v0.1.12

What's Changed

• Add vllm to gpu optional dependencies by @wizeng23 in #1614
• [HallOumi] Update inference notebook by @wizeng23 in #1613
• Update llama4 GCP jobs for non-dev environments. by @taenin in #1621
• Update transformers to 4.51.0 by @wizeng23 in #1620
• Lazy load skypilot by @taenin in #1622
• Add additional_model_kwargs and additional_trainer_kwargs to train function by @hommayushi3 in #1624
• Added 3 Pixmo vision-language datasets by @jrwana in #1523
• [GRPO] Add notebook to demonstrate GRPO & evaluation for letter counting by @wizeng23 in #1625
• [Remote Inference] Update Default Params by @kaisopos in #1630
• Update trl to 0.16 by @wizeng23 in #1631
• Support custom processor args in ModelParams by @nikg4 in #1634
• Support BerryBench evaluation by @wizeng23 in #1635
• [Remote Inference] Error checking for api_key by @kaisopos in #1638
• Rename cnn_mnist_example to cnn_mnist_tutorial by @wizeng23 in #1640
• [Remote Inference][GCP] Constructing api_url from the Project ID and Region by @kaisopos in #1636

New Contributors

• @jrwana made their first contribution in #1523

Full Changelog: v0.1.11...v0.1.12