0.1.9

What's Changed

• Add QwQ full fine-tune and QLoRA configs by @wizeng23 in #1518
• Update TRL to 0.15 and fix Liger/dataset code by @wizeng23 in #1507
• [tiny] Remove vLLM Colab link and fix Alpaca Eval quickstart by @wizeng23 in #1530
• Evaluation: Inference optimizations by @kaisopos in #1522
• Qwen2.5 VL: Replace "from source" install with transformers>=0.49 by @nikg4 in #1528
• [Evaluation] Renaming evaluation_platform → evaluation_backend by @kaisopos in #1526
• [tiny] Clean up datasets code by @wizeng23 in #1529
• Minor logging improvements in BaseMapDataset by @nikg4 in #1532
• Upload scripts used in a Weekly Walkthrough by @taenin in #1533
• Update VisionLanguageConversationFeatureGenerator by @nikg4 in #1531
• [docs] add security.md by @oelachqar in #1534
• [Evaluation] Custom evaluation notebook: a reliability classifier by @kaisopos in #1535
• Multimodal: Limit max number of images per Conversation by @nikg4 in #1536
• Auto-populate and validate params specific to vision_language_sft collator in TrainingConfig by @nikg4 in #1537
• Update Oumi Env to use Rich formatting by @taenin in #1541
• Update oumi launch to use Rich formatting by @taenin in #1543
• Update oumi evaluate to use rich formatting. by @taenin in #1544
• Update the CLI to replace all prints with Rich prints. by @taenin in #1547
• Render the oumi env command as a shell block in bug reports. by @taenin in #1548
• Define Conversation proto bufs by @nikg4 in #1550
• [Evaluation] Modifying Alpaca Eval results format to be consistent with LM Harness by @kaisopos in #1551
• Augmenting logging training/model statistics by @optas in #1545
• Misc no-op code cleanups by @nikg4 in #1553
• Add code used for the evaluation demo. by @taenin in #1556
• Add OUMI_FORCE_EDITABLE_INSTALL env var to do editable Oumi install from source in job configs by @wizeng23 in #1420
• Add letter counting GRPO example by @wizeng23 in #1539
• Remove UV install from notebooks as this breaks colab by @taenin in #1558
• [Evaluation] Updates in hallucination notebook by @kaisopos in #1552
• [Evaluations] Custom evals: Adding support for eval_kwargs by @kaisopos in #1557
• Logging message update in log_number_of_model_parameters by @nikg4 in #1560
• [Evaluation][Custom] Removing restrictions and better error checking by @kaisopos in #1561
• Support text truncation (max_length) for vision_language_sft collator by @nikg4 in #1559
• phi 4 multimodal training version 1 ( with limitations ) by @optas in #1555
• Phi-4 basic inference with native/vllm by @optas in #1563
• [minor] phi4 train improvements by @optas in #1564
• Fix printing errors in oumi env for non-string values. by @taenin in #1565

Full Changelog: v0.1.8...v0.1.9