Distributed inference via MPI #2099
ggerganov
merged
29 commits into
ggml-org:master
from
evanmiller:mpi
MPI support, first cut
f85785f6
fix warnings, update README
d05ca74d
fixes
668ba5fe
wrap includes
042c5b27
SlyEcho
commented
on 2023-07-05
Merge branch 'master' into mpi
32deabfd
PR comments
06a23934
Update CMakeLists.txt
1f0a2cfe
Add GH workflow, fix test
55207ba2
mqy
commented
on 2023-07-07
Add info to README
ef61acfb
mpi : trying to move more MPI stuff into ggml-mpi (WIP) (#2099)
3232db62
mpi : add names for layer inputs + prep ggml_mpi_graph_compute()
e339d355
mpi : move all MPI logic into ggml-mpi
01abb3b3
mpi : various fixes - communication now works but results are wrong
c717c518
mpi : fix output tensor after MPI compute (still not working)
ef37dd14
mpi : fix inference
beadbf33
mpi : minor
9da9d26c
Merge branch 'master' into mpi
0f557c2a
Add OpenMPI to GH action
4a9a4748
[mpi] continue-on-error: true
03cc12be
Merge branch 'mpi' into refactor-mpi
81c5ddd5
Merge pull request #1 from ggerganov/refactor-mpi
1c3a15c5
mpi : fix after master merge
166db36c
[mpi] Link MPI C++ libraries to fix OpenMPI
f085a57d
tests : fix new llama_backend API
00b8aa1e
Merge remote-tracking branch 'refs/remotes/origin/mpi' into mpi
666a15ae
Merge branch 'mpi' of github.com:evanmiller/llama.cpp into mpi
b18e4ad4
[mpi] use MPI_INT32_T
ada1a2aa
mpi : factor out recv / send in functions and reuse
c3c3ef11
mpi : extend API to allow usage with outer backends (e.g. Metal)
eaef2d0e
ggerganov
approved these changes
on 2023-07-10
Login to write a write a comment.
Login via GitHub