_prepare_generation_config bugfix (failed due to version update in transformers) (#25)

update readme

Patching hf bug that creates wrong cache length if only inputs_embeds are passed to the model (#19)

fixed cache over-alloc bug (#17)

DeciLMForCausalLM(DeciLMPreTrainedModel, GenerationMixin) for v4.50 (#16)

add batch_size attribute to VariableCache (#15)

nvidia-open-model-license (#14)

v4.46 support (#7)

v4.45 support (#6)

transformers>=4.44.2, backward compat