llama.cpp
quantize: Handle user-defined pruning of whole layers (blocks)
#13037
Merged
Merged
quantize: Handle user-defined pruning of whole layers (blocks) #13037
CISC
merged
22 commits into
ggml-org:master
from
EAddario:prune
Add layer remap logic
671bbee2
Merge branch 'master' into prune
532d5e9a
Add tensor pruning logic
5a805d13
Add imatrix mapping logic
63aa3f33
Add --prune-layers command line option
c128b28b
Fix LLM_KV_BLOCK_COUNT retrieval
056799f3
Add pruned metadata tag to model
4ad1f0a6
Merge branch 'master' into prune
daf7989e
Merge branch 'master' into prune
ef6d2b7f
Merge branch 'master' into prune
70842dc7
Merge branch 'master' into prune
f9c2a7c1
Merge branch 'master' into prune
595fe0fe
Merge branch 'master' into prune
c0d7fa90
Merge branch 'master' into prune
2ea44c41
Merge branch 'master' into prune
a36e0e1c
Fix blk sequence bug and incorporate CISC reccomendations
4661940e
CISC
commented
on 2025-06-21
CISC
commented
on 2025-06-21
Fix wrong block count bug and implement code readability change
f037443d
Merge branch 'master' into prune
8da828ad
CISC
approved these changes
on 2025-06-22
CISC
requested changes
on 2025-06-22
Fix typos
76218ae4
Fix typos
fa5f767b
CISC
commented
on 2025-06-22
Fix wrong split.tensors.count when pruning with --keep-split
9a751e06
Fix bug when pruning last layer
1a435246
CISC
approved these changes
on 2025-06-22
Login to write a write a comment.
Login via GitHub