trixyL/simplestories-ar-demo

refactor: default temp to 0.8

add: orange button

add: temperature slider

dump: 4k model

update: card color

refactor: faster top p sampling

fix: rolling window max size

revert: sliding window gen

add: bump max tokens

revert: new inference strat, just bumping max token

fix: max temp @ 1

add: chunking + longer text generation

fix: spaces gpu decorator

default: "attention_sdp_backend": "auto",

update: streamlined generation shenanigans

update: ui

fix; uncond prompts and max new tokens

fix: non existing flag

fix: space name and tokenizer imports

fix: missing pytz dep

dump: model/inference transformerlm code and gradio app

initial commit