@novastera-oss/llamarn 0.4.1 → 0.4.3-beta4
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.
- package/RNLlamaCpp.podspec +3 -0
- package/android/CMakeLists.txt +2 -0
- package/android/src/main/cpp/include/llama.h +44 -21
- package/android/src/main/jniLibs/arm64-v8a/libggml-base.so +0 -0
- package/android/src/main/jniLibs/arm64-v8a/libggml-cpu.so +0 -0
- package/android/src/main/jniLibs/arm64-v8a/libggml.so +0 -0
- package/android/src/main/jniLibs/arm64-v8a/libllama.so +0 -0
- package/android/src/main/jniLibs/armeabi-v7a/libggml-base.so +0 -0
- package/android/src/main/jniLibs/armeabi-v7a/libggml-cpu.so +0 -0
- package/android/src/main/jniLibs/armeabi-v7a/libggml.so +0 -0
- package/android/src/main/jniLibs/armeabi-v7a/libllama.so +0 -0
- package/android/src/main/jniLibs/x86/libggml-base.so +0 -0
- package/android/src/main/jniLibs/x86/libggml-cpu.so +0 -0
- package/android/src/main/jniLibs/x86/libggml.so +0 -0
- package/android/src/main/jniLibs/x86/libllama.so +0 -0
- package/android/src/main/jniLibs/x86_64/libggml-base.so +0 -0
- package/android/src/main/jniLibs/x86_64/libggml-cpu.so +0 -0
- package/android/src/main/jniLibs/x86_64/libggml.so +0 -0
- package/android/src/main/jniLibs/x86_64/libllama.so +0 -0
- package/cpp/build-info.cpp +2 -2
- package/cpp/llama.cpp/CMakeLists.txt +12 -0
- package/cpp/llama.cpp/CODEOWNERS +116 -10
- package/cpp/llama.cpp/CONTRIBUTING.md +30 -3
- package/cpp/llama.cpp/README.md +13 -5
- package/cpp/llama.cpp/build-xcframework.sh +5 -0
- package/cpp/llama.cpp/cmake/riscv64-spacemit-linux-gnu-gcc.cmake +29 -0
- package/cpp/llama.cpp/common/CMakeLists.txt +12 -2
- package/cpp/llama.cpp/common/arg.cpp +303 -795
- package/cpp/llama.cpp/common/arg.h +2 -3
- package/cpp/llama.cpp/common/chat-parser-xml-toolcall.cpp +861 -0
- package/cpp/llama.cpp/common/chat-parser-xml-toolcall.h +45 -0
- package/cpp/llama.cpp/common/chat-parser.cpp +156 -15
- package/cpp/llama.cpp/common/chat-parser.h +13 -0
- package/cpp/llama.cpp/common/chat.cpp +1147 -88
- package/cpp/llama.cpp/common/chat.h +16 -3
- package/cpp/llama.cpp/common/common.cpp +70 -15
- package/cpp/llama.cpp/common/common.h +57 -19
- package/cpp/llama.cpp/common/download.cpp +1072 -0
- package/cpp/llama.cpp/common/download.h +55 -0
- package/cpp/llama.cpp/common/http.h +73 -0
- package/cpp/llama.cpp/common/json-partial.cpp +70 -2
- package/cpp/llama.cpp/common/json-schema-to-grammar.cpp +61 -22
- package/cpp/llama.cpp/common/json-schema-to-grammar.h +2 -0
- package/cpp/llama.cpp/common/log.cpp +59 -2
- package/cpp/llama.cpp/common/log.h +12 -4
- package/cpp/llama.cpp/common/sampling.cpp +84 -8
- package/cpp/llama.cpp/common/sampling.h +3 -1
- package/cpp/llama.cpp/common/speculative.cpp +1 -1
- package/cpp/llama.cpp/convert_hf_to_gguf.py +1608 -233
- package/cpp/llama.cpp/convert_hf_to_gguf_update.py +6 -1
- package/cpp/llama.cpp/convert_lora_to_gguf.py +37 -5
- package/cpp/llama.cpp/ggml/CMakeLists.txt +47 -28
- package/cpp/llama.cpp/ggml/include/ggml-backend.h +19 -1
- package/cpp/llama.cpp/ggml/include/ggml-cpu.h +1 -1
- package/cpp/llama.cpp/ggml/include/ggml-hexagon.h +19 -0
- package/cpp/llama.cpp/ggml/include/ggml-metal.h +1 -6
- package/cpp/llama.cpp/ggml/include/ggml-rpc.h +7 -9
- package/cpp/llama.cpp/ggml/include/ggml-zdnn.h +2 -1
- package/cpp/llama.cpp/ggml/include/ggml.h +199 -6
- package/cpp/llama.cpp/ggml/src/CMakeLists.txt +38 -0
- package/cpp/llama.cpp/ggml/src/ggml-alloc.c +299 -130
- package/cpp/llama.cpp/ggml/src/ggml-backend-impl.h +4 -4
- package/cpp/llama.cpp/ggml/src/ggml-backend-reg.cpp +21 -5
- package/cpp/llama.cpp/ggml/src/ggml-backend.cpp +99 -2
- package/cpp/llama.cpp/ggml/src/ggml-blas/CMakeLists.txt +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-blas/ggml-blas.cpp +1 -0
- package/cpp/llama.cpp/ggml/src/ggml-cann/acl_tensor.cpp +57 -45
- package/cpp/llama.cpp/ggml/src/ggml-cann/acl_tensor.h +138 -47
- package/cpp/llama.cpp/ggml/src/ggml-cann/aclnn_ops.cpp +1584 -1773
- package/cpp/llama.cpp/ggml/src/ggml-cann/aclnn_ops.h +201 -317
- package/cpp/llama.cpp/ggml/src/ggml-cann/common.h +146 -187
- package/cpp/llama.cpp/ggml/src/ggml-cann/ggml-cann.cpp +771 -713
- package/cpp/llama.cpp/ggml/src/ggml-cpu/CMakeLists.txt +135 -77
- package/cpp/llama.cpp/ggml/src/ggml-cpu/amx/amx.cpp +5 -2
- package/cpp/llama.cpp/ggml/src/ggml-cpu/arch/arm/quants.c +428 -26
- package/cpp/llama.cpp/ggml/src/ggml-cpu/arch/loongarch/quants.c +16 -17
- package/cpp/llama.cpp/ggml/src/ggml-cpu/arch/riscv/quants.c +318 -145
- package/cpp/llama.cpp/ggml/src/ggml-cpu/arch/s390/cpu-feats.cpp +50 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/arch/s390/quants.c +155 -60
- package/cpp/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp +8 -8
- package/cpp/llama.cpp/ggml/src/ggml-cpu/arch-fallback.h +0 -1
- package/cpp/llama.cpp/ggml/src/ggml-cpu/common.h +14 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/ggml-cpu-impl.h +10 -9
- package/cpp/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.c +108 -64
- package/cpp/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.cpp +14 -4
- package/cpp/llama.cpp/ggml/src/ggml-cpu/kleidiai/kernels.cpp +530 -87
- package/cpp/llama.cpp/ggml/src/ggml-cpu/kleidiai/kernels.h +37 -45
- package/cpp/llama.cpp/ggml/src/ggml-cpu/kleidiai/kleidiai.cpp +349 -127
- package/cpp/llama.cpp/ggml/src/ggml-cpu/ops.cpp +947 -1218
- package/cpp/llama.cpp/ggml/src/ggml-cpu/ops.h +5 -4
- package/cpp/llama.cpp/ggml/src/ggml-cpu/repack.cpp +143 -29
- package/cpp/llama.cpp/ggml/src/ggml-cpu/simd-mappings.h +82 -76
- package/cpp/llama.cpp/ggml/src/ggml-cpu/spacemit/ime.cpp +1025 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/spacemit/ime.h +13 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/spacemit/ime1_kernels.cpp +3196 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/spacemit/ime_kernels.h +26 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/unary-ops.cpp +151 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/unary-ops.h +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/vec.cpp +233 -28
- package/cpp/llama.cpp/ggml/src/ggml-cpu/vec.h +326 -66
- package/cpp/llama.cpp/ggml/src/ggml-cuda/CMakeLists.txt +12 -3
- package/cpp/llama.cpp/ggml/src/ggml-cuda/argsort.cu +102 -6
- package/cpp/llama.cpp/ggml/src/ggml-cuda/binbcast.cu +110 -76
- package/cpp/llama.cpp/ggml/src/ggml-cuda/common.cuh +167 -38
- package/cpp/llama.cpp/ggml/src/ggml-cuda/conv2d.cu +6 -11
- package/cpp/llama.cpp/ggml/src/ggml-cuda/convert.cuh +12 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/cpy-utils.cuh +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-cuda/cpy.cu +245 -151
- package/cpp/llama.cpp/ggml/src/ggml-cuda/cpy.cuh +1 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-common.cuh +341 -289
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-tile.cu +49 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-tile.cuh +1233 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-vec.cuh +586 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-wmma-f16.cu +6 -6
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-wmma-f16.cuh +48 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn.cu +123 -220
- package/cpp/llama.cpp/ggml/src/ggml-cuda/getrows.cu +41 -39
- package/cpp/llama.cpp/ggml/src/ggml-cuda/ggml-cuda.cu +715 -45
- package/cpp/llama.cpp/ggml/src/ggml-cuda/im2col.cu +150 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/im2col.cuh +1 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/mma.cuh +321 -24
- package/cpp/llama.cpp/ggml/src/ggml-cuda/mmf.cu +93 -351
- package/cpp/llama.cpp/ggml/src/ggml-cuda/mmf.cuh +828 -1
- package/cpp/llama.cpp/ggml/src/ggml-cuda/mmid.cu +164 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/mmid.cuh +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/mmq.cu +3 -166
- package/cpp/llama.cpp/ggml/src/ggml-cuda/mmq.cuh +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-cuda/mmvf.cu +371 -78
- package/cpp/llama.cpp/ggml/src/ggml-cuda/mmvf.cuh +3 -2
- package/cpp/llama.cpp/ggml/src/ggml-cuda/mmvq.cu +279 -147
- package/cpp/llama.cpp/ggml/src/ggml-cuda/mmvq.cuh +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-cuda/norm.cu +97 -85
- package/cpp/llama.cpp/ggml/src/ggml-cuda/pad.cu +46 -23
- package/cpp/llama.cpp/ggml/src/ggml-cuda/pad_reflect_1d.cu +63 -54
- package/cpp/llama.cpp/ggml/src/ggml-cuda/quantize.cu +12 -10
- package/cpp/llama.cpp/ggml/src/ggml-cuda/rope.cu +192 -77
- package/cpp/llama.cpp/ggml/src/ggml-cuda/rope.cuh +2 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/scale.cu +10 -9
- package/cpp/llama.cpp/ggml/src/ggml-cuda/set-rows.cu +137 -75
- package/cpp/llama.cpp/ggml/src/ggml-cuda/set.cu +39 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/set.cuh +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq112-dv112.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq128-dv128.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq256-dv256.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq40-dv40.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq576-dv512.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq64-dv64.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq72-dv72.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq80-dv80.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq96-dv96.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-f16.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q4_0.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q4_1.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q5_0.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q5_1.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q8_0.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-f16.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q4_0.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q4_1.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q5_0.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q5_1.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q8_0.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-f16.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q4_0.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q4_1.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q5_0.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q5_1.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q8_0.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-f16.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q4_0.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q4_1.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q5_0.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q5_1.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q8_0.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-f16.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q4_0.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q4_1.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q5_0.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q5_1.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q8_0.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-f16.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q4_0.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q4_1.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q5_0.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q5_1.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q8_0.cu +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/generate_cu_files.py +40 -19
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_10.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_11.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_12.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_13.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_14.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_15.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_16.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_2.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_3.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_4.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_5.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_6.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_7.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_8.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_9.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/topk-moe.cu +336 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/topk-moe.cuh +16 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/tsembd.cu +3 -3
- package/cpp/llama.cpp/ggml/src/ggml-cuda/unary.cu +105 -11
- package/cpp/llama.cpp/ggml/src/ggml-cuda/unary.cuh +36 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/upscale.cu +87 -6
- package/cpp/llama.cpp/ggml/src/ggml-cuda/vendors/hip.h +28 -12
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/CMakeLists.txt +68 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/ggml-hexagon.cpp +3807 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/CMakeLists.txt +40 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/act-ops.c +442 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/binary-ops.c +360 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/cmake-toolchain.cmake +157 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/htp-ctx.h +40 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/htp-dma.c +69 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/htp-dma.h +119 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/htp-msg.h +156 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/htp-ops.h +64 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/htp_iface.idl +16 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/hvx-exp.c +93 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/hvx-inverse.c +60 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/hvx-sigmoid.c +49 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/hvx-utils.c +960 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/hvx-utils.h +1032 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/main.c +829 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/matmul-ops.c +2223 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/ops-utils.h +149 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/rope-ops.c +418 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/softmax-ops.c +402 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/unary-ops.c +255 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/worker-pool.c +297 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/worker-pool.h +57 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp-utils.c +448 -0
- package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp-utils.h +220 -0
- package/cpp/llama.cpp/ggml/src/ggml-hip/CMakeLists.txt +8 -13
- package/cpp/llama.cpp/ggml/src/ggml-impl.h +110 -12
- package/cpp/llama.cpp/ggml/src/ggml-metal/CMakeLists.txt +6 -5
- package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal-common.cpp +446 -0
- package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal-common.h +52 -0
- package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal-context.h +33 -0
- package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal-context.m +599 -0
- package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal-device.cpp +1662 -0
- package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal-device.h +251 -0
- package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal-device.m +1527 -0
- package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal-impl.h +244 -39
- package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal-ops.cpp +3844 -0
- package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal-ops.h +90 -0
- package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal.cpp +723 -0
- package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal.metal +3453 -1907
- package/cpp/llama.cpp/ggml/src/ggml-musa/CMakeLists.txt +3 -1
- package/cpp/llama.cpp/ggml/src/ggml-opencl/CMakeLists.txt +10 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/ggml-opencl.cpp +1331 -109
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/cvt.cl +126 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/flash_attn_f16.cl +31 -4
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/flash_attn_f32.cl +35 -7
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/flash_attn_f32_f16.cl +31 -4
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/gemm_moe_mxfp4_f32.cl +162 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/gemv_moe_mxfp4_f32.cl +156 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/get_rows.cl +36 -12
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mm_f16_f32_kq_kqv.cl +273 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mm_f16_f32_l4_lm.cl +24 -10
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mm_f32_f32_l4_lm.cl +24 -10
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mm_q8_0_f32_l4_lm.cl +154 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mv_id_mxfp4_f32_flat.cl +176 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mv_id_q8_0_f32.cl +140 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mv_id_q8_0_f32_flat.cl +222 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mv_mxfp4_f32_flat.cl +167 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mv_q8_0_f32.cl +125 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mv_q8_0_f32_flat.cl +202 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/pad.cl +29 -20
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/rms_norm.cl +25 -10
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/rope.cl +50 -24
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/set_rows.cl +123 -10
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/tsembd.cl +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-quants.c +1 -0
- package/cpp/llama.cpp/ggml/src/ggml-rpc/ggml-rpc.cpp +341 -161
- package/cpp/llama.cpp/ggml/src/ggml-sycl/backend.hpp +6 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/binbcast.cpp +6 -5
- package/cpp/llama.cpp/ggml/src/ggml-sycl/common.hpp +74 -15
- package/cpp/llama.cpp/ggml/src/ggml-sycl/concat.cpp +50 -30
- package/cpp/llama.cpp/ggml/src/ggml-sycl/conv.cpp +10 -4
- package/cpp/llama.cpp/ggml/src/ggml-sycl/convert.cpp +166 -99
- package/cpp/llama.cpp/ggml/src/ggml-sycl/count-equal.cpp +79 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/count-equal.hpp +9 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/cpy.cpp +72 -94
- package/cpp/llama.cpp/ggml/src/ggml-sycl/dmmv.cpp +67 -49
- package/cpp/llama.cpp/ggml/src/ggml-sycl/dpct/helper.hpp +21 -31
- package/cpp/llama.cpp/ggml/src/ggml-sycl/element_wise.cpp +252 -316
- package/cpp/llama.cpp/ggml/src/ggml-sycl/element_wise.hpp +6 -2
- package/cpp/llama.cpp/ggml/src/ggml-sycl/getrows.cpp +9 -6
- package/cpp/llama.cpp/ggml/src/ggml-sycl/ggml-sycl.cpp +359 -142
- package/cpp/llama.cpp/ggml/src/ggml-sycl/gla.cpp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-sycl/im2col.cpp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-sycl/mmq.cpp +80 -60
- package/cpp/llama.cpp/ggml/src/ggml-sycl/mmvq.cpp +201 -132
- package/cpp/llama.cpp/ggml/src/ggml-sycl/norm.cpp +230 -55
- package/cpp/llama.cpp/ggml/src/ggml-sycl/norm.hpp +2 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/pad.cpp +97 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/pad.hpp +24 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/pad_reflect_1d.cpp +72 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/pad_reflect_1d.hpp +8 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/presets.hpp +2 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/repeat_back.cpp +76 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/repeat_back.hpp +8 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/roll.cpp +122 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/roll.hpp +20 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/rope.cpp +50 -41
- package/cpp/llama.cpp/ggml/src/ggml-sycl/set.cpp +73 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/set.hpp +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/set_rows.cpp +45 -36
- package/cpp/llama.cpp/ggml/src/ggml-sycl/softmax.cpp +330 -165
- package/cpp/llama.cpp/ggml/src/ggml-sycl/softmax.hpp +4 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/ssm_conv.cpp +127 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/ssm_conv.hpp +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/tsembd.cpp +12 -6
- package/cpp/llama.cpp/ggml/src/ggml-sycl/wkv.cpp +16 -12
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/CMakeLists.txt +38 -18
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/ggml-vulkan.cpp +4184 -2159
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/abs.comp +21 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/acc.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/add.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/add1.comp +28 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/add_id.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/arange.comp +20 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/argmax.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/argsort.comp +33 -26
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/argsort_large.comp +114 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/ceil.comp +22 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/clamp.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/concat.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/contig_copy.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/conv2d_dw.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/conv2d_mm.comp +53 -30
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/conv_transpose_1d.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/copy.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/copy_from_quant.comp +3 -3
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/copy_to_quant.comp +13 -6
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/copy_transpose.comp +67 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/cos.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/count_equal.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_f32.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{dequant_funcs.comp → dequant_funcs.glsl} +138 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{dequant_funcs_cm2.comp → dequant_funcs_cm2.glsl} +18 -4
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{dequant_head.comp → dequant_head.glsl} +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq1_m.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq1_s.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq2_s.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq2_xs.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq2_xxs.comp +3 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq3_s.comp +7 -6
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq3_xxs.comp +5 -3
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq4_nl.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq4_xs.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_mxfp4.comp +3 -3
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q2_k.comp +3 -3
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q3_k.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q4_0.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q4_1.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q4_k.comp +3 -3
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q5_0.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q5_1.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q5_k.comp +3 -3
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q6_k.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q8_0.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/diag_mask_inf.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/div.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/exp.comp +3 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/fill.comp +19 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn.comp +52 -14
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{flash_attn_base.comp → flash_attn_base.glsl} +50 -12
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm1.comp +61 -12
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm2.comp +54 -12
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_split_k_reduce.comp +5 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/floor.comp +22 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/geglu.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/geglu_erf.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/geglu_quick.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/gelu.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/gelu_erf.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/gelu_quick.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{generic_binary_head.comp → generic_binary_head.glsl} +10 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/get_rows.comp +21 -12
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/get_rows_quant.comp +28 -18
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{glu_head.comp → glu_head.glsl} +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/group_norm.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/hardsigmoid.comp +22 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/hardswish.comp +22 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/im2col.comp +15 -7
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/im2col_3d.comp +125 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/l2_norm.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/leaky_relu.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/log.comp +18 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_base.glsl +229 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iface.glsl +33 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq1_m.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq1_s.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq2_s.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq2_xs.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq2_xxs.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq3_s.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq3_xxs.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_nc.comp +9 -7
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_p021.comp +9 -7
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q2_k.comp +3 -5
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q3_k.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q4_k.comp +3 -5
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q5_k.comp +3 -5
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q6_k.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vecq.comp +140 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm.comp +106 -634
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm_cm2.comp +118 -9
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm_funcs.glsl +556 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm_id_funcs.glsl +70 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq.comp +77 -214
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq_funcs.glsl +589 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq_shmem_types.glsl +78 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/multi_add.comp +97 -13
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/neg.comp +20 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/norm.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/opt_step_adamw.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/opt_step_sgd.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/pad.comp +25 -4
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/pool2d.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/quantize_q8_1.comp +55 -5
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/reglu.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/relu.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/repeat.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/repeat_back.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rms_norm.comp +45 -3
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rms_norm_back.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rms_norm_partials.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/roll.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rope_funcs.glsl +227 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rope_head.glsl +20 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rope_multi.comp +5 -52
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rope_neox.comp +5 -35
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rope_norm.comp +5 -35
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rope_params.glsl +27 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rope_vision.comp +5 -41
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/round.comp +29 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/scale.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/sigmoid.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/silu.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/silu_back.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/sin.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/soft_max.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/soft_max_back.comp +6 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/softplus.comp +23 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/sqrt.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/square.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/ssm_conv.comp +44 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/ssm_scan.comp +140 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/step.comp +22 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/sub.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/sum_rows.comp +1 -1
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/swiglu.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/swiglu_oai.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/tanh.comp +2 -2
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/timestep_embedding.comp +5 -4
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/topk_moe.comp +171 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/trunc.comp +22 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{types.comp → types.glsl} +79 -29
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/upscale.comp +36 -12
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp +471 -196
- package/cpp/llama.cpp/ggml/src/ggml-webgpu/CMakeLists.txt +8 -0
- package/cpp/llama.cpp/ggml/src/ggml-webgpu/ggml-webgpu.cpp +1690 -383
- package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/bin_op.tmpl.wgsl +188 -0
- package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/binary_head.tmpl +45 -0
- package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/common_decls.tmpl +930 -0
- package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/cpy.tmpl.wgsl +101 -0
- package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/embed_wgsl.py +57 -10
- package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/get_rows.tmpl.wgsl +874 -0
- package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/glu.tmpl.wgsl +323 -0
- package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat.tmpl.wgsl +25 -912
- package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_decls.tmpl +97 -0
- package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_reg_tile.tmpl.wgsl +247 -0
- package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_subgroup_matrix.tmpl.wgsl +302 -0
- package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_vec.tmpl.wgsl +267 -0
- package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/rms_norm.wgsl +123 -0
- package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/rope.tmpl.wgsl +295 -0
- package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/scale.tmpl.wgsl +90 -0
- package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/{set_rows.wgsl → set_rows.tmpl.wgsl} +38 -8
- package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/soft_max.tmpl.wgsl +345 -0
- package/cpp/llama.cpp/ggml/src/ggml-zdnn/common.hpp +59 -0
- package/cpp/llama.cpp/ggml/src/ggml-zdnn/ggml-zdnn.cpp +96 -314
- package/cpp/llama.cpp/ggml/src/ggml-zdnn/mmf.cpp +80 -0
- package/cpp/llama.cpp/ggml/src/ggml-zdnn/mmf.hpp +12 -0
- package/cpp/llama.cpp/ggml/src/ggml-zdnn/utils.cpp +79 -0
- package/cpp/llama.cpp/ggml/src/ggml-zdnn/utils.hpp +19 -0
- package/cpp/llama.cpp/ggml/src/ggml.c +440 -17
- package/cpp/llama.cpp/ggml/src/gguf.cpp +104 -29
- package/cpp/llama.cpp/gguf-py/gguf/constants.py +363 -13
- package/cpp/llama.cpp/gguf-py/gguf/gguf_writer.py +64 -0
- package/cpp/llama.cpp/gguf-py/gguf/lazy.py +8 -3
- package/cpp/llama.cpp/gguf-py/gguf/scripts/gguf_convert_endian.py +6 -0
- package/cpp/llama.cpp/gguf-py/gguf/tensor_mapping.py +156 -18
- package/cpp/llama.cpp/gguf-py/gguf/utility.py +80 -0
- package/cpp/llama.cpp/gguf-py/gguf/vocab.py +4 -4
- package/cpp/llama.cpp/include/llama.h +44 -21
- package/cpp/llama.cpp/media/llama1-icon-transparent.png +0 -0
- package/cpp/llama.cpp/media/llama1-icon-transparent.svg +77 -0
- package/cpp/llama.cpp/media/llama1-icon.png +0 -0
- package/cpp/llama.cpp/media/llama1-icon.svg +87 -0
- package/cpp/llama.cpp/requirements/requirements-all.txt +2 -0
- package/cpp/llama.cpp/requirements/requirements-convert_hf_to_gguf.txt +3 -3
- package/cpp/llama.cpp/requirements/requirements-convert_legacy_llama.txt +3 -1
- package/cpp/llama.cpp/requirements/requirements-tool_bench.txt +1 -1
- package/cpp/llama.cpp/src/CMakeLists.txt +101 -0
- package/cpp/llama.cpp/src/llama-adapter.cpp +33 -0
- package/cpp/llama.cpp/src/llama-adapter.h +3 -0
- package/cpp/llama.cpp/src/llama-arch.cpp +344 -14
- package/cpp/llama.cpp/src/llama-arch.h +50 -0
- package/cpp/llama.cpp/src/llama-batch.cpp +63 -31
- package/cpp/llama.cpp/src/llama-batch.h +13 -2
- package/cpp/llama.cpp/src/llama-chat.cpp +85 -3
- package/cpp/llama.cpp/src/llama-chat.h +4 -0
- package/cpp/llama.cpp/src/llama-context.cpp +300 -45
- package/cpp/llama.cpp/src/llama-context.h +16 -6
- package/cpp/llama.cpp/src/llama-cparams.h +2 -1
- package/cpp/llama.cpp/src/llama-grammar.cpp +17 -9
- package/cpp/llama.cpp/src/llama-graph.cpp +226 -64
- package/cpp/llama.cpp/src/llama-graph.h +27 -5
- package/cpp/llama.cpp/src/llama-hparams.cpp +53 -2
- package/cpp/llama.cpp/src/llama-hparams.h +48 -8
- package/cpp/llama.cpp/src/llama-impl.cpp +3 -3
- package/cpp/llama.cpp/src/llama-impl.h +2 -0
- package/cpp/llama.cpp/src/llama-kv-cache-iswa.cpp +13 -3
- package/cpp/llama.cpp/src/llama-kv-cache-iswa.h +2 -0
- package/cpp/llama.cpp/src/llama-kv-cache.cpp +120 -62
- package/cpp/llama.cpp/src/llama-kv-cache.h +13 -4
- package/cpp/llama.cpp/src/llama-kv-cells.h +44 -2
- package/cpp/llama.cpp/src/llama-memory-hybrid.cpp +19 -9
- package/cpp/llama.cpp/src/llama-memory-hybrid.h +2 -0
- package/cpp/llama.cpp/src/llama-memory-recurrent.cpp +38 -17
- package/cpp/llama.cpp/src/llama-memory-recurrent.h +5 -2
- package/cpp/llama.cpp/src/llama-memory.h +3 -0
- package/cpp/llama.cpp/src/llama-model-loader.cpp +2 -0
- package/cpp/llama.cpp/src/llama-model.cpp +1070 -12614
- package/cpp/llama.cpp/src/llama-model.h +40 -4
- package/cpp/llama.cpp/src/llama-quant.cpp +14 -6
- package/cpp/llama.cpp/src/llama-sampling.cpp +243 -136
- package/cpp/llama.cpp/src/llama-vocab.cpp +43 -3
- package/cpp/llama.cpp/src/llama-vocab.h +43 -39
- package/cpp/llama.cpp/src/llama.cpp +69 -10
- package/cpp/llama.cpp/src/models/afmoe.cpp +187 -0
- package/cpp/llama.cpp/src/models/apertus.cpp +125 -0
- package/cpp/llama.cpp/src/models/arcee.cpp +135 -0
- package/cpp/llama.cpp/src/models/arctic.cpp +138 -0
- package/cpp/llama.cpp/src/models/arwkv7.cpp +86 -0
- package/cpp/llama.cpp/src/models/baichuan.cpp +122 -0
- package/cpp/llama.cpp/src/models/bailingmoe.cpp +144 -0
- package/cpp/llama.cpp/src/models/bailingmoe2.cpp +135 -0
- package/cpp/llama.cpp/src/models/bert.cpp +176 -0
- package/cpp/llama.cpp/src/models/bitnet.cpp +160 -0
- package/cpp/llama.cpp/src/models/bloom.cpp +101 -0
- package/cpp/llama.cpp/src/models/chameleon.cpp +178 -0
- package/cpp/llama.cpp/src/models/chatglm.cpp +132 -0
- package/cpp/llama.cpp/src/models/codeshell.cpp +111 -0
- package/cpp/llama.cpp/src/models/cogvlm.cpp +100 -0
- package/cpp/llama.cpp/src/models/cohere2-iswa.cpp +131 -0
- package/cpp/llama.cpp/src/models/command-r.cpp +122 -0
- package/cpp/llama.cpp/src/models/dbrx.cpp +123 -0
- package/cpp/llama.cpp/src/models/deci.cpp +135 -0
- package/cpp/llama.cpp/src/models/deepseek.cpp +144 -0
- package/cpp/llama.cpp/src/models/deepseek2.cpp +237 -0
- package/cpp/llama.cpp/src/models/dots1.cpp +134 -0
- package/cpp/llama.cpp/src/models/dream.cpp +105 -0
- package/cpp/llama.cpp/src/models/ernie4-5-moe.cpp +150 -0
- package/cpp/llama.cpp/src/models/ernie4-5.cpp +110 -0
- package/cpp/llama.cpp/src/models/exaone.cpp +114 -0
- package/cpp/llama.cpp/src/models/exaone4.cpp +123 -0
- package/cpp/llama.cpp/src/models/falcon-h1.cpp +113 -0
- package/cpp/llama.cpp/src/models/falcon.cpp +120 -0
- package/cpp/llama.cpp/src/models/gemma-embedding.cpp +120 -0
- package/cpp/llama.cpp/src/models/gemma.cpp +112 -0
- package/cpp/llama.cpp/src/models/gemma2-iswa.cpp +125 -0
- package/cpp/llama.cpp/src/models/gemma3-iswa.cpp +131 -0
- package/cpp/llama.cpp/src/models/gemma3n-iswa.cpp +377 -0
- package/cpp/llama.cpp/src/models/glm4-moe.cpp +153 -0
- package/cpp/llama.cpp/src/models/glm4.cpp +127 -0
- package/cpp/llama.cpp/src/models/gpt2.cpp +105 -0
- package/cpp/llama.cpp/src/models/gptneox.cpp +144 -0
- package/cpp/llama.cpp/src/models/granite-hybrid.cpp +196 -0
- package/cpp/llama.cpp/src/models/granite.cpp +211 -0
- package/cpp/llama.cpp/src/models/graph-context-mamba.cpp +283 -0
- package/cpp/llama.cpp/src/models/grok.cpp +159 -0
- package/cpp/llama.cpp/src/models/grovemoe.cpp +141 -0
- package/cpp/llama.cpp/src/models/hunyuan-dense.cpp +132 -0
- package/cpp/llama.cpp/src/models/hunyuan-moe.cpp +154 -0
- package/cpp/llama.cpp/src/models/internlm2.cpp +120 -0
- package/cpp/llama.cpp/src/models/jais.cpp +86 -0
- package/cpp/llama.cpp/src/models/jamba.cpp +106 -0
- package/cpp/llama.cpp/src/models/lfm2.cpp +173 -0
- package/cpp/llama.cpp/src/models/llada-moe.cpp +122 -0
- package/cpp/llama.cpp/src/models/llada.cpp +99 -0
- package/cpp/llama.cpp/src/models/llama-iswa.cpp +174 -0
- package/cpp/llama.cpp/src/models/llama.cpp +155 -0
- package/cpp/llama.cpp/src/models/mamba.cpp +55 -0
- package/cpp/llama.cpp/src/models/minicpm3.cpp +199 -0
- package/cpp/llama.cpp/src/models/minimax-m2.cpp +124 -0
- package/cpp/llama.cpp/src/models/models.h +485 -0
- package/cpp/llama.cpp/src/models/mpt.cpp +126 -0
- package/cpp/llama.cpp/src/models/nemotron-h.cpp +121 -0
- package/cpp/llama.cpp/src/models/nemotron.cpp +122 -0
- package/cpp/llama.cpp/src/models/neo-bert.cpp +104 -0
- package/cpp/llama.cpp/src/models/olmo.cpp +121 -0
- package/cpp/llama.cpp/src/models/olmo2.cpp +150 -0
- package/cpp/llama.cpp/src/models/olmoe.cpp +124 -0
- package/cpp/llama.cpp/src/models/openai-moe-iswa.cpp +124 -0
- package/cpp/llama.cpp/src/models/openelm.cpp +124 -0
- package/cpp/llama.cpp/src/models/orion.cpp +123 -0
- package/cpp/llama.cpp/src/models/pangu-embedded.cpp +121 -0
- package/cpp/llama.cpp/src/models/phi2.cpp +121 -0
- package/cpp/llama.cpp/src/models/phi3.cpp +152 -0
- package/cpp/llama.cpp/src/models/plamo.cpp +110 -0
- package/cpp/llama.cpp/src/models/plamo2.cpp +316 -0
- package/cpp/llama.cpp/src/models/plm.cpp +168 -0
- package/cpp/llama.cpp/src/models/qwen.cpp +108 -0
- package/cpp/llama.cpp/src/models/qwen2.cpp +117 -0
- package/cpp/llama.cpp/src/models/qwen2moe.cpp +151 -0
- package/cpp/llama.cpp/src/models/qwen2vl.cpp +117 -0
- package/cpp/llama.cpp/src/models/qwen3.cpp +117 -0
- package/cpp/llama.cpp/src/models/qwen3moe.cpp +124 -0
- package/cpp/llama.cpp/src/models/qwen3vl-moe.cpp +149 -0
- package/cpp/llama.cpp/src/models/qwen3vl.cpp +141 -0
- package/cpp/llama.cpp/src/models/refact.cpp +94 -0
- package/cpp/llama.cpp/src/models/rwkv6-base.cpp +162 -0
- package/cpp/llama.cpp/src/models/rwkv6.cpp +94 -0
- package/cpp/llama.cpp/src/models/rwkv6qwen2.cpp +86 -0
- package/cpp/llama.cpp/src/models/rwkv7-base.cpp +135 -0
- package/cpp/llama.cpp/src/models/rwkv7.cpp +90 -0
- package/cpp/llama.cpp/src/models/seed-oss.cpp +124 -0
- package/cpp/llama.cpp/src/models/smallthinker.cpp +120 -0
- package/cpp/llama.cpp/src/models/smollm3.cpp +128 -0
- package/cpp/llama.cpp/src/models/stablelm.cpp +146 -0
- package/cpp/llama.cpp/src/models/starcoder.cpp +100 -0
- package/cpp/llama.cpp/src/models/starcoder2.cpp +121 -0
- package/cpp/llama.cpp/src/models/t5-dec.cpp +166 -0
- package/cpp/llama.cpp/src/models/t5-enc.cpp +96 -0
- package/cpp/llama.cpp/src/models/wavtokenizer-dec.cpp +149 -0
- package/cpp/llama.cpp/src/models/xverse.cpp +108 -0
- package/cpp/llama.cpp/src/unicode.cpp +77 -0
- package/cpp/llama.cpp/src/unicode.h +43 -0
- package/cpp/llama.cpp/vendor/cpp-httplib/CMakeLists.txt +94 -0
- package/cpp/llama.cpp/vendor/cpp-httplib/httplib.cpp +9339 -0
- package/cpp/llama.cpp/vendor/cpp-httplib/httplib.h +433 -8222
- package/cpp/llama.cpp/vendor/cpp-httplib/patch-boringssl.cmake +6 -0
- package/cpp/llama.cpp/vendor/miniaudio/miniaudio.h +4179 -1900
- package/cpp/llama.cpp/vendor/minja/chat-template.hpp +9 -2
- package/cpp/llama.cpp/vendor/minja/minja.hpp +101 -22
- package/ios/include/chat.h +16 -3
- package/ios/include/common/minja/chat-template.hpp +9 -2
- package/ios/include/common/minja/minja.hpp +101 -22
- package/ios/include/common.h +57 -19
- package/ios/include/json-schema-to-grammar.h +2 -0
- package/ios/include/llama.h +44 -21
- package/ios/include/log.h +12 -4
- package/ios/include/sampling.h +3 -1
- package/ios/libs/llama.xcframework/Info.plist +20 -20
- package/ios/libs/llama.xcframework/ios-arm64/dSYMs/llama.dSYM/Contents/Resources/DWARF/llama +0 -0
- package/ios/libs/llama.xcframework/ios-arm64/dSYMs/llama.dSYM/Contents/Resources/Relocations/aarch64/llama.yml +6399 -5557
- package/ios/libs/llama.xcframework/ios-arm64/llama.framework/Headers/ggml-backend.h +19 -1
- package/ios/libs/llama.xcframework/ios-arm64/llama.framework/Headers/ggml-cpu.h +1 -1
- package/ios/libs/llama.xcframework/ios-arm64/llama.framework/Headers/ggml-metal.h +1 -6
- package/ios/libs/llama.xcframework/ios-arm64/llama.framework/Headers/ggml.h +199 -6
- package/ios/libs/llama.xcframework/ios-arm64/llama.framework/Headers/llama.h +44 -21
- package/ios/libs/llama.xcframework/ios-arm64/llama.framework/llama +0 -0
- package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Resources/DWARF/llama +0 -0
- package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Resources/Relocations/aarch64/llama.yml +6362 -5520
- package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Resources/Relocations/x86_64/llama.yml +4813 -4241
- package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/llama.framework/Headers/ggml-backend.h +19 -1
- package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/llama.framework/Headers/ggml-cpu.h +1 -1
- package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/llama.framework/Headers/ggml-metal.h +1 -6
- package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/llama.framework/Headers/ggml.h +199 -6
- package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/llama.framework/Headers/llama.h +44 -21
- package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/llama.framework/llama +0 -0
- package/package.json +10 -4
- package/cpp/llama.cpp/ggml/src/ggml-cann/Doxyfile +0 -2579
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-tile-f16.cu +0 -371
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-tile-f16.cuh +0 -3
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-tile-f32.cu +0 -379
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-tile-f32.cuh +0 -3
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-vec-f16.cuh +0 -495
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-vec-f32.cuh +0 -486
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-f16.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q8_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-f16.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q8_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-f16.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q8_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-f16.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q8_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-f16.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q8_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-f16.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q8_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs256-f16-f16.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-f16.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q8_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-f16.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q8_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-f16.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q8_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-f16.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q8_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-f16.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q8_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-f16.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q8_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-f16.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q8_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs256-f16-f16.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-f16.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_1.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q8_0.cu +0 -5
- package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal.m +0 -6886
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_base.comp +0 -154
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq_funcs.comp +0 -105
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rope_head.comp +0 -55
- package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/cpy.wgsl +0 -60
- package/cpp/llama.cpp/ggml/src/ggml-zdnn/ggml-zdnn-impl.h +0 -97
- package/cpp/llama.cpp/models/ggml-vocab-aquila.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-baichuan.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-bert-bge.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-bert-bge.gguf.inp +0 -112
- package/cpp/llama.cpp/models/ggml-vocab-bert-bge.gguf.out +0 -46
- package/cpp/llama.cpp/models/ggml-vocab-command-r.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-command-r.gguf.inp +0 -112
- package/cpp/llama.cpp/models/ggml-vocab-command-r.gguf.out +0 -46
- package/cpp/llama.cpp/models/ggml-vocab-deepseek-coder.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-deepseek-coder.gguf.inp +0 -112
- package/cpp/llama.cpp/models/ggml-vocab-deepseek-coder.gguf.out +0 -46
- package/cpp/llama.cpp/models/ggml-vocab-deepseek-llm.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-deepseek-llm.gguf.inp +0 -112
- package/cpp/llama.cpp/models/ggml-vocab-deepseek-llm.gguf.out +0 -46
- package/cpp/llama.cpp/models/ggml-vocab-falcon.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-falcon.gguf.inp +0 -112
- package/cpp/llama.cpp/models/ggml-vocab-falcon.gguf.out +0 -46
- package/cpp/llama.cpp/models/ggml-vocab-gpt-2.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-gpt-2.gguf.inp +0 -112
- package/cpp/llama.cpp/models/ggml-vocab-gpt-2.gguf.out +0 -46
- package/cpp/llama.cpp/models/ggml-vocab-gpt-neox.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-llama-bpe.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-llama-bpe.gguf.inp +0 -112
- package/cpp/llama.cpp/models/ggml-vocab-llama-bpe.gguf.out +0 -46
- package/cpp/llama.cpp/models/ggml-vocab-llama-spm.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-llama-spm.gguf.inp +0 -112
- package/cpp/llama.cpp/models/ggml-vocab-llama-spm.gguf.out +0 -46
- package/cpp/llama.cpp/models/ggml-vocab-mpt.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-mpt.gguf.inp +0 -112
- package/cpp/llama.cpp/models/ggml-vocab-mpt.gguf.out +0 -46
- package/cpp/llama.cpp/models/ggml-vocab-nomic-bert-moe.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-phi-3.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-phi-3.gguf.inp +0 -112
- package/cpp/llama.cpp/models/ggml-vocab-phi-3.gguf.out +0 -46
- package/cpp/llama.cpp/models/ggml-vocab-qwen2.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-qwen2.gguf.inp +0 -112
- package/cpp/llama.cpp/models/ggml-vocab-qwen2.gguf.out +0 -46
- package/cpp/llama.cpp/models/ggml-vocab-refact.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-refact.gguf.inp +0 -112
- package/cpp/llama.cpp/models/ggml-vocab-refact.gguf.out +0 -46
- package/cpp/llama.cpp/models/ggml-vocab-starcoder.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-starcoder.gguf.inp +0 -112
- package/cpp/llama.cpp/models/ggml-vocab-starcoder.gguf.out +0 -46
- package/cpp/llama.cpp/models/templates/ByteDance-Seed-OSS.jinja +0 -171
- package/cpp/llama.cpp/models/templates/CohereForAI-c4ai-command-r-plus-tool_use.jinja +0 -202
- package/cpp/llama.cpp/models/templates/CohereForAI-c4ai-command-r7b-12-2024-tool_use.jinja +0 -156
- package/cpp/llama.cpp/models/templates/Mistral-Small-3.2-24B-Instruct-2506.jinja +0 -124
- package/cpp/llama.cpp/models/templates/NousResearch-Hermes-2-Pro-Llama-3-8B-tool_use.jinja +0 -152
- package/cpp/llama.cpp/models/templates/NousResearch-Hermes-3-Llama-3.1-8B-tool_use.jinja +0 -152
- package/cpp/llama.cpp/models/templates/Qwen-QwQ-32B.jinja +0 -62
- package/cpp/llama.cpp/models/templates/Qwen-Qwen2.5-7B-Instruct.jinja +0 -54
- package/cpp/llama.cpp/models/templates/Qwen-Qwen3-0.6B.jinja +0 -85
- package/cpp/llama.cpp/models/templates/README.md +0 -25
- package/cpp/llama.cpp/models/templates/deepseek-ai-DeepSeek-R1-Distill-Llama-8B.jinja +0 -1
- package/cpp/llama.cpp/models/templates/deepseek-ai-DeepSeek-R1-Distill-Qwen-32B.jinja +0 -1
- package/cpp/llama.cpp/models/templates/fireworks-ai-llama-3-firefunction-v2.jinja +0 -57
- package/cpp/llama.cpp/models/templates/google-gemma-2-2b-it.jinja +0 -4
- package/cpp/llama.cpp/models/templates/ibm-granite-granite-3.3-2B-Instruct.jinja +0 -59
- package/cpp/llama.cpp/models/templates/llama-cpp-deepseek-r1.jinja +0 -76
- package/cpp/llama.cpp/models/templates/llama-cpp-rwkv-world.jinja +0 -34
- package/cpp/llama.cpp/models/templates/meetkai-functionary-medium-v3.1.jinja +0 -58
- package/cpp/llama.cpp/models/templates/meetkai-functionary-medium-v3.2.jinja +0 -287
- package/cpp/llama.cpp/models/templates/meta-llama-Llama-3.1-8B-Instruct.jinja +0 -109
- package/cpp/llama.cpp/models/templates/meta-llama-Llama-3.2-3B-Instruct.jinja +0 -93
- package/cpp/llama.cpp/models/templates/meta-llama-Llama-3.3-70B-Instruct.jinja +0 -109
- package/cpp/llama.cpp/models/templates/microsoft-Phi-3.5-mini-instruct.jinja +0 -8
- package/cpp/llama.cpp/models/templates/mistralai-Mistral-Nemo-Instruct-2407.jinja +0 -87
- package/cpp/llama.cpp/models/templates/moonshotai-Kimi-K2.jinja +0 -43
- package/cpp/llama.cpp/models/templates/openai-gpt-oss-120b.jinja +0 -331
- package/cpp/llama.cpp/models/templates/unsloth-mistral-Devstral-Small-2507.jinja +0 -105
- package/cpp/llama.cpp/prompts/LLM-questions.txt +0 -49
- package/cpp/llama.cpp/prompts/alpaca.txt +0 -1
- package/cpp/llama.cpp/prompts/assistant.txt +0 -31
- package/cpp/llama.cpp/prompts/chat-with-baichuan.txt +0 -4
- package/cpp/llama.cpp/prompts/chat-with-bob.txt +0 -7
- package/cpp/llama.cpp/prompts/chat-with-qwen.txt +0 -1
- package/cpp/llama.cpp/prompts/chat-with-vicuna-v0.txt +0 -7
- package/cpp/llama.cpp/prompts/chat-with-vicuna-v1.txt +0 -7
- package/cpp/llama.cpp/prompts/chat.txt +0 -28
- package/cpp/llama.cpp/prompts/dan-modified.txt +0 -1
- package/cpp/llama.cpp/prompts/dan.txt +0 -1
- package/cpp/llama.cpp/prompts/mnemonics.txt +0 -93
- package/cpp/llama.cpp/prompts/parallel-questions.txt +0 -43
- package/cpp/llama.cpp/prompts/reason-act.txt +0 -18
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/dSYMs/llama.dSYM/Contents/Info.plist +0 -20
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/dSYMs/llama.dSYM/Contents/Resources/DWARF/llama +0 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/dSYMs/llama.dSYM/Contents/Resources/Relocations/aarch64/llama.yml +0 -5524
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/dSYMs/llama.dSYM/Contents/Resources/Relocations/x86_64/llama.yml +0 -4247
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/ggml-alloc.h +0 -76
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/ggml-backend.h +0 -354
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/ggml-blas.h +0 -25
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/ggml-cpu.h +0 -145
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/ggml-metal.h +0 -66
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/ggml-opt.h +0 -256
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/ggml.h +0 -2492
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/gguf.h +0 -202
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/llama.h +0 -1391
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Modules/module.modulemap +0 -17
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Resources/Info.plist +0 -32
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/ggml-alloc.h +0 -76
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/ggml-backend.h +0 -354
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/ggml-blas.h +0 -25
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/ggml-cpu.h +0 -145
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/ggml-metal.h +0 -66
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/ggml-opt.h +0 -256
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/ggml.h +0 -2492
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/gguf.h +0 -202
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/llama.h +0 -1391
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Modules/module.modulemap +0 -17
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Resources/Info.plist +0 -32
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/llama +0 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/ggml-alloc.h +0 -76
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/ggml-backend.h +0 -354
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/ggml-blas.h +0 -25
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/ggml-cpu.h +0 -145
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/ggml-metal.h +0 -66
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/ggml-opt.h +0 -256
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/ggml.h +0 -2492
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/gguf.h +0 -202
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/llama.h +0 -1391
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Modules/module.modulemap +0 -17
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Resources/Info.plist +0 -32
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/llama +0 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/llama +0 -0
- package/ios/libs/llama.xcframework/tvos-arm64/dSYMs/llama.dSYM/Contents/Info.plist +0 -20
- package/ios/libs/llama.xcframework/tvos-arm64/dSYMs/llama.dSYM/Contents/Resources/DWARF/llama +0 -0
- package/ios/libs/llama.xcframework/tvos-arm64/dSYMs/llama.dSYM/Contents/Resources/Relocations/aarch64/llama.yml +0 -5561
- package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/ggml-alloc.h +0 -76
- package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/ggml-backend.h +0 -354
- package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/ggml-blas.h +0 -25
- package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/ggml-cpu.h +0 -145
- package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/ggml-metal.h +0 -66
- package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/ggml-opt.h +0 -256
- package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/ggml.h +0 -2492
- package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/gguf.h +0 -202
- package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/llama.h +0 -1391
- package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Info.plist +0 -35
- package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Modules/module.modulemap +0 -17
- package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/llama +0 -0
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Info.plist +0 -20
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Resources/DWARF/llama +0 -0
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Resources/Relocations/aarch64/llama.yml +0 -5524
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Resources/Relocations/x86_64/llama.yml +0 -4246
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/ggml-alloc.h +0 -76
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/ggml-backend.h +0 -354
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/ggml-blas.h +0 -25
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/ggml-cpu.h +0 -145
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/ggml-metal.h +0 -66
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/ggml-opt.h +0 -256
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/ggml.h +0 -2492
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/gguf.h +0 -202
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/llama.h +0 -1391
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Info.plist +0 -35
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Modules/module.modulemap +0 -17
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/llama +0 -0
- package/ios/libs/llama.xcframework/xros-arm64/dSYMs/llama.dSYM/Contents/Info.plist +0 -20
- package/ios/libs/llama.xcframework/xros-arm64/dSYMs/llama.dSYM/Contents/Resources/DWARF/llama +0 -0
- package/ios/libs/llama.xcframework/xros-arm64/dSYMs/llama.dSYM/Contents/Resources/Relocations/aarch64/llama.yml +0 -5558
- package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/ggml-alloc.h +0 -76
- package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/ggml-backend.h +0 -354
- package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/ggml-blas.h +0 -25
- package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/ggml-cpu.h +0 -145
- package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/ggml-metal.h +0 -66
- package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/ggml-opt.h +0 -256
- package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/ggml.h +0 -2492
- package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/gguf.h +0 -202
- package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/llama.h +0 -1391
- package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Info.plist +0 -32
- package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Modules/module.modulemap +0 -17
- package/ios/libs/llama.xcframework/xros-arm64/llama.framework/llama +0 -0
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Info.plist +0 -20
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Resources/DWARF/llama +0 -0
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Resources/Relocations/aarch64/llama.yml +0 -5520
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Resources/Relocations/x86_64/llama.yml +0 -4243
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/ggml-alloc.h +0 -76
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/ggml-backend.h +0 -354
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/ggml-blas.h +0 -25
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/ggml-cpu.h +0 -145
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/ggml-metal.h +0 -66
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/ggml-opt.h +0 -256
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/ggml.h +0 -2492
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/gguf.h +0 -202
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/llama.h +0 -1391
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Info.plist +0 -32
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Modules/module.modulemap +0 -17
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/llama +0 -0
- /package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{test_bfloat16_support.comp → feature-tests/bfloat16.comp} +0 -0
- /package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{test_coopmat_support.comp → feature-tests/coopmat.comp} +0 -0
- /package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{test_coopmat2_support.comp → feature-tests/coopmat2.comp} +0 -0
- /package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{test_integer_dot_support.comp → feature-tests/integer_dot.comp} +0 -0
- /package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{generic_head.comp → generic_head.glsl} +0 -0
- /package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{generic_unary_head.comp → generic_unary_head.glsl} +0 -0
- /package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{glu_main.comp → glu_main.glsl} +0 -0
- /package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{rte.comp → rte.glsl} +0 -0
- /package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{utils.comp → utils.glsl} +0 -0
|
@@ -5,6 +5,7 @@
|
|
|
5
5
|
#include <map>
|
|
6
6
|
|
|
7
7
|
static const std::map<llm_arch, const char *> LLM_ARCH_NAMES = {
|
|
8
|
+
{ LLM_ARCH_CLIP, "clip" }, // dummy, only used by llama-quantize
|
|
8
9
|
{ LLM_ARCH_LLAMA, "llama" },
|
|
9
10
|
{ LLM_ARCH_LLAMA4, "llama4" },
|
|
10
11
|
{ LLM_ARCH_DECI, "deci" },
|
|
@@ -31,6 +32,8 @@ static const std::map<llm_arch, const char *> LLM_ARCH_NAMES = {
|
|
|
31
32
|
{ LLM_ARCH_QWEN2VL, "qwen2vl" },
|
|
32
33
|
{ LLM_ARCH_QWEN3, "qwen3" },
|
|
33
34
|
{ LLM_ARCH_QWEN3MOE, "qwen3moe" },
|
|
35
|
+
{ LLM_ARCH_QWEN3VL, "qwen3vl" },
|
|
36
|
+
{ LLM_ARCH_QWEN3VLMOE, "qwen3vlmoe" },
|
|
34
37
|
{ LLM_ARCH_PHI2, "phi2" },
|
|
35
38
|
{ LLM_ARCH_PHI3, "phi3" },
|
|
36
39
|
{ LLM_ARCH_PHIMOE, "phimoe" },
|
|
@@ -45,6 +48,7 @@ static const std::map<llm_arch, const char *> LLM_ARCH_NAMES = {
|
|
|
45
48
|
{ LLM_ARCH_GEMMA2, "gemma2" },
|
|
46
49
|
{ LLM_ARCH_GEMMA3, "gemma3" },
|
|
47
50
|
{ LLM_ARCH_GEMMA3N, "gemma3n" },
|
|
51
|
+
{ LLM_ARCH_GEMMA_EMBEDDING, "gemma-embedding" },
|
|
48
52
|
{ LLM_ARCH_STARCODER2, "starcoder2" },
|
|
49
53
|
{ LLM_ARCH_MAMBA, "mamba" },
|
|
50
54
|
{ LLM_ARCH_MAMBA2, "mamba2" },
|
|
@@ -83,8 +87,10 @@ static const std::map<llm_arch, const char *> LLM_ARCH_NAMES = {
|
|
|
83
87
|
{ LLM_ARCH_WAVTOKENIZER_DEC, "wavtokenizer-dec" },
|
|
84
88
|
{ LLM_ARCH_PLM, "plm" },
|
|
85
89
|
{ LLM_ARCH_BAILINGMOE, "bailingmoe" },
|
|
90
|
+
{ LLM_ARCH_BAILINGMOE2, "bailingmoe2" },
|
|
86
91
|
{ LLM_ARCH_DOTS1, "dots1" },
|
|
87
92
|
{ LLM_ARCH_ARCEE, "arcee" },
|
|
93
|
+
{ LLM_ARCH_AFMOE, "afmoe" },
|
|
88
94
|
{ LLM_ARCH_ERNIE4_5, "ernie4_5" },
|
|
89
95
|
{ LLM_ARCH_ERNIE4_5_MOE, "ernie4_5-moe" },
|
|
90
96
|
{ LLM_ARCH_HUNYUAN_MOE, "hunyuan-moe" },
|
|
@@ -92,10 +98,17 @@ static const std::map<llm_arch, const char *> LLM_ARCH_NAMES = {
|
|
|
92
98
|
{ LLM_ARCH_SMOLLM3, "smollm3" },
|
|
93
99
|
{ LLM_ARCH_OPENAI_MOE, "gpt-oss" },
|
|
94
100
|
{ LLM_ARCH_LFM2, "lfm2" },
|
|
101
|
+
{ LLM_ARCH_LFM2MOE, "lfm2moe" },
|
|
95
102
|
{ LLM_ARCH_DREAM, "dream" },
|
|
96
103
|
{ LLM_ARCH_SMALLTHINKER, "smallthinker" },
|
|
97
104
|
{ LLM_ARCH_LLADA, "llada" },
|
|
105
|
+
{ LLM_ARCH_LLADA_MOE, "llada-moe" },
|
|
98
106
|
{ LLM_ARCH_SEED_OSS, "seed_oss" },
|
|
107
|
+
{ LLM_ARCH_GROVEMOE, "grovemoe" },
|
|
108
|
+
{ LLM_ARCH_APERTUS, "apertus" },
|
|
109
|
+
{ LLM_ARCH_MINIMAX_M2, "minimax-m2" },
|
|
110
|
+
{ LLM_ARCH_COGVLM, "cogvlm" },
|
|
111
|
+
{ LLM_ARCH_PANGU_EMBED, "pangu-embedded" },
|
|
99
112
|
{ LLM_ARCH_UNKNOWN, "(unknown)" },
|
|
100
113
|
};
|
|
101
114
|
|
|
@@ -123,20 +136,28 @@ static const std::map<llm_kv, const char *> LLM_KV_NAMES = {
|
|
|
123
136
|
{ LLM_KV_FEED_FORWARD_LENGTH, "%s.feed_forward_length" },
|
|
124
137
|
{ LLM_KV_EXPERT_FEED_FORWARD_LENGTH, "%s.expert_feed_forward_length" },
|
|
125
138
|
{ LLM_KV_EXPERT_SHARED_FEED_FORWARD_LENGTH, "%s.expert_shared_feed_forward_length" },
|
|
139
|
+
{ LLM_KV_EXPERT_CHUNK_FEED_FORWARD_LENGTH, "%s.expert_chunk_feed_forward_length" },
|
|
126
140
|
{ LLM_KV_USE_PARALLEL_RESIDUAL, "%s.use_parallel_residual" },
|
|
127
141
|
{ LLM_KV_TENSOR_DATA_LAYOUT, "%s.tensor_data_layout" },
|
|
128
142
|
{ LLM_KV_EXPERT_COUNT, "%s.expert_count" },
|
|
129
143
|
{ LLM_KV_EXPERT_USED_COUNT, "%s.expert_used_count" },
|
|
130
144
|
{ LLM_KV_EXPERT_SHARED_COUNT, "%s.expert_shared_count" },
|
|
145
|
+
{ LLM_KV_EXPERT_GROUP_COUNT, "%s.expert_group_count" },
|
|
146
|
+
{ LLM_KV_EXPERT_GROUP_USED_COUNT, "%s.expert_group_used_count" },
|
|
131
147
|
{ LLM_KV_EXPERT_WEIGHTS_SCALE, "%s.expert_weights_scale" },
|
|
132
148
|
{ LLM_KV_EXPERT_WEIGHTS_NORM, "%s.expert_weights_norm" },
|
|
133
149
|
{ LLM_KV_EXPERT_GATING_FUNC, "%s.expert_gating_func" },
|
|
150
|
+
{ LLM_KV_EXPERT_GROUP_SCALE, "%s.expert_group_scale" },
|
|
151
|
+
{ LLM_KV_EXPERTS_PER_GROUP, "%s.experts_per_group" },
|
|
134
152
|
{ LLM_KV_MOE_EVERY_N_LAYERS, "%s.moe_every_n_layers" },
|
|
135
153
|
{ LLM_KV_NEXTN_PREDICT_LAYERS, "%s.nextn_predict_layers" },
|
|
154
|
+
{ LLM_KV_NUM_DEEPSTACK_LAYERS, "%s.n_deepstack_layers" },
|
|
136
155
|
{ LLM_KV_POOLING_TYPE, "%s.pooling_type" },
|
|
137
156
|
{ LLM_KV_LOGIT_SCALE, "%s.logit_scale" },
|
|
138
157
|
{ LLM_KV_DECODER_START_TOKEN_ID, "%s.decoder_start_token_id" },
|
|
158
|
+
{ LLM_KV_DECODER_BLOCK_COUNT, "%s.decoder_block_count" },
|
|
139
159
|
{ LLM_KV_ATTN_LOGIT_SOFTCAPPING, "%s.attn_logit_softcapping" },
|
|
160
|
+
{ LLM_KV_ROUTER_LOGIT_SOFTCAPPING, "%s.router_logit_softcapping" },
|
|
140
161
|
{ LLM_KV_FINAL_LOGIT_SOFTCAPPING, "%s.final_logit_softcapping" },
|
|
141
162
|
{ LLM_KV_SWIN_NORM, "%s.swin_norm" },
|
|
142
163
|
{ LLM_KV_RESCALE_EVERY_N_LAYERS, "%s.rescale_every_n_layers" },
|
|
@@ -167,19 +188,25 @@ static const std::map<llm_kv, const char *> LLM_KV_NAMES = {
|
|
|
167
188
|
{ LLM_KV_ATTENTION_RELATIVE_BUCKETS_COUNT, "%s.attention.relative_buckets_count" },
|
|
168
189
|
{ LLM_KV_ATTENTION_SLIDING_WINDOW, "%s.attention.sliding_window" },
|
|
169
190
|
{ LLM_KV_ATTENTION_SCALE, "%s.attention.scale" },
|
|
191
|
+
{ LLM_KV_ATTENTION_OUTPUT_SCALE, "%s.attention.output_scale" },
|
|
192
|
+
{ LLM_KV_ATTENTION_TEMPERATURE_LENGTH, "%s.attention.temperature_length" },
|
|
170
193
|
{ LLM_KV_ATTENTION_KEY_LENGTH_MLA, "%s.attention.key_length_mla" },
|
|
171
194
|
{ LLM_KV_ATTENTION_VALUE_LENGTH_MLA, "%s.attention.value_length_mla" },
|
|
172
195
|
|
|
173
|
-
{ LLM_KV_ROPE_DIMENSION_COUNT,
|
|
174
|
-
{ LLM_KV_ROPE_DIMENSION_SECTIONS,
|
|
175
|
-
{ LLM_KV_ROPE_FREQ_BASE,
|
|
176
|
-
{ LLM_KV_ROPE_SCALE_LINEAR,
|
|
177
|
-
{ LLM_KV_ROPE_SCALING_TYPE,
|
|
178
|
-
{ LLM_KV_ROPE_SCALING_FACTOR,
|
|
179
|
-
{ LLM_KV_ROPE_SCALING_ATTN_FACTOR,
|
|
180
|
-
{ LLM_KV_ROPE_SCALING_ORIG_CTX_LEN,
|
|
181
|
-
{ LLM_KV_ROPE_SCALING_FINETUNED,
|
|
182
|
-
{ LLM_KV_ROPE_SCALING_YARN_LOG_MUL,
|
|
196
|
+
{ LLM_KV_ROPE_DIMENSION_COUNT, "%s.rope.dimension_count" },
|
|
197
|
+
{ LLM_KV_ROPE_DIMENSION_SECTIONS, "%s.rope.dimension_sections" },
|
|
198
|
+
{ LLM_KV_ROPE_FREQ_BASE, "%s.rope.freq_base" },
|
|
199
|
+
{ LLM_KV_ROPE_SCALE_LINEAR, "%s.rope.scale_linear" },
|
|
200
|
+
{ LLM_KV_ROPE_SCALING_TYPE, "%s.rope.scaling.type" },
|
|
201
|
+
{ LLM_KV_ROPE_SCALING_FACTOR, "%s.rope.scaling.factor" },
|
|
202
|
+
{ LLM_KV_ROPE_SCALING_ATTN_FACTOR, "%s.rope.scaling.attn_factor" },
|
|
203
|
+
{ LLM_KV_ROPE_SCALING_ORIG_CTX_LEN, "%s.rope.scaling.original_context_length" },
|
|
204
|
+
{ LLM_KV_ROPE_SCALING_FINETUNED, "%s.rope.scaling.finetuned" },
|
|
205
|
+
{ LLM_KV_ROPE_SCALING_YARN_LOG_MUL, "%s.rope.scaling.yarn_log_multiplier" },
|
|
206
|
+
{ LLM_KV_ROPE_SCALING_YARN_EXT_FACTOR, "%s.rope.scaling.yarn_ext_factor" },
|
|
207
|
+
{ LLM_KV_ROPE_SCALING_YARN_ATTN_FACTOR, "%s.rope.scaling.yarn_attn_factor" },
|
|
208
|
+
{ LLM_KV_ROPE_SCALING_YARN_BETA_FAST, "%s.rope.scaling.yarn_beta_fast" },
|
|
209
|
+
{ LLM_KV_ROPE_SCALING_YARN_BETA_SLOW, "%s.rope.scaling.yarn_beta_slow" },
|
|
183
210
|
|
|
184
211
|
{ LLM_KV_SPLIT_NO, "split.no" },
|
|
185
212
|
{ LLM_KV_SPLIT_COUNT, "split.count" },
|
|
@@ -203,6 +230,11 @@ static const std::map<llm_kv, const char *> LLM_KV_NAMES = {
|
|
|
203
230
|
{ LLM_KV_CLASSIFIER_OUTPUT_LABELS, "%s.classifier.output_labels" },
|
|
204
231
|
|
|
205
232
|
{ LLM_KV_SHORTCONV_L_CACHE, "%s.shortconv.l_cache" },
|
|
233
|
+
// sentence-transformers dense modules feature dims
|
|
234
|
+
{ LLM_KV_DENSE_2_FEAT_IN, "%s.dense_2_feat_in" },
|
|
235
|
+
{ LLM_KV_DENSE_2_FEAT_OUT, "%s.dense_2_feat_out" },
|
|
236
|
+
{ LLM_KV_DENSE_3_FEAT_IN, "%s.dense_3_feat_in" },
|
|
237
|
+
{ LLM_KV_DENSE_3_FEAT_OUT, "%s.dense_3_feat_out" },
|
|
206
238
|
|
|
207
239
|
{ LLM_KV_TOKENIZER_MODEL, "tokenizer.ggml.model" },
|
|
208
240
|
{ LLM_KV_TOKENIZER_PRE, "tokenizer.ggml.pre" },
|
|
@@ -236,10 +268,16 @@ static const std::map<llm_kv, const char *> LLM_KV_NAMES = {
|
|
|
236
268
|
{ LLM_KV_TOKENIZER_FIM_REP_ID, "tokenizer.ggml.fim_rep_token_id" },
|
|
237
269
|
{ LLM_KV_TOKENIZER_FIM_SEP_ID, "tokenizer.ggml.fim_sep_token_id" },
|
|
238
270
|
|
|
239
|
-
{ LLM_KV_ADAPTER_TYPE,
|
|
240
|
-
{ LLM_KV_ADAPTER_LORA_ALPHA,
|
|
241
|
-
{ LLM_KV_ADAPTER_LORA_TASK_NAME,
|
|
242
|
-
{ LLM_KV_ADAPTER_LORA_PROMPT_PREFIX,
|
|
271
|
+
{ LLM_KV_ADAPTER_TYPE, "adapter.type" },
|
|
272
|
+
{ LLM_KV_ADAPTER_LORA_ALPHA, "adapter.lora.alpha" },
|
|
273
|
+
{ LLM_KV_ADAPTER_LORA_TASK_NAME, "adapter.lora.task_name" },
|
|
274
|
+
{ LLM_KV_ADAPTER_LORA_PROMPT_PREFIX, "adapter.lora.prompt_prefix" },
|
|
275
|
+
{ LLM_KV_ADAPTER_ALORA_INVOCATION_TOKENS, "adapter.alora.invocation_tokens" },
|
|
276
|
+
|
|
277
|
+
{ LLM_KV_XIELU_ALPHA_N, "xielu.alpha_n" },
|
|
278
|
+
{ LLM_KV_XIELU_ALPHA_P, "xielu.alpha_p" },
|
|
279
|
+
{ LLM_KV_XIELU_BETA, "xielu.beta" },
|
|
280
|
+
{ LLM_KV_XIELU_EPS, "xielu.eps" },
|
|
243
281
|
|
|
244
282
|
// deprecated
|
|
245
283
|
{ LLM_KV_TOKENIZER_PREFIX_ID, "tokenizer.ggml.prefix_token_id" },
|
|
@@ -248,6 +286,10 @@ static const std::map<llm_kv, const char *> LLM_KV_NAMES = {
|
|
|
248
286
|
};
|
|
249
287
|
|
|
250
288
|
static const std::map<llm_arch, std::map<llm_tensor, const char *>> LLM_TENSOR_NAMES = {
|
|
289
|
+
{
|
|
290
|
+
LLM_ARCH_CLIP,
|
|
291
|
+
{},
|
|
292
|
+
},
|
|
251
293
|
{
|
|
252
294
|
LLM_ARCH_LLAMA,
|
|
253
295
|
{
|
|
@@ -292,6 +334,36 @@ static const std::map<llm_arch, std::map<llm_tensor, const char *>> LLM_TENSOR_N
|
|
|
292
334
|
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
293
335
|
},
|
|
294
336
|
},
|
|
337
|
+
{
|
|
338
|
+
LLM_ARCH_AFMOE,
|
|
339
|
+
{
|
|
340
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
341
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
342
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
343
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
344
|
+
{ LLM_TENSOR_ATTN_POST_NORM, "blk.%d.post_attention_norm" },
|
|
345
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
346
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
347
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
348
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
349
|
+
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
|
|
350
|
+
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
|
|
351
|
+
{ LLM_TENSOR_ATTN_GATE, "blk.%d.attn_gate" },
|
|
352
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
353
|
+
{ LLM_TENSOR_FFN_POST_NORM, "blk.%d.post_ffw_norm" },
|
|
354
|
+
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
|
|
355
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
356
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
357
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
358
|
+
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
|
|
359
|
+
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
|
|
360
|
+
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
|
|
361
|
+
{ LLM_TENSOR_FFN_GATE_SHEXP, "blk.%d.ffn_gate_shexp" },
|
|
362
|
+
{ LLM_TENSOR_FFN_UP_SHEXP, "blk.%d.ffn_up_shexp" },
|
|
363
|
+
{ LLM_TENSOR_FFN_DOWN_SHEXP, "blk.%d.ffn_down_shexp" },
|
|
364
|
+
{ LLM_TENSOR_FFN_EXP_PROBS_B, "blk.%d.exp_probs_b" },
|
|
365
|
+
},
|
|
366
|
+
},
|
|
295
367
|
{
|
|
296
368
|
LLM_ARCH_LLAMA4,
|
|
297
369
|
{
|
|
@@ -395,12 +467,16 @@ static const std::map<llm_arch, std::map<llm_tensor, const char *>> LLM_TENSOR_N
|
|
|
395
467
|
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd" },
|
|
396
468
|
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
|
|
397
469
|
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
470
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
471
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
472
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
398
473
|
{ LLM_TENSOR_FFN_GATE_EXP, "blk.%d.ffn_gate.%d" },
|
|
399
474
|
{ LLM_TENSOR_FFN_DOWN_EXP, "blk.%d.ffn_down.%d" },
|
|
400
475
|
{ LLM_TENSOR_FFN_UP_EXP, "blk.%d.ffn_up.%d" },
|
|
401
476
|
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
|
|
402
477
|
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
|
|
403
478
|
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
|
|
479
|
+
{ LLM_TENSOR_FFN_POST_NORM, "blk.%d.post_ffw_norm" },
|
|
404
480
|
{ LLM_TENSOR_LAYER_OUT_NORM, "blk.%d.layer_output_norm" },
|
|
405
481
|
{ LLM_TENSOR_ATTN_OUT_NORM, "blk.%d.attn_output_norm" },
|
|
406
482
|
},
|
|
@@ -706,6 +782,7 @@ static const std::map<llm_arch, std::map<llm_tensor, const char *>> LLM_TENSOR_N
|
|
|
706
782
|
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
707
783
|
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
708
784
|
{ LLM_TENSOR_OUTPUT, "output" },
|
|
785
|
+
{ LLM_TENSOR_CLS_OUT, "cls.output" },
|
|
709
786
|
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
710
787
|
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
711
788
|
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
|
|
@@ -739,6 +816,45 @@ static const std::map<llm_arch, std::map<llm_tensor, const char *>> LLM_TENSOR_N
|
|
|
739
816
|
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
|
|
740
817
|
},
|
|
741
818
|
},
|
|
819
|
+
{
|
|
820
|
+
LLM_ARCH_QWEN3VL,
|
|
821
|
+
{
|
|
822
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
823
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
824
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
825
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
826
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
827
|
+
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
|
|
828
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
829
|
+
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
|
|
830
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
831
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
832
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
833
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
834
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
835
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
836
|
+
},
|
|
837
|
+
},
|
|
838
|
+
{
|
|
839
|
+
LLM_ARCH_QWEN3VLMOE,
|
|
840
|
+
{
|
|
841
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
842
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
843
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
844
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
845
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
846
|
+
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
|
|
847
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
848
|
+
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
|
|
849
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
850
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
851
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
852
|
+
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
|
|
853
|
+
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
|
|
854
|
+
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
|
|
855
|
+
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
|
|
856
|
+
},
|
|
857
|
+
},
|
|
742
858
|
{
|
|
743
859
|
LLM_ARCH_PHI2,
|
|
744
860
|
{
|
|
@@ -1038,6 +1154,29 @@ static const std::map<llm_arch, std::map<llm_tensor, const char *>> LLM_TENSOR_N
|
|
|
1038
1154
|
{ LLM_TENSOR_LAUREL_POST_NORM, "blk.%d.laurel_post_norm" },
|
|
1039
1155
|
},
|
|
1040
1156
|
},
|
|
1157
|
+
{
|
|
1158
|
+
LLM_ARCH_GEMMA_EMBEDDING,
|
|
1159
|
+
{
|
|
1160
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1161
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
1162
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
1163
|
+
{ LLM_TENSOR_DENSE_2_OUT, "dense_2" },
|
|
1164
|
+
{ LLM_TENSOR_DENSE_3_OUT, "dense_3" },
|
|
1165
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
1166
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
1167
|
+
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
|
|
1168
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
1169
|
+
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
|
|
1170
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
1171
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
1172
|
+
{ LLM_TENSOR_ATTN_POST_NORM, "blk.%d.post_attention_norm" },
|
|
1173
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
1174
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
1175
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
1176
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
1177
|
+
{ LLM_TENSOR_FFN_POST_NORM, "blk.%d.post_ffw_norm" },
|
|
1178
|
+
},
|
|
1179
|
+
},
|
|
1041
1180
|
{
|
|
1042
1181
|
LLM_ARCH_STARCODER2,
|
|
1043
1182
|
{
|
|
@@ -1886,6 +2025,38 @@ static const std::map<llm_arch, std::map<llm_tensor, const char *>> LLM_TENSOR_N
|
|
|
1886
2025
|
{ LLM_TENSOR_FFN_UP_SHEXP, "blk.%d.ffn_up_shexp" },
|
|
1887
2026
|
},
|
|
1888
2027
|
},
|
|
2028
|
+
{
|
|
2029
|
+
LLM_ARCH_BAILINGMOE2,
|
|
2030
|
+
{
|
|
2031
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
2032
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
2033
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
2034
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
2035
|
+
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
|
|
2036
|
+
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
|
|
2037
|
+
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
|
|
2038
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
2039
|
+
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
|
|
2040
|
+
{ LLM_TENSOR_FFN_EXP_PROBS_B, "blk.%d.exp_probs_b" },
|
|
2041
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
2042
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
2043
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
2044
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
2045
|
+
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
|
|
2046
|
+
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
|
|
2047
|
+
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
|
|
2048
|
+
{ LLM_TENSOR_FFN_GATE_SHEXP, "blk.%d.ffn_gate_shexp" },
|
|
2049
|
+
{ LLM_TENSOR_FFN_DOWN_SHEXP, "blk.%d.ffn_down_shexp" },
|
|
2050
|
+
{ LLM_TENSOR_FFN_UP_SHEXP, "blk.%d.ffn_up_shexp" },
|
|
2051
|
+
{ LLM_TENSOR_NEXTN_EH_PROJ, "blk.%d.nextn.eh_proj" },
|
|
2052
|
+
{ LLM_TENSOR_NEXTN_EMBED_TOKENS, "blk.%d.nextn.embed_tokens" },
|
|
2053
|
+
{ LLM_TENSOR_NEXTN_ENORM, "blk.%d.nextn.enorm" },
|
|
2054
|
+
{ LLM_TENSOR_NEXTN_HNORM, "blk.%d.nextn.hnorm" },
|
|
2055
|
+
{ LLM_TENSOR_NEXTN_SHARED_HEAD_HEAD, "blk.%d.nextn.shared_head_head" },
|
|
2056
|
+
{ LLM_TENSOR_NEXTN_SHARED_HEAD_NORM, "blk.%d.nextn.shared_head_norm" },
|
|
2057
|
+
{ LLM_TENSOR_LAYER_OUT_NORM, "blk.%d.layer_output_norm" },
|
|
2058
|
+
},
|
|
2059
|
+
},
|
|
1889
2060
|
{
|
|
1890
2061
|
LLM_ARCH_DOTS1,
|
|
1891
2062
|
{
|
|
@@ -2057,6 +2228,32 @@ static const std::map<llm_arch, std::map<llm_tensor, const char *>> LLM_TENSOR_N
|
|
|
2057
2228
|
{ LLM_TENSOR_OUTPUT, "output" },
|
|
2058
2229
|
}
|
|
2059
2230
|
},
|
|
2231
|
+
{
|
|
2232
|
+
LLM_ARCH_LFM2MOE,
|
|
2233
|
+
{
|
|
2234
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
2235
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
2236
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
2237
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
2238
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
2239
|
+
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
|
|
2240
|
+
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
|
|
2241
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
2242
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
2243
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
2244
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
2245
|
+
{ LLM_TENSOR_SHORTCONV_CONV, "blk.%d.shortconv.conv" },
|
|
2246
|
+
{ LLM_TENSOR_SHORTCONV_INPROJ, "blk.%d.shortconv.in_proj" },
|
|
2247
|
+
{ LLM_TENSOR_SHORTCONV_OUTPROJ, "blk.%d.shortconv.out_proj" },
|
|
2248
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
2249
|
+
{ LLM_TENSOR_TOKEN_EMBD_NORM, "token_embd_norm" },
|
|
2250
|
+
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
|
|
2251
|
+
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
|
|
2252
|
+
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
|
|
2253
|
+
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
|
|
2254
|
+
{ LLM_TENSOR_FFN_EXP_PROBS_B, "blk.%d.exp_probs_b" },
|
|
2255
|
+
}
|
|
2256
|
+
},
|
|
2060
2257
|
{
|
|
2061
2258
|
LLM_ARCH_SMALLTHINKER,
|
|
2062
2259
|
{
|
|
@@ -2078,6 +2275,25 @@ static const std::map<llm_arch, std::map<llm_tensor, const char *>> LLM_TENSOR_N
|
|
|
2078
2275
|
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" }
|
|
2079
2276
|
},
|
|
2080
2277
|
},
|
|
2278
|
+
{
|
|
2279
|
+
LLM_ARCH_APERTUS,
|
|
2280
|
+
{
|
|
2281
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
2282
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
2283
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
2284
|
+
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
|
|
2285
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
2286
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
2287
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
2288
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
2289
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
2290
|
+
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
|
|
2291
|
+
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
|
|
2292
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
2293
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
2294
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
2295
|
+
},
|
|
2296
|
+
},
|
|
2081
2297
|
{
|
|
2082
2298
|
LLM_ARCH_DREAM,
|
|
2083
2299
|
{
|
|
@@ -2112,6 +2328,26 @@ static const std::map<llm_arch, std::map<llm_tensor, const char *>> LLM_TENSOR_N
|
|
|
2112
2328
|
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
2113
2329
|
},
|
|
2114
2330
|
},
|
|
2331
|
+
{
|
|
2332
|
+
LLM_ARCH_LLADA_MOE,
|
|
2333
|
+
{
|
|
2334
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
2335
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
2336
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
2337
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
2338
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
2339
|
+
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
|
|
2340
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
2341
|
+
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
|
|
2342
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
2343
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
2344
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
2345
|
+
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
|
|
2346
|
+
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
|
|
2347
|
+
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
|
|
2348
|
+
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
|
|
2349
|
+
},
|
|
2350
|
+
},
|
|
2115
2351
|
{
|
|
2116
2352
|
LLM_ARCH_SEED_OSS,
|
|
2117
2353
|
{
|
|
@@ -2129,6 +2365,87 @@ static const std::map<llm_arch, std::map<llm_tensor, const char *>> LLM_TENSOR_N
|
|
|
2129
2365
|
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
2130
2366
|
},
|
|
2131
2367
|
},
|
|
2368
|
+
{
|
|
2369
|
+
LLM_ARCH_GROVEMOE,
|
|
2370
|
+
{
|
|
2371
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
2372
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
2373
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
2374
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
2375
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
2376
|
+
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
|
|
2377
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
2378
|
+
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
|
|
2379
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
2380
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
2381
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
2382
|
+
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
|
|
2383
|
+
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
|
|
2384
|
+
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
|
|
2385
|
+
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
|
|
2386
|
+
{ LLM_TENSOR_FFN_GATE_CHEXPS, "blk.%d.ffn_gate_chexps" },
|
|
2387
|
+
{ LLM_TENSOR_FFN_DOWN_CHEXPS, "blk.%d.ffn_down_chexps" },
|
|
2388
|
+
{ LLM_TENSOR_FFN_UP_CHEXPS, "blk.%d.ffn_up_chexps" },
|
|
2389
|
+
},
|
|
2390
|
+
},
|
|
2391
|
+
{
|
|
2392
|
+
LLM_ARCH_MINIMAX_M2,
|
|
2393
|
+
{
|
|
2394
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
2395
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
2396
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
2397
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
2398
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
2399
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
2400
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
2401
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
2402
|
+
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
|
|
2403
|
+
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
|
|
2404
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
2405
|
+
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
|
|
2406
|
+
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
|
|
2407
|
+
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
|
|
2408
|
+
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
|
|
2409
|
+
{ LLM_TENSOR_FFN_EXP_PROBS_B, "blk.%d.exp_probs_b" },
|
|
2410
|
+
},
|
|
2411
|
+
},
|
|
2412
|
+
{
|
|
2413
|
+
LLM_ARCH_PANGU_EMBED,
|
|
2414
|
+
{
|
|
2415
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
2416
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
2417
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
2418
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
2419
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
2420
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
2421
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
2422
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
2423
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
2424
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
2425
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
2426
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
2427
|
+
},
|
|
2428
|
+
},
|
|
2429
|
+
{
|
|
2430
|
+
LLM_ARCH_COGVLM,
|
|
2431
|
+
{
|
|
2432
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
2433
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
2434
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
2435
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
2436
|
+
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
|
|
2437
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
2438
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
2439
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
2440
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
2441
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
2442
|
+
{ LLM_TENSOR_VISEXP_ATTN_QKV, "blk.%d.vis_attn_qkv" },
|
|
2443
|
+
{ LLM_TENSOR_VISEXP_ATTN_OUT, "blk.%d.vis_attn_output" },
|
|
2444
|
+
{ LLM_TENSOR_VISEXP_FFN_GATE, "blk.%d.vis_gate" },
|
|
2445
|
+
{ LLM_TENSOR_VISEXP_FFN_DOWN, "blk.%d.vis_down" },
|
|
2446
|
+
{ LLM_TENSOR_VISEXP_FFN_UP, "blk.%d.vis_up" },
|
|
2447
|
+
},
|
|
2448
|
+
},
|
|
2132
2449
|
{
|
|
2133
2450
|
LLM_ARCH_UNKNOWN,
|
|
2134
2451
|
{
|
|
@@ -2145,6 +2462,8 @@ static const std::map<llm_tensor, llm_tensor_info> LLM_TENSOR_INFOS = {
|
|
|
2145
2462
|
{LLM_TENSOR_OUTPUT, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL_MAT}},
|
|
2146
2463
|
{LLM_TENSOR_CLS, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL_MAT}},
|
|
2147
2464
|
{LLM_TENSOR_CLS_OUT, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL_MAT}},
|
|
2465
|
+
{LLM_TENSOR_DENSE_2_OUT, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL_MAT}}, // Dense layer output
|
|
2466
|
+
{LLM_TENSOR_DENSE_3_OUT, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL_MAT}}, // Dense layer output
|
|
2148
2467
|
{LLM_TENSOR_OUTPUT_NORM, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL}},
|
|
2149
2468
|
{LLM_TENSOR_DEC_OUTPUT_NORM, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL}},
|
|
2150
2469
|
{LLM_TENSOR_ENC_OUTPUT_NORM, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL}},
|
|
@@ -2156,6 +2475,7 @@ static const std::map<llm_tensor, llm_tensor_info> LLM_TENSOR_INFOS = {
|
|
|
2156
2475
|
{LLM_TENSOR_ATTN_V, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
2157
2476
|
{LLM_TENSOR_ATTN_QKV, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
2158
2477
|
{LLM_TENSOR_ATTN_OUT, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
2478
|
+
{LLM_TENSOR_ATTN_GATE, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
2159
2479
|
{LLM_TENSOR_FFN_GATE, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
2160
2480
|
{LLM_TENSOR_FFN_DOWN, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
2161
2481
|
{LLM_TENSOR_FFN_UP, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
@@ -2261,6 +2581,9 @@ static const std::map<llm_tensor, llm_tensor_info> LLM_TENSOR_INFOS = {
|
|
|
2261
2581
|
{LLM_TENSOR_FFN_DOWN_EXPS, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT_ID}},
|
|
2262
2582
|
{LLM_TENSOR_FFN_GATE_EXPS, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT_ID}},
|
|
2263
2583
|
{LLM_TENSOR_FFN_UP_EXPS, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT_ID}},
|
|
2584
|
+
{LLM_TENSOR_FFN_DOWN_CHEXPS, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT_ID}},
|
|
2585
|
+
{LLM_TENSOR_FFN_GATE_CHEXPS, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT_ID}},
|
|
2586
|
+
{LLM_TENSOR_FFN_UP_CHEXPS, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT_ID}},
|
|
2264
2587
|
{LLM_TENSOR_FFN_EXP_PROBS_B, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_ADD}},
|
|
2265
2588
|
// altup / laurel (gemma 3n)
|
|
2266
2589
|
{LLM_TENSOR_PER_LAYER_TOKEN_EMBD, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_GET_ROWS}},
|
|
@@ -2300,6 +2623,11 @@ static const std::map<llm_tensor, llm_tensor_info> LLM_TENSOR_INFOS = {
|
|
|
2300
2623
|
{LLM_TENSOR_SHORTCONV_CONV, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_SSM_CONV}},
|
|
2301
2624
|
{LLM_TENSOR_SHORTCONV_INPROJ, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
2302
2625
|
{LLM_TENSOR_SHORTCONV_OUTPROJ, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
2626
|
+
{LLM_TENSOR_VISEXP_ATTN_QKV, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
2627
|
+
{LLM_TENSOR_VISEXP_ATTN_OUT, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
2628
|
+
{LLM_TENSOR_VISEXP_FFN_GATE, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
2629
|
+
{LLM_TENSOR_VISEXP_FFN_DOWN, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
2630
|
+
{LLM_TENSOR_VISEXP_FFN_UP, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
2303
2631
|
// NextN/MTP tensors are currently ignored (reserved for future MTP support)
|
|
2304
2632
|
// These tensors only exist in the last layer(s) and are treated as output tensors
|
|
2305
2633
|
{LLM_TENSOR_NEXTN_EH_PROJ, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL_MAT}},
|
|
@@ -2381,6 +2709,7 @@ bool llm_arch_is_hybrid(const llm_arch & arch) {
|
|
|
2381
2709
|
case LLM_ARCH_PLAMO2:
|
|
2382
2710
|
case LLM_ARCH_GRANITE_HYBRID:
|
|
2383
2711
|
case LLM_ARCH_LFM2:
|
|
2712
|
+
case LLM_ARCH_LFM2MOE:
|
|
2384
2713
|
case LLM_ARCH_NEMOTRON_H:
|
|
2385
2714
|
return true;
|
|
2386
2715
|
default:
|
|
@@ -2392,6 +2721,7 @@ bool llm_arch_is_diffusion(const llm_arch & arch) {
|
|
|
2392
2721
|
switch (arch) {
|
|
2393
2722
|
case LLM_ARCH_DREAM:
|
|
2394
2723
|
case LLM_ARCH_LLADA:
|
|
2724
|
+
case LLM_ARCH_LLADA_MOE:
|
|
2395
2725
|
return true;
|
|
2396
2726
|
default:
|
|
2397
2727
|
return false;
|