whispercpp 1.3.3 → 1.3.5
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- checksums.yaml +4 -4
- data/README.md +60 -43
- data/ext/extconf.rb +2 -2
- data/ext/ruby_whisper.c +14 -2
- data/ext/ruby_whisper.h +39 -0
- data/ext/ruby_whisper_context.c +22 -22
- data/ext/ruby_whisper_model.c +12 -12
- data/ext/ruby_whisper_params.c +79 -25
- data/ext/ruby_whisper_segment.c +84 -19
- data/ext/ruby_whisper_token.c +351 -0
- data/ext/ruby_whisper_transcribe.cpp +1 -1
- data/ext/ruby_whisper_vad_context.c +75 -0
- data/ext/ruby_whisper_vad_context_detect.cpp +50 -0
- data/ext/ruby_whisper_vad_segment.c +139 -0
- data/ext/ruby_whisper_vad_segments.c +106 -0
- data/ext/sources/CMakeLists.txt +4 -1
- data/ext/sources/bindings/javascript/package.json +1 -1
- data/ext/sources/cmake/arm64-apple-clang.cmake +16 -0
- data/ext/sources/cmake/arm64-windows-llvm.cmake +16 -0
- data/ext/sources/cmake/riscv64-spacemit-linux-gnu-gcc.cmake +29 -0
- data/ext/sources/cmake/x64-windows-llvm.cmake +5 -0
- data/ext/sources/examples/CMakeLists.txt +1 -0
- data/ext/sources/examples/addon.node/addon.cpp +19 -19
- data/ext/sources/examples/addon.node/index.js +7 -5
- data/ext/sources/examples/addon.node/vad-example.js +2 -2
- data/ext/sources/examples/bench/bench.cpp +26 -16
- data/ext/sources/examples/bench.wasm/index-tmpl.html +10 -9
- data/ext/sources/examples/cli/cli.cpp +122 -111
- data/ext/sources/examples/command/command.cpp +26 -24
- data/ext/sources/examples/command.wasm/index-tmpl.html +5 -4
- data/ext/sources/examples/common-ggml.cpp +2 -0
- data/ext/sources/examples/lsp/CMakeLists.txt +2 -1
- data/ext/sources/examples/lsp/lsp.cpp +19 -17
- data/ext/sources/examples/quantize/CMakeLists.txt +2 -1
- data/ext/sources/examples/server/server.cpp +34 -24
- data/ext/sources/examples/server.py +6 -1
- data/ext/sources/examples/stream/stream.cpp +4 -2
- data/ext/sources/examples/stream.wasm/emscripten.cpp +6 -6
- data/ext/sources/examples/stream.wasm/index-tmpl.html +82 -5
- data/ext/sources/examples/talk-llama/CMakeLists.txt +7 -3
- data/ext/sources/examples/talk-llama/llama-adapter.cpp +113 -7
- data/ext/sources/examples/talk-llama/llama-adapter.h +13 -1
- data/ext/sources/examples/talk-llama/llama-arch.cpp +2136 -1491
- data/ext/sources/examples/talk-llama/llama-arch.h +125 -3
- data/ext/sources/examples/talk-llama/llama-batch.cpp +174 -100
- data/ext/sources/examples/talk-llama/llama-batch.h +46 -20
- data/ext/sources/examples/talk-llama/llama-chat.cpp +199 -8
- data/ext/sources/examples/talk-llama/llama-chat.h +11 -0
- data/ext/sources/examples/talk-llama/llama-context.cpp +1213 -413
- data/ext/sources/examples/talk-llama/llama-context.h +99 -36
- data/ext/sources/examples/talk-llama/llama-cparams.h +5 -4
- data/ext/sources/examples/talk-llama/llama-grammar.cpp +288 -53
- data/ext/sources/examples/talk-llama/llama-grammar.h +22 -1
- data/ext/sources/examples/talk-llama/llama-graph.cpp +883 -294
- data/ext/sources/examples/talk-llama/llama-graph.h +361 -161
- data/ext/sources/examples/talk-llama/llama-hparams.cpp +144 -6
- data/ext/sources/examples/talk-llama/llama-hparams.h +100 -23
- data/ext/sources/examples/talk-llama/llama-impl.cpp +7 -3
- data/ext/sources/examples/talk-llama/llama-impl.h +3 -1
- data/ext/sources/examples/talk-llama/llama-kv-cache-iswa.cpp +328 -0
- data/ext/sources/examples/talk-llama/{llama-kv-cache-unified-iswa.h → llama-kv-cache-iswa.h} +38 -29
- data/ext/sources/examples/talk-llama/llama-kv-cache.cpp +2100 -0
- data/ext/sources/examples/talk-llama/llama-kv-cache.h +373 -27
- data/ext/sources/examples/talk-llama/llama-kv-cells.h +124 -30
- data/ext/sources/examples/talk-llama/llama-memory-hybrid.cpp +63 -41
- data/ext/sources/examples/talk-llama/llama-memory-hybrid.h +30 -29
- data/ext/sources/examples/talk-llama/llama-memory-recurrent.cpp +77 -35
- data/ext/sources/examples/talk-llama/llama-memory-recurrent.h +15 -16
- data/ext/sources/examples/talk-llama/llama-memory.h +16 -10
- data/ext/sources/examples/talk-llama/llama-mmap.cpp +172 -37
- data/ext/sources/examples/talk-llama/llama-mmap.h +8 -3
- data/ext/sources/examples/talk-llama/llama-model-loader.cpp +93 -9
- data/ext/sources/examples/talk-llama/llama-model-loader.h +9 -2
- data/ext/sources/examples/talk-llama/llama-model-saver.cpp +3 -0
- data/ext/sources/examples/talk-llama/llama-model.cpp +3369 -10145
- data/ext/sources/examples/talk-llama/llama-model.h +104 -12
- data/ext/sources/examples/talk-llama/llama-quant.cpp +53 -30
- data/ext/sources/examples/talk-llama/llama-sampling.cpp +1520 -324
- data/ext/sources/examples/talk-llama/llama-sampling.h +19 -7
- data/ext/sources/examples/talk-llama/llama-vocab.cpp +562 -39
- data/ext/sources/examples/talk-llama/llama-vocab.h +50 -0
- data/ext/sources/examples/talk-llama/llama.cpp +794 -12
- data/ext/sources/examples/talk-llama/llama.h +246 -190
- data/ext/sources/examples/talk-llama/models/afmoe.cpp +191 -0
- data/ext/sources/examples/talk-llama/models/apertus.cpp +125 -0
- data/ext/sources/examples/talk-llama/models/arcee.cpp +135 -0
- data/ext/sources/examples/talk-llama/models/arctic.cpp +138 -0
- data/ext/sources/examples/talk-llama/models/arwkv7.cpp +86 -0
- data/ext/sources/examples/talk-llama/models/baichuan.cpp +122 -0
- data/ext/sources/examples/talk-llama/models/bailingmoe.cpp +144 -0
- data/ext/sources/examples/talk-llama/models/bailingmoe2.cpp +135 -0
- data/ext/sources/examples/talk-llama/models/bert.cpp +178 -0
- data/ext/sources/examples/talk-llama/models/bitnet.cpp +160 -0
- data/ext/sources/examples/talk-llama/models/bloom.cpp +101 -0
- data/ext/sources/examples/talk-llama/models/chameleon.cpp +178 -0
- data/ext/sources/examples/talk-llama/models/chatglm.cpp +132 -0
- data/ext/sources/examples/talk-llama/models/codeshell.cpp +111 -0
- data/ext/sources/examples/talk-llama/models/cogvlm.cpp +102 -0
- data/ext/sources/examples/talk-llama/models/cohere2-iswa.cpp +134 -0
- data/ext/sources/examples/talk-llama/models/command-r.cpp +122 -0
- data/ext/sources/examples/talk-llama/models/dbrx.cpp +123 -0
- data/ext/sources/examples/talk-llama/models/deci.cpp +135 -0
- data/ext/sources/examples/talk-llama/models/deepseek.cpp +144 -0
- data/ext/sources/examples/talk-llama/models/deepseek2.cpp +259 -0
- data/ext/sources/examples/talk-llama/models/dots1.cpp +134 -0
- data/ext/sources/examples/talk-llama/models/dream.cpp +105 -0
- data/ext/sources/examples/talk-llama/models/ernie4-5-moe.cpp +150 -0
- data/ext/sources/examples/talk-llama/models/ernie4-5.cpp +110 -0
- data/ext/sources/examples/talk-llama/models/exaone.cpp +114 -0
- data/ext/sources/examples/talk-llama/models/exaone4.cpp +123 -0
- data/ext/sources/examples/talk-llama/models/falcon-h1.cpp +113 -0
- data/ext/sources/examples/talk-llama/models/falcon.cpp +120 -0
- data/ext/sources/examples/talk-llama/models/gemma-embedding.cpp +116 -0
- data/ext/sources/examples/talk-llama/models/gemma.cpp +112 -0
- data/ext/sources/examples/talk-llama/models/gemma2-iswa.cpp +128 -0
- data/ext/sources/examples/talk-llama/models/gemma3.cpp +155 -0
- data/ext/sources/examples/talk-llama/models/gemma3n-iswa.cpp +384 -0
- data/ext/sources/examples/talk-llama/models/glm4-moe.cpp +170 -0
- data/ext/sources/examples/talk-llama/models/glm4.cpp +150 -0
- data/ext/sources/examples/talk-llama/models/gpt2.cpp +105 -0
- data/ext/sources/examples/talk-llama/models/gptneox.cpp +144 -0
- data/ext/sources/examples/talk-llama/models/granite-hybrid.cpp +196 -0
- data/ext/sources/examples/talk-llama/models/granite.cpp +211 -0
- data/ext/sources/examples/talk-llama/models/graph-context-mamba.cpp +283 -0
- data/ext/sources/examples/talk-llama/models/grok.cpp +159 -0
- data/ext/sources/examples/talk-llama/models/grovemoe.cpp +141 -0
- data/ext/sources/examples/talk-llama/models/hunyuan-dense.cpp +132 -0
- data/ext/sources/examples/talk-llama/models/hunyuan-moe.cpp +154 -0
- data/ext/sources/examples/talk-llama/models/internlm2.cpp +120 -0
- data/ext/sources/examples/talk-llama/models/jais.cpp +86 -0
- data/ext/sources/examples/talk-llama/models/jamba.cpp +106 -0
- data/ext/sources/examples/talk-llama/models/lfm2.cpp +175 -0
- data/ext/sources/examples/talk-llama/models/llada-moe.cpp +122 -0
- data/ext/sources/examples/talk-llama/models/llada.cpp +99 -0
- data/ext/sources/examples/talk-llama/models/llama-iswa.cpp +178 -0
- data/ext/sources/examples/talk-llama/models/llama.cpp +168 -0
- data/ext/sources/examples/talk-llama/models/maincoder.cpp +117 -0
- data/ext/sources/examples/talk-llama/models/mamba.cpp +55 -0
- data/ext/sources/examples/talk-llama/models/mimo2-iswa.cpp +123 -0
- data/ext/sources/examples/talk-llama/models/minicpm3.cpp +199 -0
- data/ext/sources/examples/talk-llama/models/minimax-m2.cpp +124 -0
- data/ext/sources/examples/talk-llama/models/mistral3.cpp +160 -0
- data/ext/sources/examples/talk-llama/models/models.h +569 -0
- data/ext/sources/examples/talk-llama/models/modern-bert.cpp +116 -0
- data/ext/sources/examples/talk-llama/models/mpt.cpp +126 -0
- data/ext/sources/examples/talk-llama/models/nemotron-h.cpp +150 -0
- data/ext/sources/examples/talk-llama/models/nemotron.cpp +122 -0
- data/ext/sources/examples/talk-llama/models/neo-bert.cpp +104 -0
- data/ext/sources/examples/talk-llama/models/olmo.cpp +121 -0
- data/ext/sources/examples/talk-llama/models/olmo2.cpp +150 -0
- data/ext/sources/examples/talk-llama/models/olmoe.cpp +124 -0
- data/ext/sources/examples/talk-llama/models/openai-moe-iswa.cpp +127 -0
- data/ext/sources/examples/talk-llama/models/openelm.cpp +124 -0
- data/ext/sources/examples/talk-llama/models/orion.cpp +123 -0
- data/ext/sources/examples/talk-llama/models/pangu-embedded.cpp +121 -0
- data/ext/sources/examples/talk-llama/models/phi2.cpp +121 -0
- data/ext/sources/examples/talk-llama/models/phi3.cpp +152 -0
- data/ext/sources/examples/talk-llama/models/plamo.cpp +110 -0
- data/ext/sources/examples/talk-llama/models/plamo2.cpp +316 -0
- data/ext/sources/examples/talk-llama/models/plamo3.cpp +128 -0
- data/ext/sources/examples/talk-llama/models/plm.cpp +168 -0
- data/ext/sources/examples/talk-llama/models/qwen.cpp +108 -0
- data/ext/sources/examples/talk-llama/models/qwen2.cpp +126 -0
- data/ext/sources/examples/talk-llama/models/qwen2moe.cpp +151 -0
- data/ext/sources/examples/talk-llama/models/qwen2vl.cpp +117 -0
- data/ext/sources/examples/talk-llama/models/qwen3.cpp +117 -0
- data/ext/sources/examples/talk-llama/models/qwen3moe.cpp +124 -0
- data/ext/sources/examples/talk-llama/models/qwen3next.cpp +873 -0
- data/ext/sources/examples/talk-llama/models/qwen3vl-moe.cpp +149 -0
- data/ext/sources/examples/talk-llama/models/qwen3vl.cpp +141 -0
- data/ext/sources/examples/talk-llama/models/refact.cpp +94 -0
- data/ext/sources/examples/talk-llama/models/rnd1.cpp +126 -0
- data/ext/sources/examples/talk-llama/models/rwkv6-base.cpp +162 -0
- data/ext/sources/examples/talk-llama/models/rwkv6.cpp +94 -0
- data/ext/sources/examples/talk-llama/models/rwkv6qwen2.cpp +86 -0
- data/ext/sources/examples/talk-llama/models/rwkv7-base.cpp +135 -0
- data/ext/sources/examples/talk-llama/models/rwkv7.cpp +90 -0
- data/ext/sources/examples/talk-llama/models/seed-oss.cpp +124 -0
- data/ext/sources/examples/talk-llama/models/smallthinker.cpp +126 -0
- data/ext/sources/examples/talk-llama/models/smollm3.cpp +128 -0
- data/ext/sources/examples/talk-llama/models/stablelm.cpp +146 -0
- data/ext/sources/examples/talk-llama/models/starcoder.cpp +100 -0
- data/ext/sources/examples/talk-llama/models/starcoder2.cpp +121 -0
- data/ext/sources/examples/talk-llama/models/t5-dec.cpp +166 -0
- data/ext/sources/examples/talk-llama/models/t5-enc.cpp +96 -0
- data/ext/sources/examples/talk-llama/models/wavtokenizer-dec.cpp +149 -0
- data/ext/sources/examples/talk-llama/models/xverse.cpp +108 -0
- data/ext/sources/examples/talk-llama/talk-llama.cpp +9 -6
- data/ext/sources/examples/talk-llama/unicode.cpp +309 -16
- data/ext/sources/examples/talk-llama/unicode.h +45 -0
- data/ext/sources/examples/vad-speech-segments/CMakeLists.txt +1 -1
- data/ext/sources/examples/wchess/wchess.cmd/wchess.cmd.cpp +4 -2
- data/ext/sources/examples/whisper.wasm/index-tmpl.html +18 -17
- data/ext/sources/ggml/CMakeLists.txt +135 -79
- data/ext/sources/ggml/cmake/ggml-config.cmake.in +132 -93
- data/ext/sources/ggml/include/ggml-alloc.h +9 -0
- data/ext/sources/ggml/include/ggml-backend.h +21 -2
- data/ext/sources/ggml/include/ggml-cpu.h +2 -1
- data/ext/sources/ggml/include/ggml-hexagon.h +19 -0
- data/ext/sources/ggml/include/ggml-metal.h +1 -6
- data/ext/sources/ggml/include/ggml-opt.h +25 -6
- data/ext/sources/ggml/include/ggml-rpc.h +8 -11
- data/ext/sources/ggml/include/ggml-webgpu.h +19 -0
- data/ext/sources/ggml/include/ggml-zdnn.h +17 -0
- data/ext/sources/ggml/include/ggml-zendnn.h +22 -0
- data/ext/sources/ggml/include/ggml.h +406 -23
- data/ext/sources/ggml/src/CMakeLists.txt +99 -13
- data/ext/sources/ggml/src/ggml-alloc.c +368 -161
- data/ext/sources/ggml/src/ggml-backend-impl.h +5 -5
- data/ext/sources/ggml/src/ggml-backend-reg.cpp +55 -14
- data/ext/sources/ggml/src/ggml-backend.cpp +290 -57
- data/ext/sources/ggml/src/ggml-blas/CMakeLists.txt +17 -3
- data/ext/sources/ggml/src/ggml-blas/ggml-blas.cpp +10 -13
- data/ext/sources/ggml/src/ggml-cann/CMakeLists.txt +14 -0
- data/ext/sources/ggml/src/ggml-cann/acl_tensor.cpp +59 -45
- data/ext/sources/ggml/src/ggml-cann/acl_tensor.h +138 -47
- data/ext/sources/ggml/src/ggml-cann/aclnn_ops.cpp +2586 -1917
- data/ext/sources/ggml/src/ggml-cann/aclnn_ops.h +348 -309
- data/ext/sources/ggml/src/ggml-cann/common.h +350 -133
- data/ext/sources/ggml/src/ggml-cann/ggml-cann.cpp +894 -625
- data/ext/sources/ggml/src/ggml-common.h +17 -0
- data/ext/sources/ggml/src/ggml-cpu/CMakeLists.txt +167 -75
- data/ext/sources/ggml/src/ggml-cpu/amx/amx.cpp +5 -2
- data/ext/sources/ggml/src/ggml-cpu/arch/arm/cpu-feats.cpp +4 -0
- data/ext/sources/ggml/src/ggml-cpu/arch/arm/quants.c +560 -622
- data/ext/sources/ggml/src/ggml-cpu/arch/arm/repack.cpp +1002 -270
- data/ext/sources/ggml/src/ggml-cpu/arch/loongarch/quants.c +107 -587
- data/ext/sources/ggml/src/ggml-cpu/arch/powerpc/quants.c +162 -589
- data/ext/sources/ggml/src/ggml-cpu/arch/riscv/cpu-feats.cpp +38 -0
- data/ext/sources/ggml/src/ggml-cpu/arch/riscv/quants.c +373 -486
- data/ext/sources/ggml/src/ggml-cpu/arch/riscv/repack.cpp +3 -58
- data/ext/sources/ggml/src/ggml-cpu/arch/s390/cpu-feats.cpp +50 -0
- data/ext/sources/ggml/src/ggml-cpu/arch/s390/quants.c +521 -353
- data/ext/sources/ggml/src/ggml-cpu/arch/wasm/quants.c +54 -314
- data/ext/sources/ggml/src/ggml-cpu/arch/x86/quants.c +184 -675
- data/ext/sources/ggml/src/ggml-cpu/arch/x86/repack.cpp +4682 -1660
- data/ext/sources/ggml/src/ggml-cpu/arch-fallback.h +82 -4
- data/ext/sources/ggml/src/ggml-cpu/common.h +14 -0
- data/ext/sources/ggml/src/ggml-cpu/ggml-cpu-impl.h +18 -9
- data/ext/sources/ggml/src/ggml-cpu/ggml-cpu.c +263 -111
- data/ext/sources/ggml/src/ggml-cpu/ggml-cpu.cpp +39 -28
- data/ext/sources/ggml/src/ggml-cpu/kleidiai/kernels.cpp +683 -82
- data/ext/sources/ggml/src/ggml-cpu/kleidiai/kernels.h +38 -43
- data/ext/sources/ggml/src/ggml-cpu/kleidiai/kleidiai.cpp +435 -119
- data/ext/sources/ggml/src/ggml-cpu/llamafile/sgemm-ppc.h +333 -0
- data/ext/sources/ggml/src/ggml-cpu/llamafile/sgemm.cpp +1234 -1182
- data/ext/sources/ggml/src/ggml-cpu/llamafile/sgemm.h +6 -0
- data/ext/sources/ggml/src/ggml-cpu/ops.cpp +2167 -1480
- data/ext/sources/ggml/src/ggml-cpu/ops.h +10 -12
- data/ext/sources/ggml/src/ggml-cpu/quants.c +35 -0
- data/ext/sources/ggml/src/ggml-cpu/quants.h +8 -0
- data/ext/sources/ggml/src/ggml-cpu/repack.cpp +1132 -81
- data/ext/sources/ggml/src/ggml-cpu/repack.h +36 -0
- data/ext/sources/ggml/src/ggml-cpu/simd-mappings.h +120 -93
- data/ext/sources/ggml/src/ggml-cpu/spacemit/ime.cpp +1025 -0
- data/ext/sources/ggml/src/ggml-cpu/spacemit/ime.h +13 -0
- data/ext/sources/ggml/src/ggml-cpu/spacemit/ime1_kernels.cpp +3196 -0
- data/ext/sources/ggml/src/ggml-cpu/spacemit/ime_kernels.h +26 -0
- data/ext/sources/ggml/src/ggml-cpu/traits.cpp +2 -2
- data/ext/sources/ggml/src/ggml-cpu/traits.h +1 -1
- data/ext/sources/ggml/src/ggml-cpu/unary-ops.cpp +151 -0
- data/ext/sources/ggml/src/ggml-cpu/unary-ops.h +7 -0
- data/ext/sources/ggml/src/ggml-cpu/vec.cpp +294 -27
- data/ext/sources/ggml/src/ggml-cpu/vec.h +606 -48
- data/ext/sources/ggml/src/ggml-cuda/CMakeLists.txt +92 -17
- data/ext/sources/ggml/src/ggml-cuda/add-id.cu +58 -0
- data/ext/sources/ggml/src/ggml-cuda/add-id.cuh +3 -0
- data/ext/sources/ggml/src/ggml-cuda/argmax.cu +2 -2
- data/ext/sources/ggml/src/ggml-cuda/argsort.cu +123 -6
- data/ext/sources/ggml/src/ggml-cuda/argsort.cuh +16 -0
- data/ext/sources/ggml/src/ggml-cuda/binbcast.cu +330 -191
- data/ext/sources/ggml/src/ggml-cuda/binbcast.cuh +2 -0
- data/ext/sources/ggml/src/ggml-cuda/common.cuh +588 -128
- data/ext/sources/ggml/src/ggml-cuda/conv-transpose-1d.cu +1 -4
- data/ext/sources/ggml/src/ggml-cuda/conv2d.cu +166 -0
- data/ext/sources/ggml/src/ggml-cuda/conv2d.cuh +5 -0
- data/ext/sources/ggml/src/ggml-cuda/convert.cu +95 -22
- data/ext/sources/ggml/src/ggml-cuda/convert.cuh +25 -0
- data/ext/sources/ggml/src/ggml-cuda/cpy-utils.cuh +217 -0
- data/ext/sources/ggml/src/ggml-cuda/cpy.cu +335 -485
- data/ext/sources/ggml/src/ggml-cuda/cpy.cuh +1 -5
- data/ext/sources/ggml/src/ggml-cuda/cross-entropy-loss.cu +2 -14
- data/ext/sources/ggml/src/ggml-cuda/cumsum.cu +307 -0
- data/ext/sources/ggml/src/ggml-cuda/cumsum.cuh +5 -0
- data/ext/sources/ggml/src/ggml-cuda/dequantize.cuh +14 -40
- data/ext/sources/ggml/src/ggml-cuda/diag.cu +77 -0
- data/ext/sources/ggml/src/ggml-cuda/diag.cuh +5 -0
- data/ext/sources/ggml/src/ggml-cuda/fattn-common.cuh +519 -378
- data/ext/sources/ggml/src/ggml-cuda/fattn-mma-f16.cuh +750 -637
- data/ext/sources/ggml/src/ggml-cuda/fattn-tile.cu +49 -0
- data/ext/sources/ggml/src/ggml-cuda/fattn-tile.cuh +1244 -0
- data/ext/sources/ggml/src/ggml-cuda/fattn-vec.cuh +586 -0
- data/ext/sources/ggml/src/ggml-cuda/fattn-wmma-f16.cu +98 -61
- data/ext/sources/ggml/src/ggml-cuda/fattn-wmma-f16.cuh +48 -0
- data/ext/sources/ggml/src/ggml-cuda/fattn.cu +230 -197
- data/ext/sources/ggml/src/ggml-cuda/fattn.cuh +2 -0
- data/ext/sources/ggml/src/ggml-cuda/fill.cu +37 -0
- data/ext/sources/ggml/src/ggml-cuda/fill.cuh +3 -0
- data/ext/sources/ggml/src/ggml-cuda/getrows.cu +50 -39
- data/ext/sources/ggml/src/ggml-cuda/ggml-cuda.cu +1557 -294
- data/ext/sources/ggml/src/ggml-cuda/im2col.cu +196 -35
- data/ext/sources/ggml/src/ggml-cuda/im2col.cuh +1 -0
- data/ext/sources/ggml/src/ggml-cuda/mean.cu +57 -2
- data/ext/sources/ggml/src/ggml-cuda/mma.cuh +915 -69
- data/ext/sources/ggml/src/ggml-cuda/mmf.cu +171 -0
- data/ext/sources/ggml/src/ggml-cuda/mmf.cuh +835 -0
- data/ext/sources/ggml/src/ggml-cuda/mmid.cu +164 -0
- data/ext/sources/ggml/src/ggml-cuda/mmid.cuh +5 -0
- data/ext/sources/ggml/src/ggml-cuda/mmq.cu +109 -67
- data/ext/sources/ggml/src/ggml-cuda/mmq.cuh +1601 -733
- data/ext/sources/ggml/src/ggml-cuda/mmvf.cu +802 -0
- data/ext/sources/ggml/src/ggml-cuda/mmvf.cuh +12 -0
- data/ext/sources/ggml/src/ggml-cuda/mmvq.cu +286 -149
- data/ext/sources/ggml/src/ggml-cuda/mmvq.cuh +1 -1
- data/ext/sources/ggml/src/ggml-cuda/norm.cu +284 -12
- data/ext/sources/ggml/src/ggml-cuda/norm.cuh +7 -0
- data/ext/sources/ggml/src/ggml-cuda/opt-step-sgd.cu +49 -0
- data/ext/sources/ggml/src/ggml-cuda/opt-step-sgd.cuh +5 -0
- data/ext/sources/ggml/src/ggml-cuda/pad.cu +86 -32
- data/ext/sources/ggml/src/ggml-cuda/pad_reflect_1d.cu +91 -0
- data/ext/sources/ggml/src/ggml-cuda/pad_reflect_1d.cuh +5 -0
- data/ext/sources/ggml/src/ggml-cuda/quantize.cu +163 -10
- data/ext/sources/ggml/src/ggml-cuda/quantize.cuh +14 -0
- data/ext/sources/ggml/src/ggml-cuda/reduce_rows.cuh +53 -0
- data/ext/sources/ggml/src/ggml-cuda/roll.cu +67 -0
- data/ext/sources/ggml/src/ggml-cuda/roll.cuh +5 -0
- data/ext/sources/ggml/src/ggml-cuda/rope.cu +207 -98
- data/ext/sources/ggml/src/ggml-cuda/rope.cuh +2 -0
- data/ext/sources/ggml/src/ggml-cuda/scale.cu +14 -11
- data/ext/sources/ggml/src/ggml-cuda/set-rows.cu +330 -0
- data/ext/sources/ggml/src/ggml-cuda/set-rows.cuh +7 -0
- data/ext/sources/ggml/src/ggml-cuda/set.cu +39 -0
- data/ext/sources/ggml/src/ggml-cuda/set.cuh +7 -0
- data/ext/sources/ggml/src/ggml-cuda/softcap.cu +34 -0
- data/ext/sources/ggml/src/ggml-cuda/softcap.cuh +5 -0
- data/ext/sources/ggml/src/ggml-cuda/softmax.cu +325 -61
- data/ext/sources/ggml/src/ggml-cuda/solve_tri.cu +275 -0
- data/ext/sources/ggml/src/ggml-cuda/solve_tri.cuh +3 -0
- data/ext/sources/ggml/src/ggml-cuda/ssm-conv.cu +14 -12
- data/ext/sources/ggml/src/ggml-cuda/ssm-scan.cu +291 -104
- data/ext/sources/ggml/src/ggml-cuda/sum.cu +6 -10
- data/ext/sources/ggml/src/ggml-cuda/sumrows.cu +21 -4
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq112-dv112.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq128-dv128.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq256-dv256.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq40-dv40.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq576-dv512.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq64-dv64.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq72-dv72.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq80-dv80.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq96-dv96.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-f16.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q4_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q4_1.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q5_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q5_1.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q8_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-f16.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q4_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q4_1.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q5_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q5_1.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q8_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-f16.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q4_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q4_1.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q5_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q5_1.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q8_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-f16.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q4_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q4_1.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q5_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q5_1.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q8_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-f16.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q4_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q4_1.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q5_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q5_1.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q8_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-f16.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q4_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q4_1.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q5_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q5_1.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q8_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/generate_cu_files.py +40 -19
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_1.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_10.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_11.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_12.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_13.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_14.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_15.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_16.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_2.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_3.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_4.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_5.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_6.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_7.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_8.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_9.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmq-instance-mxfp4.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/top-k.cu +96 -0
- data/ext/sources/ggml/src/ggml-cuda/top-k.cuh +3 -0
- data/ext/sources/ggml/src/ggml-cuda/topk-moe.cu +351 -0
- data/ext/sources/ggml/src/ggml-cuda/topk-moe.cuh +21 -0
- data/ext/sources/ggml/src/ggml-cuda/tri.cu +136 -0
- data/ext/sources/ggml/src/ggml-cuda/tri.cuh +5 -0
- data/ext/sources/ggml/src/ggml-cuda/tsembd.cu +3 -3
- data/ext/sources/ggml/src/ggml-cuda/unary.cu +189 -5
- data/ext/sources/ggml/src/ggml-cuda/unary.cuh +44 -0
- data/ext/sources/ggml/src/ggml-cuda/upscale.cu +248 -6
- data/ext/sources/ggml/src/ggml-cuda/vecdotq.cuh +110 -22
- data/ext/sources/ggml/src/ggml-cuda/vendors/cuda.h +8 -0
- data/ext/sources/ggml/src/ggml-cuda/vendors/hip.h +70 -37
- data/ext/sources/ggml/src/ggml-cuda/vendors/musa.h +10 -3
- data/ext/sources/ggml/src/ggml-hexagon/CMakeLists.txt +80 -0
- data/ext/sources/ggml/src/ggml-hexagon/ggml-hexagon.cpp +3151 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/CMakeLists.txt +44 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/act-ops.c +682 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/binary-ops.c +360 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/cmake-toolchain.cmake +157 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/flash-attn-ops.c +566 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/get-rows-ops.c +112 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/htp-ctx.h +35 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/htp-dma.c +63 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/htp-dma.h +157 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/htp-msg.h +165 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/htp-ops.h +92 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/htp_iface.idl +16 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-exp.c +94 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-inverse.c +72 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-sigmoid.c +49 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-utils.c +1020 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-utils.h +1353 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/main.c +1001 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/matmul-ops.c +2503 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/ops-utils.h +149 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/rope-ops.c +487 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/set-rows-ops.c +168 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/softmax-ops.c +402 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/unary-ops.c +287 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/worker-pool.c +297 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/worker-pool.h +57 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp-utils.c +454 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp-utils.h +221 -0
- data/ext/sources/ggml/src/ggml-hexagon/op-desc.h +153 -0
- data/ext/sources/ggml/src/ggml-hip/CMakeLists.txt +16 -13
- data/ext/sources/ggml/src/ggml-impl.h +186 -15
- data/ext/sources/ggml/src/ggml-metal/CMakeLists.txt +10 -7
- data/ext/sources/ggml/src/ggml-metal/ggml-metal-common.cpp +446 -0
- data/ext/sources/ggml/src/ggml-metal/ggml-metal-common.h +52 -0
- data/ext/sources/ggml/src/ggml-metal/ggml-metal-context.h +33 -0
- data/ext/sources/ggml/src/ggml-metal/ggml-metal-context.m +609 -0
- data/ext/sources/ggml/src/ggml-metal/ggml-metal-device.cpp +1743 -0
- data/ext/sources/ggml/src/ggml-metal/ggml-metal-device.h +273 -0
- data/ext/sources/ggml/src/ggml-metal/ggml-metal-device.m +1686 -0
- data/ext/sources/ggml/src/ggml-metal/ggml-metal-impl.h +356 -61
- data/ext/sources/ggml/src/ggml-metal/ggml-metal-ops.cpp +4161 -0
- data/ext/sources/ggml/src/ggml-metal/ggml-metal-ops.h +94 -0
- data/ext/sources/ggml/src/ggml-metal/ggml-metal.cpp +724 -0
- data/ext/sources/ggml/src/ggml-metal/ggml-metal.metal +4495 -1876
- data/ext/sources/ggml/src/ggml-musa/CMakeLists.txt +21 -9
- data/ext/sources/ggml/src/ggml-opencl/CMakeLists.txt +29 -0
- data/ext/sources/ggml/src/ggml-opencl/ggml-opencl.cpp +4005 -427
- data/ext/sources/ggml/src/ggml-opencl/kernels/add.cl +107 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/add_id.cl +42 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/conv2d.cl +185 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/conv2d_f16_f32.cl +176 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/cvt.cl +147 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/div.cl +66 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/expm1.cl +82 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/fill.cl +17 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/flash_attn_f16.cl +370 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/flash_attn_f32.cl +371 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/flash_attn_f32_f16.cl +373 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/gelu.cl +27 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/gemm_moe_mxfp4_f32.cl +162 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/gemv_moe_mxfp4_f32.cl +156 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/get_rows.cl +36 -12
- data/ext/sources/ggml/src/ggml-opencl/kernels/glu.cl +177 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/group_norm.cl +49 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/im2col_f16.cl +1 -1
- data/ext/sources/ggml/src/ggml-opencl/kernels/im2col_f32.cl +1 -1
- data/ext/sources/ggml/src/ggml-opencl/kernels/mean.cl +39 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul.cl +73 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mat_f16_f32.cl +130 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mm_f16_f32_kq_kqv.cl +273 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mm_f16_f32_l4_lm.cl +146 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mm_f32_f32_l4_lm.cl +147 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mm_q8_0_f32_l4_lm.cl +154 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_id_mxfp4_f32.cl +189 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_id_mxfp4_f32_flat.cl +176 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_id_q8_0_f32.cl +140 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_id_q8_0_f32_flat.cl +222 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_mxfp4_f32.cl +144 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_mxfp4_f32_flat.cl +167 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q8_0_f32.cl +125 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q8_0_f32_flat.cl +202 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/norm.cl +80 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/pad.cl +29 -20
- data/ext/sources/ggml/src/ggml-opencl/kernels/rms_norm.cl +94 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/rope.cl +50 -24
- data/ext/sources/ggml/src/ggml-opencl/kernels/scale.cl +3 -2
- data/ext/sources/ggml/src/ggml-opencl/kernels/set_rows.cl +208 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/softmax_4_f16.cl +34 -13
- data/ext/sources/ggml/src/ggml-opencl/kernels/softmax_4_f32.cl +34 -13
- data/ext/sources/ggml/src/ggml-opencl/kernels/softmax_f16.cl +34 -13
- data/ext/sources/ggml/src/ggml-opencl/kernels/softmax_f32.cl +34 -13
- data/ext/sources/ggml/src/ggml-opencl/kernels/softplus.cl +88 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/sqr.cl +53 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/sqrt.cl +53 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/ssm_conv.cl +77 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/sub.cl +66 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/transpose.cl +33 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/tsembd.cl +2 -2
- data/ext/sources/ggml/src/ggml-opencl/kernels/upscale.cl +2 -3
- data/ext/sources/ggml/src/ggml-opt.cpp +97 -41
- data/ext/sources/ggml/src/ggml-quants.c +111 -16
- data/ext/sources/ggml/src/ggml-quants.h +6 -0
- data/ext/sources/ggml/src/ggml-rpc/ggml-rpc.cpp +497 -195
- data/ext/sources/ggml/src/ggml-sycl/CMakeLists.txt +48 -3
- data/ext/sources/ggml/src/ggml-sycl/add-id.cpp +77 -0
- data/ext/sources/ggml/src/ggml-sycl/add-id.hpp +8 -0
- data/ext/sources/ggml/src/ggml-sycl/backend.hpp +8 -0
- data/ext/sources/ggml/src/ggml-sycl/binbcast.cpp +6 -5
- data/ext/sources/ggml/src/ggml-sycl/common.hpp +117 -15
- data/ext/sources/ggml/src/ggml-sycl/concat.cpp +50 -30
- data/ext/sources/ggml/src/ggml-sycl/conv.cpp +10 -4
- data/ext/sources/ggml/src/ggml-sycl/convert.cpp +200 -99
- data/ext/sources/ggml/src/ggml-sycl/count-equal.cpp +79 -0
- data/ext/sources/ggml/src/ggml-sycl/count-equal.hpp +9 -0
- data/ext/sources/ggml/src/ggml-sycl/cpy.cpp +72 -309
- data/ext/sources/ggml/src/ggml-sycl/cpy.hpp +213 -1
- data/ext/sources/ggml/src/ggml-sycl/dequantize.hpp +18 -0
- data/ext/sources/ggml/src/ggml-sycl/dmmv.cpp +67 -49
- data/ext/sources/ggml/src/ggml-sycl/dpct/helper.hpp +77 -34
- data/ext/sources/ggml/src/ggml-sycl/element_wise.cpp +397 -314
- data/ext/sources/ggml/src/ggml-sycl/element_wise.hpp +12 -2
- data/ext/sources/ggml/src/ggml-sycl/gemm.hpp +14 -26
- data/ext/sources/ggml/src/ggml-sycl/getrows.cpp +9 -6
- data/ext/sources/ggml/src/ggml-sycl/ggml-sycl.cpp +643 -413
- data/ext/sources/ggml/src/ggml-sycl/gla.cpp +2 -2
- data/ext/sources/ggml/src/ggml-sycl/im2col.cpp +2 -2
- data/ext/sources/ggml/src/ggml-sycl/mmq.cpp +80 -60
- data/ext/sources/ggml/src/ggml-sycl/mmvq.cpp +223 -132
- data/ext/sources/ggml/src/ggml-sycl/norm.cpp +230 -55
- data/ext/sources/ggml/src/ggml-sycl/norm.hpp +2 -0
- data/ext/sources/ggml/src/ggml-sycl/pad.cpp +97 -0
- data/ext/sources/ggml/src/ggml-sycl/pad.hpp +24 -0
- data/ext/sources/ggml/src/ggml-sycl/pad_reflect_1d.cpp +100 -0
- data/ext/sources/ggml/src/ggml-sycl/pad_reflect_1d.hpp +10 -0
- data/ext/sources/ggml/src/ggml-sycl/presets.hpp +2 -0
- data/ext/sources/ggml/src/ggml-sycl/quantize.hpp +133 -0
- data/ext/sources/ggml/src/ggml-sycl/quants.hpp +8 -9
- data/ext/sources/ggml/src/ggml-sycl/repeat_back.cpp +76 -0
- data/ext/sources/ggml/src/ggml-sycl/repeat_back.hpp +8 -0
- data/ext/sources/ggml/src/ggml-sycl/roll.cpp +122 -0
- data/ext/sources/ggml/src/ggml-sycl/roll.hpp +20 -0
- data/ext/sources/ggml/src/ggml-sycl/rope.cpp +65 -59
- data/ext/sources/ggml/src/ggml-sycl/set.cpp +73 -0
- data/ext/sources/ggml/src/ggml-sycl/set.hpp +5 -0
- data/ext/sources/ggml/src/ggml-sycl/set_rows.cpp +234 -0
- data/ext/sources/ggml/src/ggml-sycl/set_rows.hpp +8 -0
- data/ext/sources/ggml/src/ggml-sycl/softmax.cpp +330 -165
- data/ext/sources/ggml/src/ggml-sycl/softmax.hpp +4 -0
- data/ext/sources/ggml/src/ggml-sycl/ssm_conv.cpp +127 -0
- data/ext/sources/ggml/src/ggml-sycl/ssm_conv.hpp +5 -0
- data/ext/sources/ggml/src/ggml-sycl/tsembd.cpp +12 -6
- data/ext/sources/ggml/src/ggml-sycl/vecdotq.hpp +60 -6
- data/ext/sources/ggml/src/ggml-sycl/wkv.cpp +16 -12
- data/ext/sources/ggml/src/ggml-vulkan/CMakeLists.txt +38 -18
- data/ext/sources/ggml/src/ggml-vulkan/ggml-vulkan.cpp +7398 -2635
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/abs.comp +21 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/acc.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/add.comp +43 -3
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/add1.comp +28 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/add_id.comp +42 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/arange.comp +20 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/argmax.comp +15 -6
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/argsort.comp +56 -39
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/argsort_large.comp +114 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/ceil.comp +22 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/clamp.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/concat.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/contig_copy.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/conv2d_dw.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/conv2d_mm.comp +347 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/conv_transpose_1d.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/copy.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/copy_from_quant.comp +5 -5
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/copy_to_quant.comp +67 -13
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/copy_transpose.comp +67 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/cos.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/count_equal.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/count_experts.comp +51 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/cumsum.comp +83 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/cumsum_multipass1.comp +60 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/cumsum_multipass2.comp +66 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_f32.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{dequant_funcs.comp → dequant_funcs.glsl} +158 -16
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{dequant_funcs_cm2.comp → dequant_funcs_cm2.glsl} +38 -3
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{dequant_head.comp → dequant_head.glsl} +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq1_m.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq1_s.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq2_s.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq2_xs.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq2_xxs.comp +3 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq3_s.comp +7 -6
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq3_xxs.comp +5 -3
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq4_nl.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq4_xs.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_mxfp4.comp +32 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q2_k.comp +4 -4
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q3_k.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q4_0.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q4_1.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q4_k.comp +4 -4
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q5_0.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q5_1.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q5_k.comp +4 -4
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q6_k.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q8_0.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/diag.comp +29 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/diag_mask_inf.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/div.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/exp.comp +21 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/fill.comp +19 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn.comp +103 -36
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_base.glsl +220 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm1.comp +139 -45
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm2.comp +113 -38
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_split_k_reduce.comp +75 -14
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/floor.comp +22 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/geglu.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/geglu_erf.comp +27 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/geglu_quick.comp +11 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/gelu.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/gelu_erf.comp +39 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/gelu_quick.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{generic_binary_head.comp → generic_binary_head.glsl} +19 -17
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{generic_head.comp → generic_head.glsl} +2 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{generic_unary_head.comp → generic_unary_head.glsl} +7 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/get_rows.comp +21 -12
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/get_rows_quant.comp +28 -18
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{glu_head.comp → glu_head.glsl} +4 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/group_norm.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/hardsigmoid.comp +22 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/hardswish.comp +22 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/im2col.comp +33 -17
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/im2col_3d.comp +125 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/l2_norm.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/leaky_relu.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/log.comp +18 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_base.glsl +227 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iface.glsl +35 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq1_m.comp +71 -21
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq1_s.comp +41 -25
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq2_s.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq2_xs.comp +44 -26
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq2_xxs.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq3_s.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq3_xxs.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_nc.comp +20 -14
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_p021.comp +9 -7
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q2_k.comp +4 -6
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q3_k.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q4_k.comp +4 -6
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q5_k.comp +4 -6
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q6_k.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vecq.comp +143 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vecq_funcs.glsl +494 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm.comp +144 -556
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm_cm2.comp +230 -51
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm_funcs.glsl +566 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm_id_funcs.glsl +72 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq.comp +90 -223
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq_funcs.glsl +454 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq_shmem_types.glsl +78 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/multi_add.comp +195 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/neg.comp +20 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/norm.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/opt_step_adamw.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/opt_step_sgd.comp +22 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/pad.comp +41 -5
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/pool2d.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/quantize_q8_1.comp +59 -9
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/reglu.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/relu.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/repeat.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/repeat_back.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rms_norm.comp +104 -14
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rms_norm_back.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rms_norm_partials.comp +65 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/roll.comp +46 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_funcs.glsl +234 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_head.glsl +20 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_multi.comp +6 -52
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_neox.comp +6 -35
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_norm.comp +6 -35
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_params.glsl +28 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_vision.comp +6 -39
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/round.comp +29 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rte.glsl +5 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/scale.comp +3 -3
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/sigmoid.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/silu.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/silu_back.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/sin.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/soft_max.comp +30 -8
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/soft_max_back.comp +6 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/soft_max_large1.comp +62 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/soft_max_large2.comp +79 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/soft_max_large3.comp +65 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/soft_max_large_common.glsl +53 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/softplus.comp +23 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/solve_tri.comp +81 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/sqrt.comp +17 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/square.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/ssm_conv.comp +44 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/ssm_scan.comp +124 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/step.comp +22 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/sub.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/sum_rows.comp +16 -6
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/sum_rows.glsl +25 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/swiglu.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/swiglu_oai.comp +14 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/tanh.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/timestep_embedding.comp +5 -4
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/topk_argsort.comp +118 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/topk_moe.comp +213 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/topk_nary_search.comp +246 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/tri.comp +43 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/trunc.comp +22 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{types.comp → types.glsl} +435 -24
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/upscale.comp +148 -6
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/utils.glsl +25 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp +619 -177
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/xielu.comp +35 -0
- data/ext/sources/ggml/src/ggml-webgpu/CMakeLists.txt +80 -0
- data/ext/sources/ggml/src/ggml-webgpu/ggml-webgpu-shader-lib.hpp +169 -0
- data/ext/sources/ggml/src/ggml-webgpu/ggml-webgpu.cpp +3087 -0
- data/ext/sources/ggml/src/ggml-webgpu/pre_wgsl.hpp +778 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/bin_op.tmpl.wgsl +188 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/binary_head.tmpl +45 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/common_decls.tmpl +930 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/cpy.tmpl.wgsl +101 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/embed_wgsl.py +147 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/flash_attn.wgsl +591 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/get_rows.tmpl.wgsl +874 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/glu.tmpl.wgsl +323 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/memset.wgsl +40 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat.tmpl.wgsl +907 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_decls.tmpl +97 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_reg_tile.tmpl.wgsl +247 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_subgroup_matrix.tmpl.wgsl +302 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_vec.tmpl.wgsl +267 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/rms_norm.wgsl +123 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/rope.tmpl.wgsl +295 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/scale.tmpl.wgsl +90 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/set_rows.tmpl.wgsl +112 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/set_rows.wgsl +81 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/soft_max.tmpl.wgsl +345 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/unary_op.wgsl +483 -0
- data/ext/sources/ggml/src/ggml-zdnn/CMakeLists.txt +36 -0
- data/ext/sources/ggml/src/ggml-zdnn/common.hpp +59 -0
- data/ext/sources/ggml/src/ggml-zdnn/ggml-zdnn.cpp +628 -0
- data/ext/sources/ggml/src/ggml-zdnn/mmf.cpp +80 -0
- data/ext/sources/ggml/src/ggml-zdnn/mmf.hpp +12 -0
- data/ext/sources/ggml/src/ggml-zdnn/utils.cpp +79 -0
- data/ext/sources/ggml/src/ggml-zdnn/utils.hpp +19 -0
- data/ext/sources/ggml/src/ggml-zendnn/CMakeLists.txt +92 -0
- data/ext/sources/ggml/src/ggml-zendnn/ggml-zendnn.cpp +466 -0
- data/ext/sources/ggml/src/ggml.c +901 -129
- data/ext/sources/ggml/src/gguf.cpp +8 -1
- data/ext/sources/include/whisper.h +1 -0
- data/ext/sources/src/CMakeLists.txt +3 -1
- data/ext/sources/src/whisper.cpp +124 -81
- data/ext/sources/tests/CMakeLists.txt +8 -1
- data/ext/sources/tests/test-vad-full.cpp +7 -5
- data/ext/sources/tests/test-vad.cpp +3 -3
- data/extsources.rb +1 -0
- data/lib/whisper/model/uri.rb +17 -18
- data/sig/whisper.rbs +126 -2
- data/test/test_params.rb +24 -8
- data/test/test_segment.rb +0 -1
- data/test/test_token.rb +70 -0
- data/test/test_vad.rb +1 -1
- data/test/test_vad_context.rb +50 -0
- data/test/test_vad_segment.rb +19 -0
- data/test/test_vad_segments.rb +16 -0
- data/test/test_whisper.rb +8 -1
- data/whispercpp.gemspec +1 -1
- metadata +439 -179
- data/ext/sources/build-xcframework.sh +0 -547
- data/ext/sources/examples/talk-llama/llama-kv-cache-unified-iswa.cpp +0 -279
- data/ext/sources/examples/talk-llama/llama-kv-cache-unified.cpp +0 -1841
- data/ext/sources/examples/talk-llama/llama-kv-cache-unified.h +0 -303
- data/ext/sources/ggml/include/ggml-kompute.h +0 -50
- data/ext/sources/ggml/src/ggml-amx/CMakeLists.txt +0 -107
- data/ext/sources/ggml/src/ggml-amx/common.h +0 -94
- data/ext/sources/ggml/src/ggml-amx/ggml-amx.cpp +0 -446
- data/ext/sources/ggml/src/ggml-amx/mmq.cpp +0 -2510
- data/ext/sources/ggml/src/ggml-amx/mmq.h +0 -17
- data/ext/sources/ggml/src/ggml-cann/Doxyfile +0 -2579
- data/ext/sources/ggml/src/ggml-cann/kernels/CMakeLists.txt +0 -30
- data/ext/sources/ggml/src/ggml-cann/kernels/ascendc_kernels.h +0 -19
- data/ext/sources/ggml/src/ggml-cann/kernels/dup.cpp +0 -234
- data/ext/sources/ggml/src/ggml-cann/kernels/get_row_f16.cpp +0 -197
- data/ext/sources/ggml/src/ggml-cann/kernels/get_row_f32.cpp +0 -190
- data/ext/sources/ggml/src/ggml-cann/kernels/get_row_q4_0.cpp +0 -204
- data/ext/sources/ggml/src/ggml-cann/kernels/get_row_q8_0.cpp +0 -191
- data/ext/sources/ggml/src/ggml-cann/kernels/quantize_f16_q8_0.cpp +0 -218
- data/ext/sources/ggml/src/ggml-cann/kernels/quantize_f32_q8_0.cpp +0 -216
- data/ext/sources/ggml/src/ggml-cann/kernels/quantize_float_to_q4_0.cpp +0 -295
- data/ext/sources/ggml/src/ggml-cuda/fattn-tile-f16.cu +0 -357
- data/ext/sources/ggml/src/ggml-cuda/fattn-tile-f16.cuh +0 -3
- data/ext/sources/ggml/src/ggml-cuda/fattn-tile-f32.cu +0 -365
- data/ext/sources/ggml/src/ggml-cuda/fattn-tile-f32.cuh +0 -3
- data/ext/sources/ggml/src/ggml-cuda/fattn-vec-f16.cuh +0 -482
- data/ext/sources/ggml/src/ggml-cuda/fattn-vec-f32.cuh +0 -472
- data/ext/sources/ggml/src/ggml-cuda/mmv.cu +0 -506
- data/ext/sources/ggml/src/ggml-cuda/mmv.cuh +0 -11
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs256-f16-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs256-f16-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-kompute/CMakeLists.txt +0 -166
- data/ext/sources/ggml/src/ggml-kompute/ggml-kompute.cpp +0 -2251
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/common.comp +0 -112
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_add.comp +0 -58
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_addrow.comp +0 -25
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_cpy_f16_f16.comp +0 -52
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_cpy_f16_f32.comp +0 -52
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_cpy_f32_f16.comp +0 -52
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_cpy_f32_f32.comp +0 -52
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_diagmask.comp +0 -30
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_gelu.comp +0 -22
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_getrows.comp +0 -17
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_getrows_f16.comp +0 -31
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_getrows_f32.comp +0 -31
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_getrows_q4_0.comp +0 -38
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_getrows_q4_1.comp +0 -39
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_getrows_q6_k.comp +0 -44
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul.comp +0 -52
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_f16.comp +0 -69
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_mat_f32.comp +0 -51
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q4_0.comp +0 -33
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q4_1.comp +0 -35
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q4_k.comp +0 -140
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q6_k.comp +0 -106
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q8_0.comp +0 -73
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mv_q_n.comp +0 -52
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mv_q_n_pre.comp +0 -28
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_norm.comp +0 -84
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_relu.comp +0 -21
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_rmsnorm.comp +0 -53
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_rope_neox_f16.comp +0 -52
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_rope_neox_f32.comp +0 -52
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_rope_norm_f16.comp +0 -52
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_rope_norm_f32.comp +0 -52
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_scale.comp +0 -19
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_scale_8.comp +0 -23
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_silu.comp +0 -22
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_softmax.comp +0 -72
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/rope_common.comp +0 -71
- data/ext/sources/ggml/src/ggml-metal/ggml-metal.m +0 -6280
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_base.comp +0 -162
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_base.comp +0 -118
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq_funcs.comp +0 -99
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_head.comp +0 -58
- /data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{test_bfloat16_support.comp → feature-tests/bfloat16.comp} +0 -0
- /data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{test_coopmat_support.comp → feature-tests/coopmat.comp} +0 -0
- /data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{test_coopmat2_support.comp → feature-tests/coopmat2.comp} +0 -0
- /data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{test_integer_dot_support.comp → feature-tests/integer_dot.comp} +0 -0
- /data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{glu_main.comp → glu_main.glsl} +0 -0
|
@@ -43,7 +43,7 @@ ruby_whisper_transcribe(int argc, VALUE *argv, VALUE self) {
|
|
|
43
43
|
|
|
44
44
|
int n_processors = opts[0] == Qundef ? 1 : NUM2INT(opts[0]);
|
|
45
45
|
|
|
46
|
-
|
|
46
|
+
GetContext(self, rw);
|
|
47
47
|
TypedData_Get_Struct(params, ruby_whisper_params, &ruby_whisper_params_type, rwp);
|
|
48
48
|
|
|
49
49
|
if (!rb_respond_to(wave_file_path, id_to_s)) {
|
|
@@ -0,0 +1,75 @@
|
|
|
1
|
+
#include <ruby.h>
|
|
2
|
+
#include "ruby_whisper.h"
|
|
3
|
+
|
|
4
|
+
extern ID id_to_s;
|
|
5
|
+
|
|
6
|
+
extern VALUE cVADContext;
|
|
7
|
+
|
|
8
|
+
extern VALUE ruby_whisper_vad_detect(VALUE self, VALUE file_path, VALUE params);
|
|
9
|
+
extern VALUE ruby_whisper_normalize_model_path(VALUE model_path);
|
|
10
|
+
|
|
11
|
+
static size_t
|
|
12
|
+
ruby_whisper_vad_context_memsize(const void *p)
|
|
13
|
+
{
|
|
14
|
+
const ruby_whisper_vad_context *rwvc = p;
|
|
15
|
+
size_t size = sizeof(rwvc);
|
|
16
|
+
if (!rwvc) {
|
|
17
|
+
return 0;
|
|
18
|
+
}
|
|
19
|
+
if (rwvc->context) {
|
|
20
|
+
size += sizeof(rwvc->context);
|
|
21
|
+
}
|
|
22
|
+
return size;
|
|
23
|
+
}
|
|
24
|
+
|
|
25
|
+
static void
|
|
26
|
+
ruby_whisper_vad_context_free(void *p)
|
|
27
|
+
{
|
|
28
|
+
ruby_whisper_vad_context *rwvc = (ruby_whisper_vad_context *)p;
|
|
29
|
+
if (rwvc->context) {
|
|
30
|
+
whisper_vad_free(rwvc->context);
|
|
31
|
+
rwvc->context = NULL;
|
|
32
|
+
}
|
|
33
|
+
xfree(rwvc);
|
|
34
|
+
}
|
|
35
|
+
|
|
36
|
+
const rb_data_type_t ruby_whisper_vad_context_type = {
|
|
37
|
+
"ruby_whisper_vad_context",
|
|
38
|
+
{0, ruby_whisper_vad_context_free, ruby_whisper_vad_context_memsize,},
|
|
39
|
+
0, 0,
|
|
40
|
+
0
|
|
41
|
+
};
|
|
42
|
+
|
|
43
|
+
static VALUE
|
|
44
|
+
ruby_whisper_vad_context_s_allocate(VALUE klass)
|
|
45
|
+
{
|
|
46
|
+
ruby_whisper_vad_context *rwvc;
|
|
47
|
+
VALUE obj = TypedData_Make_Struct(klass, ruby_whisper_vad_context, &ruby_whisper_vad_context_type, rwvc);
|
|
48
|
+
rwvc->context = NULL;
|
|
49
|
+
return obj;
|
|
50
|
+
}
|
|
51
|
+
|
|
52
|
+
static VALUE
|
|
53
|
+
ruby_whisper_vad_context_initialize(VALUE self, VALUE model_path)
|
|
54
|
+
{
|
|
55
|
+
ruby_whisper_vad_context *rwvc;
|
|
56
|
+
struct whisper_vad_context *context;
|
|
57
|
+
|
|
58
|
+
model_path = ruby_whisper_normalize_model_path(model_path);
|
|
59
|
+
context = whisper_vad_init_from_file_with_params(StringValueCStr(model_path), whisper_vad_default_context_params());
|
|
60
|
+
if (context == NULL) {
|
|
61
|
+
rb_raise(rb_eRuntimeError, "Failed to initialize whisper VAD context");
|
|
62
|
+
}
|
|
63
|
+
TypedData_Get_Struct(self, ruby_whisper_vad_context, &ruby_whisper_vad_context_type, rwvc);
|
|
64
|
+
rwvc->context = context;
|
|
65
|
+
|
|
66
|
+
return Qnil;
|
|
67
|
+
}
|
|
68
|
+
|
|
69
|
+
void init_ruby_whisper_vad_context(VALUE *mVAD)
|
|
70
|
+
{
|
|
71
|
+
cVADContext = rb_define_class_under(*mVAD, "Context", rb_cObject);
|
|
72
|
+
rb_define_alloc_func(cVADContext, ruby_whisper_vad_context_s_allocate);
|
|
73
|
+
rb_define_method(cVADContext, "initialize", ruby_whisper_vad_context_initialize, 1);
|
|
74
|
+
rb_define_method(cVADContext, "detect", ruby_whisper_vad_detect, 2);
|
|
75
|
+
}
|
|
@@ -0,0 +1,50 @@
|
|
|
1
|
+
#include <ruby.h>
|
|
2
|
+
#include "ruby_whisper.h"
|
|
3
|
+
#include "common-whisper.h"
|
|
4
|
+
#include <string>
|
|
5
|
+
#include <vector>
|
|
6
|
+
|
|
7
|
+
#ifdef __cplusplus
|
|
8
|
+
extern "C" {
|
|
9
|
+
#endif
|
|
10
|
+
|
|
11
|
+
extern VALUE cVADSegments;
|
|
12
|
+
|
|
13
|
+
extern const rb_data_type_t ruby_whisper_vad_context_type;
|
|
14
|
+
extern const rb_data_type_t ruby_whisper_vad_params_type;
|
|
15
|
+
extern const rb_data_type_t ruby_whisper_vad_segments_type;
|
|
16
|
+
|
|
17
|
+
extern VALUE ruby_whisper_vad_segments_s_init(struct whisper_vad_segments *segments);
|
|
18
|
+
|
|
19
|
+
VALUE
|
|
20
|
+
ruby_whisper_vad_detect(VALUE self, VALUE file_path, VALUE params) {
|
|
21
|
+
ruby_whisper_vad_context *rwvc;
|
|
22
|
+
ruby_whisper_vad_params *rwvp;
|
|
23
|
+
std::string cpp_file_path;
|
|
24
|
+
std::vector<float> pcmf32;
|
|
25
|
+
std::vector<std::vector<float>> pcmf32s;
|
|
26
|
+
whisper_vad_segments *segments;
|
|
27
|
+
|
|
28
|
+
TypedData_Get_Struct(self, ruby_whisper_vad_context, &ruby_whisper_vad_context_type, rwvc);
|
|
29
|
+
if (rwvc->context == NULL) {
|
|
30
|
+
rb_raise(rb_eRuntimeError, "Doesn't have referenxe to context internally");
|
|
31
|
+
}
|
|
32
|
+
TypedData_Get_Struct(params, ruby_whisper_vad_params, &ruby_whisper_vad_params_type, rwvp);
|
|
33
|
+
|
|
34
|
+
cpp_file_path = StringValueCStr(file_path);
|
|
35
|
+
|
|
36
|
+
if (!read_audio_data(cpp_file_path, pcmf32, pcmf32s, false)) {
|
|
37
|
+
rb_raise(rb_eRuntimeError, "Failed to open '%s' as WAV file\n", cpp_file_path.c_str());
|
|
38
|
+
}
|
|
39
|
+
|
|
40
|
+
segments = whisper_vad_segments_from_samples(rwvc->context, rwvp->params, pcmf32.data(), pcmf32.size());
|
|
41
|
+
if (segments == nullptr) {
|
|
42
|
+
rb_raise(rb_eRuntimeError, "Failed to process audio\n");
|
|
43
|
+
}
|
|
44
|
+
|
|
45
|
+
return ruby_whisper_vad_segments_s_init(segments);
|
|
46
|
+
}
|
|
47
|
+
|
|
48
|
+
#ifdef __cplusplus
|
|
49
|
+
}
|
|
50
|
+
#endif
|
|
@@ -0,0 +1,139 @@
|
|
|
1
|
+
#include <ruby.h>
|
|
2
|
+
#include "ruby_whisper.h"
|
|
3
|
+
|
|
4
|
+
#define N_KEY_NAMES 2
|
|
5
|
+
|
|
6
|
+
extern VALUE cVADSegment;
|
|
7
|
+
|
|
8
|
+
extern const rb_data_type_t ruby_whisper_vad_segments_type;
|
|
9
|
+
|
|
10
|
+
static VALUE sym_start_time;
|
|
11
|
+
static VALUE sym_end_time;
|
|
12
|
+
|
|
13
|
+
static void
|
|
14
|
+
rb_whisper_vad_segment_mark(void *p)
|
|
15
|
+
{
|
|
16
|
+
ruby_whisper_vad_segment *rwvs = (ruby_whisper_vad_segment *)p;
|
|
17
|
+
rb_gc_mark(rwvs->segments);
|
|
18
|
+
}
|
|
19
|
+
|
|
20
|
+
static size_t
|
|
21
|
+
ruby_whisper_vad_segment_memsize(const void *p)
|
|
22
|
+
{
|
|
23
|
+
const ruby_whisper_vad_segment *rwvs = p;
|
|
24
|
+
size_t size = sizeof(rwvs);
|
|
25
|
+
if (!rwvs) {
|
|
26
|
+
return 0;
|
|
27
|
+
}
|
|
28
|
+
if (rwvs->index) {
|
|
29
|
+
size += sizeof(rwvs->index);
|
|
30
|
+
}
|
|
31
|
+
return size;
|
|
32
|
+
}
|
|
33
|
+
|
|
34
|
+
static const rb_data_type_t ruby_whisper_vad_segment_type = {
|
|
35
|
+
"ruby_whisper_vad_segment",
|
|
36
|
+
{rb_whisper_vad_segment_mark, RUBY_DEFAULT_FREE, ruby_whisper_vad_segment_memsize,},
|
|
37
|
+
0, 0,
|
|
38
|
+
0
|
|
39
|
+
};
|
|
40
|
+
|
|
41
|
+
static VALUE
|
|
42
|
+
ruby_whisper_vad_segment_s_allocate(VALUE klass)
|
|
43
|
+
{
|
|
44
|
+
ruby_whisper_vad_segment *rwvs;
|
|
45
|
+
VALUE obj = TypedData_Make_Struct(klass, ruby_whisper_vad_segment, &ruby_whisper_vad_segment_type, rwvs);
|
|
46
|
+
rwvs->segments = Qnil;
|
|
47
|
+
rwvs->index = -1;
|
|
48
|
+
return obj;
|
|
49
|
+
}
|
|
50
|
+
|
|
51
|
+
VALUE
|
|
52
|
+
rb_whisper_vad_segment_s_new(VALUE segments, int index)
|
|
53
|
+
{
|
|
54
|
+
ruby_whisper_vad_segment *rwvs;
|
|
55
|
+
const VALUE segment = ruby_whisper_vad_segment_s_allocate(cVADSegment);
|
|
56
|
+
TypedData_Get_Struct(segment, ruby_whisper_vad_segment, &ruby_whisper_vad_segment_type, rwvs);
|
|
57
|
+
rwvs->segments = segments;
|
|
58
|
+
rwvs->index = index;
|
|
59
|
+
return segment;
|
|
60
|
+
}
|
|
61
|
+
|
|
62
|
+
static VALUE
|
|
63
|
+
ruby_whisper_vad_segment_get_start_time(VALUE self)
|
|
64
|
+
{
|
|
65
|
+
ruby_whisper_vad_segment *rwvs;
|
|
66
|
+
ruby_whisper_vad_segments *rwvss;
|
|
67
|
+
float t0;
|
|
68
|
+
|
|
69
|
+
TypedData_Get_Struct(self, ruby_whisper_vad_segment, &ruby_whisper_vad_segment_type, rwvs);
|
|
70
|
+
TypedData_Get_Struct(rwvs->segments, ruby_whisper_vad_segments, &ruby_whisper_vad_segments_type, rwvss);
|
|
71
|
+
t0 = whisper_vad_segments_get_segment_t0(rwvss->segments, rwvs->index);
|
|
72
|
+
return DBL2NUM(t0 * 10);
|
|
73
|
+
}
|
|
74
|
+
|
|
75
|
+
static VALUE
|
|
76
|
+
ruby_whisper_vad_segment_get_end_time(VALUE self)
|
|
77
|
+
{
|
|
78
|
+
ruby_whisper_vad_segment *rwvs;
|
|
79
|
+
ruby_whisper_vad_segments *rwvss;
|
|
80
|
+
float t1;
|
|
81
|
+
|
|
82
|
+
TypedData_Get_Struct(self, ruby_whisper_vad_segment, &ruby_whisper_vad_segment_type, rwvs);
|
|
83
|
+
TypedData_Get_Struct(rwvs->segments, ruby_whisper_vad_segments, &ruby_whisper_vad_segments_type, rwvss);
|
|
84
|
+
t1 = whisper_vad_segments_get_segment_t1(rwvss->segments, rwvs->index);
|
|
85
|
+
return DBL2NUM(t1 * 10);
|
|
86
|
+
}
|
|
87
|
+
|
|
88
|
+
static VALUE
|
|
89
|
+
ruby_whisper_vad_segment_deconstruct_keys(VALUE self, VALUE keys)
|
|
90
|
+
{
|
|
91
|
+
ruby_whisper_vad_segment *rwvs;
|
|
92
|
+
ruby_whisper_vad_segments *rwvss;
|
|
93
|
+
VALUE hash, key;
|
|
94
|
+
long n_keys;
|
|
95
|
+
int i;
|
|
96
|
+
|
|
97
|
+
TypedData_Get_Struct(self, ruby_whisper_vad_segment, &ruby_whisper_vad_segment_type, rwvs);
|
|
98
|
+
TypedData_Get_Struct(rwvs->segments, ruby_whisper_vad_segments, &ruby_whisper_vad_segments_type, rwvss);
|
|
99
|
+
|
|
100
|
+
hash = rb_hash_new();
|
|
101
|
+
if (NIL_P(keys)) {
|
|
102
|
+
keys = rb_ary_new3(
|
|
103
|
+
N_KEY_NAMES,
|
|
104
|
+
sym_start_time,
|
|
105
|
+
sym_end_time
|
|
106
|
+
);
|
|
107
|
+
n_keys = N_KEY_NAMES;
|
|
108
|
+
} else {
|
|
109
|
+
n_keys = RARRAY_LEN(keys);
|
|
110
|
+
if (n_keys > N_KEY_NAMES) {
|
|
111
|
+
return hash;
|
|
112
|
+
}
|
|
113
|
+
}
|
|
114
|
+
for (i = 0; i < n_keys; i++) {
|
|
115
|
+
key = rb_ary_entry(keys, i);
|
|
116
|
+
if (key == sym_start_time) {
|
|
117
|
+
rb_hash_aset(hash, key, ruby_whisper_vad_segment_get_start_time(self));
|
|
118
|
+
}
|
|
119
|
+
if (key == sym_end_time) {
|
|
120
|
+
rb_hash_aset(hash, key, ruby_whisper_vad_segment_get_end_time(self));
|
|
121
|
+
}
|
|
122
|
+
}
|
|
123
|
+
|
|
124
|
+
return hash;
|
|
125
|
+
}
|
|
126
|
+
|
|
127
|
+
void
|
|
128
|
+
init_ruby_whisper_vad_segment(VALUE *mVAD)
|
|
129
|
+
{
|
|
130
|
+
cVADSegment = rb_define_class_under(*mVAD, "Segment", rb_cObject);
|
|
131
|
+
|
|
132
|
+
sym_start_time = ID2SYM(rb_intern("start_time"));
|
|
133
|
+
sym_end_time = ID2SYM(rb_intern("end_time"));
|
|
134
|
+
|
|
135
|
+
rb_define_alloc_func(cVADSegment, ruby_whisper_vad_segment_s_allocate);
|
|
136
|
+
rb_define_method(cVADSegment, "start_time", ruby_whisper_vad_segment_get_start_time, 0);
|
|
137
|
+
rb_define_method(cVADSegment, "end_time", ruby_whisper_vad_segment_get_end_time, 0);
|
|
138
|
+
rb_define_method(cVADSegment, "deconstruct_keys", ruby_whisper_vad_segment_deconstruct_keys, 1);
|
|
139
|
+
}
|
|
@@ -0,0 +1,106 @@
|
|
|
1
|
+
#include <ruby.h>
|
|
2
|
+
#include "ruby_whisper.h"
|
|
3
|
+
|
|
4
|
+
extern ID id___method__;
|
|
5
|
+
extern ID id_to_enum;
|
|
6
|
+
|
|
7
|
+
extern VALUE cVADSegments;
|
|
8
|
+
|
|
9
|
+
extern VALUE rb_whisper_vad_segment_s_new(VALUE segments, int index);
|
|
10
|
+
|
|
11
|
+
static size_t
|
|
12
|
+
ruby_whisper_vad_segments_memsize(const void *p)
|
|
13
|
+
{
|
|
14
|
+
const ruby_whisper_vad_segments *rwvss = p;
|
|
15
|
+
size_t size = sizeof(rwvss);
|
|
16
|
+
if (!rwvss) {
|
|
17
|
+
return 0;
|
|
18
|
+
}
|
|
19
|
+
if (rwvss->segments) {
|
|
20
|
+
size += sizeof(rwvss->segments);
|
|
21
|
+
}
|
|
22
|
+
return size;
|
|
23
|
+
}
|
|
24
|
+
|
|
25
|
+
static void
|
|
26
|
+
ruby_whisper_vad_segments_free(void *p)
|
|
27
|
+
{
|
|
28
|
+
ruby_whisper_vad_segments *rwvss = (ruby_whisper_vad_segments *)p;
|
|
29
|
+
if (rwvss->segments) {
|
|
30
|
+
whisper_vad_free_segments(rwvss->segments);
|
|
31
|
+
rwvss->segments = NULL;
|
|
32
|
+
}
|
|
33
|
+
xfree(rwvss);
|
|
34
|
+
}
|
|
35
|
+
|
|
36
|
+
const rb_data_type_t ruby_whisper_vad_segments_type = {
|
|
37
|
+
"ruby_whisper_vad_segments",
|
|
38
|
+
{0, ruby_whisper_vad_segments_free, ruby_whisper_vad_segments_memsize,},
|
|
39
|
+
0, 0,
|
|
40
|
+
0
|
|
41
|
+
};
|
|
42
|
+
|
|
43
|
+
static VALUE
|
|
44
|
+
ruby_whisper_vad_segments_s_allocate(VALUE klass)
|
|
45
|
+
{
|
|
46
|
+
ruby_whisper_vad_segments *rwvss;
|
|
47
|
+
VALUE obj = TypedData_Make_Struct(klass, ruby_whisper_vad_segments, &ruby_whisper_vad_segments_type, rwvss);
|
|
48
|
+
rwvss->segments = NULL;
|
|
49
|
+
return obj;
|
|
50
|
+
}
|
|
51
|
+
|
|
52
|
+
VALUE
|
|
53
|
+
ruby_whisper_vad_segments_s_init(struct whisper_vad_segments *segments)
|
|
54
|
+
{
|
|
55
|
+
VALUE rb_segments;
|
|
56
|
+
ruby_whisper_vad_segments *rwvss;
|
|
57
|
+
|
|
58
|
+
rb_segments = ruby_whisper_vad_segments_s_allocate(cVADSegments);
|
|
59
|
+
TypedData_Get_Struct(rb_segments, ruby_whisper_vad_segments, &ruby_whisper_vad_segments_type, rwvss);
|
|
60
|
+
rwvss->segments = segments;
|
|
61
|
+
|
|
62
|
+
return rb_segments;
|
|
63
|
+
}
|
|
64
|
+
|
|
65
|
+
static VALUE
|
|
66
|
+
ruby_whisper_vad_segments_each(VALUE self)
|
|
67
|
+
{
|
|
68
|
+
ruby_whisper_vad_segments *rwvss;
|
|
69
|
+
VALUE method_name;
|
|
70
|
+
int n_segments, i;
|
|
71
|
+
|
|
72
|
+
if (!rb_block_given_p()) {
|
|
73
|
+
method_name = rb_funcall(self, id___method__, 0);
|
|
74
|
+
return rb_funcall(self, id_to_enum, 1, method_name);
|
|
75
|
+
}
|
|
76
|
+
|
|
77
|
+
GetVADSegments(self, rwvss);
|
|
78
|
+
n_segments = whisper_vad_segments_n_segments(rwvss->segments);
|
|
79
|
+
for (i = 0; i < n_segments; ++i) {
|
|
80
|
+
rb_yield(rb_whisper_vad_segment_s_new(self, i));
|
|
81
|
+
}
|
|
82
|
+
|
|
83
|
+
return self;
|
|
84
|
+
}
|
|
85
|
+
|
|
86
|
+
static VALUE
|
|
87
|
+
ruby_whisper_vad_segments_get_length(VALUE self)
|
|
88
|
+
{
|
|
89
|
+
ruby_whisper_vad_segments *rwvss;
|
|
90
|
+
int n_segments;
|
|
91
|
+
|
|
92
|
+
GetVADSegments(self, rwvss);
|
|
93
|
+
n_segments = whisper_vad_segments_n_segments(rwvss->segments);
|
|
94
|
+
|
|
95
|
+
return INT2NUM(n_segments);
|
|
96
|
+
}
|
|
97
|
+
|
|
98
|
+
void
|
|
99
|
+
init_ruby_whisper_vad_segments(VALUE *mVAD)
|
|
100
|
+
{
|
|
101
|
+
cVADSegments = rb_define_class_under(*mVAD, "Segments", rb_cObject);
|
|
102
|
+
rb_define_alloc_func(cVADSegments, ruby_whisper_vad_segments_s_allocate);
|
|
103
|
+
rb_define_method(cVADSegments, "each", ruby_whisper_vad_segments_each, 0);
|
|
104
|
+
rb_define_method(cVADSegments, "length", ruby_whisper_vad_segments_get_length, 0);
|
|
105
|
+
rb_include_module(cVADSegments, rb_path2class("Enumerable"));
|
|
106
|
+
}
|
data/ext/sources/CMakeLists.txt
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
cmake_minimum_required(VERSION 3.5) # for add_link_options and implicit target directories.
|
|
2
2
|
project("whisper.cpp" C CXX)
|
|
3
|
-
project("whisper.cpp" VERSION 1.
|
|
3
|
+
project("whisper.cpp" VERSION 1.8.3)
|
|
4
4
|
include(CheckIncludeFileCXX)
|
|
5
5
|
|
|
6
6
|
set(SOVERSION 1)
|
|
@@ -34,6 +34,9 @@ endif()
|
|
|
34
34
|
if (EMSCRIPTEN)
|
|
35
35
|
set(BUILD_SHARED_LIBS_DEFAULT OFF)
|
|
36
36
|
|
|
37
|
+
set(CMAKE_CXX_STANDARD 17)
|
|
38
|
+
set(CMAKE_CXX_STANDARD_REQUIRED ON)
|
|
39
|
+
|
|
37
40
|
option(WHISPER_WASM_SINGLE_FILE "whisper: embed WASM inside the generated whisper.js" ON)
|
|
38
41
|
|
|
39
42
|
# TODO: without these, we get the following error:
|
|
@@ -0,0 +1,16 @@
|
|
|
1
|
+
set( CMAKE_SYSTEM_NAME Darwin )
|
|
2
|
+
set( CMAKE_SYSTEM_PROCESSOR arm64 )
|
|
3
|
+
|
|
4
|
+
set( target arm64-apple-darwin-macho )
|
|
5
|
+
|
|
6
|
+
set( CMAKE_C_COMPILER clang )
|
|
7
|
+
set( CMAKE_CXX_COMPILER clang++ )
|
|
8
|
+
|
|
9
|
+
set( CMAKE_C_COMPILER_TARGET ${target} )
|
|
10
|
+
set( CMAKE_CXX_COMPILER_TARGET ${target} )
|
|
11
|
+
|
|
12
|
+
set( arch_c_flags "-march=armv8.4-a -fvectorize -ffp-model=fast -fno-finite-math-only" )
|
|
13
|
+
set( warn_c_flags "-Wno-format -Wno-unused-variable -Wno-unused-function" )
|
|
14
|
+
|
|
15
|
+
set( CMAKE_C_FLAGS_INIT "${arch_c_flags} ${warn_c_flags}" )
|
|
16
|
+
set( CMAKE_CXX_FLAGS_INIT "${arch_c_flags} ${warn_c_flags}" )
|
|
@@ -0,0 +1,16 @@
|
|
|
1
|
+
set( CMAKE_SYSTEM_NAME Windows )
|
|
2
|
+
set( CMAKE_SYSTEM_PROCESSOR arm64 )
|
|
3
|
+
|
|
4
|
+
set( target arm64-pc-windows-msvc )
|
|
5
|
+
|
|
6
|
+
set( CMAKE_C_COMPILER clang )
|
|
7
|
+
set( CMAKE_CXX_COMPILER clang++ )
|
|
8
|
+
|
|
9
|
+
set( CMAKE_C_COMPILER_TARGET ${target} )
|
|
10
|
+
set( CMAKE_CXX_COMPILER_TARGET ${target} )
|
|
11
|
+
|
|
12
|
+
set( arch_c_flags "-march=armv8.7-a -fvectorize -ffp-model=fast -fno-finite-math-only" )
|
|
13
|
+
set( warn_c_flags "-Wno-format -Wno-unused-variable -Wno-unused-function -Wno-gnu-zero-variadic-macro-arguments" )
|
|
14
|
+
|
|
15
|
+
set( CMAKE_C_FLAGS_INIT "${arch_c_flags} ${warn_c_flags}" )
|
|
16
|
+
set( CMAKE_CXX_FLAGS_INIT "${arch_c_flags} ${warn_c_flags}" )
|
|
@@ -0,0 +1,29 @@
|
|
|
1
|
+
set(CMAKE_SYSTEM_NAME Linux)
|
|
2
|
+
set(CMAKE_SYSTEM_PROCESSOR riscv64)
|
|
3
|
+
set(CMAKE_SYSTEM_VERSION 1)
|
|
4
|
+
|
|
5
|
+
if (CMAKE_HOST_SYSTEM_PROCESSOR MATCHES "^(riscv)")
|
|
6
|
+
message(STATUS "HOST SYSTEM ${CMAKE_HOST_SYSTEM_PROCESSOR}")
|
|
7
|
+
else()
|
|
8
|
+
set(GNU_MACHINE riscv64-unknown-linux-gnu CACHE STRING "GNU compiler triple")
|
|
9
|
+
if (DEFINED ENV{RISCV_ROOT_PATH})
|
|
10
|
+
file(TO_CMAKE_PATH $ENV{RISCV_ROOT_PATH} RISCV_ROOT_PATH)
|
|
11
|
+
else()
|
|
12
|
+
message(FATAL_ERROR "RISCV_ROOT_PATH env must be defined")
|
|
13
|
+
endif()
|
|
14
|
+
|
|
15
|
+
set(RISCV_ROOT_PATH ${RISCV_ROOT_PATH} CACHE STRING "root path to riscv toolchain")
|
|
16
|
+
set(CMAKE_C_COMPILER ${RISCV_ROOT_PATH}/bin/riscv64-unknown-linux-gnu-gcc)
|
|
17
|
+
set(CMAKE_CXX_COMPILER ${RISCV_ROOT_PATH}/bin/riscv64-unknown-linux-gnu-g++)
|
|
18
|
+
set(CMAKE_STRIP ${RISCV_ROOT_PATH}/bin/riscv64-unknown-linux-gnu-strip)
|
|
19
|
+
set(CMAKE_FIND_ROOT_PATH "${RISCV_ROOT_PATH}/riscv64-unknown-linux-gnu")
|
|
20
|
+
set(CMAKE_SYSROOT "${RISCV_ROOT_PATH}/sysroot")
|
|
21
|
+
endif()
|
|
22
|
+
|
|
23
|
+
set(CMAKE_FIND_ROOT_PATH_MODE_PROGRAM NEVER)
|
|
24
|
+
set(CMAKE_FIND_ROOT_PATH_MODE_LIBRARY ONLY)
|
|
25
|
+
set(CMAKE_FIND_ROOT_PATH_MODE_INCLUDE ONLY)
|
|
26
|
+
set(CMAKE_FIND_ROOT_PATH_MODE_PACKAGE ONLY)
|
|
27
|
+
set(CMAKE_C_FLAGS "-march=rv64gcv_zfh_zba_zicbop -mabi=lp64d ${CMAKE_C_FLAGS}")
|
|
28
|
+
set(CMAKE_CXX_FLAGS "-march=rv64gcv_zfh_zba_zicbop -mabi=lp64d ${CXX_FLAGS}")
|
|
29
|
+
set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -latomic")
|
|
@@ -207,7 +207,7 @@ class ProgressWorker : public Napi::AsyncWorker {
|
|
|
207
207
|
auto callback = [progress](Napi::Env env, Napi::Function jsCallback) {
|
|
208
208
|
jsCallback.Call({Napi::Number::New(env, progress)});
|
|
209
209
|
};
|
|
210
|
-
|
|
210
|
+
|
|
211
211
|
tsfn.BlockingCall(callback);
|
|
212
212
|
}
|
|
213
213
|
}
|
|
@@ -396,59 +396,59 @@ Napi::Value whisper(const Napi::CallbackInfo& info) {
|
|
|
396
396
|
std::string language = whisper_params.Get("language").As<Napi::String>();
|
|
397
397
|
std::string model = whisper_params.Get("model").As<Napi::String>();
|
|
398
398
|
std::string input = whisper_params.Get("fname_inp").As<Napi::String>();
|
|
399
|
-
|
|
399
|
+
|
|
400
400
|
bool use_gpu = true;
|
|
401
401
|
if (whisper_params.Has("use_gpu") && whisper_params.Get("use_gpu").IsBoolean()) {
|
|
402
402
|
use_gpu = whisper_params.Get("use_gpu").As<Napi::Boolean>();
|
|
403
403
|
}
|
|
404
|
-
|
|
404
|
+
|
|
405
405
|
bool flash_attn = false;
|
|
406
406
|
if (whisper_params.Has("flash_attn") && whisper_params.Get("flash_attn").IsBoolean()) {
|
|
407
407
|
flash_attn = whisper_params.Get("flash_attn").As<Napi::Boolean>();
|
|
408
408
|
}
|
|
409
|
-
|
|
409
|
+
|
|
410
410
|
bool no_prints = false;
|
|
411
411
|
if (whisper_params.Has("no_prints") && whisper_params.Get("no_prints").IsBoolean()) {
|
|
412
412
|
no_prints = whisper_params.Get("no_prints").As<Napi::Boolean>();
|
|
413
413
|
}
|
|
414
|
-
|
|
414
|
+
|
|
415
415
|
bool no_timestamps = false;
|
|
416
416
|
if (whisper_params.Has("no_timestamps") && whisper_params.Get("no_timestamps").IsBoolean()) {
|
|
417
417
|
no_timestamps = whisper_params.Get("no_timestamps").As<Napi::Boolean>();
|
|
418
418
|
}
|
|
419
|
-
|
|
419
|
+
|
|
420
420
|
bool detect_language = false;
|
|
421
421
|
if (whisper_params.Has("detect_language") && whisper_params.Get("detect_language").IsBoolean()) {
|
|
422
422
|
detect_language = whisper_params.Get("detect_language").As<Napi::Boolean>();
|
|
423
423
|
}
|
|
424
|
-
|
|
424
|
+
|
|
425
425
|
int32_t audio_ctx = 0;
|
|
426
426
|
if (whisper_params.Has("audio_ctx") && whisper_params.Get("audio_ctx").IsNumber()) {
|
|
427
427
|
audio_ctx = whisper_params.Get("audio_ctx").As<Napi::Number>();
|
|
428
428
|
}
|
|
429
|
-
|
|
429
|
+
|
|
430
430
|
bool comma_in_time = true;
|
|
431
431
|
if (whisper_params.Has("comma_in_time") && whisper_params.Get("comma_in_time").IsBoolean()) {
|
|
432
432
|
comma_in_time = whisper_params.Get("comma_in_time").As<Napi::Boolean>();
|
|
433
433
|
}
|
|
434
|
-
|
|
434
|
+
|
|
435
435
|
int32_t max_len = 0;
|
|
436
436
|
if (whisper_params.Has("max_len") && whisper_params.Get("max_len").IsNumber()) {
|
|
437
437
|
max_len = whisper_params.Get("max_len").As<Napi::Number>();
|
|
438
438
|
}
|
|
439
|
-
|
|
439
|
+
|
|
440
440
|
// Add support for max_context
|
|
441
441
|
int32_t max_context = -1;
|
|
442
442
|
if (whisper_params.Has("max_context") && whisper_params.Get("max_context").IsNumber()) {
|
|
443
443
|
max_context = whisper_params.Get("max_context").As<Napi::Number>();
|
|
444
444
|
}
|
|
445
|
-
|
|
445
|
+
|
|
446
446
|
// support prompt
|
|
447
447
|
std::string prompt = "";
|
|
448
448
|
if (whisper_params.Has("prompt") && whisper_params.Get("prompt").IsString()) {
|
|
449
449
|
prompt = whisper_params.Get("prompt").As<Napi::String>();
|
|
450
450
|
}
|
|
451
|
-
|
|
451
|
+
|
|
452
452
|
// Add support for print_progress
|
|
453
453
|
bool print_progress = false;
|
|
454
454
|
if (whisper_params.Has("print_progress") && whisper_params.Get("print_progress").IsBoolean()) {
|
|
@@ -465,37 +465,37 @@ Napi::Value whisper(const Napi::CallbackInfo& info) {
|
|
|
465
465
|
if (whisper_params.Has("vad") && whisper_params.Get("vad").IsBoolean()) {
|
|
466
466
|
vad = whisper_params.Get("vad").As<Napi::Boolean>();
|
|
467
467
|
}
|
|
468
|
-
|
|
468
|
+
|
|
469
469
|
std::string vad_model = "";
|
|
470
470
|
if (whisper_params.Has("vad_model") && whisper_params.Get("vad_model").IsString()) {
|
|
471
471
|
vad_model = whisper_params.Get("vad_model").As<Napi::String>();
|
|
472
472
|
}
|
|
473
|
-
|
|
473
|
+
|
|
474
474
|
float vad_threshold = 0.5f;
|
|
475
475
|
if (whisper_params.Has("vad_threshold") && whisper_params.Get("vad_threshold").IsNumber()) {
|
|
476
476
|
vad_threshold = whisper_params.Get("vad_threshold").As<Napi::Number>();
|
|
477
477
|
}
|
|
478
|
-
|
|
478
|
+
|
|
479
479
|
int vad_min_speech_duration_ms = 250;
|
|
480
480
|
if (whisper_params.Has("vad_min_speech_duration_ms") && whisper_params.Get("vad_min_speech_duration_ms").IsNumber()) {
|
|
481
481
|
vad_min_speech_duration_ms = whisper_params.Get("vad_min_speech_duration_ms").As<Napi::Number>();
|
|
482
482
|
}
|
|
483
|
-
|
|
483
|
+
|
|
484
484
|
int vad_min_silence_duration_ms = 100;
|
|
485
485
|
if (whisper_params.Has("vad_min_silence_duration_ms") && whisper_params.Get("vad_min_silence_duration_ms").IsNumber()) {
|
|
486
486
|
vad_min_silence_duration_ms = whisper_params.Get("vad_min_silence_duration_ms").As<Napi::Number>();
|
|
487
487
|
}
|
|
488
|
-
|
|
488
|
+
|
|
489
489
|
float vad_max_speech_duration_s = FLT_MAX;
|
|
490
490
|
if (whisper_params.Has("vad_max_speech_duration_s") && whisper_params.Get("vad_max_speech_duration_s").IsNumber()) {
|
|
491
491
|
vad_max_speech_duration_s = whisper_params.Get("vad_max_speech_duration_s").As<Napi::Number>();
|
|
492
492
|
}
|
|
493
|
-
|
|
493
|
+
|
|
494
494
|
int vad_speech_pad_ms = 30;
|
|
495
495
|
if (whisper_params.Has("vad_speech_pad_ms") && whisper_params.Get("vad_speech_pad_ms").IsNumber()) {
|
|
496
496
|
vad_speech_pad_ms = whisper_params.Get("vad_speech_pad_ms").As<Napi::Number>();
|
|
497
497
|
}
|
|
498
|
-
|
|
498
|
+
|
|
499
499
|
float vad_samples_overlap = 0.1f;
|
|
500
500
|
if (whisper_params.Has("vad_samples_overlap") && whisper_params.Get("vad_samples_overlap").IsNumber()) {
|
|
501
501
|
vad_samples_overlap = whisper_params.Get("vad_samples_overlap").As<Napi::Number>();
|
|
@@ -1,8 +1,10 @@
|
|
|
1
|
-
const path = require(
|
|
2
|
-
const
|
|
3
|
-
|
|
4
|
-
|
|
5
|
-
|
|
1
|
+
const path = require('path');
|
|
2
|
+
const os = require('os');
|
|
3
|
+
|
|
4
|
+
const isWindows = os.platform() === 'win32';
|
|
5
|
+
const buildPath = isWindows ? "../../build/bin/Release/addon.node" : "../../build/Release/addon.node";
|
|
6
|
+
|
|
7
|
+
const { whisper } = require(path.join(__dirname, buildPath));
|
|
6
8
|
const { promisify } = require("util");
|
|
7
9
|
|
|
8
10
|
const whisperAsync = promisify(whisper);
|
|
@@ -23,7 +23,7 @@ const vadParams = {
|
|
|
23
23
|
max_len: 0,
|
|
24
24
|
// VAD parameters
|
|
25
25
|
vad: true,
|
|
26
|
-
vad_model: path.join(__dirname, "../../models/ggml-silero-
|
|
26
|
+
vad_model: path.join(__dirname, "../../models/ggml-silero-v6.2.0.bin"), // You need to download this model
|
|
27
27
|
vad_threshold: 0.5,
|
|
28
28
|
vad_min_speech_duration_ms: 250,
|
|
29
29
|
vad_min_silence_duration_ms: 100,
|
|
@@ -63,7 +63,7 @@ async function runVADExample() {
|
|
|
63
63
|
const fs = require('fs');
|
|
64
64
|
if (!fs.existsSync(vadParams.vad_model)) {
|
|
65
65
|
console.log("⚠️ VAD model not found. Please download the VAD model first:");
|
|
66
|
-
console.log(" ./models/download-vad-model.sh silero-
|
|
66
|
+
console.log(" ./models/download-vad-model.sh silero-v6.2.0");
|
|
67
67
|
console.log(" Or run: python models/convert-silero-vad-to-ggml.py");
|
|
68
68
|
console.log("\n Falling back to traditional transcription without VAD...\n");
|
|
69
69
|
|