrad', 'rmsprop', or 'adamw'.

Common situations: The 'sgd' vs 'gd' naming trap is the most common hit: developers naturally write 'sgd' and get this error; copying optimizer names from other qlib models or sklearn-style configs; wanting AdamW for weight decay and finding it unsupported.

Related errors

AI-assisted analysis of microsoft/qlib@79633dd950 (2026-08-15). Data as JSON: /api/errors/37767fc1e0918a1b. Report an issue: GitHub.