llama.cpp
llama : add gpt-oss
#15091
Merged

llama : add gpt-oss #15091

ggerganov merged 48 commits into master from gpt-oss-mxfp4
ggerganov
ngxson oai moe
81991fcd
ngxson compat with new checkpoint
917f9233
ngxson add attn sink impl
a4ab8693
ngxson add rope scaling yarn
3801c364
ngxson logits match with latest transformers code
13f39f6b
ngxson wip chat template
b3594b30
ngxson Merge branch 'master' into xsn/oai_moe
bd571580
ngxson rm trailing space
089a7ab4
ngxson use ggml_scale_bias
4d01b36b
ngxson Merge branch 'master' into xsn/oai_moe
f271cc80
ngxson rm redundant is_swa_all
106b17e5
ngxson convert interleaved gate_up
e2c1beb3
ggerganov Merge remote-tracking branch 'gg-public/master' into xsn/oai_moe-gg
4431c823
ggerganov Merge remote-tracking branch 'gg-public/master' into xsn/oai_moe-gg
fe9b818b
ggerganov Merge remote-tracking branch 'gg-public/master' into xsn/oai_moe-gg
539c2b63
ggerganov graph : fix activation function to match reference (#7)
039a6f16
ggerganov Merge branch 'master' into xsn/oai_moe-gg
aa240b99
ggerganov Merge branch 'master' into xsn/oai_moe-gg
32a654c2
ggerganov vocab : handle o200k_harmony special tokens
13f3568c
ggerganov ggml : add attention sinks support (#1)
e59b2eb1
ngxson repack mxfp4 upon conversion
832dc26c
ngxson clean up a bit
c68069d1
ngxson enable thinking
423b1919
ngxson add quick hack to render only some special tokens
4dd479b7
ngxson fix bf16 conversion
ebc7da53
ngxson remove vocab hack
a543ddfd
ngxson webui ok
6b303729
ngxson support chat parsing for gpt-oss
44bdb752
ggerganov Merge branch 'master' into xsn/oai_moe
65b536f9
ngxson fix webui
61979176
ngxson direct mapping mxfp4, FINALLY
3c4725ba
ngxson force using mxfp4
04cfb6d2
ngxson properly use lazy tensor
4cf69dff
ggerganov ggml : add mxfp4
ec95c0e8
slaren ggml : add ggml_add_id (#13)
3ef6c8c1
slaren Merge branch 'master' into xsn/oai_moe
cd514cc3
slaren Merge branch 'xsn/oai_moe' into mxfp4-rebased
98c4be53
ggerganov ggerganov requested a review from 0cc4m 0cc4m 33 days ago
ggerganov ggerganov requested a review from JohannesGaessler JohannesGaessler 33 days ago
ggerganov ggerganov requested a review from ngxson ngxson 33 days ago
ggerganov ggerganov requested a review from slaren slaren 33 days ago
github-actions github-actions added testing
github-actions github-actions added Nvidia GPU
github-actions github-actions added Vulkan
github-actions github-actions added examples
github-actions github-actions added python
github-actions github-actions added server
github-actions github-actions added ggml
github-actions github-actions added SYCL
github-actions github-actions added Apple Metal
github-actions github-actions added Ascend NPU
github-actions github-actions added OpenCL
ngxson Merge branch 'master' into gpt-oss-mxfp4
fcb23396
ggerganov llama : fix compile error
98f34448
slaren cuda : add fallback for __nv_cvt_e8m0_to_bf16raw
df8411ed
slaren slaren force pushed from 5b6b1ffe to df8411ed 33 days ago
ggerganov
slaren cleanup
60ab08a5
slaren sycl : fix supports_op for MXFP4
256fe66c
ngxson fix Unknown reasoning format
cd8ed32b
slaren ggml-cpu : fix AVX build
a3b291e8
0cc4m
lattwood
slaren fix hip build
1ea3769f
ggerganov
SplittyDev
slaren cuda : add mxfp4 dequantization support for cuBLAS
07d781e4
slaren slaren force pushed from ee2adc79 to 07d781e4 33 days ago
ngxson
slaren ggml-cpu : fix mxfp4 fallback definitions for some architectures
b236c90f
bartowski1182
ngxson
netrunnereve
ngxson
slaren cuda : fix version required for __nv_cvt_e8m0_to_bf16raw
d9d89b42
ahmed-adly-khalil
ahmed-adly-khalil
isaac-mcfadyen
Mushoz
csabakecskemeti
slaren
slaren approved these changes on 2025-08-05
tobi
ggerganov ggerganov merged fd1234cb into master 33 days ago
ggerganov ggerganov deleted the gpt-oss-mxfp4 branch 33 days ago
csabakecskemeti
brandonj60
Sumanai
dinerburger
Green-Sky
ggerganov
semidark
ngxson
semidark
csabakecskemeti
nachoal
csabakecskemeti
jacekpoplawski
kiuckhuang
nachoal
thad0ctor
joseph777111
CHNtentes
fat-tire
nachoal
nai-kon
fat-tire
uazure
slaren
createthis
CISC
CISC commented on 2025-08-08
ericcurtin
JohannesGaessler
ericcurtin
ngxson
ericcurtin
CISC
ngxson
CISC
marvin-0042
DivyanshScore1910

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone