Support Adept Persimmon 8b #3410
Produces garbage output
7cdc3eaa
wip: correct tensors up to RoPE
4bcf412d
correct tensors thru RoPE
c9e1446f
Correct outputs through masked & softmax'd KQ
d1b40efc
fp32 works
db2181a4
Rename adept->persimmon
3f317999
Merge branch 'master' of github.com:phillip-kravtsov/llama.cpp into p…
720503ba
Produces correct outputs
d61eed0a
Merge branch 'master' of github.com:ggerganov/llama.cpp into phillip-…
d0a7143f
clean up convert scripts
fa92f6e8
remove printing logic from ggml.c
c28a6c5b
remove prints from llama.cpp & fix merge
47dcb9fc
Merge branch 'master' of github.com:ggerganov/llama.cpp into phillip-…
7473773c
trivial cleanups
d904aff0
Add offload funcs
ec0ce978
update conversion script to directly take adept artifacts rather than…
3db04db2
Fix norm eps bug
f28f52c6
Merge branch 'master' of github.com:ggerganov/llama.cpp into phillip-…
d93cf1ea
goerch
commented
on 2023-09-30
Merge branch 'master' of github.com:ggerganov/llama.cpp into phillip-…
574a9e12
Support sqr and concat on metal, persimmon-8b-q4 runs correctly
2b565916
ggerganov
approved these changes
on 2023-10-02
Small changes from review
e6bf87f7
Formatting changes
cd4d3df8
Minor changes to conversion script
422b1108
Merge branch 'master' of github.com:ggerganov/llama.cpp into phillip-…
5a0990c1
Remove old script
7a279fe5
Fix editorconfig formatting
c90ed9f1
Merge branch 'master' of github.com:ggerganov/llama.cpp into phillip-…
5d259d35
Fix build
1d518d65
Merge branch 'master' of github.com:ggerganov/llama.cpp into phillip-…
0c1a8f67
add overlooked offload code ggml-ci
485a471e
ggerganov
merged
0e797c2f
into master 2 years ago
Assignees
No one assigned
Labels
high priority
model
Login to write a write a comment.
Login via GitHub