What's Changed
* Minimal SkyPilot config for `blip2` and `llava` models for GCP with `TRL_SFT` by xrdaukar in https://github.com/oumi-ai/oumi/pull/573
* Inference Engine async writes by taenin in https://github.com/oumi-ai/oumi/pull/574
* Misc cleanups in `JsonlinesDataset` by xrdaukar in https://github.com/oumi-ai/oumi/pull/576
* Split out cloud dependencies by taenin in https://github.com/oumi-ai/oumi/pull/575
* Disable `sdpa` for `blip2` by xrdaukar in https://github.com/oumi-ai/oumi/pull/579
* Set accelerate version to fix FSDP model saving by wizeng23 in https://github.com/oumi-ai/oumi/pull/580
* Remove AWS as a default dep by taenin in https://github.com/oumi-ai/oumi/pull/582
* Update `ProfilerParams` docstrings to follow the new style by xrdaukar in https://github.com/oumi-ai/oumi/pull/583
* Minor updates in `scripts/benchmarks/minimal_multimodal_training.py` by xrdaukar in https://github.com/oumi-ai/oumi/pull/585
* Add `override` annotations to methods of few Dataset subclasses by xrdaukar in https://github.com/oumi-ai/oumi/pull/584
* Add dataset class for dolly dataset by oelachqar in https://github.com/oumi-ai/oumi/pull/586
* Refactor debugging/device utils, and add new GPU stats measurement functions by xrdaukar in https://github.com/oumi-ai/oumi/pull/587
* Add text jsonlines dataset class by oelachqar in https://github.com/oumi-ai/oumi/pull/589
* Define `DataCollationParams` by xrdaukar in https://github.com/oumi-ai/oumi/pull/581
* Misc updates to Polaris launcher scripts by xrdaukar in https://github.com/oumi-ai/oumi/pull/591
* Set up a new version of the Oumi CLI using Typer by taenin in https://github.com/oumi-ai/oumi/pull/588
* Update handling of GPU fan speed info by xrdaukar in https://github.com/oumi-ai/oumi/pull/595
* Add support for magpie dataset variants by oelachqar in https://github.com/oumi-ai/oumi/pull/594
* Rename GenerationConfig to GenerationParams by wizeng23 in https://github.com/oumi-ai/oumi/pull/592
* Fix cli infer test by wizeng23 in https://github.com/oumi-ai/oumi/pull/598
* Judge Notebook 1: default judge by kaisopos in https://github.com/oumi-ai/oumi/pull/593
* [Tiny] update missing dataset import by oelachqar in https://github.com/oumi-ai/oumi/pull/599
* Update training script to support data collators by xrdaukar in https://github.com/oumi-ai/oumi/pull/590
* Update accelerate version to 1.0.0 by wizeng23 in https://github.com/oumi-ai/oumi/pull/601
* Remove deprecated dataset code paths by oelachqar in https://github.com/oumi-ai/oumi/pull/596
* Refactor Aya & Ultrachat to use oumi dataset sft classes by oelachqar in https://github.com/oumi-ai/oumi/pull/597
* Add Llama train/eval/infer E2E integration test by wizeng23 in https://github.com/oumi-ai/oumi/pull/602
* Set docstring for `NVidiaGpuRuntimeInfo` struct by xrdaukar in https://github.com/oumi-ai/oumi/pull/603
* Add generation params to inference engines by oelachqar in https://github.com/oumi-ai/oumi/pull/600
* [bug] Fix issue loading jsonl datasets from file by oelachqar in https://github.com/oumi-ai/oumi/pull/604
* Add Llama 3B configs by wizeng23 in https://github.com/oumi-ai/oumi/pull/605
* Align pyright checks with latest Pylance version by oelachqar in https://github.com/oumi-ai/oumi/pull/611
* Fix `apply_chat_template` issue in `VisionLanguageSftDataset` by xrdaukar in https://github.com/oumi-ai/oumi/pull/609
* More robust make setup by oelachqar in https://github.com/oumi-ai/oumi/pull/610
* Fix a bug where the new CLI was improperly importing functions from top-level modules. by taenin in https://github.com/oumi-ai/oumi/pull/613
* Add support for the Launch command suite in the new CLI by taenin in https://github.com/oumi-ai/oumi/pull/612
* Support `HuggingFaceH4/llava-instruct-mix-vsft` dataset by xrdaukar in https://github.com/oumi-ai/oumi/pull/608
* [tiny] Fix .gitignore by wizeng23 in https://github.com/oumi-ai/oumi/pull/616
* [tiny] add gpt2 chat template, and update tests to use it by oelachqar in https://github.com/oumi-ai/oumi/pull/617
* Turn off pretty-printing exceptions in our CLI by taenin in https://github.com/oumi-ai/oumi/pull/618
* Cleanup dependencies by oelachqar in https://github.com/oumi-ai/oumi/pull/615
* Upgrade oumi dependencies by oelachqar in https://github.com/oumi-ai/oumi/pull/606
* Update makefile to use uv, add Jupyter target by oelachqar in https://github.com/oumi-ai/oumi/pull/614
* Add miniconda installation target, cleanup unused make commands by oelachqar in https://github.com/oumi-ai/oumi/pull/620
* Update several notebooks with the new EvaluationConfig format. by taenin in https://github.com/oumi-ai/oumi/pull/621
* Make sure conda env is registered by oelachqar in https://github.com/oumi-ai/oumi/pull/622
* Add Llama 3b sft/lora/qlora configs for Polaris by wizeng23 in https://github.com/oumi-ai/oumi/pull/626
* Add check if installation is successful by oelachqar in https://github.com/oumi-ai/oumi/pull/625
* Initial Cambrian integration by xrdaukar in https://github.com/oumi-ai/oumi/pull/557
* [tiny] alpaca - minor reproducibility boost by optas in https://github.com/oumi-ai/oumi/pull/619
* explicitly specify the model's dtype in LMH by optas in https://github.com/oumi-ai/oumi/pull/607
* [tiny] Add flops for T4 GPU by wizeng23 in https://github.com/oumi-ai/oumi/pull/628
* Use a timestamp for job directories on Polaris by taenin in https://github.com/oumi-ai/oumi/pull/627
* [tiny] Fix bug with Polaris job num by wizeng23 in https://github.com/oumi-ai/oumi/pull/629
* Update two VLLM configs. by xrdaukar in https://github.com/oumi-ai/oumi/pull/624
* Add `pip install -U uv;` to `make setup` for existing envs by xrdaukar in https://github.com/oumi-ai/oumi/pull/630
* Disable MFU logging for non-packed datasets by wizeng23 in https://github.com/oumi-ai/oumi/pull/632
* Add config example for long context fine-tuning by oelachqar in https://github.com/oumi-ai/oumi/pull/631
* Add distribution mode flag to llama_tune by wizeng23 in https://github.com/oumi-ai/oumi/pull/635
* Judge Notebook 2: Custom Judge by kaisopos in https://github.com/oumi-ai/oumi/pull/623
* Bugfixes for LLAVA by xrdaukar in https://github.com/oumi-ai/oumi/pull/634
* Update sphinx config and docs to fix misc errors and warnings by oelachqar in https://github.com/oumi-ai/oumi/pull/639
* Factor out OUMI_TOTAL_NUM_GPUS env var by wizeng23 in https://github.com/oumi-ai/oumi/pull/636
* Remove bitsandbytes from train dependencies by oelachqar in https://github.com/oumi-ai/oumi/pull/643
* Enable intershinx to allow linking to external documentation pages by oelachqar in https://github.com/oumi-ai/oumi/pull/640
* Tune few training params for LLAVA and blip2 models by xrdaukar in https://github.com/oumi-ai/oumi/pull/642
* Added support for specifying the inference engine via the InferenceConfig by taenin in https://github.com/oumi-ai/oumi/pull/638
* Add popular pre-training dataset classes by oelachqar in https://github.com/oumi-ai/oumi/pull/641
* Remove openai dependency by oelachqar in https://github.com/oumi-ai/oumi/pull/644
* Update our documentation to point to the new CLI. by taenin in https://github.com/oumi-ai/oumi/pull/645
* Enable dataloaders for VLLM-s (llava and blip2) by xrdaukar in https://github.com/oumi-ai/oumi/pull/646
* Allow gradient clipping to be optional by optas in https://github.com/oumi-ai/oumi/pull/649
* Add support for `add_generation_prompt` in LLAVA chat template by xrdaukar in https://github.com/oumi-ai/oumi/pull/648
* Add a description to the Launch CLI by taenin in https://github.com/oumi-ai/oumi/pull/651
* Add all Llama FSDP GCP configs by wizeng23 in https://github.com/oumi-ai/oumi/pull/637
* Coerce model params to correct dtype for QLoRA FSDP by wizeng23 in https://github.com/oumi-ai/oumi/pull/652
* Use uv for `pip install` commands by wizeng23 in https://github.com/oumi-ai/oumi/pull/653
* Update sphinx docs by oelachqar in https://github.com/oumi-ai/oumi/pull/654
* [Docs] Refactor docs pipeline by oelachqar in https://github.com/oumi-ai/oumi/pull/655
* [docs] swap and configure sphinx theme by oelachqar in https://github.com/oumi-ai/oumi/pull/656
* [Docs] Add documentation placeholders by oelachqar in https://github.com/oumi-ai/oumi/pull/658
* [Docs] Add sphinx-bibtex by oelachqar in https://github.com/oumi-ai/oumi/pull/659
* [Docs] fix rendering issues by oelachqar in https://github.com/oumi-ai/oumi/pull/660
* [docs] fix broken links by oelachqar in https://github.com/oumi-ai/oumi/pull/661
* Fix broken link in readme (dev_setup) by kaisopos in https://github.com/oumi-ai/oumi/pull/662
* [docs][tiny] fix minor doc typos by oelachqar in https://github.com/oumi-ai/oumi/pull/666
* [docs] add autodoc2 template by oelachqar in https://github.com/oumi-ai/oumi/pull/665
* [docs] Add content links and references by oelachqar in https://github.com/oumi-ai/oumi/pull/668
* [docs] switch to myst-nb for rendering notebooks by oelachqar in https://github.com/oumi-ai/oumi/pull/669
* [docs] Add script to generate module summaries by oelachqar in https://github.com/oumi-ai/oumi/pull/670
* [docs] Include cli reference by oelachqar in https://github.com/oumi-ai/oumi/pull/671
* Add dataset submodules by oelachqar in https://github.com/oumi-ai/oumi/pull/667
* Update notebooks to include a descriptive title by oelachqar in https://github.com/oumi-ai/oumi/pull/664
* Update tests/utils/test_device_utils.py by xrdaukar in https://github.com/oumi-ai/oumi/pull/672
* [Inference] Bug in generation config stop tokens by kaisopos in https://github.com/oumi-ai/oumi/pull/663
* Support rewriting special label values to -100 (`ignore_index`) to exclude from loss by xrdaukar in https://github.com/oumi-ai/oumi/pull/657
* Rename emails and website url to Oumi by wizeng23 in https://github.com/oumi-ai/oumi/pull/675
* Update scripts/benchmarks/minimal_multimodal_training.py to be able to test 3 other models by xrdaukar in https://github.com/oumi-ai/oumi/pull/677
* Update data collators TextCollatorWithPadding by xrdaukar in https://github.com/oumi-ai/oumi/pull/674
* Add image support to `llama3-instruct.jinja` chat template by xrdaukar in https://github.com/oumi-ai/oumi/pull/676
* Cleanup vscode launch.json by oelachqar in https://github.com/oumi-ai/oumi/pull/679
* Add Llama 3B DDP configs by wizeng23 in https://github.com/oumi-ai/oumi/pull/681
* Add cff file for citations by oelachqar in https://github.com/oumi-ai/oumi/pull/682
* Small edits to Llama configs by wizeng23 in https://github.com/oumi-ai/oumi/pull/683
* Increase `llava.7b.sft` max length from 512 to 1024 by xrdaukar in https://github.com/oumi-ai/oumi/pull/687
* Rename turn.py to conversation.py by oelachqar in https://github.com/oumi-ai/oumi/pull/685
* Pad token fix by kaisopos in https://github.com/oumi-ai/oumi/pull/678
* Add basic validation for data collators by xrdaukar in https://github.com/oumi-ai/oumi/pull/688
* Support `--image` CLI param in `infer_interactive` by xrdaukar in https://github.com/oumi-ai/oumi/pull/686
* Move sft and preference tuning datasets to their own module by oelachqar in https://github.com/oumi-ai/oumi/pull/691
* Set `model.tokenizer_pad_token` for gpt2 in Oumi Inference example in `vsconfig` by xrdaukar in https://github.com/oumi-ai/oumi/pull/692
* Update our dev set up instructions to not conflict with the new CLI by taenin in https://github.com/oumi-ai/oumi/pull/698
* Add back the polling spinner to the Launch CLI by taenin in https://github.com/oumi-ai/oumi/pull/695
* Add a dummy sft dataset for unit tests by oelachqar in https://github.com/oumi-ai/oumi/pull/697
* Include judge commands into main CLI by oelachqar in https://github.com/oumi-ai/oumi/pull/690
* Replace `logger.exception()` in `TextCollatorWithPadding` with `logger.error()` by xrdaukar in https://github.com/oumi-ai/oumi/pull/700
* [Clean-up] Remove dead code (custom eval) by kaisopos in https://github.com/oumi-ai/oumi/pull/696
* Define `oumi.core.constants` file for global constants by xrdaukar in https://github.com/oumi-ai/oumi/pull/694
* Add serialization helper methods for Conversation object by oelachqar in https://github.com/oumi-ai/oumi/pull/701
* Create pull_request_template.md by taenin in https://github.com/oumi-ai/oumi/pull/693
* Use `-100` as `ignore_index` in `VisionLanguageSftDataset` and in `build_data_collator()` by default by xrdaukar in https://github.com/oumi-ai/oumi/pull/702
* Add main entrypoint, update vscode launch.json to use new cli by oelachqar in https://github.com/oumi-ai/oumi/pull/704
* Update experimental async dataset feature by oelachqar in https://github.com/oumi-ai/oumi/pull/705
* Update new image CLI to support input image for VLLMs by xrdaukar in https://github.com/oumi-ai/oumi/pull/699
* Update sft jsonl dataset to support json files and alpaca format by oelachqar in https://github.com/oumi-ai/oumi/pull/703
* Make interactive inference continually loop by wizeng23 in https://github.com/oumi-ai/oumi/pull/689
* Update vscode launch inference example by xrdaukar in https://github.com/oumi-ai/oumi/pull/706
* Training configurations llama3.1 + ChatML-Template by optas in https://github.com/oumi-ai/oumi/pull/684
* Add help strings for `oumi train/evaluate/infer` by wizeng23 in https://github.com/oumi-ai/oumi/pull/708
* Log peak gpu memory at the end of training by oelachqar in https://github.com/oumi-ai/oumi/pull/710
* Log training git commit and tag by oelachqar in https://github.com/oumi-ai/oumi/pull/709
* Add github issue templates. by taenin in https://github.com/oumi-ai/oumi/pull/711
* Enable more ruff rules by oelachqar in https://github.com/oumi-ai/oumi/pull/712
* Add Llama inference configs by wizeng23 in https://github.com/oumi-ai/oumi/pull/713
* Enable pathlib rule by oelachqar in https://github.com/oumi-ai/oumi/pull/714
* Update index.rst - switch expanded name Open Universal Machine Intelligence and replace with "Oumi" by ciaralema in https://github.com/oumi-ai/oumi/pull/715
* Make sure registry is case agnostic by oelachqar in https://github.com/oumi-ai/oumi/pull/717
* Move config files to new directory structure by wizeng23 in https://github.com/oumi-ai/oumi/pull/720
* Replace `np.array(x)` with to `x.numpy()` to resolve test warnings by xrdaukar in https://github.com/oumi-ai/oumi/pull/724
* Update index.rst by ciaralema in https://github.com/oumi-ai/oumi/pull/719
* Add judge cli tests and bugfixes by oelachqar in https://github.com/oumi-ai/oumi/pull/722
* Update `tests/builders/test_lr_schedules.py` by xrdaukar in https://github.com/oumi-ai/oumi/pull/723
* Add dataset_path parameter by oelachqar in https://github.com/oumi-ai/oumi/pull/716
* [tiny] Update judge configs by oelachqar in https://github.com/oumi-ai/oumi/pull/725
* Rename `--generation.(input/output)_filepath` flags to `--(input/output)_path` and rename `detach` flag by wizeng23 in https://github.com/oumi-ai/oumi/pull/707
* Rename LLAVA sft_sky_job.yaml for consistency with blip2 by xrdaukar in https://github.com/oumi-ai/oumi/pull/729
* Remove deprecated preprocessing function by oelachqar in https://github.com/oumi-ai/oumi/pull/726
* Add DPO base class, initial dataset, integration tests by oelachqar in https://github.com/oumi-ai/oumi/pull/728
* [tiny] Add datasets import to enable registry discovery by oelachqar in https://github.com/oumi-ai/oumi/pull/731
* Add a helper function `is_image_text_llm()` to detect supported image+text LLMs by xrdaukar in https://github.com/oumi-ai/oumi/pull/730
* Add dpo debug dataset and integration test by oelachqar in https://github.com/oumi-ai/oumi/pull/732
* Add inference engine for vertex GCP by oelachqar in https://github.com/oumi-ai/oumi/pull/734
* [docs] Cleanup placeholder documentation pages by oelachqar in https://github.com/oumi-ai/oumi/pull/736
* Native inference with images (`model.generate`), and Processor abstraction by xrdaukar in https://github.com/oumi-ai/oumi/pull/721
* [docs] update faq section by oelachqar in https://github.com/oumi-ai/oumi/pull/739
* [docs] update advanced topics section by oelachqar in https://github.com/oumi-ai/oumi/pull/737
* [docs] update datasets section by oelachqar in https://github.com/oumi-ai/oumi/pull/738
* [docs] Update inference section by oelachqar in https://github.com/oumi-ai/oumi/pull/744
* [docs] Add evaluate and judge user guides by oelachqar in https://github.com/oumi-ai/oumi/pull/745
* [docs] Update training section by oelachqar in https://github.com/oumi-ai/oumi/pull/741
* Support `merve/vqav2-small` dataset by xrdaukar in https://github.com/oumi-ai/oumi/pull/742
* Reorder IMAGE and TEXT messages in coco and Flick datasets by xrdaukar in https://github.com/oumi-ai/oumi/pull/743
* Rename base datasets modules by oelachqar in https://github.com/oumi-ai/oumi/pull/735
* [tests] Move unit tests to unit folder by oelachqar in https://github.com/oumi-ai/oumi/pull/753
* Update training params for LLAVA and BLIP2 models by xrdaukar in https://github.com/oumi-ai/oumi/pull/754
* [docs] Add script to summarize configs by oelachqar in https://github.com/oumi-ai/oumi/pull/751
* [docs] Enable termynal cli and add example by oelachqar in https://github.com/oumi-ai/oumi/pull/750
* [docs] Update apidoc rst files with latest modules by oelachqar in https://github.com/oumi-ai/oumi/pull/749
* [docs] Add logo, analytics, and update sphinx config by oelachqar in https://github.com/oumi-ai/oumi/pull/747
* [docs] Update about section by oelachqar in https://github.com/oumi-ai/oumi/pull/748
* [docs] Add page for using skypilot by oelachqar in https://github.com/oumi-ai/oumi/pull/746
* Update base dataset class to use dataset_path by oelachqar in https://github.com/oumi-ai/oumi/pull/757
* Unit tests with sanity checks for `oumi` chat templates by xrdaukar in https://github.com/oumi-ai/oumi/pull/755
* [docs] Misc docs cleanup tasks by oelachqar in https://github.com/oumi-ai/oumi/pull/756
* And WandB integration to eval by wizeng23 in https://github.com/oumi-ai/oumi/pull/727
* Add 3 new PNG images to testdata by xrdaukar in https://github.com/oumi-ai/oumi/pull/758
* Update LLAVA integration tests to use different test images by xrdaukar in https://github.com/oumi-ai/oumi/pull/761
* [tests] Add shared mock tokenizer mixture, requires_gpus marker by oelachqar in https://github.com/oumi-ai/oumi/pull/759
* Update vllm engine to allow for continuous batching by oelachqar in https://github.com/oumi-ai/oumi/pull/762
* Added text completions collator hard-coded for Llama 3.1 support by jgreer013 in https://github.com/oumi-ai/oumi/pull/752
* Update remote inference engines to show progress, reduce native engine verbosity by oelachqar in https://github.com/oumi-ai/oumi/pull/760
* Add sample oumi launcher configs for LLAVA and blip2 by xrdaukar in https://github.com/oumi-ai/oumi/pull/763
* Create 2 other Launcher configs for LLAVA and blip2 using OUMI training loop by xrdaukar in https://github.com/oumi-ai/oumi/pull/765
* Update CI to include all dependencies, fix newly introduced vllm bugs by oelachqar in https://github.com/oumi-ai/oumi/pull/764
* Suggest using the git root directory when launching jobs via a CLI in dev-build. by taenin in https://github.com/oumi-ai/oumi/pull/766
* [bug] enable logger propagation so that pytest can capture them by oelachqar in https://github.com/oumi-ai/oumi/pull/769
* Update liger kernel integration by oelachqar in https://github.com/oumi-ai/oumi/pull/768
* Add BaseInferenceEngine supported_params to keep track of inference engine parameters by oelachqar in https://github.com/oumi-ai/oumi/pull/770
* Add READMEs for new config structure by wizeng23 in https://github.com/oumi-ai/oumi/pull/767
* Fix `*_no_cuda()` tests in `tests/unit/utils/test_device_utils.py` by xrdaukar in https://github.com/oumi-ai/oumi/pull/775
* Remove support for python 3.8, update to 3.9 by oelachqar in https://github.com/oumi-ai/oumi/pull/772
* Multimodal improvements, and basic support for `meta-llama/Llama-3.2-11B-Vision-Instruct` by xrdaukar in https://github.com/oumi-ai/oumi/pull/771
* Set vLLM max LoRA rank based on LoRA adapter by wizeng23 in https://github.com/oumi-ai/oumi/pull/776
* [WIP] Llama 3.2 Vision 11B configs by xrdaukar in https://github.com/oumi-ai/oumi/pull/777
* Tune MLLMs configs for Llama Vision and LLAVA by xrdaukar in https://github.com/oumi-ai/oumi/pull/779
* Ιnference vLLM: towards fitting big models in GPU memory by optas in https://github.com/oumi-ai/oumi/pull/774
* Adds control for an explicit quantization-compute dtype by optas in https://github.com/oumi-ai/oumi/pull/680
* Correct layer name in Llama vision accelerate config by xrdaukar in https://github.com/oumi-ai/oumi/pull/782
* Clean up configs by wizeng23 in https://github.com/oumi-ai/oumi/pull/781
* Save (pre-)processor configs by xrdaukar in https://github.com/oumi-ai/oumi/pull/780
* Tiny: force batch_size of generation parameters >= 1 by optas in https://github.com/oumi-ai/oumi/pull/783
* Switch to the regular `pip` temporarily in the old Polaris launcher by xrdaukar in https://github.com/oumi-ai/oumi/pull/787
* Define PAD token for Llama 2B model (fineweb pretrain) by xrdaukar in https://github.com/oumi-ai/oumi/pull/785
* Rename `oumi launch stop` to `oumi launch cancel` by wizeng23 in https://github.com/oumi-ai/oumi/pull/786
* [tiny] move collator test to unit folder by oelachqar in https://github.com/oumi-ai/oumi/pull/790
* Add `oumi launch stop` subcommand by wizeng23 in https://github.com/oumi-ai/oumi/pull/789
* Add sft packed dataset and tests by oelachqar in https://github.com/oumi-ai/oumi/pull/788
* CLI documentation updates. by taenin in https://github.com/oumi-ai/oumi/pull/791
* Updated default temperature from 1.0 to 0.0 for generation params by jgreer013 in https://github.com/oumi-ai/oumi/pull/792
* Add a prominent link to our docs. by taenin in https://github.com/oumi-ai/oumi/pull/793
* Fixing 2 broken links in README.md by kaisopos in https://github.com/oumi-ai/oumi/pull/794
* Switch from writing individual conversations to all conversations by jgreer013 in https://github.com/oumi-ai/oumi/pull/796
* Update pull_request_template.md by taenin in https://github.com/oumi-ai/oumi/pull/798
* Fix polling jobs and job status in the CLI by taenin in https://github.com/oumi-ai/oumi/pull/795
* Update base SFT dataset to add support for training on assistant turns only by oelachqar in https://github.com/oumi-ai/oumi/pull/797
* Small fix for building docs by taenin in https://github.com/oumi-ai/oumi/pull/799
* Update polaris user regex to handle usernames with dashes by jgreer013 in https://github.com/oumi-ai/oumi/pull/801
* Simplify `TextCollatorWithPadding` by xrdaukar in https://github.com/oumi-ai/oumi/pull/800
* Minor updates to docstrings and error messages related to `text_with_padding` collator by xrdaukar in https://github.com/oumi-ai/oumi/pull/802
* Adding more control/parameters for text generation by optas in https://github.com/oumi-ai/oumi/pull/784
* Pre-training HSDP configs for Llama 8B for A100-40GB by xrdaukar in https://github.com/oumi-ai/oumi/pull/778
* Fix pyproject.toml merge error by oelachqar in https://github.com/oumi-ai/oumi/pull/805
* Update llama_tune script to support 8B pretraining by xrdaukar in https://github.com/oumi-ai/oumi/pull/804
* Add e2e fine-tuning test by oelachqar in https://github.com/oumi-ai/oumi/pull/806
* Create Polaris and GCP launcher configs for Llama 8B pretraining by xrdaukar in https://github.com/oumi-ai/oumi/pull/807
* Enable sphinx-tabs, update OOM faq page to use it by oelachqar in https://github.com/oumi-ai/oumi/pull/808
* [docs] Start adding doctests by oelachqar in https://github.com/oumi-ai/oumi/pull/809
* Fixed broken skyssh and skycode commands by jgreer013 in https://github.com/oumi-ai/oumi/pull/810
* Enable Accelerate FSDP runs without using the Accelerate launcher by wizeng23 in https://github.com/oumi-ai/oumi/pull/803
* Disable uv in polaris launcher by xrdaukar in https://github.com/oumi-ai/oumi/pull/812
* Migrate configs to use Oumi instead of Accelerate configs by wizeng23 in https://github.com/oumi-ai/oumi/pull/811
* Update pretest.yaml by taenin in https://github.com/oumi-ai/oumi/pull/817
* Update installation.md by taenin in https://github.com/oumi-ai/oumi/pull/814
* Add smollm config by oelachqar in https://github.com/oumi-ai/oumi/pull/815
* Add extra logging in Polaris client by xrdaukar in https://github.com/oumi-ai/oumi/pull/813
* Fix a small typo in quickstart by taenin in https://github.com/oumi-ai/oumi/pull/818
* Correct names of few env vars in `llama_tune.sh` by xrdaukar in https://github.com/oumi-ai/oumi/pull/819
* [tiny] move collator to builders by oelachqar in https://github.com/oumi-ai/oumi/pull/821
* Update sample commands to run Llama 8B model on Polaris by xrdaukar in https://github.com/oumi-ai/oumi/pull/823
* [tiny] update device cleanup by oelachqar in https://github.com/oumi-ai/oumi/pull/822
* Update quickstart to use the new CLI arg format. by taenin in https://github.com/oumi-ai/oumi/pull/820
* Mock out slow tests. by taenin in https://github.com/oumi-ai/oumi/pull/825
* More changes for Llama 8B Polaris by xrdaukar in https://github.com/oumi-ai/oumi/pull/826
* Minor updates to Polaris scripts by xrdaukar in https://github.com/oumi-ai/oumi/pull/827
* Misc documentation updates for installation, quickstart, and tutorials. by taenin in https://github.com/oumi-ai/oumi/pull/828
* Add image utils to load image bytes for messages with `IMAGE_URL`, `IMAGE_PATH` types by xrdaukar in https://github.com/oumi-ai/oumi/pull/829
* Update our documentation to include modules for directories that aren't packages. by taenin in https://github.com/oumi-ai/oumi/pull/830
* Add Llama405B Qlora/Lora/FFT configs by wizeng23 in https://github.com/oumi-ai/oumi/pull/833
* Add `SGLangInferenceEngine` by xrdaukar in https://github.com/oumi-ai/oumi/pull/834
* Switch from 8-bit AdamW to AdamW for FSDP Lora/QLora by wizeng23 in https://github.com/oumi-ai/oumi/pull/835
* Add options for specifying log levels in the CLI. by taenin in https://github.com/oumi-ai/oumi/pull/836
* Request user feedback after training. by taenin in https://github.com/oumi-ai/oumi/pull/837
* Documentation fix for training. by taenin in https://github.com/oumi-ai/oumi/pull/840
* [tiny] pass dataset_path to constructor by oelachqar in https://github.com/oumi-ai/oumi/pull/841
* Move chatqa to projects, add details by oelachqar in https://github.com/oumi-ai/oumi/pull/816
* [tiny] fix typo in docs/quickstart by oelachqar in https://github.com/oumi-ai/oumi/pull/843
* Add unit tests for builders/data module by oelachqar in https://github.com/oumi-ai/oumi/pull/842
* Additional improvements to 405B by wizeng23 in https://github.com/oumi-ai/oumi/pull/838
* Add constrained decoding to vllm engine by oelachqar in https://github.com/oumi-ai/oumi/pull/844
* Add support for remote vllm inference engine by oelachqar in https://github.com/oumi-ai/oumi/pull/845
* Add `override` annotations to misc inference engines methods where missing by xrdaukar in https://github.com/oumi-ai/oumi/pull/846
* Add a debug command for dumping environment information. by taenin in https://github.com/oumi-ai/oumi/pull/839
* [tiny] fix openai remote engine auth bug by oelachqar in https://github.com/oumi-ai/oumi/pull/848
* Update RemoteInferenceEngine to support guided decoding by oelachqar in https://github.com/oumi-ai/oumi/pull/849
* Support multimodal inference with remote engines by xrdaukar in https://github.com/oumi-ai/oumi/pull/847
* Fixes for quickstart and launch commands. by taenin in https://github.com/oumi-ai/oumi/pull/851
* Support Qwen2-VL model for training by xrdaukar in https://github.com/oumi-ai/oumi/pull/850
* Fix our config parsing test by taenin in https://github.com/oumi-ai/oumi/pull/855
* Move `remote_params` from `GenerationParams` to `InferenceConfig` by xrdaukar in https://github.com/oumi-ai/oumi/pull/853
* Tune PyTorch CUDA allocation params for Llama 405B qlora by xrdaukar in https://github.com/oumi-ai/oumi/pull/854
* Update Llama 405B qlora config to use `adamw_torch_fused` optimizer by xrdaukar in https://github.com/oumi-ai/oumi/pull/856
* Update vllm config path in Polaris multi-node inference notebook by xrdaukar in https://github.com/oumi-ai/oumi/pull/857
* Change config dir structure for LLAVA and blip2 models by xrdaukar in https://github.com/oumi-ai/oumi/pull/859
* Enable LM-harness eval for multimodal, and add eval config for Llama 3.2 Vision 11B by xrdaukar in https://github.com/oumi-ai/oumi/pull/858
* Update inference engine docs. by taenin in https://github.com/oumi-ai/oumi/pull/862
* Initial port of zigzag attention by xrdaukar in https://github.com/oumi-ai/oumi/pull/861
* Add integration tests for visual SFT datasets (`VisionLanguageSftDataset`) by xrdaukar in https://github.com/oumi-ai/oumi/pull/863
* Update base config to_yaml signature by oelachqar in https://github.com/oumi-ai/oumi/pull/865
* Add support for batch inference by oelachqar in https://github.com/oumi-ai/oumi/pull/864
* Parallelize data preprocessing in `BaseMapDataset` by xrdaukar in https://github.com/oumi-ai/oumi/pull/868
* Tweak writer batch size computation in `BaseMapDataset` by xrdaukar in https://github.com/oumi-ai/oumi/pull/870
* Refresh fine-tuning notebook by oelachqar in https://github.com/oumi-ai/oumi/pull/872
* Support freezing nested model layers by xrdaukar in https://github.com/oumi-ai/oumi/pull/873
* Update dev_setup.md with minor clarification for use of SSH in GitHub by mkoukoumidis in https://github.com/oumi-ai/oumi/pull/866
* Rewrite CONTRIBUTING.md with new content, fixes and nits by mkoukoumidis in https://github.com/oumi-ai/oumi/pull/867
* Phi3 Vision (`microsoft/Phi-3-vision-128k-instruct`): Basic training support by xrdaukar in https://github.com/oumi-ai/oumi/pull/871
* Update main readme (part 1) by oelachqar in https://github.com/oumi-ai/oumi/pull/874
* Update readme sections (part 2) by oelachqar in https://github.com/oumi-ai/oumi/pull/875
* Refresh documentation pages by oelachqar in https://github.com/oumi-ai/oumi/pull/877
* `Qwen2-VL-2B-Instruct`: basic training config by xrdaukar in https://github.com/oumi-ai/oumi/pull/876
* Update recipes list in readme by oelachqar in https://github.com/oumi-ai/oumi/pull/878
* Update the list of supported VLM-s in README by xrdaukar in https://github.com/oumi-ai/oumi/pull/879
* Update documentation landing page by oelachqar in https://github.com/oumi-ai/oumi/pull/881
* Cap TRL library version to <0.12.2 by jgreer013 in https://github.com/oumi-ai/oumi/pull/882
* [tiny] fix remote inference bug for messages with image urls by oelachqar in https://github.com/oumi-ai/oumi/pull/883
* Adding support for Alpaca Eval by kaisopos in https://github.com/oumi-ai/oumi/pull/880
* Improve error messaging for the remote inference engine. by taenin in https://github.com/oumi-ai/oumi/pull/884
* Move multimodal configs from `configs/examples/vision` to `configs/recipes/vision` by xrdaukar in https://github.com/oumi-ai/oumi/pull/886
* Define internal VLM model configs by xrdaukar in https://github.com/oumi-ai/oumi/pull/885
* Alpaca Eval 2.0 notebook by kaisopos in https://github.com/oumi-ai/oumi/pull/887
* Freeze torchdata>=0.8.0,<0.10.0 by xrdaukar in https://github.com/oumi-ai/oumi/pull/890
* Create torchrun, accelerate CLI wrapper by xrdaukar in https://github.com/oumi-ai/oumi/pull/888
* Update VLM configs to use `oumi distributed torchrun` by xrdaukar in https://github.com/oumi-ai/oumi/pull/891
* Allow None batch size in `GenerationParams` by wizeng23 in https://github.com/oumi-ai/oumi/pull/895
* Evaluation: Saving metadata for future repro by kaisopos in https://github.com/oumi-ai/oumi/pull/892
* Fix missing import precommits when you create a fresh install of Oumi by taenin in https://github.com/oumi-ai/oumi/pull/898
* Update inference documentation by wizeng23 in https://github.com/oumi-ai/oumi/pull/893
* Allow simpler method of specifying LoRA Adapter for eval/inference by wizeng23 in https://github.com/oumi-ai/oumi/pull/889
* Introduce complex `Message.content` with sub-items - Phase 1 by xrdaukar in https://github.com/oumi-ai/oumi/pull/897
* Enable `Message.content` with sub-items - phase 2 by xrdaukar in https://github.com/oumi-ai/oumi/pull/901
* Update configs/notebooks to use typer CLI instead of `oumi.train/eval/infer` modules by wizeng23 in https://github.com/oumi-ai/oumi/pull/902
* Evaluation Notebooks: MT Bench by kaisopos in https://github.com/oumi-ai/oumi/pull/899
* Update chat template/conversation format - Phase 3 by xrdaukar in https://github.com/oumi-ai/oumi/pull/903
* Update `make skycode` to use Oumi launcher by wizeng23 in https://github.com/oumi-ai/oumi/pull/894
* Delete `main()` functions in train/evaluate/infer.py by wizeng23 in https://github.com/oumi-ai/oumi/pull/900
* Update Inference links in oumi README by xrdaukar in https://github.com/oumi-ai/oumi/pull/904
* Define `qwen2-vl-instruct` chat template by xrdaukar in https://github.com/oumi-ai/oumi/pull/907
* Set `max_pixels`/`min_pixels` params for Qwen2 VL model by xrdaukar in https://github.com/oumi-ai/oumi/pull/908
* Update chat template/conversation format - Phase 4 by xrdaukar in https://github.com/oumi-ai/oumi/pull/906
* Update configs to use `oumi distributed torchrun` by wizeng23 in https://github.com/oumi-ai/oumi/pull/905
* [tiny] minor docs updates by oelachqar in https://github.com/oumi-ai/oumi/pull/913
* [docs] clean-up docs pages by oelachqar in https://github.com/oumi-ai/oumi/pull/914
* Initial support for `HuggingFaceTB/SmolVLM-Instruct` by xrdaukar in https://github.com/oumi-ai/oumi/pull/910
* Catch exceptions for missing LoRA adapters by taenin in https://github.com/oumi-ai/oumi/pull/918
* Exclude custom models from Vision-Language model checks. by taenin in https://github.com/oumi-ai/oumi/pull/919
* Refactor `is_custom_model()` check into a function + Minor tests updates by xrdaukar in https://github.com/oumi-ai/oumi/pull/921
* Disable FSDP's `use_orig_params` for PEFT training by wizeng23 in https://github.com/oumi-ai/oumi/pull/869
* Mark `Message` as frozen by xrdaukar in https://github.com/oumi-ai/oumi/pull/923
* Rename `MessageContentItem` to `ContentItem` for brevity. by xrdaukar in https://github.com/oumi-ai/oumi/pull/925
* Delete `configs/examples/llama3_1` by wizeng23 in https://github.com/oumi-ai/oumi/pull/922
* Update OOM FAQ with instructions to tune CUDA alloc by xrdaukar in https://github.com/oumi-ai/oumi/pull/926
* Update `oumi launch` documentation by wizeng23 in https://github.com/oumi-ai/oumi/pull/930
* 90% speed up for the Oumi CLI by taenin in https://github.com/oumi-ai/oumi/pull/929
* Update `docs/datasets/vl_sft.md` and `docs/datasets/local_datasets.md` reflecting the latest changes in Conversation format and VL-dataset setup by xrdaukar in https://github.com/oumi-ai/oumi/pull/928
* Remove remaining references to `MessageContentItem` by xrdaukar in https://github.com/oumi-ai/oumi/pull/931
* Extend dev setup guide by wizeng23 in https://github.com/oumi-ai/oumi/pull/932
* Clean up configs by wizeng23 in https://github.com/oumi-ai/oumi/pull/933
* Add ascii art to all oumi commands in the CLI. by taenin in https://github.com/oumi-ai/oumi/pull/934
* [docs] Reorganize doc pages by oelachqar in https://github.com/oumi-ai/oumi/pull/935
* Fix broken pip targets by wizeng23 in https://github.com/oumi-ai/oumi/pull/927
* [docs] Enable doctest directive and add make doctests by oelachqar in https://github.com/oumi-ai/oumi/pull/937
* Evaluation refactoring by kaisopos in https://github.com/oumi-ai/oumi/pull/912
* Migrate Sky configs to Oumi job configs by wizeng23 in https://github.com/oumi-ai/oumi/pull/938
* [docs] fix misc doc build errors / warnings by oelachqar in https://github.com/oumi-ai/oumi/pull/936
* Update fineweb config by wizeng23 in https://github.com/oumi-ai/oumi/pull/940
* Evaluation refactoring (2nd) by kaisopos in https://github.com/oumi-ai/oumi/pull/939
* Update CLI parser to support `=` separated arguments by xrdaukar in https://github.com/oumi-ai/oumi/pull/942
* Integrating with AlpacaEval (PR 1 of 2) by kaisopos in https://github.com/oumi-ai/oumi/pull/941
* [Bug Fix] AlpacaEval by kaisopos in https://github.com/oumi-ai/oumi/pull/943
* Configure logging for `oumi.telemetry` in `train.py` by xrdaukar in https://github.com/oumi-ai/oumi/pull/944
* Only print ASCII log on main process by xrdaukar in https://github.com/oumi-ai/oumi/pull/946
* [bug] Fix issue with registered dataset loading by oelachqar in https://github.com/oumi-ai/oumi/pull/947
* Alpaca Eval: integration test by kaisopos in https://github.com/oumi-ai/oumi/pull/945
* Print telemetry stats in more compact format by xrdaukar in https://github.com/oumi-ai/oumi/pull/950
* Refactor `conversation_utils` by xrdaukar in https://github.com/oumi-ai/oumi/pull/951
* Improve rank detection for logging by xrdaukar in https://github.com/oumi-ai/oumi/pull/953
* [docs] Add core concepts page by oelachqar in https://github.com/oumi-ai/oumi/pull/952
* [docs][tiny] Update acknowledgement page by oelachqar in https://github.com/oumi-ai/oumi/pull/954
* [docs][tiny] remove unused pages by oelachqar in https://github.com/oumi-ai/oumi/pull/955
* [docs][bug] Fix rendering issue in installation page by oelachqar in https://github.com/oumi-ai/oumi/pull/956
* [docs] Refresh evaluate user guide by oelachqar in https://github.com/oumi-ai/oumi/pull/961
* [docs] Refresh datasets resource section by oelachqar in https://github.com/oumi-ai/oumi/pull/960
* [docs] Refresh models resource section by oelachqar in https://github.com/oumi-ai/oumi/pull/959
* [docs] Refresh infer user guide by oelachqar in https://github.com/oumi-ai/oumi/pull/957
* [docs] Refresh training user guide by oelachqar in https://github.com/oumi-ai/oumi/pull/958
* [docs][tiny] Update links by oelachqar in https://github.com/oumi-ai/oumi/pull/962
* [docs] Update doc generation config by oelachqar in https://github.com/oumi-ai/oumi/pull/963
* [docs] Fix misc docstring rendering issues by oelachqar in https://github.com/oumi-ai/oumi/pull/964
* [docs] Refresh judge user guide by oelachqar in https://github.com/oumi-ai/oumi/pull/965
* [docs] Add documentation page for local training by oelachqar in https://github.com/oumi-ai/oumi/pull/966
* [docs] Update documentation index page by oelachqar in https://github.com/oumi-ai/oumi/pull/967
* [Evaluation] Refactor repro info & add repro info in AlpacaEval by kaisopos in https://github.com/oumi-ai/oumi/pull/949
* [docs] Add documentation page for notebook environment by oelachqar in https://github.com/oumi-ai/oumi/pull/969
* Add more options for saving LoRA models by wizeng23 in https://github.com/oumi-ai/oumi/pull/968
* [docs] Add vscode environment doc page by oelachqar in https://github.com/oumi-ai/oumi/pull/970
* [docs] Cleanup evaluation guide page by oelachqar in https://github.com/oumi-ai/oumi/pull/971
* [docs] Update recipes page by oelachqar in https://github.com/oumi-ai/oumi/pull/972
* [docs] Update tutorials page by oelachqar in https://github.com/oumi-ai/oumi/pull/973
* [docs] Add doc page for generative evals by oelachqar in https://github.com/oumi-ai/oumi/pull/974
* [docs] misc clarifications in train, judge, dataset pages by oelachqar in https://github.com/oumi-ai/oumi/pull/976
* [docs][tiny] Remove train/common_workflows page by oelachqar in https://github.com/oumi-ai/oumi/pull/977
* [docs] Add infer common workflows page by oelachqar in https://github.com/oumi-ai/oumi/pull/978
* Update notebooks setup documentation by wizeng23 in https://github.com/oumi-ai/oumi/pull/975
* [docs] Add supported models page by oelachqar in https://github.com/oumi-ai/oumi/pull/979
* [docs] update inference engine user guide by oelachqar in https://github.com/oumi-ai/oumi/pull/980
* Refactor conversation JSON builders by xrdaukar in https://github.com/oumi-ai/oumi/pull/981
* [docs] Update CLI reference page by oelachqar in https://github.com/oumi-ai/oumi/pull/983
* [docs] Minor updates to datasets intro page by oelachqar in https://github.com/oumi-ai/oumi/pull/984
* Remove internal GCS mentions by wizeng23 in https://github.com/oumi-ai/oumi/pull/982
* Update vision/smollm config names by wizeng23 in https://github.com/oumi-ai/oumi/pull/987
* Update `VLLMInferenceEngine` to support multimodal conversations by xrdaukar in https://github.com/oumi-ai/oumi/pull/986
* Image utils refactorings by xrdaukar in https://github.com/oumi-ai/oumi/pull/985
* [docs] Add training monitoring documentation page by oelachqar in https://github.com/oumi-ai/oumi/pull/988
* Add `torchvision` dependency by xrdaukar in https://github.com/oumi-ai/oumi/pull/989
* Define `configs/recipes/vision/phi3/sft/trl_gcp_job.yaml` for Phi3 Vision by xrdaukar in https://github.com/oumi-ai/oumi/pull/990
* `data.train.text_col` cleanup for VLM-s by xrdaukar in https://github.com/oumi-ai/oumi/pull/991
* [Evaluation] Updating Documentation (1st pass) by kaisopos in https://github.com/oumi-ai/oumi/pull/993
* Fix saving merged lora model by wizeng23 in https://github.com/oumi-ai/oumi/pull/992
* Add an option to return custom datasets as `IterableDataset` by xrdaukar in https://github.com/oumi-ai/oumi/pull/994
* Fix regression in CLI speed by taenin in https://github.com/oumi-ai/oumi/pull/997
* Create alias for json/jsonl datasets by oelachqar in https://github.com/oumi-ai/oumi/pull/998
* [Evaluation] Updating Documentation (Generative Benchmarks) by kaisopos in https://github.com/oumi-ai/oumi/pull/999
* [docs][tiny] Fix broken links, update sft dataset examples by oelachqar in https://github.com/oumi-ai/oumi/pull/1002
New Contributors
* ciaralema made their first contribution in https://github.com/oumi-ai/oumi/pull/715
**Full Changelog**: https://github.com/oumi-ai/oumi/compare/v0.1-alpha...v0.1.1-alpha