Commit History

Autor SHA1 Mensaxe Data
  AlpinDale eef647deab fix: greedy decoding in TPU hai 4 meses
  AlpinDale 8d88814475 chore: reduce XLA compile times hai 4 meses
  AlpinDale 386ad8bef6 feat: tensor parallelism for TPU with ray hai 4 meses
  AlpinDale f91991f584 fix: f-string fixes hai 4 meses
  AlpinDale b6c4dfce23 chore: refactor TPU model runner and worker hai 4 meses
  AlpinDale 8c2dd39500 chore: remove multimodal stuff from TPU hai 4 meses
  AlpinDale e1475fbec7 feat: MoE support with Pallas GMM kernel for TPUs hai 4 meses
  AlpinDale 1cb06835a0 fix: TPU multimodal kwargs and outlines installation in TPU docker hai 4 meses
  AlpinDale 4f7d212b70 feat: remove vision language config hai 4 meses
  AlpinDale 4599c98f99 feat: dynamic image size support for VLMs hai 4 meses
  AlpinDale 301ec7c77d fix: pad slot id in tpu runner hai 5 meses
  AlpinDale cdff8e89f9 feat: introduce `DraftModelRunner` hai 5 meses
  AlpinDale 85ef2fe8b1 chore: clean up placeholder symbols hai 5 meses
  AlpinDale fcd58614f4 feat: support parallel sampling and swapping in TPU hai 5 meses
  AlpinDale d36b88b301 fix: raise errors if using unsupported samplers on TPU hai 5 meses
  AlpinDale 608e8e1310 chore: refactor TPU backend to make it more similar to GPU backend hai 5 meses
  AlpinDale fe21123a1c feat: TPU support (#570) hai 5 meses