rpc : do not wait for response when sending RPC_CMD_SET_TENSOR (llama/12943)
RPC_CMD_SET_TENSOR always returns an empty response and we send this 4
times per token. We can improve TG speed if we don't wait for this empty
response.
The performance impact of this change depends on the network latency.
Daniel Bevenius [Thu, 1 May 2025 08:05:24 +0000 (10:05 +0200)]
whisper : add check that target name exists (#3103)
This commit adds a check to makes sure that the target exists before
trying to add compile options to ignore warnings when using MSVC.
The motivation for this is currently the build is broken depending on
the cmake options provided. With this fix it should be possible to build
even if the targets are not actually available.
Daniel Bevenius [Thu, 1 May 2025 06:12:48 +0000 (08:12 +0200)]
ruby : ignore "Downloading" output in test_log_suppress (#3106)
This commit adds a temporary fix to the `test_log_suppress` test in the
Ruby bindings.
The motivation for this changes is that I suspect that the recent
migration of the models to HuggingFace Xet has changed the way HTTP
caching works for the models. This is causing the test in question to
fail. This is a temporary fix so that CI is not broken while we
investigate this further.
Daniel Bevenius [Tue, 29 Apr 2025 13:47:55 +0000 (15:47 +0200)]
ggml : suppress Windows compiler warnings (#3075)
* whisper: suppress Windows compiler warnings
This commit disables compiler warnings on window using MSVC.
The motivation for these changes is that some compilers generate
warnings for these conversion, for example Windows MSVC, and
there are quite a few of them. This makes it a little difficult to
spot new warnings that may be introduced and also can be difficult
for users/embedders of ggml where these warnings are hard to separate
from their own warnings.
* squash! whisper: suppress Windows compiler warnings
Move ggml related warnings into ggml. This commit also fixes the
indentation and adds a missing whitespace to the if statement.
This commit addresses a warnings that is present for Release builds:
```console
[ 30%] Building CXX object src/CMakeFiles/whisper.dir/whisper.cpp.o
In file included from /usr/include/c++/13/bits/stl_tree.h:63,
from /usr/include/c++/13/map:62,
from /home/danbev/work/ai/whisper.cpp/src/whisper-arch.h:5,
from /home/danbev/work/ai/whisper.cpp/src/whisper.cpp:2:
In static member function ‘static void std::__copy_move<false, false, std::random_access_iterator_tag>::__assign_one(_Tp*, _Up*) [with _Tp = const whisper_grammar_element*; _Up = const whisper_grammar_element* const]’,
inlined from ‘static _Up* std::__copy_move<_IsMove, true, std::random_access_iterator_tag>::__copy_m(_Tp*, _Tp*, _Up*) [with _Tp = const whisper_grammar_element* const; _Up = const whisper_grammar_element*; bool _IsMove = false]’ at /usr/include/c++/13/bits/stl_algobase.h:440:20,
inlined from ‘_OI std::__copy_move_a2(_II, _II, _OI) [with bool _IsMove = false; _II = const whisper_grammar_element* const*; _OI = const whisper_grammar_element**]’ at /usr/include/c++/13/bits/stl_algobase.h:506:30,
inlined from ‘_OI std::__copy_move_a1(_II, _II, _OI) [with bool _IsMove = false; _II = const whisper_grammar_element* const*; _OI = const whisper_grammar_element**]’ at /usr/include/c++/13/bits/stl_algobase.h:533:42,
...
```
This warning is caused by the fact that the `stack` vector is empty
when it is passed to `new_stacks.push_back(stack);`.
The suggested fix is to use `new_stacks.emplace_back();` instead of
`new_stacks.push_back(stack);`.
This commit removes the empty `.gitmodules` file from the repository.
The motivation of this is that this file is currently empty and the
project does not use any submodules at this time. Removing it mainly to
reduce clutter in the repository and any confusion when seen the file
in repo.
Daniel Bevenius [Mon, 28 Apr 2025 13:38:52 +0000 (15:38 +0200)]
ci : disable publishing of java binding [no ci] (#3086)
This commit disables the publishing of the Java binding to the Maven
repository.
The motivation for this is that this job was disabled for some time and
recently it was re-enabled, but the publishing of the Java binding
caused the build to fail and needs to be investigated further.
Jeff Bolz [Wed, 16 Apr 2025 18:37:25 +0000 (13:37 -0500)]
vulkan: enable coopmat2 FA gqa and split_k optimizations more often (llama/12931)
The grouped query attention optmization doesn't require a power of two ratio,
the only thing relying on it was the modulo operation written as bitwise &.
split_k need not depend on gqa_ratio - enable it any time there's only one
workgroup in the X dimension. The shader gets the split index from the x coord,
and multiple workgroups in the X dimension (pre-split) indicates a larger
FA operation that wouldn't need splitting.
David Huang [Tue, 15 Apr 2025 09:20:38 +0000 (17:20 +0800)]
CUDA/HIP: Share the same unified memory allocation logic. (llama/12934)
Replace compile-time `GGML_HIP_UMA` with environment variable `GGML_CUDA_ENABLE_UNIFIED_MEMORY`. This unifies the usage on NVIDIA and AMD GPUs, and allows a single binary to be shared between integrated and dedicated GPUs.
CANN: Optimize CANN buffer pool memory management (llama/12875)
Multiple optional memory pools are provided for CANN, including VMM,
priority queue-based, and traditional memory pools.
1.When the memory pool is available and GGML_CANN_DISABLE_VMM_POOL
is not defined, the VMM pool is selected by default.
2.Otherwise, if GGML_CANN_ENABLE_BUF_PRIO_POOL is defined,
the priority queue-based memory pool is used.
3.If neither condition is met, the default memory pool is used.
sycl: Support sycl_ext_oneapi_limited_graph (llama/12873)
The current usage of the SYCL-Graph extension checks for
the `sycl_ext_oneapi_graph` device aspect. However, it is also
possible to support `sycl_ext_oneapi_limied_graph` devices that
don't support update
Jeff Bolz [Wed, 9 Apr 2025 05:25:08 +0000 (00:25 -0500)]
vulkan: In coopmat2 mmq, load q4_k/q5_k scales through shared memory (llama/12833)
q4_k and q5_k had a lot of redundant global loads where the same 16B of
scale information is repeatedly loaded and decoded during each loop iteration.
This change restructures the loops to more explicitly iterate over whole
blocks in the outer loop (with unrolled inner loop) and to copy/decode the
scale data into shared memory once at the start of each outer loop. The copy
is pipelined so the scale load from global memory is relatively cheap.
This improves q4_k/q5_k model prompt processing performance by around 5-7%.
I briefly tried applying this to q6_k and q4_0, and it didn't help for q6_k
and hurt for q4_0.
The big "else" path in mul_mm_cm2.comp that had all the clamped/unclamped
variants isn't used as often as it originally was (e.g. due to the padded_N
change), so I trimmed it down to offset some of the new complexity of the
semi-manual loop unrolling.
Jeff Bolz [Sun, 6 Apr 2025 08:47:13 +0000 (03:47 -0500)]
vulkan: Use unclamped loads for flash attention mask (llama/12720)
nem1 must be a multiple of GGML_KQ_MASK_PAD, and GGML_KQ_MASK_PAD is a multiple
of the number of rows in the matrix. The KV dim is a multiple of the number of
columns for the aligned shader.
Jeff Bolz [Fri, 4 Apr 2025 05:54:35 +0000 (00:54 -0500)]
vulkan: Hybrid waitForFences/getFenceStatus to reduce fence latency (llama/12630)
There seems to be a bubble waking up from waitForFences, which costs a few
percent performance and also increased variance in performance. This change
inserts an "almost_ready" fence when the graph is about 80% complete and we
waitForFences for the almost_ready fence and then spin (with _mm_pauses) waiting
for the final fence to be signaled.
CUDA: Prefer vector flash decoding kernel for Gemma models (llama/12738)
* Prefer vector flash decoding kernel for Gemma models
Vector flash decoding kernel was not being picked for models with head dimension 256. Gemma models are in this category.
Removing this limit improves e2e performance by upto 12% in gen phase throughput for Gemm models.
* Update ggml/src/ggml-cuda/fattn.cu
Co-authored-by: Johannes Gäßler <redacted>
---------
Alan Gray [Thu, 3 Apr 2025 01:31:15 +0000 (02:31 +0100)]
Simplify and improve CUDA graphs through use of indirect copy pointers (llama/9017)
* CUDA: Simplify and improve CUDA graphs through use of indirect copy pointers
Previously there was complexity in the CUDA graphs implementation due
frequently changing parameters to copy kernels associated with K and V
cache pointers. This patch simplifies by using indirection to avoid
such parameters frequently changing, avoiding the need for frequent
graph updates.
Jeff Bolz [Wed, 2 Apr 2025 19:25:08 +0000 (14:25 -0500)]
vulkan: Implement split_k for coopmat2 flash attention. (llama/12627)
When using group query attention, we have one workgroup per KV batch and this
can be very few workgroups (e.g. just 8 in some models). Enable split_k to
spread the work across SMs. This helps a lot when the KV cache is large.
previously we would run 32 workgroups computing 1 result each, now we will
run 8 workgroups computing 4 results each.
This doesn't directly translate to better performance (at least when you have
>=32 SMs), but in a subsequent change I'll enable split_k which will scale much
better with 4x fewer workgroups.
Daniel Bevenius [Wed, 23 Apr 2025 06:24:38 +0000 (08:24 +0200)]
coreml : set convert_to="mlprogram" in convert
* coreml : skip model load in convert-whisper-to-coreml.py
This commit updates the conversion process for Whisper models to use the
"mlprogram" format instead of "neuralnetwork".
The motivation for this change is that when using the "neuralnetwork"
format the underlying model produced is based on protobuf and my
understanding is that there are limitations to this format, such as
sizes of strings and the complexity of the model.
Currently when trying to convert larger models such as large-v3 the
conversion fails but succeeds for smaller models.
The "mlprogram" format is a more recent addition to CoreML and is
designed to be more flexible and powerful, allowing for more complex
models and larger data types. This seems to work for larger and smaller
models alike and unless I'm there are considerations that I'm not aware
of I think this is what we should be using moving forward.
The error that is generated for large models is the following:
```console
Running MIL backend_neuralnetwork pipeline: 100%|█████████| 9/9 [00:00<00:00, 35.44 passes/s]
Translating MIL ==> NeuralNetwork Ops: 100%|███████████| 5641/5641 [03:31<00:00, 26.65 ops/s]
Traceback (most recent call last):
File "/Users/danbev/work/ai/whisper-work/models/convert-whisper-to-coreml.py", line 322, in <module>
encoder = convert_encoder(hparams, encoder, quantize=args.quantize)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/danbev/work/ai/whisper-work/models/convert-whisper-to-coreml.py", line 255, in convert_encoder
model = ct.convert(
^^^^^^^^^^^
File "/Users/danbev/work/ai/whisper-work/venv/lib/python3.11/site-packages/coremltools/converters/_converters_entry.py", line 635, in convert
mlmodel = mil_convert(
^^^^^^^^^^^^
File "/Users/danbev/work/ai/whisper-work/venv/lib/python3.11/site-packages/coremltools/converters/mil/converter.py", line 186, in mil_convert
return _mil_convert(
^^^^^^^^^^^^^
File "/Users/danbev/work/ai/whisper-work/venv/lib/python3.11/site-packages/coremltools/converters/mil/converter.py", line 245, in _mil_convert
return modelClass(
^^^^^^^^^^^
File "/Users/danbev/work/ai/whisper-work/venv/lib/python3.11/site-packages/coremltools/models/model.py", line 489, in __init__
self.__proxy__, self._spec, self._framework_error = self._get_proxy_and_spec(
^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/danbev/work/ai/whisper-work/venv/lib/python3.11/site-packages/coremltools/models/model.py", line 550, in _get_proxy_and_spec
_MLModelProxy(
ValueError: basic_string
```