Merge pull request #62 from OpenAccess-AI-Collective/qlora-fixes

update readme and add typehints

fix bool args according to python fire docs

qlora merge and load requires that base model isn't loaded in 4 or 8 bit

Update quickstart config

Add xformers error

Add strict yml

Fix shard config

Merge branch 'main' into winglian-patch-1

Update 4bit notes

Fix missing closing code block

Add more detail on minimum GPU

Add info on Runtime Error

Update dataset type

Add more data formats

Add other minor configs

Update scheduler configs

Update bf16 options

Update lora config

Improve Inference instruction

add discord link to #axolotl-help channel

remove dev specific remark

Add trust_remote_code config

Simplify docker instruction

Add seq length

Remove redundant formats

Add line break

Fix inference command

Update quickstart. Add common error and contribution section.

Add image. Add quickstart. Simplify dataset.

Update Docker instructions

Format dataset types

Update typo

Add all dataset types

Add extra note to Readme

Fix typo

Add new config options to Readme

Feat: Rewrite Readme

move filter to before saving so it doesn't happen everytime, update runpod manual script

Add eval_batch_size for evaluation

ignore config, add python 3.9 (#8)

cleanup empty lines, tweak env for runpod setup

quickstart instructions for starting from runpod (#5)

update readme w compat matrix

update readme

update readme

fix install to work with latest alpaca lora 4bit

4bit quantized support (wip)

cleanup, prep for 4bit quant support

config chooser, update readme instructions, device config, llama flash attention, debug out the labels, fix config key checks, other bugfixes