Fix plumbing of apply_chat_template/fewshot_as_multiturn in mergekit-…

…evolve (arcee-ai#535)

LoRA extraction fixes (arcee-ai#522)

Addresses arcee-ai#521.

Also adds:
* `--lora-merge-dtype` to specify dtype to use when applying LoRA adapters to models
* `--gpu-rich` alias for convenience
* Organize display of options in `--help`

Compute-graph based `mergekit-extract-lora` (arcee-ai#505)

Now with better embedding handling, multi-gpu execution, and lazy
loading/saving of tensors.

When extracting a LoRA from an 8B model, execution time goes from ~6
minutes down to 40 seconds with `--cuda --multi-gpu` on an 8-GPU
machine.

Additionally, the `--sv-epsilon` flag can be used to set a tolerance for
singular values to opportunistically reduce rank when the fine tuned
difference is inherently lower rank.

Also reimplement a couple of merge methods using the `@easy_define`
decorator and add some missing tests.

Update copyright (arcee-ai#497)

Also bump dependencies.

Update README.md

Reorganize slightly

Bump copyright date to 2024

Copy tokenizer