batch 3

batch size 2

batch 4

comma

16batch

train agaaa

train aga

fix train

debugging missing grad

update training script

reorder small experts

unfreeze only gate and experts

1 batch size

batch size back to 2

batch size 3 lol

batch size to 4

batch size 8

modify batch and fix tensor issue

tokenize fn

fix

add

tokenize function

debugs

key value

restore

claude attempt 2 dataset

cache diagnostics

sanity

claudeattempt dataset

alternative dataset load

dataset keep in memory

ignore mismatches

fix import

overhaul

reset modeling file

attempt fix and more prints

init expanded model after config change

config shenanigans

debug to train script

changes without config

update training script

restore import name

handle base and product architecture differences

import path

import file name

initial stuff

init no routing