llama.cpp
Refactor lora adapter support
#8332
Merged
Overview
Commits
42
Changes
View On GitHub

Refactor lora adapter support #8332

ngxson merged 42 commits into ggml-org:master from ngxson:xsn/fix_lora
ngxson
ngxson lora: load to devide buft
67c5e14d
ngxson add patch tensor function
e9d7b6c0
ngxson correct tensor patch
4e28ad40
ngxson llama_lora_adapter_apply
1b4ffbac
ngxson ngxson requested a review from slaren slaren 1 year ago
github-actions github-actions added ggml
ngxson correct ggml_backend_tensor_copy
b88ce0f8
slaren
ngxson
slaren
slaren
ngxson
ngxson add llm_build_mm
f6d090d7
ngxson
ngxson Merge branch 'master' into xsn/fix_lora
a1666aaa
ngxson fix auto merge
30faf1f3
slaren
slaren
slaren commented on 2024-07-07
ggerganov
ggerganov commented on 2024-07-08
ngxson update based on review comments
79e29827
ngxson add convert script
847135aa
ngxson no more transpose A
712fecba
ngxson add f16 convert
84288ff9
ngxson Merge branch 'master' into xsn/fix_lora
41ced241
ngxson add metadata check
0e161889
ngxson ngxson requested a review from slaren slaren 1 year ago
ngxson ngxson requested a review from ggerganov ggerganov 1 year ago
ngxson add sanity check
6c617e20
ngxson fix ftype
7a83f200
ngxson add requirements
d52455f2
ngxson fix requirements
802565ca
ngxson fix outfile
95b3eb05
ngxson Merge pull request #8 from ngxson/xsn/fix_lora_convert
03d24cae
github-actions github-actions added python
ngxson
slaren
slaren commented on 2024-07-09
amr3k
amr3k approved these changes on 2024-07-09
compilade
compilade commented on 2024-07-09
ngxson conversion: only allow selected models
ee2b35c6
ngxson fix types
713665db
slaren cuda : do not use dmmv if the tensor does not have enough cols
f15167a4
slaren llama : lora fixes
9841fbda
ngxson Merge pull request #9 from ggerganov/sl/fix_fix_lora
4fe0861a
slaren
ngxson do not disable mmap with lora
1faf7e5b
ngxson Merge branch 'master' into xsn/fix_lora
e68344cb
ngxson llm_build_lora_mm_id
916e9592
compilade convert_lora : MoE LoRA conversion support
9d96328b
mofosyne mofosyne added Review Complexity : Medium
compilade convert_hf : simplify modify_tensors for InternLM2
8956543c
compilade llama : use llm_build_lora_mm in most model graphs
87301bdd
ggerganov
ggerganov approved these changes on 2024-07-15
ngxson Merge branch 'master' into xsn/fix_lora
703573f6
ngxson auto scale
42415a48
ngxson Revert "auto scale"
5b181182
ngxson remove redundant params
f68d0924
ngxson ngxson requested a review from slaren slaren 1 year ago
ngxson Merge branch 'master' into xsn/fix_lora
b704448a
slaren
slaren
slaren commented on 2024-07-15
slaren
slaren commented on 2024-07-15
ngxson Apply suggestions from code review
9175f4b7
ngxson change kv metadata
0ba23bad
ngxson
ngxson
slaren
slaren approved these changes on 2024-07-15
ngxson ngxson added merge ready
compilade
compilade commented on 2024-07-15
ngxson ngxson removed merge ready
ngxson move add_type to __init__
b1c40695
ngxson Merge branch 'master' into xsn/fix_lora
4d9ac0f3
compilade convert_hf : move add_type to main()
d09382fa
ngxson Merge branch 'master' into xsn/fix_lora
383b6bce
compilade
compilade commented on 2024-07-15
ngxson ngxson added merge ready
compilade
compilade approved these changes on 2024-07-15
ngxson
ngxson ngxson merged 97bdd26e into master 1 year ago
zhipenghan
ltoniazzi
ltoniazzi commented on 2024-07-25

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
python Review Complexity : Medium ggml merge ready
Milestone
No milestone
Terms Of Use Privacy Policy Refund Policy Impressum