informer_training.py
Window construction and JIT/scan training for Informer.
build_windows
informer_training.build_windows
Rolling windows [n_windows, input_size+h] of y (step 1).
| Parameter | Type | Default | Description |
|---|---|---|---|
y |
jnp.ndarray |
- | (undocumented) |
input_size |
int |
- | (undocumented) |
h |
int |
- | (undocumented) |
Returns: jnp.ndarray
Raises: ValueError
build_exog_windows
informer_training.build_exog_windows
Rolling windows of an exog array [T, F].
| Parameter | Type | Default | Description |
|---|---|---|---|
arr |
jnp.ndarray |
- | (undocumented) |
input_size |
int |
- | (undocumented) |
h |
int |
- | (undocumented) |
n_windows |
int |
- | (undocumented) |
span |
str |
- | span="input" -> [n, input_size, F] (encoder window); span="full" -> [n, input_size+h, F] (future-known spanning input + horizon). |
Returns: jnp.ndarray
forward_loss
informer_training.forward_loss
Scale, forward, and reduce a point/quantile loss in scaled space.
| Parameter | Type | Default | Description |
|---|---|---|---|
net |
- | - | (undocumented) |
y_windows |
- | - | (undocumented) |
h |
- | - | (undocumented) |
input_size |
- | - | (undocumented) |
scaler |
- | - | (undocumented) |
loss_fn |
- | - | (undocumented) |
futr_windows |
- | None |
(undocumented) |
sample_key |
- | - | (undocumented) |
deterministic |
bool |
False |
(undocumented) |
train
informer_training.train
Train net in place via one nnx.scan. Returns per-step losses.
Two independent RNG streams are split off seed: batch-index keys (NF's regime-dependent window sampling, below) and ProbSparse attn_keys (one per scan step, threaded through to every attention site inside InformerNet) -- so shuffling the training batches never perturbs which ProbSparse queries get sampled for a given step.
| Parameter | Type | Default | Description |
|---|---|---|---|
net |
- | - | (undocumented) |
y |
- | - | (undocumented) |
h |
- | - | (undocumented) |
input_size |
- | - | (undocumented) |
max_steps |
- | - | (undocumented) |
windows_batch_size |
- | - | (undocumented) |
lr |
- | - | (undocumented) |
seed |
- | - | (undocumented) |
loss_fn |
- | - | (undocumented) |
scaler |
- | - | (undocumented) |
futr_exog |
- | None |
(undocumented) |
Raises: RuntimeError
predict_step
informer_training.predict_step
Forecast next h steps from the final input_size of the series.
| Parameter | Type | Default | Description |
|---|---|---|---|
net |
- | - | (undocumented) |
y_context |
- | - | (undocumented) |
h |
- | - | (undocumented) |
input_size |
- | - | (undocumented) |
scaler |
- | - | (undocumented) |
futr_full |
- | None |
futr_full is the [input_size+h, F] future-known window (history + horizon). |
Returns: jnp.ndarray (Returns [h, multiplier] in the original scale.)