0.2.0
What's Changed
- adds accelerate example by @edbeeching in #141
- Add a dry-run when calling
Pipeline.generate
by @alvarobartt in #146 - Add Notus format in
Prompt.format_as
and updateexamples/*.py
by @alvarobartt in #147 - Add
ProcessLLM
class by @gabrielmbmb in #151 - Adds
CritiqueTask
,UltraCMTask
and more by @alvarobartt in #152 - docs: add
llama.cpp
to extras by @davidberenstein1957 in #154 - Fix
_build_dataset
asprocessed_labels
were ignored by @plaguss in #158 - Add
to_argilla_{dataset,record}
methods inTextGenerationTask
by @alvarobartt in #159 - Fix
UltraFeedbackTask.to_argilla_dataset
ratings values by @alvarobartt in #160 - Align
typing
andtyping_extensions
with supported Python versions by @alvarobartt in #161 - Add
LLMPool
class by @gabrielmbmb in #156 - Add missing
CritiqueTask
andUltraCMTask
in__init__
and moveargilla_utils
toutils.argilla
by @alvarobartt in #162 - Add
test
workflow by @gabrielmbmb in #163 - Update
LLM
to returnFuture[List[List[LLMOutput]]]
by @gabrielmbmb in #164 - Add
PrometheusTask
by @alvarobartt in #165 - Randomise generations order by @gabrielmbmb in #167
- Add custom
to_argilla_{dataset,record}
toSelfInstructTask
by @alvarobartt in #169 - Fix
shuffle_before_labelling
and progress bar inPipeline.generate
by @alvarobartt in #170 - Replace
multiprocessing
withmultiprocess
by @gabrielmbmb in #171 - Refactor and improve docs by @plaguss in #134
- Fix
SelfInstructTask.{parse_output,to_argilla_record}
methods and_build_dataset
by @alvarobartt in #172 - Fix
results
didn't have same order asfutures
by @gabrielmbmb in #173 - Remove unnecesary plugin by @plaguss in #174
- Add
{generation,labelling}_model
column as metadata in Argilla by @alvarobartt in #175 - Fix exporting model name to Argilla with
LLMPool
by @gabrielmbmb in #177 - Update docs to include info about
ProcessLLM
andLLMPool
by @gabrielmbmb in #176
New Contributors
- @edbeeching made their first contribution in #141
- @davidberenstein1957 made their first contribution in #154
Full Changelog: 0.1.1...0.2.0