API Reference¶
benchmark(model_path, target='tic62124646', sector=13, exptime=120, author='SPOC', backends=None)
¶
Benchmark inference speed across available backends on a standard light curve.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
model_path
|
str
|
Path to a |
required |
target
|
str
|
TIC identifier (default: 'tic62124646'). |
'tic62124646'
|
sector
|
int
|
TESS sector to download (default: 13). |
13
|
exptime
|
int
|
Cadence in seconds (default: 120). |
120
|
author
|
str
|
Lightkurve author (default: 'SPOC'). |
'SPOC'
|
backends
|
list[str] | None
|
Specific backends to test; if None, use available ones. |
None
|
check_backend(print_summary=True)
¶
Report available Keras backends and device acceleration.
Returns a dict with keys: current, candidates, and per-backend info.
require_backend(backend=None)
¶
Validate that the selected Keras backend is installed and usable.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
backend
|
str | None
|
Backend to require ('jax' or 'torch'). If None, uses the active backend from KERAS_BACKEND or keras.backend.backend(). |
None
|
Raises:
| Type | Description |
|---|---|
RuntimeError
|
If the required backend is not installed or has no available devices. |
swap_backend(backend, accelerator=None, restart=False)
¶
Prepare environment for a different Keras backend and accelerator.
Note: Keras backend is selected at import time. If keras is already
imported in this process, you must restart the interpreter for the change
to take effect. Setting restart=True will perform an in-place re-exec.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
backend
|
str
|
One of 'jax' or 'torch'. |
required |
accelerator
|
str | None
|
Optional accelerator hint: 'cpu', 'cuda'/'gpu', or 'mps' (Apple Metal). |
None
|
restart
|
bool
|
If True and keras is already imported, re-exec the current process. |
False
|