llama.cpp
server : speculative checkpointing
#19493
Merged
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Overview
Commits
39
Changes
View On
GitHub
server : speculative checkpointing
#19493
ggerganov
merged 39 commits into
ggml-org:master
from
srogmann:feature/speculative-checkpointing
srogmann
requested a review
from
ngxson
159 days ago
srogmann
requested a review
from
ggerganov
159 days ago
github-actions
added
examples
github-actions
added
server
ggerganov
commented on 2026-02-11
srogmann
force pushed
from
c5911892
to
0fa66c27
153 days ago
srogmann
force pushed
from
0fa66c27
to
665893a0
145 days ago
srogmann
force pushed
from
58d612a2
to
edc9b889
139 days ago
ggerganov
commented on 2026-03-09
srogmann
force pushed
from
278bb0a9
to
e0c2f928
132 days ago
srogmann
force pushed
from
e0c2f928
to
f39bb5db
131 days ago
srogmann
force pushed
from
1f629668
to
e0a61a5f
127 days ago
server : speculative decoding using checkpoints
a4237ea0
server : fix draft check with checkpoints
3723f8e5
server : rename spec vars
e002b095
server : log levels
01763e80
server : refactored spec logic to speculative.cpp
e994c4ec
server : renamed spec checkpoints option
bd2f7f2d
server : fix spec checkpoints, logging
af3b630e
speculative : checkpoints with draft model, logging
fe4f859a
server : n_tokens_cur and create_checkpoint in draft
91932ae0
server : fix server_speculative_callback (slot.id)
b5b3ac3b
spec : fix ngram-map/begin idx_last_check
7b7ce932
spec : init ckpt (begin() wasn't called)
21921a1e
chore: update webui build output
7d2814a9
srogmann
force pushed
from
e0a61a5f
to
7d2814a9
121 days ago
srogmann
requested a review
121 days ago
srogmann
requested a review
121 days ago
server : restore sampler in spec checkpoint and clear mem
d0a85689
Merge branch 'master' into pr/19493
63c66f15
ggerganov
commented on 2026-04-01
cont : avoid --spec-use-checkpoints argument
8491e154
cont : remove server_prompt_checkpoint_with_size
e1141d1c
Merge branch 'master' into pr/19493
00c0c4c1
spec : rename (leave_draft_state)
66f0ce2c
Merge branch 'master' into pr/19493
cfb6c4c7
cont : clean-up
ca9e370b
cont : do not ignore partial drafts even if the are short
5b9b0f02
cont : spec callback owned by session
c1adaf99
cont : simplify
dc59f5b1
cont : avoid empty speculative session
50a15510
Merge branch 'master' into pr/19493
cec07350
cont : simplify
ee248c68
cont : simplify
e385affc
cont : enable mtmd speculative decoding
0d4c9118
cont : keep the spec sampler alive
257bb7cc
cont : simplify
2132f818
cont : fix nullptr deref + draft checkpoints
89cf768c
cont : remove common_speculative_accept_response
edf2af7c
cont : remove callback
73f6658c
cont : simplify
2a0a0c74
cont : minor
0c6b7a0d
cont : simplify
66e0c9a2
Merge branch 'master' into pr/19493
ada21d30
cont : fix accepted number
e1f8a1f0
ggerganov
approved these changes on 2026-04-18
ggerganov
added
merge ready
ggerganov
merged
455d8e4b
into master
91 days ago
Login to write a write a comment.
Login via GitHub
Reviewers
ggerganov
ngxson
Assignees
No one assigned
Labels
examples
server
merge ready
Milestone
No milestone
Login to write a write a comment.
Login via GitHub