]>
git.djapps.eu Git - pkg/ggml/sources/whisper.cpp/log
KITAITI Makoto [Wed, 18 Dec 2024 09:00:50 +0000 (18:00 +0900)]
ruby : Add no_speech_thold (#2641)
* Remove Whisper::Model.[]
* Fix Whisper::Model::URI#request
* Make Whisper::Context#initialize accept pre-converted model name
* Use downloading pre-converted model feature for testing
* Update README
* Remove unnecessary task
* Move whisper/model.rb -> whisper/model/uri.rb
* Update document comment of Whisper::Context#initialize
* Don't show download progress when not tty
* Pass String to raise
* Use cache model file if download fails
* Add test for auto download
* Specify required Ruby version
* Fix a typo
* Remove unnecessary flags
* Initialize Whisper::Params#diarize explicitely
* Remove redundant code from README for simplicity
* Add Whisper::Params#no_speech_thold attribute
* Add test for Whisper::Params#no_speech_thold
crummyh [Wed, 18 Dec 2024 06:43:48 +0000 (00:43 -0600)]
stream : improve consistency in README (#2642)
Karthick [Tue, 17 Dec 2024 17:15:47 +0000 (22:45 +0530)]
whisper : support no_speech_thold (#2625)
* Implement no_speech_thold
no_speech_thold functionality is on par with OpenAI's whisper
* Addressed review comments
Karthick [Tue, 17 Dec 2024 17:07:08 +0000 (22:37 +0530)]
whisper : add single-timestamp logic (#2629)
* Fix hallucinations during silence
When the predicted tokens end with a single timestamp the the entire 30 segment should be considered as done, to avoid hallucinations for the remaining part of segment.
This behaviour is on par with openai's whisper. Refer to logic related to `single_timestamp_ending` in https://github.com/openai/whisper/blob/main/whisper/transcribe.py
* Accept review comments related to formatting.
Co-authored-by: Georgi Gerganov <redacted>
---------
Co-authored-by: Georgi Gerganov <redacted>
crummyh [Tue, 17 Dec 2024 17:05:35 +0000 (11:05 -0600)]
readme : fix typo (#2637)
Georgi Gerganov [Tue, 17 Dec 2024 16:34:32 +0000 (18:34 +0200)]
cmake : fix "amd64" processor string (#2638)
gn64 [Mon, 16 Dec 2024 10:34:38 +0000 (19:34 +0900)]
vulkan : fix soft_max.comp division by zero (#2633)
This change prevents a division by zero error when p.KY is 0.
Georgi Gerganov [Mon, 16 Dec 2024 06:57:04 +0000 (08:57 +0200)]
common : add cstdio header
Georgi Gerganov [Sun, 15 Dec 2024 19:55:36 +0000 (21:55 +0200)]
stream : update build instructions
Thamster [Sat, 14 Dec 2024 15:25:53 +0000 (10:25 -0500)]
android : fix build and ci (#2624)
* Adding missing CMakeLists.txt include for ggm-cpu needed by whisper.android
* attempt to re-enable CI for JNI android
---------
Co-authored-by: Your Name <redacted>
Michael Rienstra [Thu, 12 Dec 2024 16:02:00 +0000 (08:02 -0800)]
models : fix typo in download-ggml-model.sh (#2623)
Introduced in #2589
KITAITI Makoto [Mon, 9 Dec 2024 11:17:50 +0000 (20:17 +0900)]
ruby : Sync whisper.cpp and model download feature (#2617)
* Use C++17
* Add test for Pathname of model
* Make Whisper::Context#initialize accept Pathname
* Add shorthand for pre-converted models
* Update documents
* Add headings to API section in README [skip ci]
* Remove unused function
* Don't care about no longer included file
* Cosmetic fix
* Use conditional get when get model files
Georgi Gerganov [Mon, 9 Dec 2024 09:30:16 +0000 (11:30 +0200)]
scripts : update to new build system
Georgi Gerganov [Sun, 8 Dec 2024 21:09:26 +0000 (23:09 +0200)]
devops : add cmake
Georgi Gerganov [Sun, 8 Dec 2024 21:07:29 +0000 (23:07 +0200)]
devops : update make commands
Georgi Gerganov [Sun, 8 Dec 2024 21:04:26 +0000 (23:04 +0200)]
ggml : remove old files (skip) (#0)
Georgi Gerganov [Sun, 8 Dec 2024 20:48:25 +0000 (22:48 +0200)]
ggml : sync remnants (skip) (#0)
Georgi Gerganov [Sun, 8 Dec 2024 20:48:14 +0000 (22:48 +0200)]
scripts : remove amx from sync
Georgi Gerganov [Sun, 8 Dec 2024 13:52:57 +0000 (15:52 +0200)]
ci : disable freeBSD builds [no ci]
Georgi Gerganov [Sun, 8 Dec 2024 13:48:14 +0000 (15:48 +0200)]
readme : update build instructions
Georgi Gerganov [Sun, 8 Dec 2024 13:36:01 +0000 (15:36 +0200)]
ci : disable CUDA and Android builds
Georgi Gerganov [Sun, 8 Dec 2024 11:35:35 +0000 (13:35 +0200)]
ci : disable Obj-C build + fixes
Georgi Gerganov [Fri, 6 Dec 2024 13:34:53 +0000 (15:34 +0200)]
make : shim cmake
Georgi Gerganov [Thu, 5 Dec 2024 12:30:33 +0000 (14:30 +0200)]
talk-llama : sync llama.cpp
Georgi Gerganov [Thu, 5 Dec 2024 12:29:18 +0000 (14:29 +0200)]
sync : ggml
Diego Devesa [Wed, 4 Dec 2024 13:45:40 +0000 (14:45 +0100)]
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
Diego Devesa [Wed, 4 Dec 2024 13:40:44 +0000 (14:40 +0100)]
ggml-cpu : fix HWCAP2_I8MM value (llama/10646)
Jeff Bolz [Wed, 4 Dec 2024 07:28:59 +0000 (01:28 -0600)]
vulkan: Implement "fast divide" (mul+shift) for unary ops like copy (llama/10642)
Nicolò Scipione [Wed, 4 Dec 2024 01:29:20 +0000 (02:29 +0100)]
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 <redacted>
* Formatting
* Address PR comments to increase readibility
---------
Signed-off-by: nscipione <redacted>
Frankie Robertson [Wed, 4 Dec 2024 00:41:37 +0000 (02:41 +0200)]
Avoid using __fp16 on ARM with old nvcc (llama/10616)
Jeff Bolz [Tue, 3 Dec 2024 19:29:54 +0000 (13:29 -0600)]
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.
PAB [Wed, 4 Dec 2024 08:19:30 +0000 (09:19 +0100)]
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
PAB [Tue, 3 Dec 2024 19:20:04 +0000 (20:20 +0100)]
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
Georgi Gerganov [Tue, 3 Dec 2024 18:29:32 +0000 (20:29 +0200)]
files : remove make artifacts
Georgi Gerganov [Tue, 3 Dec 2024 18:25:37 +0000 (20:25 +0200)]
common : fix compile warning
ggml-ci
Diego Devesa [Tue, 3 Dec 2024 18:22:12 +0000 (20:22 +0200)]
ggml : move AMX to the CPU backend (llama/10570)
ggml : automatic selection of best CPU backend (llama/10606)
Georgi Gerganov [Tue, 3 Dec 2024 09:52:33 +0000 (11:52 +0200)]
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
Akarshan Biswas [Mon, 2 Dec 2024 07:04:11 +0000 (12:34 +0530)]
SYCL: Fix and switch to GGML_LOG system instead of fprintf (llama/10579)
* Switched to GGML_LOG
* Fix missing semicolon
Adrien Gallouët [Sat, 30 Nov 2024 17:13:18 +0000 (18:13 +0100)]
ggml-cpu: replace AArch64 NEON assembly with intrinsics in ggml_gemv_q4_0_4x4_q8_0() (llama/10567)
Signed-off-by: Adrien Gallouët <redacted>
Eve [Sat, 30 Nov 2024 07:00:02 +0000 (07:00 +0000)]
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
Georgi Gerganov [Fri, 29 Nov 2024 14:25:39 +0000 (16:25 +0200)]
ggml : fix I8MM Q4_1 scaling factor conversion (llama/10562)
ggml-ci
Shupei Fan [Fri, 29 Nov 2024 13:49:02 +0000 (21:49 +0800)]
ggml-cpu: fix typo in gemv/gemm iq4_nl_4_4 (llama/10580)
Alberto Cabrera Pérez [Fri, 29 Nov 2024 12:38:45 +0000 (12:38 +0000)]
sycl : offload of get_rows set to 0 (llama/10432)
Alberto Cabrera Pérez [Fri, 29 Nov 2024 09:49:43 +0000 (09:49 +0000)]
sycl : Reroute permuted mul_mats through oneMKL (llama/10408)
This PR fixes the failing MUL_MAT tests for the sycl backend.
Chenguang Li [Fri, 29 Nov 2024 06:46:55 +0000 (14:46 +0800)]
CANN: RoPE operator optimization (llama/10563)
* [cann] RoPE operator optimization
* [CANN]Code Formatting
---------
Co-authored-by: noemotiovon <redacted>
Jeff Bolz [Fri, 29 Nov 2024 06:18:02 +0000 (00:18 -0600)]
vulkan: get the first command buffer submitted sooner (llama/10499)
This is an incremental improvement over #9118 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.
Georgi Gerganov [Thu, 28 Nov 2024 18:46:40 +0000 (20:46 +0200)]
ggml : remove redundant copyright notice + update authors
Georgi Gerganov [Thu, 28 Nov 2024 12:56:37 +0000 (14:56 +0200)]
ggml : fix row condition for i8mm kernels (llama/10561)
ggml-ci
Georgi Gerganov [Thu, 28 Nov 2024 12:56:23 +0000 (14:56 +0200)]
cmake : fix ARM feature detection (llama/10543)
ggml-ci
Shupei Fan [Thu, 28 Nov 2024 12:52:03 +0000 (20:52 +0800)]
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
Sergio López [Thu, 28 Nov 2024 11:51:38 +0000 (12:51 +0100)]
kompute : improve backend to pass test_backend_ops (llama/10542)
* kompute: op_unary: reject unsupported parameters
Signed-off-by: Sergio Lopez <redacted>
* kompute: softmax: implement ALiBi support
Signed-off-by: Sergio Lopez <redacted>
* kompute: rope: implement neox and phi3 support
Signed-off-by: Sergio Lopez <redacted>
* kompute: op_mul_mat_q4_k permutted support
Signed-off-by: Sergio Lopez <redacted>
* kompute: op_mul_mat_[q4_0|q4_1|q8_0] permutted support
Signed-off-by: Sergio Lopez <redacted>
* kompute: op_mul_mat_f16 permutted support
Signed-off-by: Sergio Lopez <redacted>
* kompute: op_mul_mat_q6_k permutted support
Signed-off-by: Sergio Lopez <redacted>
---------
Signed-off-by: Sergio Lopez <redacted>
leo-pony [Thu, 28 Nov 2024 07:25:24 +0000 (15:25 +0800)]
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
Chenguang Li [Thu, 28 Nov 2024 06:24:46 +0000 (14:24 +0800)]
CANN: ROPE operator optimization (llama/10540)
* [cann] ROPE operator optimization
Co-authored-by: noemotiovon <redacted>
uvos [Wed, 27 Nov 2024 16:10:08 +0000 (17:10 +0100)]
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
Georgi Gerganov [Wed, 27 Nov 2024 09:22:14 +0000 (11:22 +0200)]
metal : fix group_norm support condition (llama/0)
Jeff Bolz [Wed, 27 Nov 2024 07:32:54 +0000 (01:32 -0600)]
vulkan: define all quant data structures in types.comp (llama/10440)
Jeff Bolz [Wed, 27 Nov 2024 07:30:27 +0000 (01:30 -0600)]
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. #10037). 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.
Jeff Bolz [Wed, 27 Nov 2024 07:21:59 +0000 (01:21 -0600)]
vulkan: further optimize q5_k mul_mat_vec (llama/10479)
Jeff Bolz [Wed, 27 Nov 2024 07:08:54 +0000 (01:08 -0600)]
vulkan: skip integer div/mod in get_offsets for batch_idx==0 (llama/10506)
Jeff Bolz [Wed, 27 Nov 2024 07:00:50 +0000 (01:00 -0600)]
vulkan: optimize Q2_K and Q3_K mul_mat_vec (llama/10459)
R0CKSTAR [Tue, 26 Nov 2024 16:00:41 +0000 (00:00 +0800)]
mtgpu: Add MUSA_DOCKER_ARCH in Dockerfiles && update cmake and make (llama/10516)
Signed-off-by: Xiaodong Ye <redacted>
Jeff Bolz [Tue, 26 Nov 2024 15:45:05 +0000 (09:45 -0600)]
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.
Georgi Gerganov [Tue, 26 Nov 2024 12:18:08 +0000 (14:18 +0200)]
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
Charles Xu [Tue, 26 Nov 2024 11:37:05 +0000 (12:37 +0100)]
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
Shanshan Shen [Tue, 26 Nov 2024 10:08:37 +0000 (18:08 +0800)]
CANN: Improve the Inferencing Performance for Ascend NPU Device (llama/10454)
* improve inferencing performance for ascend npu.
Co-authored-by: Frank Mai <redacted>
* some modification after review
* some modifications after review
* restore some modifications
* restore some modifications
---------
Co-authored-by: shanshan shen <redacted>
Co-authored-by: Frank Mai <redacted>
Chenguang Li [Tue, 26 Nov 2024 09:31:05 +0000 (17:31 +0800)]
CANN: RoPE and CANCAT operator optimization (llama/10488)
Co-authored-by: noemotiovon <redacted>
Junil Kim [Tue, 26 Nov 2024 01:47:20 +0000 (10:47 +0900)]
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.
Georgi Gerganov [Mon, 25 Nov 2024 19:49:31 +0000 (21:49 +0200)]
metal : enable mat-vec kernels for bs <= 4 (llama/10491)
Diego Devesa [Mon, 25 Nov 2024 18:30:06 +0000 (19:30 +0100)]
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
Diego Devesa [Mon, 25 Nov 2024 14:13:39 +0000 (15:13 +0100)]
ggml : add support for dynamic loading of backends (llama/10469)
* ggml : add support for dynamic loading of backends
---------
Co-authored-by: Georgi Gerganov <redacted>
Georgi Gerganov [Mon, 25 Nov 2024 13:08:04 +0000 (15:08 +0200)]
metal : minor code formatting
Diego Devesa [Sat, 23 Nov 2024 13:41:12 +0000 (14:41 +0100)]
ggml : do not use ARM features not included in the build (llama/10457)
leo-pony [Fri, 22 Nov 2024 06:07:20 +0000 (14:07 +0800)]
CANN: Support Ascend310P to accelerate F32 and F16 Model (llama/10216)
* CANN Support Ascend310P to accelerate F32 and F16 Model
* Add compile option soc type macro ASCEND_310P to ggml-cann lib
* Remove unused code
* Remove the ascend soc_type hard code compile option in CMakelist.txt
Diego Devesa [Thu, 21 Nov 2024 17:18:50 +0000 (18:18 +0100)]
cuda : optimize argmax (llama/10441)
* cuda : optimize argmax
* remove unused parameter
ggml-ci
* fixup : use full warps
ggml-ci
* Apply suggestions from code review
Co-authored-by: Johannes Gäßler <redacted>
* fix ub
* ggml : check ne00 <= INT32_MAX in argmax and argsort
---------
Co-authored-by: Johannes Gäßler <redacted>
Jeff Bolz [Wed, 20 Nov 2024 19:47:36 +0000 (13:47 -0600)]
vulkan: predicate max operation in soft_max shaders/soft_max (llama/10437)
Fixes #10434
Jeff Bolz [Wed, 20 Nov 2024 07:40:18 +0000 (01:40 -0600)]
vulkan: copy iq4_nl LUT into shared memory (llama/10409)
Jeff Bolz [Wed, 20 Nov 2024 07:11:00 +0000 (01:11 -0600)]
vulkan: further optimize mul_mat_vec using larger loads (llama/10387)
* vulkan: Use pipeline_robustness to disable robustness in mul_mat_vec.
Add some early returns for nonexistent rows in mul_mat_vec shaders. These
can only be hit when dispatching a 2D grid of workgroups. Fix the logic
for the 2D grid of workgroups to round up.
Enable the pipeline robustness extension if it's available, and use it to
disable robustness for these pipelines. The instructions to do the bounds
checking contend for the same ALU resources as the bit twiddling dequant
instructions.
* vulkan: Add GLSL structure aliases for quant types to allow larger loads
In Vulkan it's not possible to cast pointer types, so instead you have to
declare an aliased binding for the memory with a different type. This
commit adds aliases for the quant formats using 16b ints, and in a few
places where the struct size is a multiple of 4 also using 32b ints.
Currently only q4_k's aliases are used, but others will be used in
subsequent commits.
* vulkan: use larger loads in q5_k and q6_k shaders.
Similar to the optimization I did in q4_k recently, this vectorizes some loads
and reduces the number of bit twiddling instructions.
* vulkan: use larger K step per iteration in mul_mat_vec.
Add vec4 dequantization functions, and use them to do K=8 per iteration in
mul_mat_vec. This uses 16b loads for the quant values and 128b loads for B
which helps reduce the load on the memory system.
The K_PER_ITER==2 logic is still there, just for F16/F32, and really only
because they support unaligned sizes.
Tweak the num_iters/unrolling logic to be simpler and catch a couple missed
unrolling opportunities.
haopeng [Tue, 19 Nov 2024 20:10:31 +0000 (04:10 +0800)]
add cmake rvv support (llama/10411)
mahorozte [Tue, 3 Dec 2024 13:11:43 +0000 (21:11 +0800)]
CUDA: remove unnecessary warp reduce in FA (ggml/1032)
* kqmax_new_j in every thread within warp is same after operate at line 199,this reduce can be omit
* same problem in vec32
---------
Co-authored-by: ZhaoXiaoYu <redacted>
PAB [Mon, 2 Dec 2024 18:27:24 +0000 (19:27 +0100)]
feat: add `GGML_UNARY_OP_ARGMAX` Metal kernel (ggml/1019)
* implemented argmax kernel
* tpig -> tgpig
* change to strides
* contiguous assertions
* kernel working and tested
* argmax simd parallel implementation
* added 2 new tests for argmax in test-backend-ops
* cosmit
* added 3 tests cases for perf eval
* add test_argmax in make_test_cases_perf
* Update test-backend-ops.cpp
Co-authored-by: Diego Devesa <redacted>
---------
Co-authored-by: Diego Devesa <redacted>
PAB [Thu, 28 Nov 2024 08:25:06 +0000 (09:25 +0100)]
metal : add `GGML_OP_CONV_TRANSPOSE_1D` kernels (ggml/1026)
* wip
* wip implementation f32
* kernel conv transpose 1d f32 working
* initial commit
Frankie Robertson [Tue, 26 Nov 2024 13:50:26 +0000 (15:50 +0200)]
Do not include arm_neon.h when compiling CUDA code (ggml/1028)
Johannes Gäßler [Wed, 20 Nov 2024 13:56:04 +0000 (14:56 +0100)]
ggml-opt: fix data corruption (ggml/1022)
KITAITI Makoto [Thu, 28 Nov 2024 08:33:07 +0000 (17:33 +0900)]
ruby : Add low-level methods to transcribe (#2585)
* Add tests for Whisper::Context#full
* Add Whisper::Context#full
* Add tests for Whisper::Error
* Add document of Whisper::Context#full [skip ci]
* Add additional signature for Whisper::Context#full
* Add description to Whisper::Context#full
* Add test for Whisper::Context#full_parallel
* Add Whisper::Context#full_parallel
* Hide Whisper's instance methods from Ruby code
* Add class to test MemoryView
* Build test class before running test
* Add test for MemoryView
* Make Whisper::Context#full and #full_parallel accept MemoryView
* Use Ruby 3.1 on CI
* Add comment on samples data type
* Update README
* Update README
* Remove unused code
Michael Rienstra [Thu, 28 Nov 2024 08:31:54 +0000 (00:31 -0800)]
models : add `q8_0` models to `download-ggml-model.sh` (#2589)
KITAITI Makoto [Thu, 21 Nov 2024 15:04:29 +0000 (00:04 +0900)]
ruby : Follow source tree change (#2580)
* Follow whisper.cpp source tree change
* Update whispercpp.gemspec
* Follow whisper.cpp log level change
* Fix paths in GitHub workflow for Ruby bindings
* Use GitHub workflow setting for dependency definition
* Use ternary operator
Georgi Gerganov [Wed, 20 Nov 2024 13:32:34 +0000 (15:32 +0200)]
whisper : use backend registry (#0)
slaren [Wed, 20 Nov 2024 12:25:08 +0000 (13:25 +0100)]
ggml/sched : do not skip views in pre-assignments
Georgi Gerganov [Tue, 19 Nov 2024 17:09:07 +0000 (19:09 +0200)]
whisper : adapt to new ggml (wip)
Georgi Gerganov [Tue, 19 Nov 2024 17:08:57 +0000 (19:08 +0200)]
talk-llama : sync llama.cpp
Georgi Gerganov [Tue, 19 Nov 2024 17:04:21 +0000 (19:04 +0200)]
sync : ggml
Georgi Gerganov [Tue, 19 Nov 2024 17:03:47 +0000 (19:03 +0200)]
ggml : sync resolve (skip) (#0)
bandoti [Tue, 19 Nov 2024 16:10:30 +0000 (12:10 -0400)]
Add required ggml-base and backend libs to cmake pkg (llama/10407)
Diego Devesa [Tue, 19 Nov 2024 13:29:38 +0000 (14:29 +0100)]
cuda : fix CUDA_FLAGS not being applied (llama/10403)
Romain Biessy [Tue, 19 Nov 2024 08:02:23 +0000 (09:02 +0100)]
sycl : Add option to set the SYCL architecture for all targets (llama/10266)
* Add option to set the SYCL architecture for all targets
* Convert GGML_SYCL_HIP_TARGET to the more generic GGML_SYCL_ARCH option
* Document that setting GGML_SYCL_ARCH can improve the performance
Jeff Bolz [Tue, 19 Nov 2024 07:25:17 +0000 (01:25 -0600)]
vulkan: Optimize soft_max (llama/10301)
* vulkan: Optimize soft_max
Large soft_max could already saturate memory, but small/medium sizes were
pretty slow. The bulk of the gains for them comes from using a smaller
workgroup size, and making the workgroup size match the subgroup size also
makes the barriers much cheaper.
Cache some values in locals to avoid refetching/recomputing. And stamp
out a few "template instantiations" so smaller cases will fully unroll.
Add a missing early return for OOB rows. This happens when there are more
than 512 rows and the dispatch is 512 x H.
* vulkan: Further soft_max optimizations
Restore the workgroup size of 512 case, use it for >1024.
Use unrollable loops for more iteration counts.
Alberto Cabrera Pérez [Tue, 19 Nov 2024 00:50:04 +0000 (00:50 +0000)]
sycl: Revert MUL_MAT_OP support changes (llama/10385)
Diego Devesa [Mon, 18 Nov 2024 17:43:40 +0000 (18:43 +0100)]
cuda : only use native when supported by cmake (llama/10389)
Jeff Bolz [Mon, 18 Nov 2024 14:28:42 +0000 (08:28 -0600)]
vulkan: remove use of null initializer (llama/10372)
Seems like this isn't working for vulkan-over-metal when the array is sized
by a spec constant. Maybe a spirv-cross limitation?
Plamen Minev [Mon, 18 Nov 2024 13:02:27 +0000 (15:02 +0200)]
metal : fox offset integer overflows in im2col (ggml/1015)
-- While running StableDiffusion.cpp locally with Metal some offsets overflow and results in incorrect calculations