sync : llama.cpp #1160
Merged
sync : llama.cpp #1160
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…247) This commit updates the compilation of default.metallib to skip the intermediate .air (Apple Intermediate Representation) file. The motivation for this change is to simplify the custom command a little and avoid generating and then removing the .air file.
This patch nudges the llama.cpp a bit to be supported on PoCL which doesn't support OpenCL C CL2.0. The issue is solved by querying the device for the supported OpenCL C versions and using the highest one available.
Signed-off-by: Xiaodong Ye <[email protected]>
* Fix backend search path * replace .native() with '/' * reverted .native()
…per block between host and device code. (llama/12177) refactor mmqv to unify the calculation of nwarps and rows per block between host and device code. --------- Co-authored-by: Johannes Gäßler <[email protected]>
* tests: run mul_mat_id with a larger N * vulkan: fix bug in coopmat1 mul_mat_id
When fattn-wmma was ported over to warp64 various bits that also touch fattn-vec where converted to selectable warp size, however the fattn-vec kernels dont work with 64 wide warps for now, so we need to avoid launching them with parameters for warp64
…llama/12399) * sycl : support non-contiguous tensors in binary ops * sycl : silence unused variable warning --------- Co-authored-by: Stanisław Szymczyk <[email protected]>
* SYCL: set extras only on GGML_TYPE_Q4_0 * release tensor_extras in reset buffer interface
* cmake: Factor out compiler flag function from ggml llama.cpps's build requires it, too, and we may want to make use of it without add_subdirectory(ggml). * cmake: Enable building against system ggml This facilitates package maintenance for Linux distributions, where the libggml library most likely will be shipped as an individual package upon which a llama.cpp package depends.
…s checking (llama/12273) * vulkan: Pad N dimension of B matrix for coopmat2 perf, to avoid bounds checking
* vulkan: subgroup size test * Vulkan: Add device architecture enum and logic to recognize AMD generations * vulkan: use new architecture logic to specify subgroup size * Initial vulkan subgroup size tuning for RDNA3 * vulkan: commonize RDNA subgroup tuning * vulkan: override subgroup size if required_subgroup_size = 0 * vulkan: disable warp 32 for RDNA3 * vulkan: fine tuned RDNA1 subgroup sizes * vulkan: adjusted subgroup size map * vulkan: fixed RDNA2 subgroup map --------- Co-authored-by: 0cc4m <[email protected]>
It's already found by FindVulkan.cmake in the parent CMakeLists
* Enable CUDA Graph on CTK < 12.x `cudaGraphExecUpdate` API was changed on 12.x. For this reason CUDA graph support was disabled on older CUDA toolkit. This change enables CUDA support in CTK version < 12.x by using older API if CTK < 12.x. * Fix compilation errors with MUSA * Disable CUDA Graph for MUSA
* ggml: Add op l2_norm Signed-off-by: Molly Sophia <[email protected]> * ggml: Add op rwkv_wkv7 Signed-off-by: Molly Sophia <[email protected]> * llama: Add support for RWKV7 and ARWKV7 models Signed-off-by: Molly Sophia <[email protected]> * llama: fix inference with RWKV6Qwen2 Signed-off-by: Molly Sophia <[email protected]> * llama: add more (a)rwkv7 variants in size Signed-off-by: Molly Sophia <[email protected]> * Apply code-format changes Signed-off-by: Molly Sophia <[email protected]> * fix MUSA build Signed-off-by: Molly Sophia <[email protected]> * llama: fix shape error with rwkv using llama-parallel Signed-off-by: Molly Sophia <[email protected]> --------- Signed-off-by: Molly Sophia <[email protected]>
…ion and driver issues (llama/12434)
…e option (llama/12371) * alberto changes * enable sycl graphs by env variable * fixed compilation warnings in ggml-sycl.cpp * renamed graph variables * fix markdown in docs/backend/SYCL.md Co-authored-by: Romain Biessy <[email protected]> * fix markdown in docs/backend/SYCL.md again * compiling graphs by default, renamed graph_enable to graph_disable --------- Co-authored-by: Romain Biessy <[email protected]>
Signed-off-by: Xiaodong Ye <[email protected]>
* opencl: more profiling timing * opencl: generate trace for profiling * opencl: reduce profiling overhead * Populate profiling timing info at the end rather than after each kernel run * opencl: fix for chrome tracing
I've been seeing significantly worse performance for tg with flash attention enabled vs disabled, and it seems to be related to the submit heuristic. Change the heuristic to check how many bytes worth of weight matrix are used and flush every 100MB, and ramp up after the first few submits. This seems to resolve the issue, and also increases perf for non-FA a bit.
* ci: add visionOS build workflow Add a new GitHub Actions workflow for building on visionOS with CMake and Xcode. * ggml: Define _DARWIN_C_SOURCE for visionOS to fix missing u_xxx typedefs * ci: remove define hacks for u_xxx system types --------- Co-authored-by: Giovanni Petrantoni <[email protected]>
…a/12183) - Find out active blocks per SM using cudaOccupancyMaxActiveBlocksPerMultiprocessor API. Use this value to determine the optimal parallel_blocks value. - Prefer vector flash attention kernels over MMA kernel for BS=1 Fixes Issue: #12182 --------- Co-authored-by: Johannes Gäßler <[email protected]>
…architecture (llama/12332) * Add block interleaving support for Q4_K quantization * Remove whitespaces and fix CI/CD issues * Update pointer of bsums from int16_t to const int16_t * Add vector version of quantize_q8_K_4x8 function * Update code formatting based on review comments
* [SYCL] Fix build on Windows when ccache enabled (llama/9954) * take effect only on windows and force it to icl --------- Co-authored-by: Romain Biessy <[email protected]>
* Vulkan: RTE rounding for cpy to quant Co-Authored-By: Jeff Bolz <[email protected]> * remove trailing whitespace * avoid duplicating pipeline_cpy_f32_quant * fix copypasting issue * remove duplicated code --------- Co-authored-by: Jeff Bolz <[email protected]>
* tests: add mul_mat perf/functional tests for p021/nc vulkan shaders * vulkan: Optimize mul_mat_vec p021 and nc shaders. These shaders are used in attention calculations, and when the KV cache grows large they start to dominate the run time. For the nc shader (which is called with large 'k' dimension), use unrolling and vector loads. For the p021 shader (which is called with large 'm' and small 'k' dimensions), take advantage of grouped query attention to reuse loads from the A matrix for the whole group, and reduce the number of workgroups (too much overhead from tiny dispatches). Using subgroupAdd in the p021 shader also helps, use that conditionally.
* musa: refine compute capability Signed-off-by: Xiaodong Ye <[email protected]> * Address review comments Signed-off-by: Xiaodong Ye <[email protected]> --------- Signed-off-by: Xiaodong Ye <[email protected]>
* ggml : fix quantized cpy op ggml-ci * tests : add cpy tests for all types ggml-ci * tests : add BF16 copy tests ggml-ci * tests : fix loop for same-type copy ggml-ci * tests : add option to permute the dst tensor ggml-ci
The OOB calculation could be wrong if the last iteration was during one of the unrolled loops. Adjust the unrolling counts to avoid this. Add a couple new backend tests that hit this failure on NVIDIA GPUs.
Signed-off-by: Xiaodong Ye <[email protected]>
Co-authored-by: Max Krasnyansky <[email protected]>
ggml-cpu : bug fix related to KleidiAI LHS packing Signed-off-by: Dan Johansson <[email protected]>
* ggml : fix MUL_MAT_ID repack with Q8_K ggml-ci * ggml : improve repack templates ggml-ci
* metal : refactor mat-vec code ggml-ci * metal : rename all_sum -> sum_all ggml-ci * metal : fix comments [no ci] * metal : fix nr constant [no ci] * metal : mv q6_K support nr0 > 1 ggml-ci * metal : reduce register pressure ggml-ci * metal : fix typo [no ci] * metal : reduce register pressure ggml-ci
* SYCL: implement memset ggml backend buffer interface * use GGML_ABORT macro * Do not wait for all queues to finish for memset operation
This change upstreams llamafile's cpu matrix multiplication kernels for ppc64le ISA using MMA builtins. This patch handles matrix multiplication between quantised datatypes, block_q4_0 and block_q8_0. This change results in 5% - 50% improvement in total speed(ie all tokens/total time), across various batch sizes. The patch is tested with Meta-Lllama-3-8B, Mistral-7B, Llama-2-7B-chat-hf models on a IBM POWER10 machine. Signed-off-by: Amrita H S <[email protected]>
ggml-ci
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.