WIP: start implementing LLaVA
59aa1acf
rm scratch buf for now, will revert after cleanup
0f0e7c64
LLaVA image encoder is working. will combine with llama
7e9120f7
Add llava inference code, but it's buggy. debugging
d37ed475
LLaVA is working e2e, needs to optimize memory allocation + cleanup
8690f425
Use ggml_allocr + rm unnecessary code
94eeac35
fix: crlf -> lf
0c2bd797
fix: new line at EoF
204d08be
fix: trailing whitespace
95da79e7
Merge branch 'master' into llava
2a04d0b5
Add readme
444dbce8
Update readme
8af7e210
Some cleanup
54495c94
Are you happy editorconfig?
9b0ec4d2
rm unused batch image preprocessing
8278a736
rm unused import
d78e8163
fix: rm designated initializers
4759bfd6
introduce pad-to-square mode for non-square images
325d2400
are you happy editorconfig?
d75a0315
gitignore /llava
ae01c859
Handle cases where image file does not exist
5009ae90
add llava target to Makefile
96171de5
add support for 13b model variant
d640aae7
Maybe seed is unlucky?
587bde8e
Merge branch 'master' into llava
f1564bb2
Check if apples are compared to apples
ab215879
are you happy editorconfig?
0409ae00
Green-Sky
commented
on 2023-10-11
Use temperature = 0.1 by default
f0f78345
command line: use gpt_params_parse()
2bc1710e
Merge master and fix conflicts
1403d87c
minor
dc913ea3
handle default n_predict
56ccf97b
doomed151
commented
on 2023-10-12
fix typo
e9534ea6
Merge branch 'master' into llava
346e3c16
llava : code formatting, rename files, fix compile warnings
4bc5c9c5
ggerganov
approved these changes
on 2023-10-12
do not use Wno-cast-qual for MSVC
0bd7e69d
Login to write a write a comment.
Login via GitHub