Commit Graph

387 Commits

Author SHA1 Message Date
Georgi Gerganov
94e7da1ff2
cmake : fix "amd64" processor string () 2024-12-17 18:34:32 +02:00
gn64
c4aed6831e
vulkan : fix soft_max.comp division by zero ()
This change prevents a division by zero error when p.KY is 0.
2024-12-16 12:34:38 +02:00
Georgi Gerganov
7d134e3737
ggml : remove old files (skip) () 2024-12-08 23:04:26 +02:00
Georgi Gerganov
9df53b357e
ggml : sync remnants (skip) () 2024-12-08 22:48:25 +02:00
Diego Devesa
a815940e0e ggml : add predefined list of CPU backend variants to build (llama/10626)
* ggml : add predefined list of CPU backend variants to build

* update CPU dockerfiles
2024-12-08 20:14:35 +02:00
Diego Devesa
904e307bce ggml-cpu : fix HWCAP2_I8MM value (llama/10646) 2024-12-08 20:14:35 +02:00
Jeff Bolz
491ec076b4 vulkan: Implement "fast divide" (mul+shift) for unary ops like copy (llama/10642) 2024-12-08 20:14:35 +02:00
Nicolò Scipione
966433fdf2 SYCL : Move to compile time oneMKL interface backend selection for NVIDIA backend (llama/10584)
* [SYCL] Move to Compile Time backend selection on oneMKL Interface for NVIDIA backend

Move to compile time selection to backend to avoid latency at run time.
Add it to all mkl gemm calls and only for NVIDIA backend.

Signed-off-by: nscipione <nicolo.scipione@codeplay.com>

* Formatting

* Address PR comments to increase readibility

---------

Signed-off-by: nscipione <nicolo.scipione@codeplay.com>
2024-12-08 20:14:35 +02:00
Frankie Robertson
6f1ba9d82d Avoid using __fp16 on ARM with old nvcc (llama/10616) 2024-12-08 20:14:35 +02:00
Jeff Bolz
015ecd0001 vulkan: optimize and reenable split_k (llama/10637)
Use vector loads when possible in mul_mat_split_k_reduce. Use split_k
when there aren't enough workgroups to fill the shaders.
2024-12-08 20:14:35 +02:00
PAB
b7c64a4352 ggml: add GGML_SET Metal kernel + i32 CPU kernel (ggml/1037)
* implemented cpu kernel

* add i32 test cases in test-backend-ops

* typedef `ggml_metal_kargs_set`

* implemented `kernel_set`

* memcpy
2024-12-08 20:14:35 +02:00
PAB
7895d39508 ggml : add GGML_PAD_REFLECT_1D operation (ggml/1034)
* ggml_pad_reflect_1d defined in header

* implemented on CPU

* called the forward pass

* impl Metal kernel

* added Metal kernel

* added OP_PAD_REFLECT_1D in test-backend-ops.cpp

* add test-pad-reflect-1d test case

* test case support multiple backend
2024-12-08 20:14:35 +02:00
Georgi Gerganov
22616f00f9 files : remove make artifacts 2024-12-08 20:14:35 +02:00
Diego Devesa
3daeacad24 ggml : move AMX to the CPU backend (llama/10570)
ggml : automatic selection of best CPU backend (llama/10606)
2024-12-08 20:14:35 +02:00
Georgi Gerganov
4d73962da4 metal : small-batch mat-mul kernels (llama/10581)
* metal : small-batch mat-mul kernels

ggml-ci

* metal : add rest of types

ggml-ci

* metal : final adjustments

ggml-ci

* metal : add comments

ggml-ci
2024-12-08 20:14:35 +02:00
Akarshan Biswas
068812650e SYCL: Fix and switch to GGML_LOG system instead of fprintf (llama/10579)
* Switched to GGML_LOG

* Fix missing semicolon
2024-12-08 20:14:35 +02:00
Adrien Gallouët
4b7e059e15 ggml-cpu: replace AArch64 NEON assembly with intrinsics in ggml_gemv_q4_0_4x4_q8_0() (llama/10567)
Signed-off-by: Adrien Gallouët <angt@huggingface.co>
2024-12-08 20:14:35 +02:00
Eve
30e35d7271 vulkan: Dynamic subgroup size support for Q6_K mat_vec (llama/10536)
* subgroup 64 version with subgroup add. 15% faster

scalable version

tested for subgroup sizes 16-128

* check for subgroup multiple of 16 and greater than 16

* subgroup sizes are always a power of 2 (https://github.com/KhronosGroup/GLSL/issues/45)

* force 16 sequential threads per block

* make 16 subgroup size a constant
2024-12-08 20:14:35 +02:00
Georgi Gerganov
3623bd58f2 ggml : fix I8MM Q4_1 scaling factor conversion (llama/10562)
ggml-ci
2024-12-08 20:14:35 +02:00
Shupei Fan
cb847c20a7 ggml-cpu: fix typo in gemv/gemm iq4_nl_4_4 (llama/10580) 2024-12-08 20:14:35 +02:00
Alberto Cabrera Pérez
964b154a2a sycl : offload of get_rows set to 0 (llama/10432) 2024-12-08 20:14:35 +02:00
Alberto Cabrera Pérez
d7c2a04bce sycl : Reroute permuted mul_mats through oneMKL (llama/10408)
This PR fixes the failing MUL_MAT tests for the sycl backend.
2024-12-08 20:14:35 +02:00
Chenguang Li
2bb4ca9cba CANN: RoPE operator optimization (llama/10563)
* [cann] RoPE operator optimization

* [CANN]Code Formatting

---------

Co-authored-by: noemotiovon <noemotiovon@gmail.com>
2024-12-08 20:14:35 +02:00
Jeff Bolz
a753a82462 vulkan: get the first command buffer submitted sooner (llama/10499)
This is an incremental improvement over  to get work to the GPU a bit
sooner. The first part is to start with a smaller number of nodes before
the first submit, and ramp it up to the current 100 nodes/submit. The
second part is to reduce the dryrun overhead for all the nodes that just
need to request descriptor space.

With these changes I get around 1-2% speedup on RTX 4070 combined with my
old Haswell-era CPU.
2024-12-08 20:14:35 +02:00
Georgi Gerganov
276b08d8f0 ggml : remove redundant copyright notice + update authors 2024-12-08 20:14:35 +02:00
Georgi Gerganov
4ca1e72fe0 ggml : fix row condition for i8mm kernels (llama/10561)
ggml-ci
2024-12-08 20:14:35 +02:00
Georgi Gerganov
16a66f103f cmake : fix ARM feature detection (llama/10543)
ggml-ci
2024-12-08 20:14:35 +02:00
Shupei Fan
330273901f ggml-cpu: support IQ4_NL_4_4 by runtime repack (llama/10541)
* ggml-cpu: support IQ4_NL_4_4 by runtime repack

* ggml-cpu: add __ARM_FEATURE_DOTPROD guard
2024-12-08 20:14:35 +02:00
Sergio López
42099a9342 kompute : improve backend to pass test_backend_ops (llama/10542)
* kompute: op_unary: reject unsupported parameters

Signed-off-by: Sergio Lopez <slp@redhat.com>

* kompute: softmax: implement ALiBi support

Signed-off-by: Sergio Lopez <slp@redhat.com>

* kompute: rope: implement neox and phi3 support

Signed-off-by: Sergio Lopez <slp@redhat.com>

* kompute: op_mul_mat_q4_k permutted support

Signed-off-by: Sergio Lopez <slp@redhat.com>

* kompute: op_mul_mat_[q4_0|q4_1|q8_0] permutted support

Signed-off-by: Sergio Lopez <slp@redhat.com>

* kompute: op_mul_mat_f16 permutted support

Signed-off-by: Sergio Lopez <slp@redhat.com>

* kompute: op_mul_mat_q6_k permutted support

Signed-off-by: Sergio Lopez <slp@redhat.com>

---------

Signed-off-by: Sergio Lopez <slp@redhat.com>
2024-12-08 20:14:35 +02:00
leo-pony
90dd5fca9c CANN: Fix SOC_TYPE compile bug (llama/10519)
* CANN: Fix the bug build fail on Ascend310P under two cases:
1) Manual specify SOC_TYPE
2) Under some unusual compile environment

* Update the cann backend News content: Support F16 and F32 data type model for Ascend 310P NPU.

* fix CANN  compile fail bug: the assert in ascend kernel function doesn't supportted on some CANN version
2024-12-08 20:14:35 +02:00
Chenguang Li
2490f2a7f8 CANN: ROPE operator optimization (llama/10540)
* [cann] ROPE operator optimization

Co-authored-by: noemotiovon <noemotiovon@gmail.com>
2024-12-08 20:14:35 +02:00
uvos
230e985633 Add some minimal optimizations for CDNA (llama/10498)
* Add some minimal optimizations for CDNA

* ggml_cuda: set launch bounds also for GCN as it helps there too
2024-12-08 20:14:35 +02:00
Georgi Gerganov
ae24083f23 metal : fix group_norm support condition (llama/0) 2024-12-08 20:14:35 +02:00
Jeff Bolz
6463e36369 vulkan: define all quant data structures in types.comp (llama/10440) 2024-12-08 20:14:35 +02:00
Jeff Bolz
b3301f7d82 vulkan: Handle GPUs with less shared memory (llama/10468)
There have been reports of failure to compile on systems with <= 32KB
of shared memory (e.g. ). This change makes the large tile size
fall back to a smaller size if necessary, and makes mul_mat_id fall
back to CPU if there's only 16KB of shared memory.
2024-12-08 20:14:35 +02:00
Jeff Bolz
ab5d4d93ec vulkan: further optimize q5_k mul_mat_vec (llama/10479) 2024-12-08 20:14:35 +02:00
Jeff Bolz
2d6e9dd723 vulkan: skip integer div/mod in get_offsets for batch_idx==0 (llama/10506) 2024-12-08 20:14:35 +02:00
Jeff Bolz
2f16e51553 vulkan: optimize Q2_K and Q3_K mul_mat_vec (llama/10459) 2024-12-08 20:14:35 +02:00
R0CKSTAR
0f0994902f mtgpu: Add MUSA_DOCKER_ARCH in Dockerfiles && update cmake and make (llama/10516)
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
2024-12-08 20:14:35 +02:00
Jeff Bolz
5e1fcc1780 vulkan: fix group_norm (llama/10496)
Fix bad calculation of the end of the range. Add a backend test that
covers the bad case (taken from stable diffusion).

Fixes https://github.com/leejet/stable-diffusion.cpp/issues/439.
2024-12-08 20:14:35 +02:00
Georgi Gerganov
48f421de23 cmake : enable warnings in llama (llama/10474)
* cmake : enable warnings in llama

ggml-ci

* cmake : add llama_get_flags and respect LLAMA_FATAL_WARNINGS

* cmake : get_flags -> ggml_get_flags

* speculative-simple : fix warnings

* cmake : reuse ggml_get_flags

ggml-ci

* speculative-simple : fix compile warning

ggml-ci
2024-12-08 20:14:35 +02:00
Charles Xu
e7afb2b991 ggml-cpu: cmake add arm64 cpu feature check for macos (llama/10487)
* ggml-cpu: cmake add arm64 cpu feature check for macos

* use vmmlaq_s32 for compile option i8mm check
2024-12-08 20:14:35 +02:00
Shanshan Shen
9a5ef7b169 CANN: Improve the Inferencing Performance for Ascend NPU Device (llama/10454)
* improve inferencing performance for ascend npu.

Co-authored-by: Frank Mai <thxCode@thxcode0824@gmail.com>

* some modification after review

* some modifications after review

* restore some modifications

* restore some modifications

---------

Co-authored-by: shanshan shen <shanshanshen333@gmail.com>
Co-authored-by: Frank Mai <thxCode@thxcode0824@gmail.com>
2024-12-08 20:14:35 +02:00
Chenguang Li
453cc0fcf1 CANN: RoPE and CANCAT operator optimization (llama/10488)
Co-authored-by: noemotiovon <noemotiovon@gmail.com>
2024-12-08 20:14:35 +02:00
Junil Kim
78dfec6bc5 vulkan: Fix a vulkan-shaders-gen arugment parsing error (llama/10484)
The vulkan-shaders-gen was not parsing the --no-clean argument correctly.
Because the previous code was parsing the arguments which have a value only
and the --no-clean argument does not have a value, it was not being parsed
correctly. This commit can now correctly parse arguments that don't have values.
2024-12-08 20:14:35 +02:00
Georgi Gerganov
f6d518fc4c metal : enable mat-vec kernels for bs <= 4 (llama/10491) 2024-12-08 20:14:35 +02:00
Diego Devesa
ac33379a35 llama : accept a list of devices to use to offload a model (llama/10497)
* llama : accept a list of devices to use to offload a model

* accept `--dev none` to completely disable offloading

* fix dev list with dl backends

* rename env parameter to LLAMA_ARG_DEVICE for consistency
2024-12-08 20:14:35 +02:00
Diego Devesa
77e3e4a090 ggml : add support for dynamic loading of backends (llama/10469)
* ggml : add support for dynamic loading of backends

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-12-08 20:14:35 +02:00
Georgi Gerganov
b840bb09be metal : minor code formatting 2024-12-08 20:14:35 +02:00
Diego Devesa
8b1c1c30a7 ggml : do not use ARM features not included in the build (llama/10457) 2024-12-08 20:14:35 +02:00