What's Changed

• Peft LoRA support by @BlackSamorez in #68

Full Changelog: v1.2.2...v1.2.3

What's Changed

• Torch distributed hotfix by @BlackSamorez in #70

Full Changelog: v1.2.1...v1.2.2

What's Changed

• Set seed for tests reproducibility by @BlackSamorez in #64
• Mention linear speedup in Readme by @BlackSamorez in #65
• LLaMa models by @BlackSamorez in #53

Full Changelog: v1.2.0...v1.2.1

Config refactoring

state_actions are now interfaces:

class StateAction(ABC):
    @abstractclassmethod
    def __call__(self, tensor: Tensor, rank: int) -> Tensor:
        pass

    @abstractclassmethod
    def undo(self, tensors: Sequence[Tensor]) -> Tensor:
        pass

Callables and tuples of callables are still allowed but will produce deprecation warning.

Full Changelog: v1.1.4...v1.2.0

What's Changed

• CodeGen config by @BlackSamorez in #61
• Added int8 LLMs demo link by @BlackSamorez in #62
• Converting state dicts without model creation by @BlackSamorez in #60

Full Changelog: v1.1.3...v1.1.4

What's Changed

• Unpersistent buffers meta loading fix by @BlackSamorez in #57
• _reorder_cache fix for generation utils by @BlackSamorez in #56
• Shard parameters initial dispatch fix by @BlackSamorez in #58
• New version for dispatch hotfix by @BlackSamorez in #59

Full Changelog: v1.1.1...v1.1.3

What's Changed

• GPT NeoX config by @BlackSamorez in #55
• Removing accelerate hooks before splitting the model by @BlackSamorez in #52
• Meta devices support by @BlackSamorez in #54

Full Changelog: v1.1.0...v1.1.1

What's Changed

• Fixed PyPi link in readme by @BlackSamorez in #40
• Adding support for more model architectures by @BlackSamorez in #47
• Replace architecture with model_type by @BlackSamorez in #44
• Saving utilities by @BlackSamorez in #49
• Version update by @BlackSamorez in #50

Full Changelog: v1.0.25...v1.1.0

What's Changed

• hotfix canonic torch.device by @justheuristic in #37
• _TensorParallelWrapper attribute forwarding by @BlackSamorez in #38
• Version update by @BlackSamorez in #39

Full Changelog: v1.0.24...v1.0.25

What's Changed

• got rid of some unnecessary lines of code by @IaroslavLisniak in #13
• T5 example by @justheuristic in #30
• Hugging Face encoder-decoder architecture support by @BlackSamorez in #14

Full Changelog: https://github.com/BlackSamorez/tensor_parallel/commits/v1.0.24