v0.8.5

[justinxzhao]

What's Changed

• Add function to free GPU memory by @Infernaught in Add function to free GPU memory #3643
• ❗ Enable LLM fine-tuning tests when no quantization is specified by @arnavgarg1 in ❗ Enable LLM fine-tuning tests when no quantization is specified #3626
• Add check to ensure selected backend works with quantization for LLMs by @arnavgarg1 in Add check to ensure selected backend works with quantization for LLMs  #3646
• [CI] Use a torch-nightly-compatible version of torchaudio by @justinxzhao in [CI] Use a torch-nightly-compatible version of torchaudio #3644
• Set do_sample default to True by @Infernaught in Set do_sample default to True #3641
• FIX: Failure in audio feature related test by @jimthompson5802 in FIX: Failure in audio feature related test #3651
• Remove unnecessary peft config updating by @Infernaught in Remove unnecessary peft config updating #3642
• FIX: docker build error for ludwig-gpu by @jimthompson5802 in FIX: docker build error for ludwig-gpu #3658
• Exclude getdaft on Windows by @carlogrisetti in Exclude getdaft on Windows #3629
• Add daft back for windows since the wheels are now officially published by @arnavgarg1 in Add daft back for windows since the wheels are now officially published #3663
• fix: The final batch of an epoch is skipped when batch size is 1 by @jeffkinnison in fix: The final batch of an epoch is skipped when batch size is 1 #3653
• Place metric functions for BLEU and Rogue on correct devices when using multiple GPUs by @arnavgarg1 in Place metric functions for BLEU and Rogue on correct devices when using multiple GPUs #3671
• Remove duplicate metrics by @Infernaught in Remove duplicate metrics #3670
• Increment epochs based on last_batch() instead of at the end of the train loop. by @justinxzhao in Increment epochs based on last_batch() instead of at the end of the train loop. #3668
• [FEATURE] Support Merging LoRA Weights Into Base Model (Issue-3603) by @alexsherstinsky in [FEATURE] Support Merging LoRA Weights Into Base Model (Issue-3603) #3649
• [FEATURE] Include Mistral-7B model in list of supported base models by @alexsherstinsky in [FEATURE] Include Mistral-7B model in list of supported base models #3674
• [MAINTENANCE] Partially reconcile type hints, fix some warnings, and fix comments in parts of the codebase. by @alexsherstinsky in [MAINTENANCE] Partially reconcile type hints, fix some warnings, and fix comments in parts of the codebase. #3673
• Improve error message for when an LLM base model can't be loaded. by @justinxzhao in Improve error message for when an LLM base model can't be loaded. #3675
• Fix eos_token and pad_token issue by @Infernaught in Fix eos_token and pad_token issue #3667
• FIX: error with nightly CI tests for test_resize_image by @jimthompson5802 in FIX: error with nightly CI tests for test_resize_image #3678
• [BUGFIX] Remove spurious test directory at the end of the test_llm.py::test_local_path_loading test run by @alexsherstinsky in [BUGFIX] Remove spurious test directory at the end of the test_llm.py::test_local_path_loading test run #3680
• Add per-device logging to tensorboard by @Infernaught in Add per-device logging to tensorboard #3677
• Fix dynamic generation config load during model.predict by @geoffreyangus in Fix dynamic generation config load during model.predict #3666
• [CI] Ensure that mlflow callback cleans up background-saving threads on trainer teardown. by @justinxzhao in [CI] Ensure that mlflow callback cleans up background-saving threads on trainer teardown. #3683
• fix: temporarily remove config validation check for backend by @geoffreyangus in fix: temporarily remove config validation check for backend #3688
• fix: Failing test for backend with quantization by @arnavgarg1 in fix: Failing test for backend with quantization #3689
• [BUGFIX] Ensure that full base models and not only adapter weights get saved when merge_and_unload is set by @alexsherstinsky in [BUGFIX] Ensure that full base models and not only adapter weights get saved when merge_and_unload is set #3679
• Add Ludwig Star History to README by @arnavgarg1 in Add Ludwig Star History to README #3696
• Use sphinx for all docstrings in api.py by @justinxzhao in Use sphinx for all docstrings in api.py #3693
• Fix binary variables being visualized as 0 and 1 by @Infernaught in Fix binary variables being visualized as 0 and 1 #3691
• [MAINTENANCE] Fix the linting warnings in two backend component classes. by @alexsherstinsky in [MAINTENANCE] Fix the linting warnings in two backend component classes. #3698
• [BUGFIX] Pin deepspeed<0.11, skip Horovod tests by @alexsherstinsky in [BUGFIX] Pin deepspeed<0.11, skip Horovod tests #3700
• Unpin deepspeed following fix in v0.11.1 by @tgaddair in Unpin deepspeed following fix in v0.11.1 #3706
• Move on_epoch_end and epoch increment to after run_evaluation loop. by @justinxzhao in Move on_epoch_end and epoch increment to after run_evaluation loop. #3690
• Remove model_load_path from experiment by @Infernaught in Remove model_load_path from experiment #3707
• [FEATURE] Allow typehints without the quotes. by @alexsherstinsky in [FEATURE] Allow typehints without the quotes. #3699

New Contributors

• @alexsherstinsky made their first contribution in [FEATURE] Support Merging LoRA Weights Into Base Model (Issue-3603) #3649

Full Changelog: v0.8.4...v0.8.5

---

This discussion was created from the release v0.8.5.