Releases: Labbeti/aac-metrics
Releases · Labbeti/aac-metrics
Version 0.4.1
[0.4.1] 2023-04-13
Deleted
- Old unused files
package_tree.rst
,fluency_error.py
,sbert.py
andspider_err.py
.
Version 0.4.0
[0.4.0] 2023-04-13
Added
- Argument
return_probs
for fluency error metric.
Changed
- Rename
SPIDErErr
toSPIDErFL
to match DCASE2023 metric name. - Rename
SBERT
toSBERTSim
to avoid confusion with SBERT model name. - Rename
FluencyError
toFluErr
. - Check if Java executable version between 8 and 11. (#1)
Fixed
SPIDErFL
sentences scores outputs when usingreturn_all_scores=True
.- Argument
reset_state
inSPIDErFL
,SBERTSim
,FluErr
andFENSE
when using their functional interface. - Classes and functions factories now support SPICE and CIDEr-D metrics.
SBERTSim
class instantiation.
Version 0.3.0
[0.3.0] 2023-02-27
Added
- Parameters
timeout
andseparate_cache_dir
inSPICE
function and class. - Documentation pages with sphinx.
- Parameter
language
inMETEOR
function and class. - Options to download only
PTBTokenizer
,METEOR
,SPICE
orFENSE
indownload.py
. SBERT
andFluencyError
metrics extracted fromFENSE
.SPIDErErr
metric which combinesSPIDEr
withFluencyError
.- Parameter
reset_state
inSBERT
,FluencyError
,SPIDErErr
andFENSE
functions and classes.
Changed
- Fix README typo and SPIDEr-max tables.
Fixed
- Workflow badge with Github changes. (badges/shields#8671)
Version 0.2.0
[0.2.0] 2022-12-14
Added
FENSE
class and function metric, with fluency error rate and raw output probabilities.- Unittest with
fense
repository. load_metric
function in init to match huggingface evaluation package.
Changed
- Rename
global_scores
tocorpus_scores
andlocal_scores
tosents_scores
. - Rename
CustomEvaluate
toEvaluate
andcustom_evaluate
toevaluate
. - Set default cache path to
$HOME/.cache
. - Remove 'coco' prefix to file, functions and classes names to have cleaner names.
Fixed
FENSE
metric error when computing scores with less thanbatch_size
sentences.
Version 0.1.2
[0.1.2] 2022-10-31
Added
- All candidates scores option
return_all_cands_scores
for SPIDEr-max. - Functions
is_mono_sents
andis_mult_sents
to detectlist[str]
sentences andlist[list[str]]
multiples sentences. - Functions
flat_list
andunflat_list
to flat multiples sentences to sentences.
Changed
- Update default value used for
return_all_scores
in cider and rouge functions. - Update internal metric factory with functions instead of classes to avoid cyclic dependency.
Fixed
- Fix SPIDEr-max local scores output shape.
Version 0.1.1
[0.1.1] 2022-09-30
Added
- Documentation for metric functions and classes.
- A second larger example for unit testing.
Changed
- Update README information, references and description.
Fixed
- SPIDEr-max computation with correct global and local outputs.
- Unit testing for computing SPICE metric from caption-evaluation-tools.
Version 0.1.0
[0.1.0] 2022-09-28
Added
- BLEU, METEOR, ROUGE-l, SPICE, CIDEr and SPIDEr metrics functions and modules.
- SPIDEr-max experimental implementation.
- Installation script in download.py.
- Evaluation script in evaluate.py.
- Unittest with
caption-evaluation-tools
repository.