v0.1.5

oumi-ai/oumiv0.1.5Feb 20, 2025by wizeng23

AI Summary

This release adds support for several new datasets (Docmatix, the_cauldron, Tulu3), integrates SambaNova for inference, and adds SGLang multi-image support.

Key Highlights

  • Added support for Docmatix and the_cauldron multimodal datasets
  • Integrated SambaNova Systems for oumi inference
  • Added multi-image support in SGLang inference engine
  • Adopted new Llama 3.1 HuggingFace names

New Features

  • Tulu3 SFT Mixture Dataset Support
  • WildChat-50M Reproduction support
  • CoALM (formerly CALM) recipe
  • Added `oumi://` prefix support to fetch and infer commands
  • DeepSeek R1 Distill Llama configs

Full Release Notes

## What's Changed
* Fix the remainder of our configs by @wizeng23 in https://github.com/oumi-ai/oumi/pull/1356
* Adopt new Llama 3.1 HF names by @wizeng23 in https://github.com/oumi-ai/oumi/pull/1357
* Define `OUMI_USE_SPOT_VM` env var and start using it to override `use_spot` param  by @xrdaukar in https://github.com/oumi-ai/oumi/pull/1359
* Support HuggingFaceM4/Docmatix dataset by @vishwamartur in https://github.com/oumi-ai/oumi/pull/1342
* [nit] update default issue names by @oelachqar in https://github.com/oumi-ai/oumi/pull/1367
* Update sft_datasets.md by @penfever in https://github.com/oumi-ai/oumi/pull/1349
* Have GitHub Trending image hyperlink to GitHub Trending page by @wizeng23 in https://github.com/oumi-ai/oumi/pull/1370
* Update the link for the trending banner. by @taenin in https://github.com/oumi-ai/oumi/pull/1371
* Move code to disable caching in `model.config` to a helper function by @xrdaukar in https://github.com/oumi-ai/oumi/pull/1378
* Update transformers version to 4.48 by @wizeng23 in https://github.com/oumi-ai/oumi/pull/1372
* Update notebooks to improve their Colab experience by @wizeng23 in https://github.com/oumi-ai/oumi/pull/1380
* Add proper labels and types to new Bugs and Feature Requests. by @taenin in https://github.com/oumi-ai/oumi/pull/1383
* Upgrade omegaconf to 2.4.0dev3 by @wizeng23 in https://github.com/oumi-ai/oumi/pull/1384
* Support HuggingFaceM4/the_cauldron dataset by @vishwamartur in https://github.com/oumi-ai/oumi/pull/1366
* Update our FAQ for tips about installing oumi on Windows by @taenin in https://github.com/oumi-ai/oumi/pull/1385
* Cleanup `HuggingFaceM4/Docmatix` and `HuggingFaceM4/the_cauldron` multimodal datasets by @xrdaukar in https://github.com/oumi-ai/oumi/pull/1387
* Remove uneeded env vars from job configs by @wizeng23 in https://github.com/oumi-ai/oumi/pull/1390
* Remove transformer version override for `HuggingFaceTB/SmolVLM-Instruct` in launcher script by @xrdaukar in https://github.com/oumi-ai/oumi/pull/1388
* [Small Refactor] Moving the inference engine def outside the inference config by @kaisopos in https://github.com/oumi-ai/oumi/pull/1395
* Evaluation - LM Harness: Adding vLLM support by @kaisopos in https://github.com/oumi-ai/oumi/pull/1379
* Remove `Docmatix` dataset references from docstrings VLM config examples by @xrdaukar in https://github.com/oumi-ai/oumi/pull/1397
* Fixed broken link in Oumi - A Tour.ipynb notebook by @ciaralema in https://github.com/oumi-ai/oumi/pull/1398
* Fix broken links in notebooks. by @taenin in https://github.com/oumi-ai/oumi/pull/1402
* Create a client for communicating with a Slurm node via SSH. by @taenin in https://github.com/oumi-ai/oumi/pull/1389
* [tiny] Remove references to missing job configs in README by @wizeng23 in https://github.com/oumi-ai/oumi/pull/1404
* Train+Inference with Qwen 2.5 VL (3B) by @optas in https://github.com/oumi-ai/oumi/pull/1396
* Add a Slurm cluster and cloud to the oumi launcher. by @taenin in https://github.com/oumi-ai/oumi/pull/1406
* Move `pretokenize` script from `scripts/pretokenize/` to `scripts/datasets/pretokenize/` by @xrdaukar in https://github.com/oumi-ai/oumi/pull/1412
* Create a script to save `Conversation`-s from SFT datasets into  `.jsonl` file by @xrdaukar in https://github.com/oumi-ai/oumi/pull/1413
* [Evaluation] LM Harness refactor by @kaisopos in https://github.com/oumi-ai/oumi/pull/1410
* Update `save_conversations` tool by @xrdaukar in https://github.com/oumi-ai/oumi/pull/1421
* [SambaNova] Integrate SambaNova Systems to oumi inference by @ctseng777 in https://github.com/oumi-ai/oumi/pull/1415
* [Μinor] Equating Qwen's 2.5 chat-template to version's 2.0 by @optas in https://github.com/oumi-ai/oumi/pull/1419
* Add requirements header to configs and clean them up by @wizeng23 in https://github.com/oumi-ai/oumi/pull/1411
* Updated oumi infer to support CLI argument for system prompt by @Spaarsh in https://github.com/oumi-ai/oumi/pull/1422
* [Evaluation] LM Harness remote server support by @kaisopos in https://github.com/oumi-ai/oumi/pull/1414
* [Feature] Add Tulu3 SFT Mixture Dataset Support by @bwalshe in https://github.com/oumi-ai/oumi/pull/1381
* Support Multimodal inference with multiple images and PDF-s in `NATIVE` engine by @xrdaukar in https://github.com/oumi-ai/oumi/pull/1424
* Update notebooks to run on Colab by @wizeng23 in https://github.com/oumi-ai/oumi/pull/1423
* Add calm recipe. by @taenin in https://github.com/oumi-ai/oumi/pull/1425
* Update VLM sample `oumi infer -i` commands by @xrdaukar in https://github.com/oumi-ai/oumi/pull/1428
* Provide example show to start SGLang server using Docker by @xrdaukar in https://github.com/oumi-ai/oumi/pull/1429
* Multi-image support in SGLang inference engine by @xrdaukar in https://github.com/oumi-ai/oumi/pull/1426
* Calm readme by @emrecanacikgoz in https://github.com/oumi-ai/oumi/pull/1432
* WildChat-50M Reproduction by @penfever in https://github.com/oumi-ai/oumi/pull/1433
* Add WildChat support by @penfever in https://github.com/oumi-ai/oumi/pull/1348
* Create `pad_to_max_dim_and_stack()` function in torch_utils by @xrdaukar in https://github.com/oumi-ai/oumi/pull/1435
* use deterministic by @penfever in https://github.com/oumi-ai/oumi/pull/1434
* Additional HF trainer parameters for config by @penfever in https://github.com/oumi-ai/oumi/pull/1436
* Set a better default for vllm inference GPU usage. by @taenin in https://github.com/oumi-ai/oumi/pull/1437
* Added fetch command and modified infer command to resolve oumi:// by @Spaarsh in https://github.com/oumi-ai/oumi/pull/1439
* Require an inference config for oumi infer. by @taenin in https://github.com/oumi-ai/oumi/pull/1443
* Make the tulu3 unit tests hermetic. by @taenin in https://github.com/oumi-ai/oumi/pull/1446
* Add 2 more sample PDF-s with 1 and 2 pages under `testdata/pdfs` by @xrdaukar in https://github.com/oumi-ai/oumi/pull/1427
* Enable ability to override list values in config via CLI by @wizeng23 in https://github.com/oumi-ai/oumi/pull/1430
* Renamed CALM to CoALM by @jgreer013 in https://github.com/oumi-ai/oumi/pull/1450
* Add support for Docmatix dataset to multimodal training script by @xrdaukar in https://github.com/oumi-ai/oumi/pull/1449
* Update oumi launch status to show clusters with no running jobs. by @taenin in https://github.com/oumi-ai/oumi/pull/1451

## New Contributors
* @vishwamartur made their first contribution in https://github.com/oumi-ai/oumi/pull/1342
* @ctseng777 made their first contribution in https://github.com/oumi-ai/oumi/pull/1415
* @Spaarsh made their first contribution in https://github.com/oumi-ai/oumi/pull/1422
* @bwalshe made their first contribution in https://github.com/oumi-ai/oumi/pull/1381
* @emrecanacikgoz made their first contribution in https://github.com/oumi-ai/oumi/pull/1432

**Full Changelog**: https://github.com/oumi-ai/oumi/compare/v0.1.4...v0.1.5