openaccess-ai-collective/ggml-ui

fix workflow since we use tag instead of releasE

fix docker run

fix docker tag

ctransformers inconsistencise

fix paths

build in /build root

more fixes

fix pipeline names

more fixes to actions

Fix ggml

move v in version to part of the arg

support for ctransformers

upgrade to v0.1.53

update to latest 0.1.52 llama.cpp

remove revision from yml

various fixes post-release

fix for ZWSP

fix chat prompts for vicuna

use stable vicuna as the base for everyone to demo

fix streaming on instruct

use new wizard mega ggml models

fix stop tokens to match new prompt formatting, stream instruct response, add comments about concurrency to config

remove chat and instruct, fix instruct prompt format

fix prompt for sharegpt chat

shout out to @TheBloke

fix layout, max size back to 1, llama.cpp doesn't like parallel calls

loading torch doesn't play nice, just use config for now

try to fix combining gr.interface with blocks, try to increase concurrency on larger gpus

support tabbed interface and fix instruct to use GPU

patience

update copy

more fixes to config file based llama

update instruct to use gpu, add verbose description

default to instruct for now

cleanup instruct app. add queue so UI notifies users

single concurrency is probably safest

base model in config is unnecessary, and fix grammar

move byline around

tweak the system prompt

add prediction settings to ui

docker cp doesn't support wildcards

link model attributions, use config.yml for some of the chat settings, increase context size

rm docker implementation, add llama-cpp-python builder github actions, update copy to identify model in ui

fix chat history, update settings to use GPU

require torch and nvidia-cublas

rollback to gradio

try out docker sdk

Update requirements.txt

let's see what happens on a GPU

retry on failed hub fetch