2 Releases published by 1 person

• v0.1.9 0.1.9

  published Mar 24, 2025

• v0.1.10

  published Mar 25, 2025

12 Pull requests merged by 5 people

• [Evaluations][Custom] Simplify custom eval function output

  #1573 merged Mar 27, 2025

• Minor updates to e2e eternal tests scripts

  #1572 merged Mar 27, 2025

• [GRPO] Reorganize letter counting configs

  #1570 merged Mar 26, 2025

• test_eval_e2e.py: Provide an option to use oumi evaluate command for some configs

  #1571 merged Mar 25, 2025

• e2e eval integration tests updates

  #1569 merged Mar 25, 2025

• Refactor dataset creation to take DataParams as input instead of TrainingConfig

  #1568 merged Mar 25, 2025

• [Evaluation] Bug: Multiple GPUs attempt to save in the same folder

  #1567 merged Mar 25, 2025

• Increment pillow version for compatibility with Python 3.13

  #1566 merged Mar 24, 2025

• Fix printing errors in oumi env for non-string values.

  #1565 merged Mar 24, 2025

• [minor] phi4 train improvements

  #1564 merged Mar 23, 2025

• Phi-4 basic inference with native/vllm

  #1563 merged Mar 23, 2025

• phi 4 multimodal training version 1 ( with limitations )

  #1555 merged Mar 22, 2025

3 Pull requests opened by 2 people

• [Remote Inference] Supported params are ignored

  #1562 opened Mar 21, 2025

• [GRPO][Eval] Add letter counting eval

  #1574 opened Mar 27, 2025

• [Evaluation] Custom functions: support for 2 output types

  #1575 opened Mar 27, 2025

1 Issue closed by 1 person

• Pillow 10.3.0 not supported for Python 3.13

  #1525 closed Mar 24, 2025

5 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• Add s1k dataset implementation

  #1527 commented on Mar 21, 2025 • 2 new comments

• [Feature] Unsloth fine-tuning support

  #1322 commented on Mar 24, 2025 • 0 new comments

• Add tokenized example debugging during training

  #1520 commented on Mar 27, 2025 • 0 new comments

• Added 3 Pixmo vision-language datasets

  #1523 commented on Mar 27, 2025 • 0 new comments

• Add KTO support for preference tuning

  #1538 commented on Mar 25, 2025 • 0 new comments