remove bitnet handling completely

Update utils/models.py

Roll back interruption changes

Update interruption method

Update inference interruption

Update utils/models.py

restore to live status

remove all explicit device assigns

disable bitnet and set models to previous status

try remote_code=False and device_map=auto

try using pipe but explicitly set model to be BitNetForCausalLM

remove explicit tokenizer device assignment

separate bitnet from generic pipeline

try bitnet with transformers 4.52

remove some unneeded lines, fix pipe issue

reactivate gemma models, add some ice cream, support peft

fixed qwen, disabled gemma3-1b and minicpm, re-enabled cogito

maybe fixed qwen3

use chat temploate try 2

remove minicpm3 for incompatibility reasons

switch to applying chat template

fix argument passing

add gen prompt and kwargs dicts

hopefully fixed qwen

change where trust_remote is applied

trust_remote_code=True

add a ton of models and update arena dataset

upgraded highlighting, added bitnet

remove debugging

format prompts and add debug log

remove cogito for gpu-hour reasons and add approved llama models

add additional models

fix typo

gemma 3 4b misaligned bug, so removing that as an option for now

update model list

revert to sequential processing

typo fix

prepare for zeroGPU

specified device

vllm backend swap v1

proper threaded generation interrupt

Handled interruption (#10)

ready hyperlinks for leaderboard

first pass at async handling

switched models to using full context

fix models

remove gated models for now

v1 inference code

add UI and data processing