Releases: ecmwf/anemoi-inference
Releases · ecmwf/anemoi-inference
v0.8.0
0.8.0 (2025-10-15)
Features
- Add requirements subcommand (#245) (fd0877c)
- Better support for dynamic forcings and coupling in prepml (#262) (fd0877c)
- Lazy torch loading (#277) (fd0877c)
- Replace no grad with inference mode (#331) (4b7c495)
- Support empty input files (#342) (8078a89)
Bug Fixes
- Consolidate path behaviour on outputs (#339) (ba8faf1)
- metadata: Patching when receiving entry is not a dict (#338) (ff222e2)
- Netcdf incorrect shape when using extract post-processors (#346) (95c23ae), closes #341
- parallel runner: Support base class with main_argument decorator (#333) (7ec5cce)
- Update cutout due to #262 (#334) (35c1cdb)
Miscellaneous Chores
- Bump 0.8.0 (ec25c99)
v0.7.3
v0.7.2
v0.7.1
v0.7.0
0.7.0 (2025-08-04)
This release brings a change to the default accumulation behaviour. Prior to this release, accumulated fields were accumulated from the beginning of the forecast. Now, the default is to write accumulated fields unchanged as output by the model.
Users that wish to keep the old behaviour and accumulate fields from the beginning of the forecast, need to add the accumulate_from_start_of_forecast post-processor to the config, like so:
post_processors:
- accumulate_from_start_of_forecast⚠ BREAKING CHANGES
- Stop accumulating from start of forecast by default (#265)
- Drop python 3.9 support
Features
- Add logging control to mars (#268) (e95f184)
- Add Zarr Output (#275) (6c04b44)
- Allow for
Runner.runto return torch (#263) (77330f7) - Extend GribOutput class to write to FileLike Objects (#269) (b9770e2)
- Inner-level processors (#260) (59664cb)
- Move anemoi-inference metadata command to anemoi-utils (#257) (d735be5)
- Option to pass extra kwargs to
predict_step(#283) (1d9eb02) - outputs: Extend tee to enable postprocessors (#294) (2684293)
- post-processors: Add
assign_maskpost-processor (#287) (0313909) - post-processors: Extraction post-processors (#285) (7205af1)
- Remove python 3.9 from pyproject.toml (#290) (0adbddd)
- Stop accumulating from start of forecast by default (#265) (21826fb)
- Temporal interpolation runner (#227) (74048d9)
- waves: Add ability to update
typed_variablesfrom config (#202) (c02c45a)
Bug Fixes
- Add area to template lookup dictionary (#284) (0c5c812)
- Allow input preprocessors to patch data request (#286) (833cb6f)
- Be less helpful (#295) (a134f78)
- Checkpoint patching (#203) (77b90c0)
- grib: Ocean grib encoding (#282) (b6afaac)
- plot output: Cast numpy values to float32 (#288) (3cc6915), closes #276
- Provenance git dict reference issue (#259) (2d70411)
- Tensor not detached in debug mode (#279) (d9efac5)
- Use data frequency in interpolator inference to be consistent with training (#266) (feac2a4)
v0.6.3
v0.6.2
v0.6.1
0.6.1 (2025-06-04)
Features
- Add
assign_maskoutput (#221) (053e750) - Dynamic parallel runner (run any runner in parallel) (#242) (2411873)
- grib output: Support string format specification for output filename (#233) (be00edf)
- Inference with external graph. (#216) (6765316)
- Set debug log level on verbosity > 2 (#239) (532d070)