@novastera-oss/llamarn 0.0.1-alpha.4
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/INTERFACE.md +389 -0
- package/LICENSE +201 -0
- package/README.md +235 -0
- package/RNLlamaCpp.podspec +69 -0
- package/android/CMakeLists.txt +107 -0
- package/android/build.gradle +111 -0
- package/android/generated/java/com/novastera/llamarn/NativeRNLlamaCppSpec.java +47 -0
- package/android/generated/jni/CMakeLists.txt +36 -0
- package/android/generated/jni/RNLlamaCppSpec-generated.cpp +44 -0
- package/android/generated/jni/RNLlamaCppSpec.h +31 -0
- package/android/generated/jni/react/renderer/components/RNLlamaCppSpec/RNLlamaCppSpecJSI-generated.cpp +42 -0
- package/android/generated/jni/react/renderer/components/RNLlamaCppSpec/RNLlamaCppSpecJSI.h +336 -0
- package/android/gradle.properties +5 -0
- package/android/src/main/AndroidManifest.xml +3 -0
- package/android/src/main/AndroidManifestNew.xml +2 -0
- package/android/src/main/cpp/include/llama-cpp.h +30 -0
- package/android/src/main/cpp/include/llama.h +1440 -0
- package/android/src/main/java/com/novastera/llamarn/RNLlamaCppPackage.kt +21 -0
- package/android/src/main/jniLibs/arm64-v8a/libOpenCL.so +0 -0
- 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/x86_64/libOpenCL.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/LlamaCppModel.cpp +984 -0
- package/cpp/LlamaCppModel.h +162 -0
- package/cpp/PureCppImpl.cpp +308 -0
- package/cpp/PureCppImpl.h +59 -0
- package/cpp/SystemUtils.cpp +180 -0
- package/cpp/SystemUtils.h +74 -0
- package/cpp/build-info.cpp +4 -0
- package/cpp/llama.cpp/AUTHORS +1106 -0
- package/cpp/llama.cpp/CMakeLists.txt +254 -0
- package/cpp/llama.cpp/CMakePresets.json +84 -0
- package/cpp/llama.cpp/CODEOWNERS +11 -0
- package/cpp/llama.cpp/CONTRIBUTING.md +127 -0
- package/cpp/llama.cpp/LICENSE +21 -0
- package/cpp/llama.cpp/Makefile +1608 -0
- package/cpp/llama.cpp/README.md +575 -0
- package/cpp/llama.cpp/SECURITY.md +68 -0
- package/cpp/llama.cpp/build-xcframework.sh +540 -0
- package/cpp/llama.cpp/cmake/arm64-apple-clang.cmake +16 -0
- package/cpp/llama.cpp/cmake/arm64-windows-llvm.cmake +16 -0
- package/cpp/llama.cpp/cmake/build-info.cmake +64 -0
- package/cpp/llama.cpp/cmake/common.cmake +35 -0
- package/cpp/llama.cpp/cmake/git-vars.cmake +22 -0
- package/cpp/llama.cpp/cmake/llama-config.cmake.in +30 -0
- package/cpp/llama.cpp/cmake/llama.pc.in +10 -0
- package/cpp/llama.cpp/cmake/x64-windows-llvm.cmake +5 -0
- package/cpp/llama.cpp/common/CMakeLists.txt +170 -0
- package/cpp/llama.cpp/common/arg.cpp +3337 -0
- package/cpp/llama.cpp/common/arg.h +89 -0
- package/cpp/llama.cpp/common/base64.hpp +392 -0
- package/cpp/llama.cpp/common/build-info.cpp.in +4 -0
- package/cpp/llama.cpp/common/chat.cpp +1781 -0
- package/cpp/llama.cpp/common/chat.h +135 -0
- package/cpp/llama.cpp/common/cmake/build-info-gen-cpp.cmake +24 -0
- package/cpp/llama.cpp/common/common.cpp +1567 -0
- package/cpp/llama.cpp/common/common.h +668 -0
- package/cpp/llama.cpp/common/console.cpp +504 -0
- package/cpp/llama.cpp/common/console.h +19 -0
- package/cpp/llama.cpp/common/json-schema-to-grammar.cpp +1027 -0
- package/cpp/llama.cpp/common/json-schema-to-grammar.h +21 -0
- package/cpp/llama.cpp/common/json.hpp +24766 -0
- package/cpp/llama.cpp/common/llguidance.cpp +254 -0
- package/cpp/llama.cpp/common/log.cpp +393 -0
- package/cpp/llama.cpp/common/log.h +103 -0
- package/cpp/llama.cpp/common/minja/chat-template.hpp +537 -0
- package/cpp/llama.cpp/common/minja/minja.hpp +2941 -0
- package/cpp/llama.cpp/common/ngram-cache.cpp +286 -0
- package/cpp/llama.cpp/common/ngram-cache.h +101 -0
- package/cpp/llama.cpp/common/sampling.cpp +580 -0
- package/cpp/llama.cpp/common/sampling.h +107 -0
- package/cpp/llama.cpp/common/speculative.cpp +278 -0
- package/cpp/llama.cpp/common/speculative.h +28 -0
- package/cpp/llama.cpp/common/stb_image.h +7988 -0
- package/cpp/llama.cpp/convert_hf_to_gguf.py +6195 -0
- package/cpp/llama.cpp/convert_hf_to_gguf_update.py +393 -0
- package/cpp/llama.cpp/convert_llama_ggml_to_gguf.py +450 -0
- package/cpp/llama.cpp/convert_lora_to_gguf.py +461 -0
- package/cpp/llama.cpp/flake.lock +58 -0
- package/cpp/llama.cpp/flake.nix +185 -0
- package/cpp/llama.cpp/ggml/CMakeLists.txt +388 -0
- package/cpp/llama.cpp/ggml/cmake/GitVars.cmake +22 -0
- package/cpp/llama.cpp/ggml/cmake/common.cmake +26 -0
- package/cpp/llama.cpp/ggml/cmake/ggml-config.cmake.in +152 -0
- package/cpp/llama.cpp/ggml/include/ggml-alloc.h +76 -0
- package/cpp/llama.cpp/ggml/include/ggml-backend.h +354 -0
- package/cpp/llama.cpp/ggml/include/ggml-blas.h +25 -0
- package/cpp/llama.cpp/ggml/include/ggml-cann.h +123 -0
- package/cpp/llama.cpp/ggml/include/ggml-cpp.h +39 -0
- package/cpp/llama.cpp/ggml/include/ggml-cpu.h +143 -0
- package/cpp/llama.cpp/ggml/include/ggml-cuda.h +47 -0
- package/cpp/llama.cpp/ggml/include/ggml-kompute.h +50 -0
- package/cpp/llama.cpp/ggml/include/ggml-metal.h +66 -0
- package/cpp/llama.cpp/ggml/include/ggml-opencl.h +26 -0
- package/cpp/llama.cpp/ggml/include/ggml-opt.h +216 -0
- package/cpp/llama.cpp/ggml/include/ggml-rpc.h +33 -0
- package/cpp/llama.cpp/ggml/include/ggml-sycl.h +49 -0
- package/cpp/llama.cpp/ggml/include/ggml-vulkan.h +29 -0
- package/cpp/llama.cpp/ggml/include/ggml.h +2192 -0
- package/cpp/llama.cpp/ggml/include/gguf.h +202 -0
- package/cpp/llama.cpp/ggml/src/CMakeLists.txt +345 -0
- package/cpp/llama.cpp/ggml/src/ggml-alloc.c +1042 -0
- package/cpp/llama.cpp/ggml/src/ggml-backend-impl.h +255 -0
- package/cpp/llama.cpp/ggml/src/ggml-backend-reg.cpp +586 -0
- package/cpp/llama.cpp/ggml/src/ggml-backend.cpp +2008 -0
- package/cpp/llama.cpp/ggml/src/ggml-blas/CMakeLists.txt +87 -0
- package/cpp/llama.cpp/ggml/src/ggml-blas/ggml-blas.cpp +517 -0
- package/cpp/llama.cpp/ggml/src/ggml-cann/CMakeLists.txt +74 -0
- package/cpp/llama.cpp/ggml/src/ggml-cann/Doxyfile +2579 -0
- package/cpp/llama.cpp/ggml/src/ggml-cann/acl_tensor.cpp +179 -0
- package/cpp/llama.cpp/ggml/src/ggml-cann/acl_tensor.h +258 -0
- package/cpp/llama.cpp/ggml/src/ggml-cann/aclnn_ops.cpp +2589 -0
- package/cpp/llama.cpp/ggml/src/ggml-cann/aclnn_ops.h +1083 -0
- package/cpp/llama.cpp/ggml/src/ggml-cann/common.h +420 -0
- package/cpp/llama.cpp/ggml/src/ggml-cann/ggml-cann.cpp +2554 -0
- package/cpp/llama.cpp/ggml/src/ggml-common.h +1857 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/CMakeLists.txt +495 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/amx/amx.cpp +221 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/amx/amx.h +8 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/amx/common.h +91 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/amx/mmq.cpp +2511 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/amx/mmq.h +10 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/binary-ops.cpp +158 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/binary-ops.h +16 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/cmake/FindSIMD.cmake +100 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/common.h +72 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/cpu-feats-x86.cpp +327 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/ggml-cpu-aarch64.cpp +6431 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/ggml-cpu-aarch64.h +8 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/ggml-cpu-hbm.cpp +55 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/ggml-cpu-hbm.h +8 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/ggml-cpu-impl.h +512 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/ggml-cpu-quants.c +13131 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/ggml-cpu-quants.h +63 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/ggml-cpu-traits.cpp +36 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/ggml-cpu-traits.h +38 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.c +3492 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.cpp +671 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/kleidiai/kernels.cpp +254 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/kleidiai/kernels.h +60 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/kleidiai/kleidiai.cpp +287 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/kleidiai/kleidiai.h +17 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/llamafile/sgemm.cpp +3544 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/llamafile/sgemm.h +14 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/ops.cpp +8796 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/ops.h +110 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/simd-mappings.h +892 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/unary-ops.cpp +186 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/unary-ops.h +28 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/vec.cpp +252 -0
- package/cpp/llama.cpp/ggml/src/ggml-cpu/vec.h +802 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/CMakeLists.txt +184 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/acc.cu +47 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/acc.cuh +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/arange.cu +34 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/arange.cuh +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/argmax.cu +91 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/argmax.cuh +3 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/argsort.cu +104 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/argsort.cuh +3 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/binbcast.cu +363 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/binbcast.cuh +9 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/clamp.cu +45 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/clamp.cuh +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/common.cuh +828 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/concat.cu +221 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/concat.cuh +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/conv-transpose-1d.cu +89 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/conv-transpose-1d.cuh +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/convert.cu +730 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/convert.cuh +26 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/count-equal.cu +64 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/count-equal.cuh +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/cp-async.cuh +57 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/cpy.cu +695 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/cpy.cuh +11 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/cross-entropy-loss.cu +189 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/cross-entropy-loss.cuh +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/dequantize.cuh +103 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/diagmask.cu +40 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/diagmask.cuh +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-common.cuh +873 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-mma-f16.cuh +1269 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-tile-f16.cu +357 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-tile-f16.cuh +3 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-tile-f32.cu +365 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-tile-f32.cuh +3 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-vec-f16.cuh +437 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-vec-f32.cuh +428 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-wmma-f16.cu +634 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-wmma-f16.cuh +3 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn.cu +345 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn.cuh +3 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/getrows.cu +275 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/getrows.cuh +15 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/ggml-cuda.cu +3501 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/gla.cu +93 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/gla.cuh +3 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/im2col.cu +103 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/im2col.cuh +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/mma.cuh +396 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/mmq.cu +322 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/mmq.cuh +3217 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/mmv.cu +336 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/mmv.cuh +12 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/mmvq.cu +595 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/mmvq.cuh +12 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/norm.cu +458 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/norm.cuh +11 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/opt-step-adamw.cu +78 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/opt-step-adamw.cuh +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/out-prod.cu +68 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/out-prod.cuh +3 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/pad.cu +49 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/pad.cuh +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/pool2d.cu +94 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/pool2d.cuh +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/quantize.cu +189 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/quantize.cuh +27 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/rope.cu +456 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/rope.cuh +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/scale.cu +31 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/scale.cuh +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/softmax.cu +283 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/softmax.cuh +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/ssm-conv.cu +148 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/ssm-conv.cuh +3 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/ssm-scan.cu +153 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/ssm-scan.cuh +3 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/sum.cu +45 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/sum.cuh +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/sumrows.cu +39 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/sumrows.cuh +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_1-ncols2_16.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_1-ncols2_8.cu +10 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_1.cu +10 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_2.cu +10 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_4.cu +10 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_16.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_4.cu +10 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_8.cu +10 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_32-ncols2_1.cu +10 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_32-ncols2_2.cu +10 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_16.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_2.cu +10 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_4.cu +10 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_8.cu +10 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_64-ncols2_1.cu +10 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_1.cu +10 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_2.cu +10 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_4.cu +10 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_8.cu +10 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-f16.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q8_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-f16.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q8_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-f16.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q8_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-f16.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q8_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-f16.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q8_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-f16.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q8_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs256-f16-f16.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-f16.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q8_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-f16.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q8_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-f16.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q8_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-f16.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q8_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-f16.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q8_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-f16.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q8_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-f16.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q8_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs256-f16-f16.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-f16.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q8_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/generate_cu_files.py +78 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq1_s.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_s.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_xs.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_xxs.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq3_s.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq3_xxs.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_nl.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_xs.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q2_k.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q3_k.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q4_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q4_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q4_k.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q5_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q5_1.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q5_k.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q6_k.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q8_0.cu +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/tsembd.cu +47 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/tsembd.cuh +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/unary.cu +279 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/unary.cuh +57 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/upscale.cu +51 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/upscale.cuh +5 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/vecdotq.cuh +1135 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/vendors/cuda.h +15 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/vendors/hip.h +243 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/vendors/musa.h +140 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/wkv.cu +199 -0
- package/cpp/llama.cpp/ggml/src/ggml-cuda/wkv.cuh +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-hip/CMakeLists.txt +131 -0
- package/cpp/llama.cpp/ggml/src/ggml-impl.h +601 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/CMakeLists.txt +166 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/ggml-kompute.cpp +2251 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/common.comp +112 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_add.comp +58 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_addrow.comp +25 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_cpy_f16_f16.comp +52 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_cpy_f16_f32.comp +52 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_cpy_f32_f16.comp +52 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_cpy_f32_f32.comp +52 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_diagmask.comp +30 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_gelu.comp +22 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_getrows.comp +17 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_getrows_f16.comp +31 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_getrows_f32.comp +31 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_getrows_q4_0.comp +38 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_getrows_q4_1.comp +39 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_getrows_q6_k.comp +44 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_mul.comp +52 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_f16.comp +69 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_mat_f32.comp +51 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q4_0.comp +33 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q4_1.comp +35 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q4_k.comp +140 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q6_k.comp +106 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q8_0.comp +73 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_mul_mv_q_n.comp +52 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_mul_mv_q_n_pre.comp +28 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_norm.comp +84 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_relu.comp +21 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_rmsnorm.comp +53 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_rope_neox_f16.comp +52 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_rope_neox_f32.comp +52 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_rope_norm_f16.comp +52 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_rope_norm_f32.comp +52 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_scale.comp +19 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_scale_8.comp +23 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_silu.comp +22 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/op_softmax.comp +72 -0
- package/cpp/llama.cpp/ggml/src/ggml-kompute/kompute-shaders/rope_common.comp +71 -0
- package/cpp/llama.cpp/ggml/src/ggml-metal/CMakeLists.txt +120 -0
- package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal-impl.h +618 -0
- package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal.m +5916 -0
- package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal.metal +6891 -0
- package/cpp/llama.cpp/ggml/src/ggml-musa/CMakeLists.txt +107 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/CMakeLists.txt +96 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/ggml-opencl.cpp +4966 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/add.cl +83 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/clamp.cl +20 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/cpy.cl +184 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/cvt.cl +118 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/diag_mask_inf.cl +58 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/embed_kernel.py +26 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/gelu.cl +62 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/gemv_noshuffle.cl +268 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/gemv_noshuffle_general.cl +274 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/get_rows.cl +163 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/im2col_f16.cl +57 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/im2col_f32.cl +57 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul.cl +79 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mat_Ab_Bi_8x4.cl +139 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mv_f16_f16.cl +118 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mv_f16_f32.cl +118 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mv_f16_f32_1row.cl +94 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mv_f16_f32_l4.cl +84 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mv_f32_f32.cl +118 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mv_q4_0_f32.cl +192 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mv_q4_0_f32_1d_16x_flat.cl +307 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mv_q4_0_f32_1d_8x_flat.cl +265 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mv_q4_0_f32_8x_flat.cl +272 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mv_q4_0_f32_v.cl +254 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mv_q6_k.cl +190 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/norm.cl +81 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/relu.cl +16 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/rms_norm.cl +96 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/rope.cl +721 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/scale.cl +16 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/silu.cl +30 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/softmax_4_f16.cl +87 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/softmax_4_f32.cl +87 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/softmax_f16.cl +86 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/softmax_f32.cl +86 -0
- package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/transpose.cl +84 -0
- package/cpp/llama.cpp/ggml/src/ggml-opt.cpp +854 -0
- package/cpp/llama.cpp/ggml/src/ggml-quants.c +5232 -0
- package/cpp/llama.cpp/ggml/src/ggml-quants.h +100 -0
- package/cpp/llama.cpp/ggml/src/ggml-rpc/CMakeLists.txt +9 -0
- package/cpp/llama.cpp/ggml/src/ggml-rpc/ggml-rpc.cpp +1813 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/CMakeLists.txt +183 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/backend.hpp +37 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/binbcast.cpp +350 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/binbcast.hpp +39 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/common.cpp +83 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/common.hpp +493 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/concat.cpp +197 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/concat.hpp +20 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/conv.cpp +100 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/conv.hpp +20 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/convert.cpp +596 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/convert.hpp +34 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/cpy.cpp +701 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/cpy.hpp +11 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/dequantize.hpp +753 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/dmmv.cpp +1154 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/dmmv.hpp +27 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/dpct/helper.hpp +2957 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/element_wise.cpp +1559 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/element_wise.hpp +75 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/gemm.hpp +70 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/getrows.cpp +311 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/getrows.hpp +20 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/ggml-sycl.cpp +4302 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/gla.cpp +105 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/gla.hpp +8 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/im2col.cpp +136 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/im2col.hpp +21 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/mmq.cpp +3030 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/mmq.hpp +33 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/mmvq.cpp +1081 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/mmvq.hpp +27 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/norm.cpp +474 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/norm.hpp +26 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/outprod.cpp +46 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/outprod.hpp +10 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/presets.hpp +74 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/quants.hpp +61 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/rope.cpp +362 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/rope.hpp +20 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/softmax.cpp +264 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/softmax.hpp +20 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/sycl_hw.cpp +13 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/sycl_hw.hpp +23 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/tsembd.cpp +73 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/tsembd.hpp +20 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/vecdotq.hpp +1189 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/wkv.cpp +305 -0
- package/cpp/llama.cpp/ggml/src/ggml-sycl/wkv.hpp +10 -0
- package/cpp/llama.cpp/ggml/src/ggml-threading.cpp +12 -0
- package/cpp/llama.cpp/ggml/src/ggml-threading.h +14 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/CMakeLists.txt +202 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/cmake/host-toolchain.cmake.in +15 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/ggml-vulkan.cpp +10502 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/CMakeLists.txt +22 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/acc.comp +29 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/add.comp +29 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/argmax.comp +51 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/argsort.comp +69 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/clamp.comp +17 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/concat.comp +41 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/contig_copy.comp +49 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/conv2d_dw.comp +105 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/copy.comp +23 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/copy_from_quant.comp +51 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/copy_to_quant.comp +242 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/cos.comp +17 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/count_equal.comp +31 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_f32.comp +20 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_funcs.comp +462 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_funcs_cm2.comp +699 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_head.comp +13 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq1_m.comp +42 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq1_s.comp +35 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq2_s.comp +44 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq2_xs.comp +43 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq2_xxs.comp +48 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq3_s.comp +39 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq3_xxs.comp +49 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq4_nl.comp +32 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq4_xs.comp +34 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q2_k.comp +34 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q3_k.comp +42 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q4_0.comp +30 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q4_1.comp +32 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q4_k.comp +68 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q5_0.comp +34 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q5_1.comp +35 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q5_k.comp +70 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q6_k.comp +33 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q8_0.comp +31 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/diag_mask_inf.comp +34 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/div.comp +27 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn.comp +483 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm2.comp +383 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_split_k_reduce.comp +59 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/gelu.comp +25 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/gelu_quick.comp +23 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/generic_binary_head.comp +64 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/generic_head.comp +9 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/generic_unary_head.comp +76 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/get_rows.comp +33 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/get_rows_quant.comp +41 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/group_norm.comp +66 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/im2col.comp +100 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/l2_norm.comp +41 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/leaky_relu.comp +22 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul.comp +27 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_split_k_reduce.comp +48 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec.comp +169 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_base.comp +118 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq1_m.comp +82 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq1_s.comp +79 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq2_s.comp +90 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq2_xs.comp +87 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq2_xxs.comp +87 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq3_s.comp +90 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq3_xxs.comp +88 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_nc.comp +118 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_p021.comp +154 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q2_k.comp +130 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q3_k.comp +132 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q4_k.comp +136 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q5_k.comp +167 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q6_k.comp +130 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm.comp +868 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm_cm2.comp +441 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq.comp +442 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq_funcs.comp +99 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/norm.comp +44 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/opt_step_adamw.comp +42 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/pad.comp +28 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/pool2d.comp +74 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/quantize_q8_1.comp +77 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/relu.comp +21 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/repeat.comp +26 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/repeat_back.comp +37 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rms_norm.comp +52 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rms_norm_back.comp +55 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rope_head.comp +58 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rope_multi.comp +60 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rope_neox.comp +43 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rope_norm.comp +43 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rope_vision.comp +47 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/scale.comp +24 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/sigmoid.comp +20 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/silu.comp +22 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/silu_back.comp +26 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/sin.comp +17 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/soft_max.comp +173 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/soft_max_back.comp +50 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/square.comp +17 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/sub.comp +29 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/sum_rows.comp +37 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/tanh.comp +20 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/test_bfloat16_support.comp +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/test_coopmat2_support.comp +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/test_coopmat_support.comp +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/test_integer_dot_support.comp +7 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/timestep_embedding.comp +41 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/types.comp +1373 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/upscale.comp +36 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp +740 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/wkv6.comp +87 -0
- package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/wkv7.comp +91 -0
- package/cpp/llama.cpp/ggml/src/ggml.c +6499 -0
- package/cpp/llama.cpp/ggml/src/gguf.cpp +1330 -0
- package/cpp/llama.cpp/gguf-py/LICENSE +21 -0
- package/cpp/llama.cpp/gguf-py/README.md +99 -0
- package/cpp/llama.cpp/gguf-py/examples/reader.py +49 -0
- package/cpp/llama.cpp/gguf-py/examples/writer.py +39 -0
- package/cpp/llama.cpp/gguf-py/gguf/__init__.py +9 -0
- package/cpp/llama.cpp/gguf-py/gguf/constants.py +2296 -0
- package/cpp/llama.cpp/gguf-py/gguf/gguf.py +15 -0
- package/cpp/llama.cpp/gguf-py/gguf/gguf_reader.py +367 -0
- package/cpp/llama.cpp/gguf-py/gguf/gguf_writer.py +1041 -0
- package/cpp/llama.cpp/gguf-py/gguf/lazy.py +223 -0
- package/cpp/llama.cpp/gguf-py/gguf/metadata.py +642 -0
- package/cpp/llama.cpp/gguf-py/gguf/py.typed +0 -0
- package/cpp/llama.cpp/gguf-py/gguf/quants.py +1269 -0
- package/cpp/llama.cpp/gguf-py/gguf/scripts/gguf_convert_endian.py +182 -0
- package/cpp/llama.cpp/gguf-py/gguf/scripts/gguf_dump.py +454 -0
- package/cpp/llama.cpp/gguf-py/gguf/scripts/gguf_editor_gui.py +1610 -0
- package/cpp/llama.cpp/gguf-py/gguf/scripts/gguf_hash.py +102 -0
- package/cpp/llama.cpp/gguf-py/gguf/scripts/gguf_new_metadata.py +207 -0
- package/cpp/llama.cpp/gguf-py/gguf/scripts/gguf_set_metadata.py +95 -0
- package/cpp/llama.cpp/gguf-py/gguf/tensor_mapping.py +1172 -0
- package/cpp/llama.cpp/gguf-py/gguf/utility.py +264 -0
- package/cpp/llama.cpp/gguf-py/gguf/vocab.py +492 -0
- package/cpp/llama.cpp/gguf-py/pyproject.toml +43 -0
- package/cpp/llama.cpp/gguf-py/tests/__init__.py +1 -0
- package/cpp/llama.cpp/gguf-py/tests/test_metadata.py +238 -0
- package/cpp/llama.cpp/gguf-py/tests/test_quants.py +238 -0
- package/cpp/llama.cpp/grammars/README.md +382 -0
- package/cpp/llama.cpp/grammars/arithmetic.gbnf +6 -0
- package/cpp/llama.cpp/grammars/c.gbnf +42 -0
- package/cpp/llama.cpp/grammars/chess.gbnf +13 -0
- package/cpp/llama.cpp/grammars/english.gbnf +6 -0
- package/cpp/llama.cpp/grammars/japanese.gbnf +7 -0
- package/cpp/llama.cpp/grammars/json.gbnf +25 -0
- package/cpp/llama.cpp/grammars/json_arr.gbnf +34 -0
- package/cpp/llama.cpp/grammars/list.gbnf +4 -0
- package/cpp/llama.cpp/include/llama-cpp.h +30 -0
- package/cpp/llama.cpp/include/llama.h +1440 -0
- package/cpp/llama.cpp/licenses/LICENSE-curl +9 -0
- package/cpp/llama.cpp/licenses/LICENSE-httplib +21 -0
- package/cpp/llama.cpp/licenses/LICENSE-jsonhpp +21 -0
- package/cpp/llama.cpp/licenses/LICENSE-linenoise +26 -0
- package/cpp/llama.cpp/media/llama0-banner.png +0 -0
- package/cpp/llama.cpp/media/llama0-logo.png +0 -0
- package/cpp/llama.cpp/media/llama1-banner.png +0 -0
- package/cpp/llama.cpp/media/llama1-logo.png +0 -0
- package/cpp/llama.cpp/media/llama1-logo.svg +34 -0
- package/cpp/llama.cpp/media/matmul.png +0 -0
- package/cpp/llama.cpp/media/matmul.svg +1238 -0
- 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 +112 -0
- package/cpp/llama.cpp/models/ggml-vocab-bert-bge.gguf.out +46 -0
- package/cpp/llama.cpp/models/ggml-vocab-chameleon.gguf.inp +112 -0
- package/cpp/llama.cpp/models/ggml-vocab-chameleon.gguf.out +46 -0
- package/cpp/llama.cpp/models/ggml-vocab-command-r.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-command-r.gguf.inp +112 -0
- package/cpp/llama.cpp/models/ggml-vocab-command-r.gguf.out +46 -0
- package/cpp/llama.cpp/models/ggml-vocab-deepseek-coder.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-deepseek-coder.gguf.inp +112 -0
- package/cpp/llama.cpp/models/ggml-vocab-deepseek-coder.gguf.out +46 -0
- package/cpp/llama.cpp/models/ggml-vocab-deepseek-llm.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-deepseek-llm.gguf.inp +112 -0
- package/cpp/llama.cpp/models/ggml-vocab-deepseek-llm.gguf.out +46 -0
- package/cpp/llama.cpp/models/ggml-vocab-deepseek-r1-qwen.gguf.inp +112 -0
- package/cpp/llama.cpp/models/ggml-vocab-deepseek-r1-qwen.gguf.out +46 -0
- package/cpp/llama.cpp/models/ggml-vocab-falcon.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-falcon.gguf.inp +112 -0
- package/cpp/llama.cpp/models/ggml-vocab-falcon.gguf.out +46 -0
- package/cpp/llama.cpp/models/ggml-vocab-gpt-2.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-gpt-2.gguf.inp +112 -0
- package/cpp/llama.cpp/models/ggml-vocab-gpt-2.gguf.out +46 -0
- package/cpp/llama.cpp/models/ggml-vocab-gpt-4o.gguf.inp +112 -0
- package/cpp/llama.cpp/models/ggml-vocab-gpt-4o.gguf.out +46 -0
- 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 +112 -0
- package/cpp/llama.cpp/models/ggml-vocab-llama-bpe.gguf.out +46 -0
- package/cpp/llama.cpp/models/ggml-vocab-llama-spm.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-llama-spm.gguf.inp +112 -0
- package/cpp/llama.cpp/models/ggml-vocab-llama-spm.gguf.out +46 -0
- package/cpp/llama.cpp/models/ggml-vocab-llama4.gguf.inp +112 -0
- package/cpp/llama.cpp/models/ggml-vocab-llama4.gguf.out +46 -0
- package/cpp/llama.cpp/models/ggml-vocab-mpt.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-mpt.gguf.inp +112 -0
- package/cpp/llama.cpp/models/ggml-vocab-mpt.gguf.out +46 -0
- package/cpp/llama.cpp/models/ggml-vocab-phi-3.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-phi-3.gguf.inp +112 -0
- package/cpp/llama.cpp/models/ggml-vocab-phi-3.gguf.out +46 -0
- package/cpp/llama.cpp/models/ggml-vocab-pixtral.gguf.inp +112 -0
- package/cpp/llama.cpp/models/ggml-vocab-pixtral.gguf.out +46 -0
- package/cpp/llama.cpp/models/ggml-vocab-qwen2.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-qwen2.gguf.inp +112 -0
- package/cpp/llama.cpp/models/ggml-vocab-qwen2.gguf.out +46 -0
- package/cpp/llama.cpp/models/ggml-vocab-refact.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-refact.gguf.inp +112 -0
- package/cpp/llama.cpp/models/ggml-vocab-refact.gguf.out +46 -0
- package/cpp/llama.cpp/models/ggml-vocab-roberta-bpe.gguf.inp +112 -0
- package/cpp/llama.cpp/models/ggml-vocab-roberta-bpe.gguf.out +46 -0
- package/cpp/llama.cpp/models/ggml-vocab-starcoder.gguf +0 -0
- package/cpp/llama.cpp/models/ggml-vocab-starcoder.gguf.inp +112 -0
- package/cpp/llama.cpp/models/ggml-vocab-starcoder.gguf.out +46 -0
- package/cpp/llama.cpp/models/templates/CohereForAI-c4ai-command-r-plus-tool_use.jinja +202 -0
- package/cpp/llama.cpp/models/templates/CohereForAI-c4ai-command-r7b-12-2024-tool_use.jinja +156 -0
- package/cpp/llama.cpp/models/templates/NousResearch-Hermes-2-Pro-Llama-3-8B-tool_use.jinja +152 -0
- package/cpp/llama.cpp/models/templates/NousResearch-Hermes-3-Llama-3.1-8B-tool_use.jinja +152 -0
- package/cpp/llama.cpp/models/templates/Qwen-Qwen2.5-7B-Instruct.jinja +54 -0
- package/cpp/llama.cpp/models/templates/README.md +22 -0
- package/cpp/llama.cpp/models/templates/deepseek-ai-DeepSeek-R1-Distill-Llama-8B.jinja +1 -0
- package/cpp/llama.cpp/models/templates/deepseek-ai-DeepSeek-R1-Distill-Qwen-32B.jinja +1 -0
- package/cpp/llama.cpp/models/templates/fireworks-ai-llama-3-firefunction-v2.jinja +57 -0
- package/cpp/llama.cpp/models/templates/google-gemma-2-2b-it.jinja +4 -0
- package/cpp/llama.cpp/models/templates/llama-cpp-deepseek-r1.jinja +76 -0
- package/cpp/llama.cpp/models/templates/meetkai-functionary-medium-v3.1.jinja +58 -0
- package/cpp/llama.cpp/models/templates/meetkai-functionary-medium-v3.2.jinja +287 -0
- package/cpp/llama.cpp/models/templates/meta-llama-Llama-3.1-8B-Instruct.jinja +109 -0
- package/cpp/llama.cpp/models/templates/meta-llama-Llama-3.2-3B-Instruct.jinja +93 -0
- package/cpp/llama.cpp/models/templates/meta-llama-Llama-3.3-70B-Instruct.jinja +109 -0
- package/cpp/llama.cpp/models/templates/microsoft-Phi-3.5-mini-instruct.jinja +8 -0
- package/cpp/llama.cpp/models/templates/mistralai-Mistral-Nemo-Instruct-2407.jinja +87 -0
- package/cpp/llama.cpp/mypy.ini +7 -0
- package/cpp/llama.cpp/pocs/CMakeLists.txt +14 -0
- package/cpp/llama.cpp/pocs/vdot/CMakeLists.txt +9 -0
- package/cpp/llama.cpp/pocs/vdot/q8dot.cpp +173 -0
- package/cpp/llama.cpp/pocs/vdot/vdot.cpp +311 -0
- package/cpp/llama.cpp/poetry.lock +1197 -0
- package/cpp/llama.cpp/prompts/LLM-questions.txt +49 -0
- package/cpp/llama.cpp/prompts/alpaca.txt +1 -0
- package/cpp/llama.cpp/prompts/assistant.txt +31 -0
- package/cpp/llama.cpp/prompts/chat-with-baichuan.txt +4 -0
- package/cpp/llama.cpp/prompts/chat-with-bob.txt +7 -0
- package/cpp/llama.cpp/prompts/chat-with-qwen.txt +1 -0
- package/cpp/llama.cpp/prompts/chat-with-vicuna-v0.txt +7 -0
- package/cpp/llama.cpp/prompts/chat-with-vicuna-v1.txt +7 -0
- package/cpp/llama.cpp/prompts/chat.txt +28 -0
- package/cpp/llama.cpp/prompts/dan-modified.txt +1 -0
- package/cpp/llama.cpp/prompts/dan.txt +1 -0
- package/cpp/llama.cpp/prompts/mnemonics.txt +93 -0
- package/cpp/llama.cpp/prompts/parallel-questions.txt +43 -0
- package/cpp/llama.cpp/prompts/reason-act.txt +18 -0
- package/cpp/llama.cpp/pyproject.toml +45 -0
- package/cpp/llama.cpp/pyrightconfig.json +22 -0
- package/cpp/llama.cpp/requirements/requirements-all.txt +15 -0
- package/cpp/llama.cpp/requirements/requirements-compare-llama-bench.txt +2 -0
- package/cpp/llama.cpp/requirements/requirements-convert_hf_to_gguf.txt +3 -0
- package/cpp/llama.cpp/requirements/requirements-convert_hf_to_gguf_update.txt +3 -0
- package/cpp/llama.cpp/requirements/requirements-convert_legacy_llama.txt +5 -0
- package/cpp/llama.cpp/requirements/requirements-convert_llama_ggml_to_gguf.txt +1 -0
- package/cpp/llama.cpp/requirements/requirements-convert_lora_to_gguf.txt +2 -0
- package/cpp/llama.cpp/requirements/requirements-gguf_editor_gui.txt +3 -0
- package/cpp/llama.cpp/requirements/requirements-pydantic.txt +3 -0
- package/cpp/llama.cpp/requirements/requirements-test-tokenizer-random.txt +1 -0
- package/cpp/llama.cpp/requirements/requirements-tool_bench.txt +12 -0
- package/cpp/llama.cpp/requirements.txt +13 -0
- package/cpp/llama.cpp/src/CMakeLists.txt +45 -0
- package/cpp/llama.cpp/src/llama-adapter.cpp +388 -0
- package/cpp/llama.cpp/src/llama-adapter.h +76 -0
- package/cpp/llama.cpp/src/llama-arch.cpp +1743 -0
- package/cpp/llama.cpp/src/llama-arch.h +437 -0
- package/cpp/llama.cpp/src/llama-batch.cpp +372 -0
- package/cpp/llama.cpp/src/llama-batch.h +89 -0
- package/cpp/llama.cpp/src/llama-chat.cpp +663 -0
- package/cpp/llama.cpp/src/llama-chat.h +58 -0
- package/cpp/llama.cpp/src/llama-context.cpp +2459 -0
- package/cpp/llama.cpp/src/llama-context.h +246 -0
- package/cpp/llama.cpp/src/llama-cparams.cpp +1 -0
- package/cpp/llama.cpp/src/llama-cparams.h +39 -0
- package/cpp/llama.cpp/src/llama-grammar.cpp +1219 -0
- package/cpp/llama.cpp/src/llama-grammar.h +173 -0
- package/cpp/llama.cpp/src/llama-graph.cpp +1713 -0
- package/cpp/llama.cpp/src/llama-graph.h +595 -0
- package/cpp/llama.cpp/src/llama-hparams.cpp +79 -0
- package/cpp/llama.cpp/src/llama-hparams.h +161 -0
- package/cpp/llama.cpp/src/llama-impl.cpp +167 -0
- package/cpp/llama.cpp/src/llama-impl.h +61 -0
- package/cpp/llama.cpp/src/llama-io.cpp +15 -0
- package/cpp/llama.cpp/src/llama-io.h +35 -0
- package/cpp/llama.cpp/src/llama-kv-cache.cpp +2486 -0
- package/cpp/llama.cpp/src/llama-kv-cache.h +405 -0
- package/cpp/llama.cpp/src/llama-memory.cpp +1 -0
- package/cpp/llama.cpp/src/llama-memory.h +31 -0
- package/cpp/llama.cpp/src/llama-mmap.cpp +600 -0
- package/cpp/llama.cpp/src/llama-mmap.h +68 -0
- package/cpp/llama.cpp/src/llama-model-loader.cpp +1133 -0
- package/cpp/llama.cpp/src/llama-model-loader.h +169 -0
- package/cpp/llama.cpp/src/llama-model.cpp +13453 -0
- package/cpp/llama.cpp/src/llama-model.h +420 -0
- package/cpp/llama.cpp/src/llama-quant.cpp +964 -0
- package/cpp/llama.cpp/src/llama-quant.h +1 -0
- package/cpp/llama.cpp/src/llama-sampling.cpp +2575 -0
- package/cpp/llama.cpp/src/llama-sampling.h +32 -0
- package/cpp/llama.cpp/src/llama-vocab.cpp +3313 -0
- package/cpp/llama.cpp/src/llama-vocab.h +125 -0
- package/cpp/llama.cpp/src/llama.cpp +340 -0
- package/cpp/llama.cpp/src/unicode-data.cpp +7034 -0
- package/cpp/llama.cpp/src/unicode-data.h +20 -0
- package/cpp/llama.cpp/src/unicode.cpp +849 -0
- package/cpp/llama.cpp/src/unicode.h +66 -0
- package/cpp/rn-completion.cpp +431 -0
- package/cpp/rn-llama.hpp +60 -0
- package/cpp/rn-utils.hpp +331 -0
- package/ios/OnLoad.mm +22 -0
- package/ios/generated/RNLlamaCppSpec/RNLlamaCppSpec-generated.mm +64 -0
- package/ios/generated/RNLlamaCppSpec/RNLlamaCppSpec.h +251 -0
- package/ios/generated/RNLlamaCppSpecJSI-generated.cpp +42 -0
- package/ios/generated/RNLlamaCppSpecJSI.h +336 -0
- package/ios/include/chat.h +135 -0
- package/ios/include/common/base64.hpp +392 -0
- package/ios/include/common/json.hpp +24766 -0
- package/ios/include/common/minja/chat-template.hpp +537 -0
- package/ios/include/common/minja/minja.hpp +2941 -0
- package/ios/include/common.h +668 -0
- package/ios/include/json-schema-to-grammar.h +21 -0
- package/ios/include/llama-cpp.h +30 -0
- package/ios/include/llama.h +1440 -0
- package/ios/include/log.h +103 -0
- package/ios/include/ngram-cache.h +101 -0
- package/ios/include/sampling.h +107 -0
- package/ios/include/speculative.h +28 -0
- package/ios/libs/llama.xcframework/Info.plist +135 -0
- package/ios/libs/llama.xcframework/ios-arm64/dSYMs/llama.dSYM/Contents/Info.plist +20 -0
- 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 +4492 -0
- package/ios/libs/llama.xcframework/ios-arm64/llama.framework/Headers/ggml-alloc.h +76 -0
- package/ios/libs/llama.xcframework/ios-arm64/llama.framework/Headers/ggml-backend.h +354 -0
- package/ios/libs/llama.xcframework/ios-arm64/llama.framework/Headers/ggml-blas.h +25 -0
- package/ios/libs/llama.xcframework/ios-arm64/llama.framework/Headers/ggml-cpu.h +143 -0
- package/ios/libs/llama.xcframework/ios-arm64/llama.framework/Headers/ggml-metal.h +66 -0
- package/ios/libs/llama.xcframework/ios-arm64/llama.framework/Headers/ggml.h +2192 -0
- package/ios/libs/llama.xcframework/ios-arm64/llama.framework/Headers/gguf.h +202 -0
- package/ios/libs/llama.xcframework/ios-arm64/llama.framework/Headers/llama.h +1440 -0
- package/ios/libs/llama.xcframework/ios-arm64/llama.framework/Info.plist +36 -0
- package/ios/libs/llama.xcframework/ios-arm64/llama.framework/Modules/module.modulemap +17 -0
- 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/Info.plist +20 -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 +4513 -0
- package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Resources/Relocations/x86_64/llama.yml +3440 -0
- package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/llama.framework/Headers/ggml-alloc.h +76 -0
- package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/llama.framework/Headers/ggml-backend.h +354 -0
- package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/llama.framework/Headers/ggml-blas.h +25 -0
- package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/llama.framework/Headers/ggml-cpu.h +143 -0
- package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/llama.framework/Headers/ggml-metal.h +66 -0
- package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/llama.framework/Headers/ggml.h +2192 -0
- package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/llama.framework/Headers/gguf.h +202 -0
- package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/llama.framework/Headers/llama.h +1440 -0
- package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/llama.framework/Info.plist +36 -0
- package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/llama.framework/Modules/module.modulemap +17 -0
- package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/llama.framework/llama +0 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/dSYMs/llama.dSYM/Contents/Info.plist +20 -0
- 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 +4513 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/dSYMs/llama.dSYM/Contents/Resources/Relocations/x86_64/llama.yml +3442 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/ggml-alloc.h +76 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/ggml-backend.h +354 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/ggml-blas.h +25 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/ggml-cpu.h +143 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/ggml-metal.h +66 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/ggml.h +2192 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/gguf.h +202 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/llama.h +1440 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Modules/module.modulemap +17 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Resources/Info.plist +32 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/ggml-alloc.h +76 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/ggml-backend.h +354 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/ggml-blas.h +25 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/ggml-cpu.h +143 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/ggml-metal.h +66 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/ggml.h +2192 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/gguf.h +202 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/llama.h +1440 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Modules/module.modulemap +17 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Resources/Info.plist +32 -0
- 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 +76 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/ggml-backend.h +354 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/ggml-blas.h +25 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/ggml-cpu.h +143 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/ggml-metal.h +66 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/ggml.h +2192 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/gguf.h +202 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/llama.h +1440 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Modules/module.modulemap +17 -0
- package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Resources/Info.plist +32 -0
- 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 +20 -0
- 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 +4492 -0
- package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/ggml-alloc.h +76 -0
- package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/ggml-backend.h +354 -0
- package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/ggml-blas.h +25 -0
- package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/ggml-cpu.h +143 -0
- package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/ggml-metal.h +66 -0
- package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/ggml.h +2192 -0
- package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/gguf.h +202 -0
- package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/llama.h +1440 -0
- package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Info.plist +35 -0
- package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Modules/module.modulemap +17 -0
- 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 +20 -0
- 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 +4513 -0
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Resources/Relocations/x86_64/llama.yml +3440 -0
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/ggml-alloc.h +76 -0
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/ggml-backend.h +354 -0
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/ggml-blas.h +25 -0
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/ggml-cpu.h +143 -0
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/ggml-metal.h +66 -0
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/ggml.h +2192 -0
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/gguf.h +202 -0
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/llama.h +1440 -0
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Info.plist +35 -0
- package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Modules/module.modulemap +17 -0
- 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 +20 -0
- 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 +4528 -0
- package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/ggml-alloc.h +76 -0
- package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/ggml-backend.h +354 -0
- package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/ggml-blas.h +25 -0
- package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/ggml-cpu.h +143 -0
- package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/ggml-metal.h +66 -0
- package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/ggml.h +2192 -0
- package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/gguf.h +202 -0
- package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/llama.h +1440 -0
- package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Info.plist +32 -0
- package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Modules/module.modulemap +17 -0
- 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 +20 -0
- 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 +4549 -0
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Resources/Relocations/x86_64/llama.yml +3470 -0
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/ggml-alloc.h +76 -0
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/ggml-backend.h +354 -0
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/ggml-blas.h +25 -0
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/ggml-cpu.h +143 -0
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/ggml-metal.h +66 -0
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/ggml.h +2192 -0
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/gguf.h +202 -0
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/llama.h +1440 -0
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Info.plist +32 -0
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Modules/module.modulemap +17 -0
- package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/llama +0 -0
- package/lib/module/NativeRNLlamaCpp.js +35 -0
- package/lib/module/NativeRNLlamaCpp.js.map +1 -0
- package/lib/module/index.js +20 -0
- package/lib/module/index.js.map +1 -0
- package/lib/module/package.json +1 -0
- package/lib/typescript/package.json +1 -0
- package/lib/typescript/src/NativeRNLlamaCpp.d.ts +222 -0
- package/lib/typescript/src/NativeRNLlamaCpp.d.ts.map +1 -0
- package/lib/typescript/src/index.d.ts +5 -0
- package/lib/typescript/src/index.d.ts.map +1 -0
- package/package.json +161 -0
- package/react-native.config.js +15 -0
- package/src/NativeRNLlamaCpp.ts +282 -0
- package/src/index.tsx +54 -0
|
@@ -0,0 +1,1743 @@
|
|
|
1
|
+
#include "llama-arch.h"
|
|
2
|
+
|
|
3
|
+
#include "llama-impl.h"
|
|
4
|
+
|
|
5
|
+
#include <map>
|
|
6
|
+
|
|
7
|
+
static const std::map<llm_arch, const char *> LLM_ARCH_NAMES = {
|
|
8
|
+
{ LLM_ARCH_LLAMA, "llama" },
|
|
9
|
+
{ LLM_ARCH_LLAMA4, "llama4" },
|
|
10
|
+
{ LLM_ARCH_DECI, "deci" },
|
|
11
|
+
{ LLM_ARCH_FALCON, "falcon" },
|
|
12
|
+
{ LLM_ARCH_GROK, "grok" },
|
|
13
|
+
{ LLM_ARCH_GPT2, "gpt2" },
|
|
14
|
+
{ LLM_ARCH_GPTJ, "gptj" },
|
|
15
|
+
{ LLM_ARCH_GPTNEOX, "gptneox" },
|
|
16
|
+
{ LLM_ARCH_MPT, "mpt" },
|
|
17
|
+
{ LLM_ARCH_BAICHUAN, "baichuan" },
|
|
18
|
+
{ LLM_ARCH_STARCODER, "starcoder" },
|
|
19
|
+
{ LLM_ARCH_REFACT, "refact" },
|
|
20
|
+
{ LLM_ARCH_BERT, "bert" },
|
|
21
|
+
{ LLM_ARCH_NOMIC_BERT, "nomic-bert" },
|
|
22
|
+
{ LLM_ARCH_NOMIC_BERT_MOE, "nomic-bert-moe" },
|
|
23
|
+
{ LLM_ARCH_JINA_BERT_V2, "jina-bert-v2" },
|
|
24
|
+
{ LLM_ARCH_BLOOM, "bloom" },
|
|
25
|
+
{ LLM_ARCH_STABLELM, "stablelm" },
|
|
26
|
+
{ LLM_ARCH_QWEN, "qwen" },
|
|
27
|
+
{ LLM_ARCH_QWEN2, "qwen2" },
|
|
28
|
+
{ LLM_ARCH_QWEN2MOE, "qwen2moe" },
|
|
29
|
+
{ LLM_ARCH_QWEN2VL, "qwen2vl" },
|
|
30
|
+
{ LLM_ARCH_QWEN3, "qwen3" },
|
|
31
|
+
{ LLM_ARCH_QWEN3MOE, "qwen3moe" },
|
|
32
|
+
{ LLM_ARCH_PHI2, "phi2" },
|
|
33
|
+
{ LLM_ARCH_PHI3, "phi3" },
|
|
34
|
+
{ LLM_ARCH_PHIMOE, "phimoe" },
|
|
35
|
+
{ LLM_ARCH_PLAMO, "plamo" },
|
|
36
|
+
{ LLM_ARCH_CODESHELL, "codeshell" },
|
|
37
|
+
{ LLM_ARCH_ORION, "orion" },
|
|
38
|
+
{ LLM_ARCH_INTERNLM2, "internlm2" },
|
|
39
|
+
{ LLM_ARCH_MINICPM, "minicpm" },
|
|
40
|
+
{ LLM_ARCH_MINICPM3, "minicpm3" },
|
|
41
|
+
{ LLM_ARCH_GEMMA, "gemma" },
|
|
42
|
+
{ LLM_ARCH_GEMMA2, "gemma2" },
|
|
43
|
+
{ LLM_ARCH_GEMMA3, "gemma3" },
|
|
44
|
+
{ LLM_ARCH_STARCODER2, "starcoder2" },
|
|
45
|
+
{ LLM_ARCH_MAMBA, "mamba" },
|
|
46
|
+
{ LLM_ARCH_XVERSE, "xverse" },
|
|
47
|
+
{ LLM_ARCH_COMMAND_R, "command-r" },
|
|
48
|
+
{ LLM_ARCH_COHERE2, "cohere2" },
|
|
49
|
+
{ LLM_ARCH_DBRX, "dbrx" },
|
|
50
|
+
{ LLM_ARCH_OLMO, "olmo" },
|
|
51
|
+
{ LLM_ARCH_OLMO2, "olmo2" },
|
|
52
|
+
{ LLM_ARCH_OLMOE, "olmoe" },
|
|
53
|
+
{ LLM_ARCH_OPENELM, "openelm" },
|
|
54
|
+
{ LLM_ARCH_ARCTIC, "arctic" },
|
|
55
|
+
{ LLM_ARCH_DEEPSEEK, "deepseek" },
|
|
56
|
+
{ LLM_ARCH_DEEPSEEK2, "deepseek2" },
|
|
57
|
+
{ LLM_ARCH_CHATGLM, "chatglm" },
|
|
58
|
+
{ LLM_ARCH_GLM4, "glm4" },
|
|
59
|
+
{ LLM_ARCH_BITNET, "bitnet" },
|
|
60
|
+
{ LLM_ARCH_T5, "t5" },
|
|
61
|
+
{ LLM_ARCH_T5ENCODER, "t5encoder" },
|
|
62
|
+
{ LLM_ARCH_JAIS, "jais" },
|
|
63
|
+
{ LLM_ARCH_NEMOTRON, "nemotron" },
|
|
64
|
+
{ LLM_ARCH_EXAONE, "exaone" },
|
|
65
|
+
{ LLM_ARCH_RWKV6, "rwkv6" },
|
|
66
|
+
{ LLM_ARCH_RWKV6QWEN2, "rwkv6qwen2" },
|
|
67
|
+
{ LLM_ARCH_RWKV7, "rwkv7" },
|
|
68
|
+
{ LLM_ARCH_ARWKV7, "arwkv7" },
|
|
69
|
+
{ LLM_ARCH_GRANITE, "granite" },
|
|
70
|
+
{ LLM_ARCH_GRANITE_MOE, "granitemoe" },
|
|
71
|
+
{ LLM_ARCH_CHAMELEON, "chameleon" },
|
|
72
|
+
{ LLM_ARCH_WAVTOKENIZER_DEC, "wavtokenizer-dec" },
|
|
73
|
+
{ LLM_ARCH_PLM, "plm" },
|
|
74
|
+
{ LLM_ARCH_BAILINGMOE, "bailingmoe" },
|
|
75
|
+
{ LLM_ARCH_UNKNOWN, "(unknown)" },
|
|
76
|
+
};
|
|
77
|
+
|
|
78
|
+
static const std::map<llm_kv, const char *> LLM_KV_NAMES = {
|
|
79
|
+
{ LLM_KV_GENERAL_TYPE, "general.type" },
|
|
80
|
+
{ LLM_KV_GENERAL_ARCHITECTURE, "general.architecture" },
|
|
81
|
+
{ LLM_KV_GENERAL_QUANTIZATION_VERSION, "general.quantization_version" },
|
|
82
|
+
{ LLM_KV_GENERAL_ALIGNMENT, "general.alignment" },
|
|
83
|
+
{ LLM_KV_GENERAL_FILE_TYPE, "general.file_type" },
|
|
84
|
+
{ LLM_KV_GENERAL_NAME, "general.name" },
|
|
85
|
+
{ LLM_KV_GENERAL_AUTHOR, "general.author" },
|
|
86
|
+
{ LLM_KV_GENERAL_VERSION, "general.version" },
|
|
87
|
+
{ LLM_KV_GENERAL_URL, "general.url" },
|
|
88
|
+
{ LLM_KV_GENERAL_DESCRIPTION, "general.description" },
|
|
89
|
+
{ LLM_KV_GENERAL_LICENSE, "general.license" },
|
|
90
|
+
{ LLM_KV_GENERAL_SOURCE_URL, "general.source.url" },
|
|
91
|
+
{ LLM_KV_GENERAL_SOURCE_HF_REPO, "general.source.huggingface.repository" },
|
|
92
|
+
|
|
93
|
+
{ LLM_KV_VOCAB_SIZE, "%s.vocab_size" },
|
|
94
|
+
{ LLM_KV_CONTEXT_LENGTH, "%s.context_length" },
|
|
95
|
+
{ LLM_KV_EMBEDDING_LENGTH, "%s.embedding_length" },
|
|
96
|
+
{ LLM_KV_FEATURES_LENGTH, "%s.features_length" },
|
|
97
|
+
{ LLM_KV_BLOCK_COUNT, "%s.block_count" },
|
|
98
|
+
{ LLM_KV_LEADING_DENSE_BLOCK_COUNT, "%s.leading_dense_block_count" },
|
|
99
|
+
{ LLM_KV_FEED_FORWARD_LENGTH, "%s.feed_forward_length" },
|
|
100
|
+
{ LLM_KV_EXPERT_FEED_FORWARD_LENGTH, "%s.expert_feed_forward_length" },
|
|
101
|
+
{ LLM_KV_EXPERT_SHARED_FEED_FORWARD_LENGTH, "%s.expert_shared_feed_forward_length" },
|
|
102
|
+
{ LLM_KV_USE_PARALLEL_RESIDUAL, "%s.use_parallel_residual" },
|
|
103
|
+
{ LLM_KV_TENSOR_DATA_LAYOUT, "%s.tensor_data_layout" },
|
|
104
|
+
{ LLM_KV_EXPERT_COUNT, "%s.expert_count" },
|
|
105
|
+
{ LLM_KV_EXPERT_USED_COUNT, "%s.expert_used_count" },
|
|
106
|
+
{ LLM_KV_EXPERT_SHARED_COUNT, "%s.expert_shared_count" },
|
|
107
|
+
{ LLM_KV_EXPERT_WEIGHTS_SCALE, "%s.expert_weights_scale" },
|
|
108
|
+
{ LLM_KV_EXPERT_WEIGHTS_NORM, "%s.expert_weights_norm" },
|
|
109
|
+
{ LLM_KV_EXPERT_GATING_FUNC, "%s.expert_gating_func" },
|
|
110
|
+
{ LLM_KV_MOE_EVERY_N_LAYERS, "%s.moe_every_n_layers" },
|
|
111
|
+
{ LLM_KV_POOLING_TYPE, "%s.pooling_type" },
|
|
112
|
+
{ LLM_KV_LOGIT_SCALE, "%s.logit_scale" },
|
|
113
|
+
{ LLM_KV_DECODER_START_TOKEN_ID, "%s.decoder_start_token_id" },
|
|
114
|
+
{ LLM_KV_ATTN_LOGIT_SOFTCAPPING, "%s.attn_logit_softcapping" },
|
|
115
|
+
{ LLM_KV_FINAL_LOGIT_SOFTCAPPING, "%s.final_logit_softcapping" },
|
|
116
|
+
{ LLM_KV_SWIN_NORM, "%s.swin_norm" },
|
|
117
|
+
{ LLM_KV_RESCALE_EVERY_N_LAYERS, "%s.rescale_every_n_layers" },
|
|
118
|
+
{ LLM_KV_TIME_MIX_EXTRA_DIM, "%s.time_mix_extra_dim" },
|
|
119
|
+
{ LLM_KV_TIME_DECAY_EXTRA_DIM, "%s.time_decay_extra_dim" },
|
|
120
|
+
{ LLM_KV_RESIDUAL_SCALE, "%s.residual_scale" },
|
|
121
|
+
{ LLM_KV_EMBEDDING_SCALE, "%s.embedding_scale" },
|
|
122
|
+
{ LLM_KV_TOKEN_SHIFT_COUNT, "%s.token_shift_count" },
|
|
123
|
+
{ LLM_KV_INTERLEAVE_MOE_LAYER_STEP, "%s.interleave_moe_layer_step" },
|
|
124
|
+
|
|
125
|
+
{ LLM_KV_ATTENTION_HEAD_COUNT, "%s.attention.head_count" },
|
|
126
|
+
{ LLM_KV_ATTENTION_HEAD_COUNT_KV, "%s.attention.head_count_kv" },
|
|
127
|
+
{ LLM_KV_ATTENTION_MAX_ALIBI_BIAS, "%s.attention.max_alibi_bias" },
|
|
128
|
+
{ LLM_KV_ATTENTION_CLAMP_KQV, "%s.attention.clamp_kqv" },
|
|
129
|
+
{ LLM_KV_ATTENTION_KEY_LENGTH, "%s.attention.key_length" },
|
|
130
|
+
{ LLM_KV_ATTENTION_VALUE_LENGTH, "%s.attention.value_length" },
|
|
131
|
+
{ LLM_KV_ATTENTION_LAYERNORM_EPS, "%s.attention.layer_norm_epsilon" },
|
|
132
|
+
{ LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, "%s.attention.layer_norm_rms_epsilon" },
|
|
133
|
+
{ LLM_KV_ATTENTION_GROUPNORM_EPS, "%s.attention.group_norm_epsilon" },
|
|
134
|
+
{ LLM_KV_ATTENTION_GROUPNORM_GROUPS, "%s.attention.group_norm_groups" },
|
|
135
|
+
{ LLM_KV_ATTENTION_CAUSAL, "%s.attention.causal" },
|
|
136
|
+
{ LLM_KV_ATTENTION_Q_LORA_RANK, "%s.attention.q_lora_rank" },
|
|
137
|
+
{ LLM_KV_ATTENTION_KV_LORA_RANK, "%s.attention.kv_lora_rank" },
|
|
138
|
+
{ LLM_KV_ATTENTION_DECAY_LORA_RANK, "%s.attention.decay_lora_rank" },
|
|
139
|
+
{ LLM_KV_ATTENTION_ICLR_LORA_RANK, "%s.attention.iclr_lora_rank" },
|
|
140
|
+
{ LLM_KV_ATTENTION_VALUE_RESIDUAL_MIX_LORA_RANK, "%s.attention.value_residual_mix_lora_rank" },
|
|
141
|
+
{ LLM_KV_ATTENTION_GATE_LORA_RANK, "%s.attention.gate_lora_rank" },
|
|
142
|
+
{ LLM_KV_ATTENTION_RELATIVE_BUCKETS_COUNT, "%s.attention.relative_buckets_count" },
|
|
143
|
+
{ LLM_KV_ATTENTION_SLIDING_WINDOW, "%s.attention.sliding_window" },
|
|
144
|
+
{ LLM_KV_ATTENTION_SCALE, "%s.attention.scale" },
|
|
145
|
+
{ LLM_KV_ATTENTION_KEY_LENGTH_MLA, "%s.attention.key_length_mla" },
|
|
146
|
+
{ LLM_KV_ATTENTION_VALUE_LENGTH_MLA, "%s.attention.value_length_mla" },
|
|
147
|
+
|
|
148
|
+
{ LLM_KV_ROPE_DIMENSION_COUNT, "%s.rope.dimension_count" },
|
|
149
|
+
{ LLM_KV_ROPE_DIMENSION_SECTIONS, "%s.rope.dimension_sections" },
|
|
150
|
+
{ LLM_KV_ROPE_FREQ_BASE, "%s.rope.freq_base" },
|
|
151
|
+
{ LLM_KV_ROPE_SCALE_LINEAR, "%s.rope.scale_linear" },
|
|
152
|
+
{ LLM_KV_ROPE_SCALING_TYPE, "%s.rope.scaling.type" },
|
|
153
|
+
{ LLM_KV_ROPE_SCALING_FACTOR, "%s.rope.scaling.factor" },
|
|
154
|
+
{ LLM_KV_ROPE_SCALING_ATTN_FACTOR, "%s.rope.scaling.attn_factor" },
|
|
155
|
+
{ LLM_KV_ROPE_SCALING_ORIG_CTX_LEN, "%s.rope.scaling.original_context_length" },
|
|
156
|
+
{ LLM_KV_ROPE_SCALING_FINETUNED, "%s.rope.scaling.finetuned" },
|
|
157
|
+
{ LLM_KV_ROPE_SCALING_YARN_LOG_MUL, "%s.rope.scaling.yarn_log_multiplier" },
|
|
158
|
+
|
|
159
|
+
{ LLM_KV_SPLIT_NO, "split.no" },
|
|
160
|
+
{ LLM_KV_SPLIT_COUNT, "split.count" },
|
|
161
|
+
{ LLM_KV_SPLIT_TENSORS_COUNT, "split.tensors.count" },
|
|
162
|
+
|
|
163
|
+
{ LLM_KV_SSM_CONV_KERNEL, "%s.ssm.conv_kernel" },
|
|
164
|
+
{ LLM_KV_SSM_INNER_SIZE, "%s.ssm.inner_size" },
|
|
165
|
+
{ LLM_KV_SSM_STATE_SIZE, "%s.ssm.state_size" },
|
|
166
|
+
{ LLM_KV_SSM_TIME_STEP_RANK, "%s.ssm.time_step_rank" },
|
|
167
|
+
{ LLM_KV_SSM_DT_B_C_RMS, "%s.ssm.dt_b_c_rms" },
|
|
168
|
+
|
|
169
|
+
{ LLM_KV_WKV_HEAD_SIZE, "%s.wkv.head_size" },
|
|
170
|
+
|
|
171
|
+
{ LLM_KV_POSNET_EMBEDDING_LENGTH, "%s.posnet.embedding_length" },
|
|
172
|
+
{ LLM_KV_POSNET_BLOCK_COUNT, "%s.posnet.block_count" },
|
|
173
|
+
|
|
174
|
+
{ LLM_KV_CONVNEXT_EMBEDDING_LENGTH, "%s.convnext.embedding_length" },
|
|
175
|
+
{ LLM_KV_CONVNEXT_BLOCK_COUNT, "%s.convnext.block_count" },
|
|
176
|
+
|
|
177
|
+
{ LLM_KV_TOKENIZER_MODEL, "tokenizer.ggml.model" },
|
|
178
|
+
{ LLM_KV_TOKENIZER_PRE, "tokenizer.ggml.pre" },
|
|
179
|
+
{ LLM_KV_TOKENIZER_LIST, "tokenizer.ggml.tokens" },
|
|
180
|
+
{ LLM_KV_TOKENIZER_TOKEN_TYPE, "tokenizer.ggml.token_type" },
|
|
181
|
+
{ LLM_KV_TOKENIZER_TOKEN_TYPE_COUNT, "tokenizer.ggml.token_type_count" },
|
|
182
|
+
{ LLM_KV_TOKENIZER_SCORES, "tokenizer.ggml.scores" },
|
|
183
|
+
{ LLM_KV_TOKENIZER_MERGES, "tokenizer.ggml.merges" },
|
|
184
|
+
{ LLM_KV_TOKENIZER_BOS_ID, "tokenizer.ggml.bos_token_id" },
|
|
185
|
+
{ LLM_KV_TOKENIZER_EOS_ID, "tokenizer.ggml.eos_token_id" },
|
|
186
|
+
{ LLM_KV_TOKENIZER_EOT_ID, "tokenizer.ggml.eot_token_id" },
|
|
187
|
+
{ LLM_KV_TOKENIZER_EOM_ID, "tokenizer.ggml.eom_token_id" },
|
|
188
|
+
{ LLM_KV_TOKENIZER_UNK_ID, "tokenizer.ggml.unknown_token_id" },
|
|
189
|
+
{ LLM_KV_TOKENIZER_SEP_ID, "tokenizer.ggml.seperator_token_id" },
|
|
190
|
+
{ LLM_KV_TOKENIZER_PAD_ID, "tokenizer.ggml.padding_token_id" },
|
|
191
|
+
{ LLM_KV_TOKENIZER_CLS_ID, "tokenizer.ggml.cls_token_id" },
|
|
192
|
+
{ LLM_KV_TOKENIZER_MASK_ID, "tokenizer.ggml.mask_token_id" },
|
|
193
|
+
{ LLM_KV_TOKENIZER_ADD_BOS, "tokenizer.ggml.add_bos_token" },
|
|
194
|
+
{ LLM_KV_TOKENIZER_ADD_EOS, "tokenizer.ggml.add_eos_token" },
|
|
195
|
+
{ LLM_KV_TOKENIZER_ADD_PREFIX, "tokenizer.ggml.add_space_prefix" },
|
|
196
|
+
{ LLM_KV_TOKENIZER_REMOVE_EXTRA_WS, "tokenizer.ggml.remove_extra_whitespaces" },
|
|
197
|
+
{ LLM_KV_TOKENIZER_PRECOMPILED_CHARSMAP, "tokenizer.ggml.precompiled_charsmap" },
|
|
198
|
+
{ LLM_KV_TOKENIZER_HF_JSON, "tokenizer.huggingface.json" },
|
|
199
|
+
{ LLM_KV_TOKENIZER_RWKV, "tokenizer.rwkv.world" },
|
|
200
|
+
{ LLM_KV_TOKENIZER_CHAT_TEMPLATE, "tokenizer.chat_template" },
|
|
201
|
+
{ LLM_KV_TOKENIZER_CHAT_TEMPLATE_N, "tokenizer.chat_template.%s" },
|
|
202
|
+
{ LLM_KV_TOKENIZER_FIM_PRE_ID, "tokenizer.ggml.fim_pre_token_id" },
|
|
203
|
+
{ LLM_KV_TOKENIZER_FIM_SUF_ID, "tokenizer.ggml.fim_suf_token_id" },
|
|
204
|
+
{ LLM_KV_TOKENIZER_FIM_MID_ID, "tokenizer.ggml.fim_mid_token_id" },
|
|
205
|
+
{ LLM_KV_TOKENIZER_FIM_PAD_ID, "tokenizer.ggml.fim_pad_token_id" },
|
|
206
|
+
{ LLM_KV_TOKENIZER_FIM_REP_ID, "tokenizer.ggml.fim_rep_token_id" },
|
|
207
|
+
{ LLM_KV_TOKENIZER_FIM_SEP_ID, "tokenizer.ggml.fim_sep_token_id" },
|
|
208
|
+
|
|
209
|
+
{ LLM_KV_ADAPTER_TYPE, "adapter.type" },
|
|
210
|
+
{ LLM_KV_ADAPTER_LORA_ALPHA, "adapter.lora.alpha" },
|
|
211
|
+
|
|
212
|
+
// deprecated
|
|
213
|
+
{ LLM_KV_TOKENIZER_PREFIX_ID, "tokenizer.ggml.prefix_token_id" },
|
|
214
|
+
{ LLM_KV_TOKENIZER_SUFFIX_ID, "tokenizer.ggml.suffix_token_id" },
|
|
215
|
+
{ LLM_KV_TOKENIZER_MIDDLE_ID, "tokenizer.ggml.middle_token_id" },
|
|
216
|
+
};
|
|
217
|
+
|
|
218
|
+
static const std::map<llm_arch, std::map<llm_tensor, const char *>> LLM_TENSOR_NAMES = {
|
|
219
|
+
{
|
|
220
|
+
LLM_ARCH_LLAMA,
|
|
221
|
+
{
|
|
222
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
223
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
224
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
225
|
+
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
|
|
226
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
227
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
228
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
229
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
230
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
231
|
+
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd" },
|
|
232
|
+
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
|
|
233
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
234
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
235
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
236
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
237
|
+
{ LLM_TENSOR_FFN_GATE_EXP, "blk.%d.ffn_gate.%d" },
|
|
238
|
+
{ LLM_TENSOR_FFN_DOWN_EXP, "blk.%d.ffn_down.%d" },
|
|
239
|
+
{ LLM_TENSOR_FFN_UP_EXP, "blk.%d.ffn_up.%d" },
|
|
240
|
+
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
|
|
241
|
+
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
|
|
242
|
+
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
|
|
243
|
+
},
|
|
244
|
+
},
|
|
245
|
+
{
|
|
246
|
+
LLM_ARCH_LLAMA4,
|
|
247
|
+
{
|
|
248
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
249
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
250
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
251
|
+
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
|
|
252
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
253
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
254
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
255
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
256
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
257
|
+
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd" },
|
|
258
|
+
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
|
|
259
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
260
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
261
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
262
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
263
|
+
{ LLM_TENSOR_FFN_GATE_EXP, "blk.%d.ffn_gate.%d" },
|
|
264
|
+
{ LLM_TENSOR_FFN_DOWN_EXP, "blk.%d.ffn_down.%d" },
|
|
265
|
+
{ LLM_TENSOR_FFN_UP_EXP, "blk.%d.ffn_up.%d" },
|
|
266
|
+
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
|
|
267
|
+
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
|
|
268
|
+
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
|
|
269
|
+
{ LLM_TENSOR_FFN_GATE_SHEXP, "blk.%d.ffn_gate_shexp" },
|
|
270
|
+
{ LLM_TENSOR_FFN_DOWN_SHEXP, "blk.%d.ffn_down_shexp" },
|
|
271
|
+
{ LLM_TENSOR_FFN_UP_SHEXP, "blk.%d.ffn_up_shexp" },
|
|
272
|
+
},
|
|
273
|
+
},
|
|
274
|
+
{
|
|
275
|
+
LLM_ARCH_DECI,
|
|
276
|
+
{
|
|
277
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
278
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
279
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
280
|
+
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
|
|
281
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
282
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
283
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
284
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
285
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
286
|
+
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd" },
|
|
287
|
+
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
|
|
288
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
289
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
290
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
291
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
292
|
+
{ LLM_TENSOR_FFN_GATE_EXP, "blk.%d.ffn_gate.%d" },
|
|
293
|
+
{ LLM_TENSOR_FFN_DOWN_EXP, "blk.%d.ffn_down.%d" },
|
|
294
|
+
{ LLM_TENSOR_FFN_UP_EXP, "blk.%d.ffn_up.%d" },
|
|
295
|
+
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
|
|
296
|
+
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
|
|
297
|
+
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
|
|
298
|
+
},
|
|
299
|
+
},
|
|
300
|
+
{
|
|
301
|
+
LLM_ARCH_BAICHUAN,
|
|
302
|
+
{
|
|
303
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
304
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
305
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
306
|
+
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
|
|
307
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
308
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
309
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
310
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
311
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
312
|
+
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd" },
|
|
313
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
314
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
315
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
316
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
317
|
+
},
|
|
318
|
+
},
|
|
319
|
+
{
|
|
320
|
+
LLM_ARCH_FALCON,
|
|
321
|
+
{
|
|
322
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
323
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
324
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
325
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
326
|
+
{ LLM_TENSOR_ATTN_NORM_2, "blk.%d.attn_norm_2" },
|
|
327
|
+
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
|
|
328
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
329
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
330
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
331
|
+
},
|
|
332
|
+
},
|
|
333
|
+
{
|
|
334
|
+
LLM_ARCH_GROK,
|
|
335
|
+
{
|
|
336
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
337
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
338
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
339
|
+
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
|
|
340
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
341
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
342
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
343
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
344
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
345
|
+
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd" },
|
|
346
|
+
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
|
|
347
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
348
|
+
{ LLM_TENSOR_FFN_GATE_EXP, "blk.%d.ffn_gate.%d" },
|
|
349
|
+
{ LLM_TENSOR_FFN_DOWN_EXP, "blk.%d.ffn_down.%d" },
|
|
350
|
+
{ LLM_TENSOR_FFN_UP_EXP, "blk.%d.ffn_up.%d" },
|
|
351
|
+
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
|
|
352
|
+
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
|
|
353
|
+
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
|
|
354
|
+
{ LLM_TENSOR_LAYER_OUT_NORM, "blk.%d.layer_output_norm" },
|
|
355
|
+
{ LLM_TENSOR_ATTN_OUT_NORM, "blk.%d.attn_output_norm" },
|
|
356
|
+
},
|
|
357
|
+
},
|
|
358
|
+
{
|
|
359
|
+
LLM_ARCH_GPT2,
|
|
360
|
+
{
|
|
361
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
362
|
+
{ LLM_TENSOR_POS_EMBD, "position_embd" },
|
|
363
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
364
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
365
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
366
|
+
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
|
|
367
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
368
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
369
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
370
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
371
|
+
},
|
|
372
|
+
},
|
|
373
|
+
{
|
|
374
|
+
LLM_ARCH_GPTJ,
|
|
375
|
+
{
|
|
376
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
377
|
+
},
|
|
378
|
+
},
|
|
379
|
+
{
|
|
380
|
+
LLM_ARCH_GPTNEOX,
|
|
381
|
+
{
|
|
382
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
383
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
384
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
385
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
386
|
+
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
|
|
387
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
388
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
389
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
390
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
391
|
+
},
|
|
392
|
+
},
|
|
393
|
+
{
|
|
394
|
+
LLM_ARCH_MPT,
|
|
395
|
+
{
|
|
396
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
397
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
398
|
+
{ LLM_TENSOR_OUTPUT, "output"},
|
|
399
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
400
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
401
|
+
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
|
|
402
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
403
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
404
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
405
|
+
{ LLM_TENSOR_FFN_ACT, "blk.%d.ffn.act" },
|
|
406
|
+
{ LLM_TENSOR_POS_EMBD, "position_embd" },
|
|
407
|
+
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm"},
|
|
408
|
+
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm"},
|
|
409
|
+
},
|
|
410
|
+
},
|
|
411
|
+
{
|
|
412
|
+
LLM_ARCH_STARCODER,
|
|
413
|
+
{
|
|
414
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
415
|
+
{ LLM_TENSOR_POS_EMBD, "position_embd" },
|
|
416
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
417
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
418
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
419
|
+
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
|
|
420
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
421
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
422
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
423
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
424
|
+
},
|
|
425
|
+
},
|
|
426
|
+
{
|
|
427
|
+
LLM_ARCH_REFACT,
|
|
428
|
+
{
|
|
429
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
430
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
431
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
432
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
433
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
434
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
435
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
436
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
437
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
438
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
439
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
440
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
441
|
+
},
|
|
442
|
+
},
|
|
443
|
+
{
|
|
444
|
+
LLM_ARCH_BERT,
|
|
445
|
+
{
|
|
446
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
447
|
+
{ LLM_TENSOR_TOKEN_EMBD_NORM, "token_embd_norm" },
|
|
448
|
+
{ LLM_TENSOR_TOKEN_TYPES, "token_types" },
|
|
449
|
+
{ LLM_TENSOR_POS_EMBD, "position_embd" },
|
|
450
|
+
{ LLM_TENSOR_ATTN_OUT_NORM, "blk.%d.attn_output_norm" },
|
|
451
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
452
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
453
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
454
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
455
|
+
{ LLM_TENSOR_LAYER_OUT_NORM, "blk.%d.layer_output_norm" },
|
|
456
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
457
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
458
|
+
{ LLM_TENSOR_CLS, "cls" },
|
|
459
|
+
{ LLM_TENSOR_CLS_OUT, "cls.output" },
|
|
460
|
+
},
|
|
461
|
+
},
|
|
462
|
+
{
|
|
463
|
+
LLM_ARCH_NOMIC_BERT,
|
|
464
|
+
{
|
|
465
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
466
|
+
{ LLM_TENSOR_TOKEN_EMBD_NORM, "token_embd_norm" },
|
|
467
|
+
{ LLM_TENSOR_TOKEN_TYPES, "token_types" },
|
|
468
|
+
{ LLM_TENSOR_ATTN_OUT_NORM, "blk.%d.attn_output_norm" },
|
|
469
|
+
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
|
|
470
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
471
|
+
{ LLM_TENSOR_LAYER_OUT_NORM, "blk.%d.layer_output_norm" },
|
|
472
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
473
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
474
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
475
|
+
},
|
|
476
|
+
},
|
|
477
|
+
{
|
|
478
|
+
LLM_ARCH_NOMIC_BERT_MOE,
|
|
479
|
+
{
|
|
480
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
481
|
+
{ LLM_TENSOR_TOKEN_EMBD_NORM, "token_embd_norm" },
|
|
482
|
+
{ LLM_TENSOR_TOKEN_TYPES, "token_types" },
|
|
483
|
+
{ LLM_TENSOR_ATTN_OUT_NORM, "blk.%d.attn_output_norm" },
|
|
484
|
+
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
|
|
485
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
486
|
+
{ LLM_TENSOR_LAYER_OUT_NORM, "blk.%d.layer_output_norm" },
|
|
487
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
488
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
489
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
490
|
+
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
|
|
491
|
+
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
|
|
492
|
+
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
|
|
493
|
+
},
|
|
494
|
+
},
|
|
495
|
+
{
|
|
496
|
+
LLM_ARCH_JINA_BERT_V2,
|
|
497
|
+
{
|
|
498
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
499
|
+
{ LLM_TENSOR_TOKEN_EMBD_NORM, "token_embd_norm" },
|
|
500
|
+
{ LLM_TENSOR_TOKEN_TYPES, "token_types" },
|
|
501
|
+
{ LLM_TENSOR_ATTN_NORM_2, "blk.%d.attn_norm_2" },
|
|
502
|
+
{ LLM_TENSOR_ATTN_OUT_NORM, "blk.%d.attn_output_norm" },
|
|
503
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
504
|
+
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
|
|
505
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
506
|
+
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
|
|
507
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
508
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
509
|
+
{ LLM_TENSOR_LAYER_OUT_NORM, "blk.%d.layer_output_norm" },
|
|
510
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
511
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
512
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
513
|
+
{ LLM_TENSOR_CLS, "cls" },
|
|
514
|
+
},
|
|
515
|
+
},
|
|
516
|
+
{
|
|
517
|
+
LLM_ARCH_BLOOM,
|
|
518
|
+
{
|
|
519
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
520
|
+
{ LLM_TENSOR_TOKEN_EMBD_NORM, "token_embd_norm" },
|
|
521
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
522
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
523
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
524
|
+
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
|
|
525
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
526
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
527
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
528
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
529
|
+
},
|
|
530
|
+
},
|
|
531
|
+
{
|
|
532
|
+
LLM_ARCH_STABLELM,
|
|
533
|
+
{
|
|
534
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
535
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
536
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
537
|
+
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
|
|
538
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
539
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
540
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
541
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
542
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
543
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
544
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
545
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
546
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
547
|
+
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
|
|
548
|
+
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
|
|
549
|
+
},
|
|
550
|
+
},
|
|
551
|
+
{
|
|
552
|
+
LLM_ARCH_QWEN,
|
|
553
|
+
{
|
|
554
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
555
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
556
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
557
|
+
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
|
|
558
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
559
|
+
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
|
|
560
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
561
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
562
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
563
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
564
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
565
|
+
},
|
|
566
|
+
},
|
|
567
|
+
{
|
|
568
|
+
LLM_ARCH_QWEN2,
|
|
569
|
+
{
|
|
570
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
571
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
572
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
573
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
574
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
575
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
576
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
577
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
578
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
579
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
580
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
581
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
582
|
+
},
|
|
583
|
+
},
|
|
584
|
+
{
|
|
585
|
+
LLM_ARCH_QWEN2VL,
|
|
586
|
+
{
|
|
587
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
588
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
589
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
590
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
591
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
592
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
593
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
594
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
595
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
596
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
597
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
598
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
599
|
+
},
|
|
600
|
+
},
|
|
601
|
+
{
|
|
602
|
+
LLM_ARCH_QWEN2MOE,
|
|
603
|
+
{
|
|
604
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
605
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
606
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
607
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
608
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
609
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
610
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
611
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
612
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
613
|
+
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
|
|
614
|
+
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
|
|
615
|
+
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
|
|
616
|
+
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
|
|
617
|
+
{ LLM_TENSOR_FFN_GATE_INP_SHEXP, "blk.%d.ffn_gate_inp_shexp" },
|
|
618
|
+
{ LLM_TENSOR_FFN_GATE_SHEXP, "blk.%d.ffn_gate_shexp" },
|
|
619
|
+
{ LLM_TENSOR_FFN_DOWN_SHEXP, "blk.%d.ffn_down_shexp" },
|
|
620
|
+
{ LLM_TENSOR_FFN_UP_SHEXP, "blk.%d.ffn_up_shexp" },
|
|
621
|
+
},
|
|
622
|
+
},
|
|
623
|
+
{
|
|
624
|
+
LLM_ARCH_QWEN3,
|
|
625
|
+
{
|
|
626
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
627
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
628
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
629
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
630
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
631
|
+
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
|
|
632
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
633
|
+
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
|
|
634
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
635
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
636
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
637
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
638
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
639
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
640
|
+
},
|
|
641
|
+
},
|
|
642
|
+
{
|
|
643
|
+
LLM_ARCH_QWEN3MOE,
|
|
644
|
+
{
|
|
645
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
646
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
647
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
648
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
649
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
650
|
+
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
|
|
651
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
652
|
+
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
|
|
653
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
654
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
655
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
656
|
+
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
|
|
657
|
+
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
|
|
658
|
+
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
|
|
659
|
+
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
|
|
660
|
+
},
|
|
661
|
+
},
|
|
662
|
+
{
|
|
663
|
+
LLM_ARCH_PHI2,
|
|
664
|
+
{
|
|
665
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
666
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
667
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
668
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
669
|
+
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
|
|
670
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
671
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
672
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
673
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
674
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
675
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
676
|
+
},
|
|
677
|
+
},
|
|
678
|
+
{
|
|
679
|
+
LLM_ARCH_PHI3,
|
|
680
|
+
{
|
|
681
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
682
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
683
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
684
|
+
{ LLM_TENSOR_ROPE_FACTORS_LONG, "rope_factors_long" },
|
|
685
|
+
{ LLM_TENSOR_ROPE_FACTORS_SHORT, "rope_factors_short" },
|
|
686
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
687
|
+
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
|
|
688
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
689
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
690
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
691
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
692
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
693
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
694
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
695
|
+
},
|
|
696
|
+
},
|
|
697
|
+
{
|
|
698
|
+
LLM_ARCH_PHIMOE,
|
|
699
|
+
{
|
|
700
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
701
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
702
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
703
|
+
{ LLM_TENSOR_ROPE_FACTORS_LONG, "rope_factors_long" },
|
|
704
|
+
{ LLM_TENSOR_ROPE_FACTORS_SHORT, "rope_factors_short" },
|
|
705
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
706
|
+
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
|
|
707
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
708
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
709
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
710
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
711
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
712
|
+
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
|
|
713
|
+
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
|
|
714
|
+
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
|
|
715
|
+
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
|
|
716
|
+
},
|
|
717
|
+
},
|
|
718
|
+
{
|
|
719
|
+
LLM_ARCH_PLAMO,
|
|
720
|
+
{
|
|
721
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
722
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
723
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
724
|
+
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
|
|
725
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
726
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
727
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
728
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
729
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
730
|
+
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd" },
|
|
731
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
732
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
733
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
734
|
+
},
|
|
735
|
+
},
|
|
736
|
+
{
|
|
737
|
+
LLM_ARCH_CODESHELL,
|
|
738
|
+
{
|
|
739
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
740
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
741
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
742
|
+
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
|
|
743
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
744
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
745
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
746
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
747
|
+
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
|
|
748
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
749
|
+
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd" },
|
|
750
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
751
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
752
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
753
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
754
|
+
},
|
|
755
|
+
},
|
|
756
|
+
{
|
|
757
|
+
LLM_ARCH_ORION,
|
|
758
|
+
{
|
|
759
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
760
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
761
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
762
|
+
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
|
|
763
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
764
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
765
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
766
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
767
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
768
|
+
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd" },
|
|
769
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
770
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
771
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
772
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
773
|
+
},
|
|
774
|
+
},
|
|
775
|
+
{
|
|
776
|
+
LLM_ARCH_INTERNLM2,
|
|
777
|
+
{
|
|
778
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
779
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
780
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
781
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
782
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
783
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
784
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
785
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
786
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
787
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
788
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
789
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
790
|
+
},
|
|
791
|
+
},
|
|
792
|
+
{
|
|
793
|
+
LLM_ARCH_MINICPM,
|
|
794
|
+
{
|
|
795
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
796
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
797
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
798
|
+
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
|
|
799
|
+
{ LLM_TENSOR_ROPE_FACTORS_LONG, "rope_factors_long" },
|
|
800
|
+
{ LLM_TENSOR_ROPE_FACTORS_SHORT, "rope_factors_short" },
|
|
801
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
802
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
803
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
804
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
805
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
806
|
+
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd" },
|
|
807
|
+
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
|
|
808
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
809
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
810
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
811
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
812
|
+
{ LLM_TENSOR_FFN_GATE_EXP, "blk.%d.ffn_gate.%d" },
|
|
813
|
+
{ LLM_TENSOR_FFN_DOWN_EXP, "blk.%d.ffn_down.%d" },
|
|
814
|
+
{ LLM_TENSOR_FFN_UP_EXP, "blk.%d.ffn_up.%d" },
|
|
815
|
+
},
|
|
816
|
+
},
|
|
817
|
+
{
|
|
818
|
+
LLM_ARCH_MINICPM3,
|
|
819
|
+
{
|
|
820
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
821
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
822
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
823
|
+
{ LLM_TENSOR_ROPE_FACTORS_LONG, "rope_factors_long" },
|
|
824
|
+
{ LLM_TENSOR_ROPE_FACTORS_SHORT, "rope_factors_short" },
|
|
825
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
826
|
+
{ LLM_TENSOR_ATTN_Q_A_NORM, "blk.%d.attn_q_a_norm" },
|
|
827
|
+
{ LLM_TENSOR_ATTN_KV_A_NORM, "blk.%d.attn_kv_a_norm" },
|
|
828
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
829
|
+
{ LLM_TENSOR_ATTN_Q_A, "blk.%d.attn_q_a" },
|
|
830
|
+
{ LLM_TENSOR_ATTN_Q_B, "blk.%d.attn_q_b" },
|
|
831
|
+
{ LLM_TENSOR_ATTN_KV_A_MQA, "blk.%d.attn_kv_a_mqa" },
|
|
832
|
+
{ LLM_TENSOR_ATTN_KV_B, "blk.%d.attn_kv_b" },
|
|
833
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
834
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
835
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
836
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
837
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
838
|
+
},
|
|
839
|
+
},
|
|
840
|
+
{
|
|
841
|
+
LLM_ARCH_GEMMA,
|
|
842
|
+
{
|
|
843
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
844
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
845
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
846
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
847
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
848
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
849
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
850
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
851
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
852
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
853
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
854
|
+
},
|
|
855
|
+
},
|
|
856
|
+
{
|
|
857
|
+
LLM_ARCH_GEMMA2,
|
|
858
|
+
{
|
|
859
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
860
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
861
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
862
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
863
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
864
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
865
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
866
|
+
{ LLM_TENSOR_ATTN_POST_NORM, "blk.%d.post_attention_norm" },
|
|
867
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
868
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
869
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
870
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
871
|
+
{ LLM_TENSOR_FFN_POST_NORM, "blk.%d.post_ffw_norm" },
|
|
872
|
+
},
|
|
873
|
+
},
|
|
874
|
+
{
|
|
875
|
+
LLM_ARCH_GEMMA3,
|
|
876
|
+
{
|
|
877
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
878
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
879
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
880
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
881
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
882
|
+
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
|
|
883
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
884
|
+
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
|
|
885
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
886
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
887
|
+
{ LLM_TENSOR_ATTN_POST_NORM, "blk.%d.post_attention_norm" },
|
|
888
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
889
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
890
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
891
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
892
|
+
{ LLM_TENSOR_FFN_POST_NORM, "blk.%d.post_ffw_norm" },
|
|
893
|
+
},
|
|
894
|
+
},
|
|
895
|
+
{
|
|
896
|
+
LLM_ARCH_STARCODER2,
|
|
897
|
+
{
|
|
898
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
899
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
900
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
901
|
+
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
|
|
902
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
903
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
904
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
905
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
906
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
907
|
+
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd" },
|
|
908
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
909
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
910
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
911
|
+
},
|
|
912
|
+
},
|
|
913
|
+
{
|
|
914
|
+
LLM_ARCH_MAMBA,
|
|
915
|
+
{
|
|
916
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
917
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
918
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
919
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
920
|
+
{ LLM_TENSOR_SSM_IN, "blk.%d.ssm_in" },
|
|
921
|
+
{ LLM_TENSOR_SSM_CONV1D, "blk.%d.ssm_conv1d" },
|
|
922
|
+
{ LLM_TENSOR_SSM_X, "blk.%d.ssm_x" },
|
|
923
|
+
{ LLM_TENSOR_SSM_DT, "blk.%d.ssm_dt" },
|
|
924
|
+
{ LLM_TENSOR_SSM_A, "blk.%d.ssm_a" },
|
|
925
|
+
{ LLM_TENSOR_SSM_D, "blk.%d.ssm_d" },
|
|
926
|
+
{ LLM_TENSOR_SSM_OUT, "blk.%d.ssm_out" },
|
|
927
|
+
},
|
|
928
|
+
},
|
|
929
|
+
{
|
|
930
|
+
LLM_ARCH_XVERSE,
|
|
931
|
+
{
|
|
932
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
933
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
934
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
935
|
+
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
|
|
936
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
937
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
938
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
939
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
940
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
941
|
+
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd" },
|
|
942
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
943
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
944
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
945
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
946
|
+
},
|
|
947
|
+
},
|
|
948
|
+
{
|
|
949
|
+
LLM_ARCH_COMMAND_R,
|
|
950
|
+
{
|
|
951
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
952
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
953
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
954
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
955
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
956
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
957
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
958
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
959
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
960
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
961
|
+
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
|
|
962
|
+
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
|
|
963
|
+
},
|
|
964
|
+
},
|
|
965
|
+
{
|
|
966
|
+
LLM_ARCH_COHERE2,
|
|
967
|
+
{
|
|
968
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
969
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
970
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
971
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
972
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
973
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
974
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
975
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
976
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
977
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
978
|
+
},
|
|
979
|
+
},
|
|
980
|
+
{
|
|
981
|
+
LLM_ARCH_DBRX,
|
|
982
|
+
{
|
|
983
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
984
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
985
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
986
|
+
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
|
|
987
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
988
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
989
|
+
{ LLM_TENSOR_ATTN_OUT_NORM, "blk.%d.attn_output_norm" },
|
|
990
|
+
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
|
|
991
|
+
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
|
|
992
|
+
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
|
|
993
|
+
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
|
|
994
|
+
},
|
|
995
|
+
},
|
|
996
|
+
{
|
|
997
|
+
LLM_ARCH_OLMO,
|
|
998
|
+
{
|
|
999
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1000
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
1001
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
1002
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
1003
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
1004
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
1005
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
1006
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
1007
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
1008
|
+
},
|
|
1009
|
+
},
|
|
1010
|
+
{
|
|
1011
|
+
LLM_ARCH_OLMO2,
|
|
1012
|
+
{
|
|
1013
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1014
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
1015
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
1016
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
1017
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
1018
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
1019
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
1020
|
+
{ LLM_TENSOR_ATTN_POST_NORM, "blk.%d.post_attention_norm" },
|
|
1021
|
+
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
|
|
1022
|
+
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
|
|
1023
|
+
{ LLM_TENSOR_FFN_POST_NORM, "blk.%d.post_ffw_norm" },
|
|
1024
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
1025
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
1026
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
1027
|
+
},
|
|
1028
|
+
},
|
|
1029
|
+
{
|
|
1030
|
+
LLM_ARCH_OLMOE,
|
|
1031
|
+
{
|
|
1032
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1033
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
1034
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
1035
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
1036
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
1037
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
1038
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
1039
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
1040
|
+
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
|
|
1041
|
+
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
|
|
1042
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
1043
|
+
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
|
|
1044
|
+
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
|
|
1045
|
+
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
|
|
1046
|
+
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
|
|
1047
|
+
},
|
|
1048
|
+
},
|
|
1049
|
+
{
|
|
1050
|
+
LLM_ARCH_OPENELM,
|
|
1051
|
+
{
|
|
1052
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1053
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
1054
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
1055
|
+
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
|
|
1056
|
+
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
|
|
1057
|
+
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
|
|
1058
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
1059
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
1060
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
1061
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
1062
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
1063
|
+
},
|
|
1064
|
+
},
|
|
1065
|
+
{
|
|
1066
|
+
LLM_ARCH_ARCTIC,
|
|
1067
|
+
{
|
|
1068
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1069
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
1070
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
1071
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
1072
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
1073
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
1074
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
1075
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
1076
|
+
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
|
|
1077
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
1078
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
1079
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
1080
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
1081
|
+
{ LLM_TENSOR_FFN_NORM_EXPS, "blk.%d.ffn_norm_exps" },
|
|
1082
|
+
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
|
|
1083
|
+
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
|
|
1084
|
+
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
|
|
1085
|
+
},
|
|
1086
|
+
},
|
|
1087
|
+
{
|
|
1088
|
+
LLM_ARCH_DEEPSEEK,
|
|
1089
|
+
{
|
|
1090
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1091
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
1092
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
1093
|
+
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
|
|
1094
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
1095
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
1096
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
1097
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
1098
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
1099
|
+
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd" },
|
|
1100
|
+
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
|
|
1101
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
1102
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
1103
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
1104
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
1105
|
+
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
|
|
1106
|
+
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
|
|
1107
|
+
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
|
|
1108
|
+
{ LLM_TENSOR_FFN_GATE_INP_SHEXP, "blk.%d.ffn_gate_inp_shexp" },
|
|
1109
|
+
{ LLM_TENSOR_FFN_GATE_SHEXP, "blk.%d.ffn_gate_shexp" },
|
|
1110
|
+
{ LLM_TENSOR_FFN_DOWN_SHEXP, "blk.%d.ffn_down_shexp" },
|
|
1111
|
+
{ LLM_TENSOR_FFN_UP_SHEXP, "blk.%d.ffn_up_shexp" },
|
|
1112
|
+
},
|
|
1113
|
+
},
|
|
1114
|
+
{
|
|
1115
|
+
LLM_ARCH_DEEPSEEK2,
|
|
1116
|
+
{
|
|
1117
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1118
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
1119
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
1120
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
1121
|
+
{ LLM_TENSOR_ATTN_Q_A_NORM, "blk.%d.attn_q_a_norm" },
|
|
1122
|
+
{ LLM_TENSOR_ATTN_KV_A_NORM, "blk.%d.attn_kv_a_norm" },
|
|
1123
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
1124
|
+
{ LLM_TENSOR_ATTN_Q_A, "blk.%d.attn_q_a" },
|
|
1125
|
+
{ LLM_TENSOR_ATTN_Q_B, "blk.%d.attn_q_b" },
|
|
1126
|
+
{ LLM_TENSOR_ATTN_KV_A_MQA, "blk.%d.attn_kv_a_mqa" },
|
|
1127
|
+
{ LLM_TENSOR_ATTN_KV_B, "blk.%d.attn_kv_b" },
|
|
1128
|
+
{ LLM_TENSOR_ATTN_K_B, "blk.%d.attn_k_b" },
|
|
1129
|
+
{ LLM_TENSOR_ATTN_V_B, "blk.%d.attn_v_b" },
|
|
1130
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
1131
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
1132
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
1133
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
1134
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
1135
|
+
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
|
|
1136
|
+
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
|
|
1137
|
+
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
|
|
1138
|
+
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
|
|
1139
|
+
{ LLM_TENSOR_FFN_GATE_INP_SHEXP, "blk.%d.ffn_gate_inp_shexp" },
|
|
1140
|
+
{ LLM_TENSOR_FFN_GATE_SHEXP, "blk.%d.ffn_gate_shexp" },
|
|
1141
|
+
{ LLM_TENSOR_FFN_DOWN_SHEXP, "blk.%d.ffn_down_shexp" },
|
|
1142
|
+
{ LLM_TENSOR_FFN_UP_SHEXP, "blk.%d.ffn_up_shexp" },
|
|
1143
|
+
{ LLM_TENSOR_FFN_EXP_PROBS_B, "blk.%d.exp_probs_b" },
|
|
1144
|
+
},
|
|
1145
|
+
},
|
|
1146
|
+
{
|
|
1147
|
+
LLM_ARCH_PLM,
|
|
1148
|
+
{
|
|
1149
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1150
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
1151
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
1152
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
1153
|
+
{ LLM_TENSOR_ATTN_KV_A_MQA, "blk.%d.attn_kv_a_mqa" },
|
|
1154
|
+
{ LLM_TENSOR_ATTN_KV_A_NORM, "blk.%d.attn_kv_a_norm" },
|
|
1155
|
+
{ LLM_TENSOR_ATTN_KV_B, "blk.%d.attn_kv_b" },
|
|
1156
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
1157
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
1158
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
1159
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
1160
|
+
},
|
|
1161
|
+
},
|
|
1162
|
+
{
|
|
1163
|
+
LLM_ARCH_CHATGLM,
|
|
1164
|
+
{
|
|
1165
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1166
|
+
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
|
|
1167
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
1168
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
1169
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
1170
|
+
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
|
|
1171
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
1172
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
1173
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
1174
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
1175
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
1176
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
1177
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
1178
|
+
},
|
|
1179
|
+
},
|
|
1180
|
+
{
|
|
1181
|
+
LLM_ARCH_GLM4,
|
|
1182
|
+
{
|
|
1183
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1184
|
+
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
|
|
1185
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
1186
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
1187
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
1188
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
1189
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
1190
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
1191
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
1192
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
1193
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
1194
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
1195
|
+
{ LLM_TENSOR_ATTN_POST_NORM, "blk.%d.post_attention_norm" },
|
|
1196
|
+
{ LLM_TENSOR_FFN_POST_NORM, "blk.%d.post_ffw_norm" },
|
|
1197
|
+
},
|
|
1198
|
+
},
|
|
1199
|
+
{
|
|
1200
|
+
LLM_ARCH_BITNET,
|
|
1201
|
+
{
|
|
1202
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1203
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
1204
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
1205
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
1206
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
1207
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
1208
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
1209
|
+
{ LLM_TENSOR_ATTN_SUB_NORM, "blk.%d.attn_sub_norm" },
|
|
1210
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
1211
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
1212
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
1213
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
1214
|
+
{ LLM_TENSOR_FFN_SUB_NORM, "blk.%d.ffn_sub_norm" },
|
|
1215
|
+
},
|
|
1216
|
+
},
|
|
1217
|
+
{
|
|
1218
|
+
LLM_ARCH_T5,
|
|
1219
|
+
{
|
|
1220
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1221
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
1222
|
+
{ LLM_TENSOR_DEC_OUTPUT_NORM, "dec.output_norm" },
|
|
1223
|
+
{ LLM_TENSOR_DEC_ATTN_NORM, "dec.blk.%d.attn_norm" },
|
|
1224
|
+
{ LLM_TENSOR_DEC_ATTN_Q, "dec.blk.%d.attn_q" },
|
|
1225
|
+
{ LLM_TENSOR_DEC_ATTN_K, "dec.blk.%d.attn_k" },
|
|
1226
|
+
{ LLM_TENSOR_DEC_ATTN_V, "dec.blk.%d.attn_v" },
|
|
1227
|
+
{ LLM_TENSOR_DEC_ATTN_OUT, "dec.blk.%d.attn_o" },
|
|
1228
|
+
{ LLM_TENSOR_DEC_ATTN_REL_B, "dec.blk.%d.attn_rel_b" },
|
|
1229
|
+
{ LLM_TENSOR_DEC_CROSS_ATTN_NORM, "dec.blk.%d.cross_attn_norm" },
|
|
1230
|
+
{ LLM_TENSOR_DEC_CROSS_ATTN_Q, "dec.blk.%d.cross_attn_q" },
|
|
1231
|
+
{ LLM_TENSOR_DEC_CROSS_ATTN_K, "dec.blk.%d.cross_attn_k" },
|
|
1232
|
+
{ LLM_TENSOR_DEC_CROSS_ATTN_V, "dec.blk.%d.cross_attn_v" },
|
|
1233
|
+
{ LLM_TENSOR_DEC_CROSS_ATTN_OUT, "dec.blk.%d.cross_attn_o" },
|
|
1234
|
+
{ LLM_TENSOR_DEC_CROSS_ATTN_REL_B, "dec.blk.%d.cross_attn_rel_b" },
|
|
1235
|
+
{ LLM_TENSOR_DEC_FFN_NORM, "dec.blk.%d.ffn_norm" },
|
|
1236
|
+
{ LLM_TENSOR_DEC_FFN_GATE, "dec.blk.%d.ffn_gate" },
|
|
1237
|
+
{ LLM_TENSOR_DEC_FFN_DOWN, "dec.blk.%d.ffn_down" },
|
|
1238
|
+
{ LLM_TENSOR_DEC_FFN_UP, "dec.blk.%d.ffn_up" },
|
|
1239
|
+
{ LLM_TENSOR_ENC_OUTPUT_NORM, "enc.output_norm" },
|
|
1240
|
+
{ LLM_TENSOR_ENC_ATTN_NORM, "enc.blk.%d.attn_norm" },
|
|
1241
|
+
{ LLM_TENSOR_ENC_ATTN_Q, "enc.blk.%d.attn_q" },
|
|
1242
|
+
{ LLM_TENSOR_ENC_ATTN_K, "enc.blk.%d.attn_k" },
|
|
1243
|
+
{ LLM_TENSOR_ENC_ATTN_V, "enc.blk.%d.attn_v" },
|
|
1244
|
+
{ LLM_TENSOR_ENC_ATTN_OUT, "enc.blk.%d.attn_o" },
|
|
1245
|
+
{ LLM_TENSOR_ENC_ATTN_REL_B, "enc.blk.%d.attn_rel_b" },
|
|
1246
|
+
{ LLM_TENSOR_ENC_FFN_NORM, "enc.blk.%d.ffn_norm" },
|
|
1247
|
+
{ LLM_TENSOR_ENC_FFN_GATE, "enc.blk.%d.ffn_gate" },
|
|
1248
|
+
{ LLM_TENSOR_ENC_FFN_DOWN, "enc.blk.%d.ffn_down" },
|
|
1249
|
+
{ LLM_TENSOR_ENC_FFN_UP, "enc.blk.%d.ffn_up" },
|
|
1250
|
+
},
|
|
1251
|
+
},
|
|
1252
|
+
{
|
|
1253
|
+
LLM_ARCH_T5ENCODER,
|
|
1254
|
+
{
|
|
1255
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1256
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
1257
|
+
{ LLM_TENSOR_ENC_OUTPUT_NORM, "enc.output_norm" },
|
|
1258
|
+
{ LLM_TENSOR_ENC_ATTN_NORM, "enc.blk.%d.attn_norm" },
|
|
1259
|
+
{ LLM_TENSOR_ENC_ATTN_Q, "enc.blk.%d.attn_q" },
|
|
1260
|
+
{ LLM_TENSOR_ENC_ATTN_K, "enc.blk.%d.attn_k" },
|
|
1261
|
+
{ LLM_TENSOR_ENC_ATTN_V, "enc.blk.%d.attn_v" },
|
|
1262
|
+
{ LLM_TENSOR_ENC_ATTN_OUT, "enc.blk.%d.attn_o" },
|
|
1263
|
+
{ LLM_TENSOR_ENC_ATTN_REL_B, "enc.blk.%d.attn_rel_b" },
|
|
1264
|
+
{ LLM_TENSOR_ENC_FFN_NORM, "enc.blk.%d.ffn_norm" },
|
|
1265
|
+
{ LLM_TENSOR_ENC_FFN_GATE, "enc.blk.%d.ffn_gate" },
|
|
1266
|
+
{ LLM_TENSOR_ENC_FFN_DOWN, "enc.blk.%d.ffn_down" },
|
|
1267
|
+
{ LLM_TENSOR_ENC_FFN_UP, "enc.blk.%d.ffn_up" },
|
|
1268
|
+
},
|
|
1269
|
+
},
|
|
1270
|
+
{
|
|
1271
|
+
LLM_ARCH_JAIS,
|
|
1272
|
+
{
|
|
1273
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1274
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
1275
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
1276
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
1277
|
+
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
|
|
1278
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
1279
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
1280
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
1281
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
1282
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
1283
|
+
},
|
|
1284
|
+
},
|
|
1285
|
+
{
|
|
1286
|
+
LLM_ARCH_NEMOTRON,
|
|
1287
|
+
{
|
|
1288
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1289
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
1290
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
1291
|
+
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
|
|
1292
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
1293
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
1294
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
1295
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
1296
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
1297
|
+
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd" },
|
|
1298
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
1299
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
1300
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
1301
|
+
},
|
|
1302
|
+
},
|
|
1303
|
+
{
|
|
1304
|
+
LLM_ARCH_EXAONE,
|
|
1305
|
+
{
|
|
1306
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1307
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
1308
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
1309
|
+
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
|
|
1310
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
1311
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
1312
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
1313
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
1314
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
1315
|
+
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd" },
|
|
1316
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
1317
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
1318
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
1319
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
1320
|
+
},
|
|
1321
|
+
},
|
|
1322
|
+
{
|
|
1323
|
+
LLM_ARCH_RWKV6,
|
|
1324
|
+
{
|
|
1325
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1326
|
+
{ LLM_TENSOR_TOKEN_EMBD_NORM, "token_embd_norm" },
|
|
1327
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
1328
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
1329
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
1330
|
+
{ LLM_TENSOR_ATTN_NORM_2, "blk.%d.attn_norm_2" },
|
|
1331
|
+
{ LLM_TENSOR_TIME_MIX_W1, "blk.%d.time_mix_w1" },
|
|
1332
|
+
{ LLM_TENSOR_TIME_MIX_W2, "blk.%d.time_mix_w2" },
|
|
1333
|
+
{ LLM_TENSOR_TIME_MIX_LERP_X, "blk.%d.time_mix_lerp_x" },
|
|
1334
|
+
{ LLM_TENSOR_TIME_MIX_LERP_W, "blk.%d.time_mix_lerp_w" },
|
|
1335
|
+
{ LLM_TENSOR_TIME_MIX_LERP_K, "blk.%d.time_mix_lerp_k" },
|
|
1336
|
+
{ LLM_TENSOR_TIME_MIX_LERP_V, "blk.%d.time_mix_lerp_v" },
|
|
1337
|
+
{ LLM_TENSOR_TIME_MIX_LERP_R, "blk.%d.time_mix_lerp_r" },
|
|
1338
|
+
{ LLM_TENSOR_TIME_MIX_LERP_G, "blk.%d.time_mix_lerp_g" },
|
|
1339
|
+
{ LLM_TENSOR_TIME_MIX_LERP_FUSED, "blk.%d.time_mix_lerp_fused" },
|
|
1340
|
+
{ LLM_TENSOR_TIME_MIX_FIRST, "blk.%d.time_mix_first" },
|
|
1341
|
+
{ LLM_TENSOR_TIME_MIX_DECAY, "blk.%d.time_mix_decay" },
|
|
1342
|
+
{ LLM_TENSOR_TIME_MIX_DECAY_W1, "blk.%d.time_mix_decay_w1" },
|
|
1343
|
+
{ LLM_TENSOR_TIME_MIX_DECAY_W2, "blk.%d.time_mix_decay_w2" },
|
|
1344
|
+
{ LLM_TENSOR_TIME_MIX_KEY, "blk.%d.time_mix_key" },
|
|
1345
|
+
{ LLM_TENSOR_TIME_MIX_VALUE, "blk.%d.time_mix_value" },
|
|
1346
|
+
{ LLM_TENSOR_TIME_MIX_RECEPTANCE, "blk.%d.time_mix_receptance" },
|
|
1347
|
+
{ LLM_TENSOR_TIME_MIX_GATE, "blk.%d.time_mix_gate" },
|
|
1348
|
+
{ LLM_TENSOR_TIME_MIX_LN, "blk.%d.time_mix_ln" },
|
|
1349
|
+
{ LLM_TENSOR_TIME_MIX_OUTPUT, "blk.%d.time_mix_output" },
|
|
1350
|
+
{ LLM_TENSOR_CHANNEL_MIX_LERP_K, "blk.%d.channel_mix_lerp_k" },
|
|
1351
|
+
{ LLM_TENSOR_CHANNEL_MIX_LERP_R, "blk.%d.channel_mix_lerp_r" },
|
|
1352
|
+
{ LLM_TENSOR_CHANNEL_MIX_KEY, "blk.%d.channel_mix_key" },
|
|
1353
|
+
{ LLM_TENSOR_CHANNEL_MIX_VALUE, "blk.%d.channel_mix_value" },
|
|
1354
|
+
{ LLM_TENSOR_CHANNEL_MIX_RECEPTANCE, "blk.%d.channel_mix_receptance" },
|
|
1355
|
+
},
|
|
1356
|
+
},
|
|
1357
|
+
{
|
|
1358
|
+
LLM_ARCH_RWKV6QWEN2,
|
|
1359
|
+
{
|
|
1360
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1361
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
1362
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
1363
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
1364
|
+
{ LLM_TENSOR_TIME_MIX_W1, "blk.%d.time_mix_w1" },
|
|
1365
|
+
{ LLM_TENSOR_TIME_MIX_W2, "blk.%d.time_mix_w2" },
|
|
1366
|
+
{ LLM_TENSOR_TIME_MIX_LERP_X, "blk.%d.time_mix_lerp_x" },
|
|
1367
|
+
{ LLM_TENSOR_TIME_MIX_LERP_FUSED, "blk.%d.time_mix_lerp_fused" },
|
|
1368
|
+
{ LLM_TENSOR_TIME_MIX_FIRST, "blk.%d.time_mix_first" },
|
|
1369
|
+
{ LLM_TENSOR_TIME_MIX_DECAY, "blk.%d.time_mix_decay" },
|
|
1370
|
+
{ LLM_TENSOR_TIME_MIX_DECAY_W1, "blk.%d.time_mix_decay_w1" },
|
|
1371
|
+
{ LLM_TENSOR_TIME_MIX_DECAY_W2, "blk.%d.time_mix_decay_w2" },
|
|
1372
|
+
{ LLM_TENSOR_TIME_MIX_KEY, "blk.%d.time_mix_key" },
|
|
1373
|
+
{ LLM_TENSOR_TIME_MIX_VALUE, "blk.%d.time_mix_value" },
|
|
1374
|
+
{ LLM_TENSOR_TIME_MIX_RECEPTANCE, "blk.%d.time_mix_receptance" },
|
|
1375
|
+
{ LLM_TENSOR_TIME_MIX_GATE, "blk.%d.time_mix_gate" },
|
|
1376
|
+
{ LLM_TENSOR_TIME_MIX_OUTPUT, "blk.%d.time_mix_output" },
|
|
1377
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
1378
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
1379
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
1380
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
1381
|
+
},
|
|
1382
|
+
},
|
|
1383
|
+
{
|
|
1384
|
+
LLM_ARCH_RWKV7,
|
|
1385
|
+
{
|
|
1386
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1387
|
+
{ LLM_TENSOR_TOKEN_EMBD_NORM, "token_embd_norm" },
|
|
1388
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
1389
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
1390
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
1391
|
+
{ LLM_TENSOR_ATTN_NORM_2, "blk.%d.attn_norm_2" },
|
|
1392
|
+
{ LLM_TENSOR_TIME_MIX_W0, "blk.%d.time_mix_w0" },
|
|
1393
|
+
{ LLM_TENSOR_TIME_MIX_W1, "blk.%d.time_mix_w1" },
|
|
1394
|
+
{ LLM_TENSOR_TIME_MIX_W2, "blk.%d.time_mix_w2" },
|
|
1395
|
+
{ LLM_TENSOR_TIME_MIX_A0, "blk.%d.time_mix_a0" },
|
|
1396
|
+
{ LLM_TENSOR_TIME_MIX_A1, "blk.%d.time_mix_a1" },
|
|
1397
|
+
{ LLM_TENSOR_TIME_MIX_A2, "blk.%d.time_mix_a2" },
|
|
1398
|
+
{ LLM_TENSOR_TIME_MIX_V0, "blk.%d.time_mix_v0" },
|
|
1399
|
+
{ LLM_TENSOR_TIME_MIX_V1, "blk.%d.time_mix_v1" },
|
|
1400
|
+
{ LLM_TENSOR_TIME_MIX_V2, "blk.%d.time_mix_v2" },
|
|
1401
|
+
{ LLM_TENSOR_TIME_MIX_G1, "blk.%d.time_mix_g1" },
|
|
1402
|
+
{ LLM_TENSOR_TIME_MIX_G2, "blk.%d.time_mix_g2" },
|
|
1403
|
+
{ LLM_TENSOR_TIME_MIX_K_K, "blk.%d.time_mix_k_k" },
|
|
1404
|
+
{ LLM_TENSOR_TIME_MIX_K_A, "blk.%d.time_mix_k_a" },
|
|
1405
|
+
{ LLM_TENSOR_TIME_MIX_R_K, "blk.%d.time_mix_r_k" },
|
|
1406
|
+
{ LLM_TENSOR_TIME_MIX_LERP_FUSED, "blk.%d.time_mix_lerp_fused" },
|
|
1407
|
+
{ LLM_TENSOR_TIME_MIX_KEY, "blk.%d.time_mix_key" },
|
|
1408
|
+
{ LLM_TENSOR_TIME_MIX_VALUE, "blk.%d.time_mix_value" },
|
|
1409
|
+
{ LLM_TENSOR_TIME_MIX_RECEPTANCE, "blk.%d.time_mix_receptance" },
|
|
1410
|
+
{ LLM_TENSOR_TIME_MIX_LN, "blk.%d.time_mix_ln" },
|
|
1411
|
+
{ LLM_TENSOR_TIME_MIX_OUTPUT, "blk.%d.time_mix_output" },
|
|
1412
|
+
{ LLM_TENSOR_CHANNEL_MIX_LERP_K, "blk.%d.channel_mix_lerp_k" },
|
|
1413
|
+
{ LLM_TENSOR_CHANNEL_MIX_KEY, "blk.%d.channel_mix_key" },
|
|
1414
|
+
{ LLM_TENSOR_CHANNEL_MIX_VALUE, "blk.%d.channel_mix_value" },
|
|
1415
|
+
},
|
|
1416
|
+
},
|
|
1417
|
+
{
|
|
1418
|
+
LLM_ARCH_ARWKV7,
|
|
1419
|
+
{
|
|
1420
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1421
|
+
{ LLM_TENSOR_TOKEN_EMBD_NORM, "token_embd_norm" },
|
|
1422
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
1423
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
1424
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
1425
|
+
{ LLM_TENSOR_TIME_MIX_W0, "blk.%d.time_mix_w0" },
|
|
1426
|
+
{ LLM_TENSOR_TIME_MIX_W1, "blk.%d.time_mix_w1" },
|
|
1427
|
+
{ LLM_TENSOR_TIME_MIX_W2, "blk.%d.time_mix_w2" },
|
|
1428
|
+
{ LLM_TENSOR_TIME_MIX_A0, "blk.%d.time_mix_a0" },
|
|
1429
|
+
{ LLM_TENSOR_TIME_MIX_A1, "blk.%d.time_mix_a1" },
|
|
1430
|
+
{ LLM_TENSOR_TIME_MIX_A2, "blk.%d.time_mix_a2" },
|
|
1431
|
+
{ LLM_TENSOR_TIME_MIX_V0, "blk.%d.time_mix_v0" },
|
|
1432
|
+
{ LLM_TENSOR_TIME_MIX_V1, "blk.%d.time_mix_v1" },
|
|
1433
|
+
{ LLM_TENSOR_TIME_MIX_V2, "blk.%d.time_mix_v2" },
|
|
1434
|
+
{ LLM_TENSOR_TIME_MIX_G1, "blk.%d.time_mix_g1" },
|
|
1435
|
+
{ LLM_TENSOR_TIME_MIX_G2, "blk.%d.time_mix_g2" },
|
|
1436
|
+
{ LLM_TENSOR_TIME_MIX_K_K, "blk.%d.time_mix_k_k" },
|
|
1437
|
+
{ LLM_TENSOR_TIME_MIX_K_A, "blk.%d.time_mix_k_a" },
|
|
1438
|
+
{ LLM_TENSOR_TIME_MIX_R_K, "blk.%d.time_mix_r_k" },
|
|
1439
|
+
{ LLM_TENSOR_TIME_MIX_LERP_FUSED, "blk.%d.time_mix_lerp_fused" },
|
|
1440
|
+
{ LLM_TENSOR_TIME_MIX_KEY, "blk.%d.time_mix_key" },
|
|
1441
|
+
{ LLM_TENSOR_TIME_MIX_VALUE, "blk.%d.time_mix_value" },
|
|
1442
|
+
{ LLM_TENSOR_TIME_MIX_RECEPTANCE, "blk.%d.time_mix_receptance" },
|
|
1443
|
+
{ LLM_TENSOR_TIME_MIX_LN, "blk.%d.time_mix_ln" },
|
|
1444
|
+
{ LLM_TENSOR_TIME_MIX_OUTPUT, "blk.%d.time_mix_output" },
|
|
1445
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
1446
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
1447
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
1448
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
1449
|
+
},
|
|
1450
|
+
},
|
|
1451
|
+
{
|
|
1452
|
+
LLM_ARCH_GRANITE,
|
|
1453
|
+
{
|
|
1454
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1455
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
1456
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
1457
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
1458
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
1459
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
1460
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
1461
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
1462
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
1463
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
1464
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
1465
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
1466
|
+
},
|
|
1467
|
+
},
|
|
1468
|
+
{
|
|
1469
|
+
LLM_ARCH_GRANITE_MOE,
|
|
1470
|
+
{
|
|
1471
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1472
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
1473
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
1474
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
1475
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
1476
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
1477
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
1478
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
1479
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
1480
|
+
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
|
|
1481
|
+
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
|
|
1482
|
+
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
|
|
1483
|
+
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
|
|
1484
|
+
},
|
|
1485
|
+
},
|
|
1486
|
+
{
|
|
1487
|
+
LLM_ARCH_CHAMELEON,
|
|
1488
|
+
{
|
|
1489
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1490
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
1491
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
1492
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
1493
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
1494
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
1495
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
1496
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
1497
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
1498
|
+
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
|
1499
|
+
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
|
1500
|
+
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
|
1501
|
+
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
|
|
1502
|
+
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
|
|
1503
|
+
},
|
|
1504
|
+
},
|
|
1505
|
+
{
|
|
1506
|
+
LLM_ARCH_WAVTOKENIZER_DEC,
|
|
1507
|
+
{
|
|
1508
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1509
|
+
{ LLM_TENSOR_TOKEN_EMBD_NORM, "token_embd_norm" },
|
|
1510
|
+
{ LLM_TENSOR_CONV1D, "conv1d" },
|
|
1511
|
+
{ LLM_TENSOR_CONVNEXT_DW, "convnext.%d.dw" },
|
|
1512
|
+
{ LLM_TENSOR_CONVNEXT_NORM, "convnext.%d.norm" },
|
|
1513
|
+
{ LLM_TENSOR_CONVNEXT_PW1, "convnext.%d.pw1" },
|
|
1514
|
+
{ LLM_TENSOR_CONVNEXT_PW2, "convnext.%d.pw2" },
|
|
1515
|
+
{ LLM_TENSOR_CONVNEXT_GAMMA, "convnext.%d.gamma" },
|
|
1516
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
1517
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
1518
|
+
{ LLM_TENSOR_POS_NET_CONV1, "posnet.%d.conv1" },
|
|
1519
|
+
{ LLM_TENSOR_POS_NET_CONV2, "posnet.%d.conv2" },
|
|
1520
|
+
{ LLM_TENSOR_POS_NET_NORM, "posnet.%d.norm" },
|
|
1521
|
+
{ LLM_TENSOR_POS_NET_NORM1, "posnet.%d.norm1" },
|
|
1522
|
+
{ LLM_TENSOR_POS_NET_NORM2, "posnet.%d.norm2" },
|
|
1523
|
+
{ LLM_TENSOR_POS_NET_ATTN_NORM, "posnet.%d.attn_norm" },
|
|
1524
|
+
{ LLM_TENSOR_POS_NET_ATTN_Q, "posnet.%d.attn_q" },
|
|
1525
|
+
{ LLM_TENSOR_POS_NET_ATTN_K, "posnet.%d.attn_k" },
|
|
1526
|
+
{ LLM_TENSOR_POS_NET_ATTN_V, "posnet.%d.attn_v" },
|
|
1527
|
+
{ LLM_TENSOR_POS_NET_ATTN_OUT, "posnet.%d.attn_output" },
|
|
1528
|
+
},
|
|
1529
|
+
},
|
|
1530
|
+
{
|
|
1531
|
+
LLM_ARCH_BAILINGMOE,
|
|
1532
|
+
{
|
|
1533
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1534
|
+
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
|
1535
|
+
{ LLM_TENSOR_OUTPUT, "output" },
|
|
1536
|
+
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
|
|
1537
|
+
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
|
1538
|
+
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
|
|
1539
|
+
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
|
|
1540
|
+
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
|
|
1541
|
+
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
|
1542
|
+
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
|
|
1543
|
+
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
|
1544
|
+
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
|
|
1545
|
+
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
|
|
1546
|
+
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
|
|
1547
|
+
{ LLM_TENSOR_FFN_GATE_INP_SHEXP, "blk.%d.ffn_gate_inp_shexp" },
|
|
1548
|
+
{ LLM_TENSOR_FFN_GATE_SHEXP, "blk.%d.ffn_gate_shexp" },
|
|
1549
|
+
{ LLM_TENSOR_FFN_DOWN_SHEXP, "blk.%d.ffn_down_shexp" },
|
|
1550
|
+
{ LLM_TENSOR_FFN_UP_SHEXP, "blk.%d.ffn_up_shexp" },
|
|
1551
|
+
},
|
|
1552
|
+
},
|
|
1553
|
+
{
|
|
1554
|
+
LLM_ARCH_UNKNOWN,
|
|
1555
|
+
{
|
|
1556
|
+
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
|
1557
|
+
},
|
|
1558
|
+
},
|
|
1559
|
+
};
|
|
1560
|
+
|
|
1561
|
+
static const std::map<llm_tensor, llm_tensor_info> LLM_TENSOR_INFOS = {
|
|
1562
|
+
{LLM_TENSOR_TOKEN_EMBD, {LLM_TENSOR_LAYER_INPUT, GGML_OP_GET_ROWS}},
|
|
1563
|
+
{LLM_TENSOR_POS_EMBD, {LLM_TENSOR_LAYER_INPUT, GGML_OP_GET_ROWS}},
|
|
1564
|
+
{LLM_TENSOR_TOKEN_EMBD_NORM, {LLM_TENSOR_LAYER_INPUT, GGML_OP_GET_ROWS}},
|
|
1565
|
+
{LLM_TENSOR_TOKEN_TYPES, {LLM_TENSOR_LAYER_INPUT, GGML_OP_GET_ROWS}},
|
|
1566
|
+
{LLM_TENSOR_OUTPUT, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL_MAT}},
|
|
1567
|
+
{LLM_TENSOR_CLS, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL_MAT}},
|
|
1568
|
+
{LLM_TENSOR_CLS_OUT, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL_MAT}},
|
|
1569
|
+
{LLM_TENSOR_OUTPUT_NORM, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL}},
|
|
1570
|
+
{LLM_TENSOR_DEC_OUTPUT_NORM, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL}},
|
|
1571
|
+
{LLM_TENSOR_ENC_OUTPUT_NORM, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL}},
|
|
1572
|
+
{LLM_TENSOR_ROPE_FREQS, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_ROPE}},
|
|
1573
|
+
{LLM_TENSOR_ROPE_FACTORS_LONG, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_ROPE}},
|
|
1574
|
+
{LLM_TENSOR_ROPE_FACTORS_SHORT, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_ROPE}},
|
|
1575
|
+
{LLM_TENSOR_ATTN_Q, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1576
|
+
{LLM_TENSOR_ATTN_K, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1577
|
+
{LLM_TENSOR_ATTN_V, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1578
|
+
{LLM_TENSOR_ATTN_QKV, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1579
|
+
{LLM_TENSOR_ATTN_OUT, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1580
|
+
{LLM_TENSOR_FFN_GATE, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1581
|
+
{LLM_TENSOR_FFN_DOWN, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1582
|
+
{LLM_TENSOR_FFN_UP, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1583
|
+
{LLM_TENSOR_FFN_DOWN_SHEXP, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1584
|
+
{LLM_TENSOR_FFN_GATE_SHEXP, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1585
|
+
{LLM_TENSOR_FFN_UP_SHEXP, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1586
|
+
{LLM_TENSOR_ATTN_Q_A, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1587
|
+
{LLM_TENSOR_ATTN_Q_B, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1588
|
+
{LLM_TENSOR_ATTN_KV_A_MQA, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1589
|
+
{LLM_TENSOR_ATTN_KV_B, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1590
|
+
{LLM_TENSOR_ATTN_K_B, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1591
|
+
{LLM_TENSOR_ATTN_V_B, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1592
|
+
{LLM_TENSOR_DEC_ATTN_Q, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1593
|
+
{LLM_TENSOR_DEC_ATTN_K, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1594
|
+
{LLM_TENSOR_DEC_ATTN_V, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1595
|
+
{LLM_TENSOR_DEC_ATTN_OUT, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1596
|
+
{LLM_TENSOR_DEC_CROSS_ATTN_Q, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1597
|
+
{LLM_TENSOR_DEC_CROSS_ATTN_K, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1598
|
+
{LLM_TENSOR_DEC_CROSS_ATTN_V, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1599
|
+
{LLM_TENSOR_DEC_CROSS_ATTN_OUT, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1600
|
+
{LLM_TENSOR_DEC_FFN_GATE, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1601
|
+
{LLM_TENSOR_DEC_FFN_DOWN, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1602
|
+
{LLM_TENSOR_DEC_FFN_UP, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1603
|
+
{LLM_TENSOR_ENC_ATTN_Q, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1604
|
+
{LLM_TENSOR_ENC_ATTN_K, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1605
|
+
{LLM_TENSOR_ENC_ATTN_V, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1606
|
+
{LLM_TENSOR_ENC_ATTN_OUT, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1607
|
+
{LLM_TENSOR_ENC_FFN_GATE, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1608
|
+
{LLM_TENSOR_ENC_FFN_DOWN, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1609
|
+
{LLM_TENSOR_ENC_FFN_UP, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1610
|
+
{LLM_TENSOR_FFN_GATE_INP_SHEXP, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1611
|
+
{LLM_TENSOR_FFN_GATE_INP, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1612
|
+
{LLM_TENSOR_SSM_IN, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1613
|
+
{LLM_TENSOR_SSM_X, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1614
|
+
{LLM_TENSOR_SSM_DT, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1615
|
+
{LLM_TENSOR_SSM_OUT, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1616
|
+
{LLM_TENSOR_TIME_MIX_W1, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1617
|
+
{LLM_TENSOR_TIME_MIX_W2, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1618
|
+
{LLM_TENSOR_TIME_MIX_A1, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1619
|
+
{LLM_TENSOR_TIME_MIX_A2, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1620
|
+
{LLM_TENSOR_TIME_MIX_V1, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1621
|
+
{LLM_TENSOR_TIME_MIX_V2, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1622
|
+
{LLM_TENSOR_TIME_MIX_G1, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1623
|
+
{LLM_TENSOR_TIME_MIX_G2, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1624
|
+
{LLM_TENSOR_TIME_MIX_DECAY_W1, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1625
|
+
{LLM_TENSOR_TIME_MIX_DECAY_W2, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1626
|
+
{LLM_TENSOR_TIME_MIX_KEY, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1627
|
+
{LLM_TENSOR_TIME_MIX_VALUE, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1628
|
+
{LLM_TENSOR_TIME_MIX_RECEPTANCE, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1629
|
+
{LLM_TENSOR_TIME_MIX_GATE, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1630
|
+
{LLM_TENSOR_TIME_MIX_OUTPUT, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1631
|
+
{LLM_TENSOR_CHANNEL_MIX_KEY, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1632
|
+
{LLM_TENSOR_CHANNEL_MIX_RECEPTANCE, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1633
|
+
{LLM_TENSOR_CHANNEL_MIX_VALUE, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1634
|
+
{LLM_TENSOR_FFN_ACT, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_DIV}},
|
|
1635
|
+
{LLM_TENSOR_SSM_CONV1D, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_SSM_CONV}},
|
|
1636
|
+
{LLM_TENSOR_SSM_A, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_SSM_SCAN}},
|
|
1637
|
+
{LLM_TENSOR_SSM_D, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1638
|
+
{LLM_TENSOR_TIME_MIX_LERP_X, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1639
|
+
{LLM_TENSOR_TIME_MIX_LN, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1640
|
+
{LLM_TENSOR_CHANNEL_MIX_LERP_K, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1641
|
+
{LLM_TENSOR_CHANNEL_MIX_LERP_R, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1642
|
+
{LLM_TENSOR_TIME_MIX_K_K, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1643
|
+
{LLM_TENSOR_TIME_MIX_K_A, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1644
|
+
{LLM_TENSOR_TIME_MIX_R_K, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1645
|
+
{LLM_TENSOR_TIME_MIX_LERP_W, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_ADD}},
|
|
1646
|
+
{LLM_TENSOR_TIME_MIX_LERP_K, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_ADD}},
|
|
1647
|
+
{LLM_TENSOR_TIME_MIX_LERP_V, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_ADD}},
|
|
1648
|
+
{LLM_TENSOR_TIME_MIX_LERP_R, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_ADD}},
|
|
1649
|
+
{LLM_TENSOR_TIME_MIX_LERP_G, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_ADD}},
|
|
1650
|
+
{LLM_TENSOR_TIME_MIX_LERP_FUSED, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_ADD}},
|
|
1651
|
+
{LLM_TENSOR_TIME_MIX_DECAY, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_ADD}},
|
|
1652
|
+
{LLM_TENSOR_TIME_MIX_W0, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_ADD}},
|
|
1653
|
+
{LLM_TENSOR_TIME_MIX_A0, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_ADD}},
|
|
1654
|
+
{LLM_TENSOR_TIME_MIX_V0, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_ADD}},
|
|
1655
|
+
{LLM_TENSOR_TIME_MIX_FIRST, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_RWKV_WKV6}},
|
|
1656
|
+
{LLM_TENSOR_ATTN_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1657
|
+
{LLM_TENSOR_ATTN_NORM_2, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1658
|
+
{LLM_TENSOR_ATTN_OUT_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1659
|
+
{LLM_TENSOR_ATTN_POST_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1660
|
+
{LLM_TENSOR_FFN_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1661
|
+
{LLM_TENSOR_FFN_POST_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1662
|
+
{LLM_TENSOR_FFN_NORM_EXPS, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1663
|
+
{LLM_TENSOR_ATTN_Q_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1664
|
+
{LLM_TENSOR_ATTN_K_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1665
|
+
{LLM_TENSOR_LAYER_OUT_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1666
|
+
{LLM_TENSOR_ATTN_Q_A_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1667
|
+
{LLM_TENSOR_ATTN_KV_A_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1668
|
+
{LLM_TENSOR_ATTN_SUB_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1669
|
+
{LLM_TENSOR_FFN_SUB_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1670
|
+
{LLM_TENSOR_DEC_ATTN_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1671
|
+
{LLM_TENSOR_DEC_CROSS_ATTN_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1672
|
+
{LLM_TENSOR_DEC_FFN_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1673
|
+
{LLM_TENSOR_ENC_ATTN_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1674
|
+
{LLM_TENSOR_ENC_FFN_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1675
|
+
{LLM_TENSOR_DEC_ATTN_REL_B, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_GET_ROWS}},
|
|
1676
|
+
{LLM_TENSOR_ENC_ATTN_REL_B, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_GET_ROWS}},
|
|
1677
|
+
{LLM_TENSOR_FFN_DOWN_EXPS, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT_ID}},
|
|
1678
|
+
{LLM_TENSOR_FFN_GATE_EXPS, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT_ID}},
|
|
1679
|
+
{LLM_TENSOR_FFN_UP_EXPS, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT_ID}},
|
|
1680
|
+
{LLM_TENSOR_FFN_EXP_PROBS_B, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_ADD}},
|
|
1681
|
+
// this tensor is loaded for T5, but never used
|
|
1682
|
+
{LLM_TENSOR_DEC_CROSS_ATTN_REL_B, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_NONE}},
|
|
1683
|
+
{LLM_TENSOR_CONV1D, {LLM_TENSOR_LAYER_INPUT, GGML_OP_IM2COL}},
|
|
1684
|
+
{LLM_TENSOR_POS_NET_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1685
|
+
{LLM_TENSOR_POS_NET_NORM1, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1686
|
+
{LLM_TENSOR_POS_NET_NORM2, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1687
|
+
{LLM_TENSOR_POS_NET_CONV1, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_IM2COL}},
|
|
1688
|
+
{LLM_TENSOR_POS_NET_CONV2, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_IM2COL}},
|
|
1689
|
+
{LLM_TENSOR_POS_NET_ATTN_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1690
|
+
{LLM_TENSOR_POS_NET_ATTN_Q, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1691
|
+
{LLM_TENSOR_POS_NET_ATTN_K, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1692
|
+
{LLM_TENSOR_POS_NET_ATTN_V, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1693
|
+
{LLM_TENSOR_POS_NET_ATTN_OUT, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1694
|
+
{LLM_TENSOR_CONVNEXT_DW, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_IM2COL}},
|
|
1695
|
+
{LLM_TENSOR_CONVNEXT_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1696
|
+
{LLM_TENSOR_CONVNEXT_PW1, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1697
|
+
{LLM_TENSOR_CONVNEXT_PW2, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
|
|
1698
|
+
{LLM_TENSOR_CONVNEXT_GAMMA, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
|
|
1699
|
+
};
|
|
1700
|
+
|
|
1701
|
+
LLM_KV::LLM_KV(llm_arch arch, const char * suffix) : arch(arch), suffix(suffix) {}
|
|
1702
|
+
|
|
1703
|
+
std::string LLM_KV::operator()(llm_kv kv) const {
|
|
1704
|
+
return suffix ? ::format(LLM_KV_NAMES.at(kv), LLM_ARCH_NAMES.at(arch), suffix)
|
|
1705
|
+
: ::format(LLM_KV_NAMES.at(kv), LLM_ARCH_NAMES.at(arch));
|
|
1706
|
+
}
|
|
1707
|
+
|
|
1708
|
+
std::string LLM_TN_IMPL::str() const {
|
|
1709
|
+
if (LLM_TENSOR_NAMES.at(arch).find(tensor) == LLM_TENSOR_NAMES.at(arch).end()) {
|
|
1710
|
+
return "__missing__";
|
|
1711
|
+
}
|
|
1712
|
+
|
|
1713
|
+
std::string name = ::format(LLM_TENSOR_NAMES.at(arch).at(tensor), bid, xid);
|
|
1714
|
+
|
|
1715
|
+
if (suffix != nullptr) {
|
|
1716
|
+
name += ".";
|
|
1717
|
+
name += suffix;
|
|
1718
|
+
}
|
|
1719
|
+
|
|
1720
|
+
return name;
|
|
1721
|
+
}
|
|
1722
|
+
|
|
1723
|
+
const char * llm_arch_name(llm_arch arch) {
|
|
1724
|
+
auto it = LLM_ARCH_NAMES.find(arch);
|
|
1725
|
+
if (it == LLM_ARCH_NAMES.end()) {
|
|
1726
|
+
return "unknown";
|
|
1727
|
+
}
|
|
1728
|
+
return it->second;
|
|
1729
|
+
}
|
|
1730
|
+
|
|
1731
|
+
llm_arch llm_arch_from_string(const std::string & name) {
|
|
1732
|
+
for (const auto & kv : LLM_ARCH_NAMES) { // NOLINT
|
|
1733
|
+
if (kv.second == name) {
|
|
1734
|
+
return kv.first;
|
|
1735
|
+
}
|
|
1736
|
+
}
|
|
1737
|
+
|
|
1738
|
+
return LLM_ARCH_UNKNOWN;
|
|
1739
|
+
}
|
|
1740
|
+
|
|
1741
|
+
const llm_tensor_info & llm_tensor_info_for(llm_tensor tensor) {
|
|
1742
|
+
return LLM_TENSOR_INFOS.at(tensor);
|
|
1743
|
+
}
|