llama.cpp
llama : simplify Mamba with advanced batch splits
#8526
Merged
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Overview
Commits
19
Changes
View On
GitHub
llama : simplify Mamba with advanced batch splits
#8526
compilade
merged 19 commits into
master
from
compilade/batch-splits
llama : advanced batch splits
c51daefc
Merge branch 'master' into compilade/batch-splits
22504ec6
github-actions
added
ggml
compilade
added
refactoring
compilade
added
Review Complexity : Medium
compilade
marked this pull request as draft
1 year ago
llama : fix integer signedness mixing
2e4adb47
llama : logits_all has priority over batch->logits
7b7db0bb
ggerganov
commented on 2024-07-17
github-actions
added
testing
ggerganov
force pushed
to
7b7db0bb
1 year ago
llama : apply suggestions
1fb5d4fd
llama : fix t5 segfault
1725de76
Merge branch 'master' into compilade/batch-splits
9c0a61f8
Merge branch 'master' into compilade/batch-splits
0dea4263
llama : fix Mamba session save and restore
704a3033
llama : minor cosmetic changes
952ed35b
Merge branch 'master' into compilade/batch-splits
5679a3bd
llama : rename llama_reorder_outputs to llama_output_reorder
cfd5a113
compilade
marked this pull request as ready for review
1 year ago
minor : add struct members for clarity
0596a99f
ggerganov
approved these changes on 2024-08-09
Merge branch 'master' into compilade/batch-splits
702e1995
llama : fix T5 segfault again
652e9b0d
llama : fix Mamba pooled embeddings with multiple sequences
b264eddb
llama : add llama_model_is_recurrent to simplify figuring that out
1be5ea7d
compilade
added
merge ready
Merge branch 'master' into compilade/batch-splits
80d9d2a5
ggerganov
approved these changes on 2024-08-21
llama : fix simple splits when the batch contains embeddings
80626503
compilade
merged
a1631e53
into master
1 year ago
Login to write a write a comment.
Login via GitHub
Reviewers
ggerganov
Assignees
No one assigned
Labels
testing
refactoring
Review Complexity : Medium
ggml
merge ready
Milestone
No milestone
Login to write a write a comment.
Login via GitHub