0.4.3

What's Changed

• Implement abstract function in prompt template factories by @nkaenzig in #938
• bump version to 0.4.2 [skip ci] by @github-actions[bot] in #937
• Fix resize_to_max_bytes function & move to vision by @nkaenzig in #939
• Clone model only when fit stage is involved by @nkaenzig in #941
• Make image position in multimodal conversations configurable by @MaxFeucht in #942
• Revert special output token change by @MaxFeucht in #945
• Adjust example answer formatting in raw FFQA template by @MaxFeucht in #946
• Fix/litellm text only inputs by @sebastianffx in #948
• Update README.md by @sebastianffx in #949
• Improve xml parsing robustness to special characters by @MaxFeucht in #951
• Render role correctly in string format by @MaxFeucht in #952
• Enable logging of input text for HF and LiteLLM models by @MaxFeucht in #953
• Increase MAX_NEW_TOKENS for model outputs by @MaxFeucht in #959
• Add VllmModel wrapper to eva.multimodal by @nkaenzig in #958
• Enable passing context as string or sequence across templates by @MaxFeucht in #964
• Update vLLM default params to improve stability by @nkaenzig in #966
• Correct GleasonArvaniti link in Datasets Markdown. Fixes #962 by @ahmedhassayoune in #963
• Align HuggingFaceModel in eva.language with multimodal by @nkaenzig in #961
• Fix unpack model logic for ModelFromRegistry models by @nkaenzig in #967
• Make huggingface wrapper attributes public again by @nkaenzig in #970
• Support multi-image datasets for multimodal eval by @MaxFeucht in #969
• Disable fast processors for qwen 2.5 models by @nkaenzig in #971
• Fix system prompt propagation in multimodal huggingface model by @nkaenzig in #974
• Make chat template configurable by @nkaenzig in #975
• added boxed formatting by @jonaserb-k in #973
• Fix inconsistency in implementation of extract classes by @nkaenzig in #977
• Reinstate int conversion of missing_answer in discrete extraction by @MaxFeucht in #978
• fixed logging bug in ExtractDiscreteAnswerFromStructuredOutput by @jonaserb-k in #979
• Expose compute_groups and separator via MetricsSchema by @nkaenzig in #982

New Contributors

• @sebastianffx made their first contribution in #948
• @ahmedhassayoune made their first contribution in #963
• @jonaserb-k made their first contribution in #973

Full Changelog: 0.4.2...0.4.3