fallback ov to vulkan - #47

Open

zhaixuejun1993 wants to merge 253 commits into

ravi9:dev_backend_openvinofrom

zhaixuejun1993:xuejun/fallback-ov-to-gpu-v1

This pull request is big! We're only showing the most recent 250 commits

Commits on Jan 15, 2026

Solve the issue of abnormal model output caused by using OpenVINO ADD operator

zhanmyz
authored and
cavusmustafa
committed
faa4a7d
Add OpenVINO MUL operator to GGML of Llama.cpp.

zhanmyz
authored and
cavusmustafa
committed
adc2c70
Add compile options

zhanmyz
authored and
cavusmustafa
committed
0a81aa1
add OpenVINO frontend convert process steps

zhanmyz
authored and
cavusmustafa
committed
77d6814
add get openvino available ops function

zhanmyz
authored and
cavusmustafa
committed
ee31dc1
Add PoC of integration of openvino frontend. Main changes: ggml-ov-frontend-utils, GraphIterator, Decoder

yumengbo
authored and
cavusmustafa
committed
171c468
Implement GgmlOvDecoder. Add dump functions.

yumengbo
authored and
cavusmustafa
committed
34e826a
Convert subgraph with add, sub, mul, div op to ov model and do infer on openvino device

yumengbo
authored and
cavusmustafa
committed
9b7b63d
Add GGML_OV_FRONTEND option. Add readme.

yumengbo
authored and
cavusmustafa
committed
31bd816
Change output for infer request to set output tensor. Support scale, view op.

yumengbo
authored and
cavusmustafa
committed
5b46dc2
add GET_ROWS operator of OpenVINO to GGML of llama.cpp

zhanmyz
authored and
cavusmustafa
committed
49804f4
Update build.md and add operation mapping(GGML to OpenVINO)

zhanmyz
authored and
cavusmustafa
committed
80c330a
add the rms_norm operator implemented using OpenVINO to the GGML backend of llama.cpp

zhanmyz
authored and
cavusmustafa
committed
8c5a609
Fix issue for output memory copy of infer request

yumengbo
authored and
cavusmustafa
committed
e95f29c
Change to implementation following pytorch frontend

yumengbo
authored and
cavusmustafa
committed
b100f89
Add support for UNARY SILU op . Fix pytorch impl bugs.

yumengbo
authored and
cavusmustafa
committed
590f587
Support Softmax op

yumengbo
authored and
cavusmustafa
committed
d218c61
Support Softmax op

yumengbo
authored and
cavusmustafa
committed
8aba03b
Support ROPE op.

yumengbo
authored and
cavusmustafa
committed
2353c73
Add support for RMS_NORM OP

zhanmyz
authored and
cavusmustafa
committed
0f7d07d
Add MUL_MAT,CPY,CONT as operators implemented in OpenVINO for GGML backend

zhanmyz
authored and
cavusmustafa
committed
2b04bd4
Move CPY from GGML OV Backend to OV Frontend

zhanmyz
authored and
cavusmustafa
committed
cb2729b
add implementation of MUL_MAT, CPY, CONT of GGML ops using OV ops

zhanmyz
authored and
cavusmustafa
committed
8484769
add implementation of CPY when the output tensor is non-contiguous

zhanmyz
authored and
cavusmustafa
committed
57582fd
add tmp source code files

zhanmyz
authored and
cavusmustafa
committed
afb8594
Execute singel CONT operator is OK

zhanmyz
authored and
cavusmustafa
committed
081b526
Execute CONT & VIEW operators in OV Frontend is OK

zhanmyz
authored and
cavusmustafa
committed
901f734
OV Frontend supports GET_ROWS/RMS_NORM/MUL/MUL_MAT graph conversion of consecutive OPs

zhanmyz
authored and
cavusmustafa
committed
95ae982
OV Frontend supports GET_ROWS/RMS_NORM/MUL/MUL_MAT/ROPE/SCALE/SOFTMAX/ADD adjacent op graph conversion

zhanmyz
authored and
cavusmustafa
committed
9a7b7d8
Change the input parameter shape of CONT operator

zhanmyz
authored and
cavusmustafa
committed
f98d215
Change the input and ouput node shape of MUL_MAT operator

zhanmyz
authored and
cavusmustafa
committed
f37fa21
Change the input and ouput node shape of MUL_MAT operator

zhanmyz
authored and
cavusmustafa
committed
246a2d1
change CONT and MULMAT input node shape

zhanmyz
authored and
cavusmustafa
committed
d05c458
All adjacent ops can conversion but calculation result is wrong and need debugging

zhanmyz
authored and
cavusmustafa
committed
e08a7fd
1. All operators implemented using OpenVINO can be successfully executed individually.

zhanmyz
authored and
cavusmustafa
committed
cff473a
1. Update the implementation of CPY node when it's non-contiguous

zhanmyz
authored and
cavusmustafa
committed
467a5dd
Minor Update

zhanmyz
authored and
cavusmustafa
committed
b14b49d
Try to add VIEW node to OV Frontend and have some issues that need to be dealt with

zhanmyz
authored and
cavusmustafa
committed
19ec9b6
1. In the Prompt process and predict first token stage, the PERMUTE node needs to be integrated into the OV Frontend

zhanmyz
authored and
cavusmustafa
committed
b02265a
add debug info

zhanmyz
authored and
cavusmustafa
committed
8020138
Process Prompt and predict first token is OK

zhanmyz
authored and
cavusmustafa
committed
8ae700a
1. Solve the AC issue of Permute+VIEW and MULMAL issue in the phase of “1. Process Prompt and predict the first token”.

zhanmyz
authored and
cavusmustafa
committed
eac9a99
1. Delete some comments

zhanmyz
authored and
cavusmustafa
committed
84be5c6
* Use find_package in CMake to configure OpenVINO

wine99
authored and
cavusmustafa
committed
651b2c0
change op mappings to list in openvino_supports_op

wine99
authored and
cavusmustafa
committed
91d2a19
2nd+ token correct by fix CPY in OV, remove single op backend compute code

wine99
authored and
cavusmustafa
committed
8d263bd
Arbitrary token len (>32) work; Fix bug in mulmat

wine99
authored and
cavusmustafa
committed
8b40886
FEAT: do PERMUTE eagerly

wine99
authored and
cavusmustafa
committed
6ed44a3
FEAT: Add interleaved mode for ROPE

wine99
authored and
cavusmustafa
committed
0c7b026
REFACTOR: support weigts as constant

wine99
authored and
cavusmustafa
committed
c04966c
STYLE: minor refactor

wine99
authored and
cavusmustafa
committed
96ba47d
PERF: share const nodes for weights for diff infer

wine99
authored and
cavusmustafa
committed
d3bdca2
BUILD: update build doc, add cmake preset, add CACHE_DIR env var

wine99
authored and
cavusmustafa
committed
0a8cc9a
FEAT: improve debug capability

wine99
authored and
cavusmustafa
committed
7d5e234
PERF: compile once (dynamic graph + cache)

wine99
authored and
cavusmustafa
committed
a8e5efa
Rebase - Bring up to date and fix build process

virajwad
authored and
cavusmustafa
committed
ffabe95
fix build error

wine99
authored and
cavusmustafa
committed
4c905b2
FIX: backend buffer type issue

wine99
authored and
cavusmustafa
committed
a0b3052
STYLE: clang-format

wine99
authored and
cavusmustafa
committed
f15a2cc
FEAT: Add all conversion code from ov side

wine99
authored and
cavusmustafa
committed
0d009fe
PERF: favor low precision matmul

wine99
authored and
cavusmustafa
committed
cdf5370
STYLE and minor REFACTOR

wine99
authored and
cavusmustafa
committed
0d505b4
FIX: Re-add tensor names in cgraph, Add another case for RESHAPE

wine99
authored and
cavusmustafa
committed
041d220
FIX: input shape of KQ_mask

wine99
authored and
cavusmustafa
committed
c57f614
PERF: add weight constant in parallel

wine99
authored and
cavusmustafa
committed
a30dc6e
FIX: set_max_token_len

wine99
authored and
cavusmustafa
committed
8ac5c22
PERF: use Slice+Concat in writing cache_v

wine99
authored and
cavusmustafa
committed
d7cc802
Update build doc

wine99
authored and
cavusmustafa
committed
fd32436
Add cgraph tensor output name to OV op name

wine99
authored and
cavusmustafa
committed
8ce5cc5
Update openvino build instructions

ravi9
authored and
cavusmustafa
committed
3051d5a
Add initial NPU support

wine99
authored and
cavusmustafa
committed
7fec223
draft NPU support version 2: prefill + kvcache

wine99
authored and
cavusmustafa
committed
34531ab
NPU support version 2: prefill + kvcache

wine99
authored and
cavusmustafa
committed
d9ca8f5
Change due to ggml cgraph changes, not correct yet

wine99
authored and
cavusmustafa
committed
f7ad779
Change due to ggml cgraph changes, llama-3.2 CPU work

wine99
authored and
cavusmustafa
committed
592d7f8
Add AMD64 to CMakeLists

wine99
authored and
cavusmustafa
committed
e27738a
Change due to ggml cgraph changes, all device work

wine99
authored and
cavusmustafa
committed
42d4240
Refactor: clean, fix warning

wine99
authored and
cavusmustafa
committed
593484c
Update clang-format

wine99
authored and
cavusmustafa
committed
8afee79
Statful transformation for CPU GPU

wine99
authored and
cavusmustafa
committed
4c582ac
Add SwiGLU

wine99
authored and
cavusmustafa
committed
73ee84f
Fuse to SDPA

wine99
authored and
cavusmustafa
committed
ebc4fc9
Replace Concat with Broadcast in MulMat for GQA

wine99
authored and
cavusmustafa
committed
bf5414c
Pull out indices creation for kv cache update

wine99
authored and
cavusmustafa
committed
acf358d
Refactor: remove past_token_len from extra_inputs

wine99
authored and
cavusmustafa
committed
0fa7a5e
Fix Phi3 SwiGLU and SoftMax

wine99
authored and
cavusmustafa
committed
3533c14
Pull out sin cos from rope

wine99
authored and
cavusmustafa
committed
a80da69
Reduce memory: free ov weights node after graph conversion

wine99
authored and
cavusmustafa
committed
f3c0519
Fix CPY due to cgraph change

wine99
authored and
cavusmustafa
committed
d61f83c
Added OpenVINO CI/CD. Updated docs

ravi9
authored and
cavusmustafa
committed
ea75772
Fix llama-cli

wine99
authored and
cavusmustafa
committed
1ed49bb
Fix Phi3 ROPE; Add test-backend-ops

wine99
authored and
cavusmustafa
committed
44f4cf3
Fix NPU

wine99
authored and
cavusmustafa
committed
6dc4b90
Fix llama-bench; Clang-format

wine99
authored and
cavusmustafa
committed
75eec62
Fix llama-perplexity

wine99
authored and
cavusmustafa
committed
4e7f04a
temp. changes for mark decomp
cavusmustafa
committed
9cf56d6
matmul in fp32

wine99
authored and
cavusmustafa
committed
01cdf4a
mulmat input conversion fix
cavusmustafa
committed
e2fdc1b
mulmat type conversion update
cavusmustafa
committed
93b2d09
add mark decomp pass
cavusmustafa
committed
1a19566
Revert changes in fuse_to_sdpa

wine99
authored and
cavusmustafa
committed
43489bb
Update build.md

ravi9
authored and
cavusmustafa
committed
2f99135
Fix test-backend-ops

wine99
authored and
cavusmustafa
committed
fc86534
Skip test-thread-safety; Run ctest only in ci/run.sh

wine99
authored and
cavusmustafa
committed
1141350
Use CiD for NPU

wine99
authored and
cavusmustafa
committed
37ff226
Optimize tensor conversion, improve TTFT

wine99
authored and
cavusmustafa
committed
9a91ca6
Support op SET_ROWS

wine99
authored and
cavusmustafa
committed
63d000b
Fix NPU

wine99
authored and
cavusmustafa
committed
7bda502
Remove CPY

wine99
authored and
cavusmustafa
committed
839f8c6
Fix test-backend-ops

wine99
authored and
cavusmustafa
committed
f4123be
Minor updates for raising PR

wine99
authored and
cavusmustafa
committed
a7b611b
Perf: RMS fused to OV internal RMS op

wine99
authored and
cavusmustafa
committed
14c8a85
Fix after rebasing

wine99
authored and
cavusmustafa
committed
65e1b1a
Change openvino device_type to GPU; Enable flash_attn

wine99
authored and
cavusmustafa
committed
56d5967
Update supports_buft and supports_op for quantized models

wine99
authored and
cavusmustafa
committed
3e897df
Add quant weight conversion functions from genai gguf reader

wine99
authored and
cavusmustafa
committed
d4ca760
Quant models run with accuracy issue

wine99
authored and
cavusmustafa
committed
663a0b8
Fix accuracy: disable cpu_repack

wine99
authored and
cavusmustafa
committed
6ab76ed
Fix CI; Disable test-backend-ops

wine99
authored and
cavusmustafa
committed
dd80b04
Fix Q4_1

wine99
authored and
cavusmustafa
committed
a1ce428
Fix test-backend-ops: Treat quantized tensors as weights

wine99
authored and
cavusmustafa
committed
9900245
Add NPU Q4_0 support

wine99
authored and
cavusmustafa
committed
9ca53c7
NPU perf: eliminate zp

wine99
authored and
cavusmustafa
committed
82c9833
Dequantize q4_1 q4_k q6_k for NPU

wine99
authored and
cavusmustafa
committed
b593428
Add custom quant type: q8_1_c, q4_0_128

wine99
authored and
cavusmustafa
committed
6926655
Set m_is_static=false as default in decoder

wine99
authored and
cavusmustafa
committed
c5231a2
Simpilfy translation of get_rows

wine99
authored and
cavusmustafa
committed
810eb48
Fix after rebasing

wine99
authored and
cavusmustafa
committed
0f7b253
Improve debug util; Eliminate nop ReshapeReshape

wine99
authored and
cavusmustafa
committed
2ad1147
STYLE: make get_types_to_requant a function

wine99
authored and
cavusmustafa
committed
dc77cbb
Support BF16 model

wine99
authored and
cavusmustafa
committed
bcc343a
Fix NPU compile

wine99
authored and
cavusmustafa
committed
434059a
WA for npu 1st token acc issue

wine99
authored and
cavusmustafa
committed
da2cc99
Apply EliminateZP only for npu

wine99
authored and
cavusmustafa
committed
be07073
Add GeGLU

wine99
authored and
cavusmustafa
committed
5975612
Fix Hunyuan

wine99
authored and
cavusmustafa
committed
7d81861
Support iSWA

wine99
authored and
cavusmustafa
committed
9de874c
Fix NPU accuracy

wine99
authored and
cavusmustafa
committed
602f9ca
Fix ROPE accuracy when freq_scale != 1

wine99
authored and
cavusmustafa
committed
1a38339
Minor: not add attention_size_swa for non-swa model

wine99
authored and
cavusmustafa
committed
67e178a
Minor refactor

wine99
authored and
cavusmustafa
committed
2f1d50f
Add Q5_K to support phi-3-q4_k_m

wine99
authored and
cavusmustafa
committed
e4bfe5a
Requantize Q6_K (gs16) to gs32 on GPU

wine99
authored and
cavusmustafa
committed
f3afa7b
Fix after rebasing

wine99
authored and
cavusmustafa
committed
fdadca1
Always apply Eliminate_ZP to fix GPU compile issue on some platforms

wine99
authored and
cavusmustafa
committed
973a80f
kvcachefusion support
cavusmustafa
committed
c112bc4
env variable GGML_OPENVINO_DISABLE_SDPA_OPTIMIZATION added
cavusmustafa
committed
e725292
Fix for Phi3
cavusmustafa
committed
05d7aba
Fix llama-cli (need to run with --no-warmup)

wine99
authored and
cavusmustafa
committed
a9371ea
Fix add_sliced_mask; Revert mulmat, softmax; Remove input attention_size, iSWA model not working

wine99
authored and
cavusmustafa
committed
8b82d11
fix after rebasing

wine99
authored and
cavusmustafa
committed
299f492
Fix llama-3-8b and phi3-mini q4_0 NPU

wine99
authored and
cavusmustafa
committed
2d2f00a
Update to OV-2025.3 and CMakeLists.txt

ravi9
authored and
cavusmustafa
committed
841d673
Add OV CI cache

wine99
authored and
cavusmustafa
committed
4c8406e
Apply CISC review and update CI to OV2025.3

ravi9
authored and
cavusmustafa
committed
38e8a19
Update CI to run OV dep install before build

ravi9
authored and
cavusmustafa
committed
45af912
Update OV dockerfile to use OV2025.3 and update build docs

ravi9
authored and
cavusmustafa
committed
3a1129e
Style: use switch in supports_ops

wine99
authored and
cavusmustafa
committed
bd3093f
Style: middle ptr and ref align, omit optional struct keyword

wine99
authored and
cavusmustafa
committed
eba8113
NPU Unify PD (#14)

wine99
authored and
cavusmustafa
committed
b8690bc
Clean placeholders in ggml-openvino.cpp

wine99
authored and
cavusmustafa
committed
303923a
NPU unify PD (handled internally)

wine99
authored and
cavusmustafa
committed
ea2c99b
change graph to 4d, support multi sequences

wine99
authored and
cavusmustafa
committed
072dde0
Fix llama-bench

wine99
authored and
cavusmustafa
committed
ae404f7
Fix NPU

wine99
authored and
cavusmustafa
committed
531941b
Update ggml-decoder.cpp

I-N-T-E-L
authored and
cavusmustafa
committed
047bfb5
Update ggml-decoder.cpp

I-N-T-E-L
authored and
cavusmustafa
committed
11b4cc5
Update ggml-decoder.cpp

I-N-T-E-L
authored and
cavusmustafa
committed
bed4952
Update ggml-decoder.cpp

I-N-T-E-L
authored and
cavusmustafa
committed
4a57b37
Update ggml-decoder.cpp

I-N-T-E-L
authored and
cavusmustafa
committed
98396b2
Update ggml-decoder.cpp

I-N-T-E-L
authored and
cavusmustafa
committed
4400b5c
Remove the second decoder for node. Moving the function into the model decoder

zhaixuejun1993
authored and
cavusmustafa
committed
ae93651
Fix error for naive

zhaixuejun1993
authored and
cavusmustafa
committed
992dea7
NPU prefill chunking

wine99
authored and
cavusmustafa
committed
38254cf
NPU fix llama-bench

wine99
authored and
cavusmustafa
committed
59e7e7c
fallback naive run with accuracy issue

wine99
authored and
cavusmustafa
committed
65348b5
NPU support llma-perplexity -b 512 --no-warmup

wine99
authored and
cavusmustafa
committed
808619e
Refactor: split ov_graph_compute for dynamic and static

wine99
authored and
cavusmustafa
committed
2a9d4ca
remove unused API GgmlOvDecoder::get_output_stride(const std::string & name)

zhaixuejun1993
authored and
cavusmustafa
committed
0ea8238
minor update due to ov 2025.4

wine99
authored and
cavusmustafa
committed
8f4ee4e
remove unused API GgmlOvDecoder::get_output_names()

zhaixuejun1993
authored and
cavusmustafa
committed
497964a
remove unused API get_output_shape(const std::string & name)

zhaixuejun1993
authored and
cavusmustafa
committed
f516db1
Modified API GgmlOvDecoder::get_output_type(const std::string & name)

zhaixuejun1993
authored and
cavusmustafa
committed
6d7a0d6
Removed API GgmlOvDecoder::get_output_op_params(const std::string & name)

zhaixuejun1993
authored and
cavusmustafa
committed
ba852f2
Removed API get_output_ggml_tensor(const std::string & name)

zhaixuejun1993
authored and
cavusmustafa
committed
111c96c
Removed API m_outputs

zhaixuejun1993
authored and
cavusmustafa
committed
8ff73e5
Removed m_output_names

zhaixuejun1993
authored and
cavusmustafa
committed
197ed99
Removed API GgmlOvDecoder::get_input_names()

zhaixuejun1993
authored and
cavusmustafa
committed
95c3071
Removed API GgmlOvDecoder::get_input_stride(const std::string& name)

zhaixuejun1993
authored and
cavusmustafa
committed
cd61178
Removed API get_input_type

zhaixuejun1993
authored and
cavusmustafa
committed
891a3be
Removed API get_input_type

zhaixuejun1993
authored and
cavusmustafa
committed
42ca27f
Removed API GgmlOvDecoder::get_input_shape(const std::string & name)

zhaixuejun1993
authored and
cavusmustafa
committed
acb8a01
Removed API GgmlOvDecoder::get_input_op_params(const std::string & name)

zhaixuejun1993
authored and
cavusmustafa
committed
47c91db
Fix error for decoder cache

zhaixuejun1993
authored and
cavusmustafa
committed
91a1b20
Reuse cached decoder

wine99
authored and
cavusmustafa
committed
28da9a9
GPU remove Q6_K requantization

wine99
authored and
cavusmustafa
committed
469325c
NPU fix wrong model output shape

wine99
authored and
cavusmustafa
committed
ae01322
NPU fix q4 perf regression

wine99
authored and
cavusmustafa
committed
c9234b4
Remove unused variable nodes

zhaixuejun1993
authored and
cavusmustafa
committed
9e3163e
Fix decoder can_reuse for llama-bench

wine99
authored and
cavusmustafa
committed
0ef2e5e
Update build.md for Windows

I-N-T-E-L
authored and
cavusmustafa
committed
ae53363
backend buffer: allocate on host

wine99
authored and
cavusmustafa
committed
22d9c17
Use shared_buffer for GPU NPU; Refactor

wine99
authored and
cavusmustafa
committed
72bba82
Add ov_backend_host_buffer; Use cached remote context

wine99
authored and
cavusmustafa
committed
3fdcb6a
Put kvcache on GPU

wine99
authored and
cavusmustafa
committed
d757849
Use ggml_aligned_malloc

wine99
authored and
cavusmustafa
committed
8273a7c
only use remote tensor for kvcache

wine99
authored and
cavusmustafa
committed
88d1d17
only use remote tensor for kvcache for GPU

wine99
authored and
cavusmustafa
committed
a356b44
FIX: use remote tensor from singleton

wine99
authored and
cavusmustafa
committed
cfc4713
Update build.md to include OpenCL

wine99
authored and
cavusmustafa
committed
52a4401
NPU always requant to q4_0_128

wine99
authored and
cavusmustafa
committed
c1142dd
Optimize symmetric quant weight extraction: use single zp

wine99
authored and
cavusmustafa
committed
67c9720
Use Q8_0_C in token embd, lm_head, and for 5 and 6 bits quant

wine99
authored and
cavusmustafa
committed
4e45177
Update build.md

wine99
authored and
cavusmustafa
committed
f5c71e3
Support -ctk f32

wine99
authored and
cavusmustafa
committed
0d6f253
Initial stateful graph support
cavusmustafa
committed
5f30eac
Update ggml/src/ggml-openvino/ggml-decoder.cpp

cavusmustafa
and
ynimmaga
committed
d2fc152
code cleanup
cavusmustafa
committed
981ec65
npu perf fix
cavusmustafa
committed
a40a5df
requant to f16 for Q6 embed on NPU
cavusmustafa
committed
a81b202
Update ggml/src/ggml-openvino/ggml-decoder.cpp
cavusmustafa
committed
a92ecee
Update ggml/src/ggml-openvino/ggml-openvino-extra.cpp
cavusmustafa
committed
599335c
Create OPENVINO.md in llama.cpp backend docs

ynimmaga
authored and
cavusmustafa
committed
416556a
Update OPENVINO.md

ynimmaga
authored and
cavusmustafa
committed
25e6525
Update OPENVINO.md

ynimmaga
authored and
cavusmustafa
committed
9ba3247
Update OPENVINO.md

ynimmaga
authored and
cavusmustafa
committed
61552e4
Update build.md

ynimmaga
authored and
cavusmustafa
committed
63eed0d
Update OPENVINO.md

ynimmaga
authored and
cavusmustafa
committed
f44c60e
Update OPENVINO.md

ynimmaga
authored and
cavusmustafa
committed
e9ed5c4
Update OPENVINO.md

ynimmaga
authored and
cavusmustafa
committed
d3649c1
kq_mask naming fix
cavusmustafa
committed
d7dccf8

Commits on Jan 16, 2026

Syntax correction for workflows build file
cavusmustafa
committed
aa4bc90

Commits on Jan 21, 2026

Change ov backend buffer is_host to false
wine99
committed
9a15c8b

Commits on Feb 11, 2026

Fix llama-bench -p -n where p<=256
wine99
committed
8fb20b2
Fix --direct-io 0
wine99
committed
1c0a47a
Don't put kvcache on GPU in stateful mode
wine99
committed
c840210
Remove hardcode names
wine99
committed
d398214
Fix stateful shapes
wine99
committed
26328fe
Simplification for stateful and update output shape processing

cavusmustafa
authored and
wine99
committed
3259921
Remove hardcode names
wine99
committed
18ab0f5
Avoid re-compilation in llama-bench
wine99
committed
b6c0697
Extract zp directly instead of bias
wine99
committed
0ee7e05
Refactor weight tensor processing
wine99
committed
900dd76
Merge branch 'master' into dev_backend_openvino
wine99
committed
7b3b65b
create_weight_node accept non-ov backend buffer
wine99
committed
1d4ec1b
remove changes in llama-graph.cpp
wine99
committed
e059015

Commits on Feb 12, 2026

stateful masking fix (#38)
cavusmustafa
authored
0d74aba
Fix test-backend-ops crash glu, get_rows, scale, rms_norm, add
wine99
committed
d5d673c
Add ov falback to CPU machinisim & verified with OV without ROPE support

zhaixuejun1993
authored and
cavusmustafa
committed
046ba55

Commits on Feb 15, 2026

fallback to vulkan, verified with ROPE
cavusmustafa
committed
f7cf4e9