Distribute and run LLMs with a single file.
C++ cross-platformggufllama-cpplocal-ailocal-inferencelocal-llmopen-source-aisingle-file-executablespeech-to-text
Release llamafile v0.8.10
Justine Tunneyf7c6ef42650f
diff --git a/build/config.mk b/build/config.mkindex 333ccfeb..18d81a4a 100644--- a/build/config.mk+++ b/build/config.mk@@ -2,7 +2,7 @@#── vi: set noet ft=make ts=8 sw=8 fenc=utf-8 :vi ────────────────────┘PREFIX = /usr/local-COSMOCC = .cosmocc/3.5.9+COSMOCC = .cosmocc/3.6.0TOOLCHAIN = $(COSMOCC)/bin/cosmoAR = $(TOOLCHAIN)ar@@ -52,5 +52,5 @@ clean:; rm -rf o.PHONY: distcleandistclean:; rm -rf o .cosmocc-.cosmocc/3.5.9:- build/download-cosmocc.sh $@ 3.5.9 1f66831de4bf2d82e138d8993e9ee84a7559afc47aeeb2e2de51872401790a0a+.cosmocc/3.6.0:+ build/download-cosmocc.sh $@ 3.6.0 4918c45ac3e0972ff260e2a249e25716881e39fb679d5e714ae216a2ef6c3f7ediff --git a/llama.cpp/ggml-cuda.cu b/llama.cpp/ggml-cuda.cuindex ef45cd5b..b33072ec 100644--- a/llama.cpp/ggml-cuda.cu+++ b/llama.cpp/ggml-cuda.cu@@ -9709,6 +9709,7 @@ static __device__ void rope_yarn(}*cos_theta = cosf(theta) * mscale;*sin_theta = sinf(theta) * mscale;+ (void)ggml_cuda_hmax;}// rope == RoPE == rotary positional embedding@@ -13856,7 +13857,6 @@ GGML_API GGML_CALL int ggml_backend_cuda_reg_devices() {(void)exit_;(void)fast_fp16_available;(void)fp16_mma_available;- (void)ggml_cuda_hmax;(void)set_ggml_graph_node_properties;(void)ggml_graph_node_has_matching_properties;}diff --git a/llamafile/version.h b/llamafile/version.hindex 07706b67..af69042b 100644--- a/llamafile/version.h+++ b/llamafile/version.h@@ -2,7 +2,7 @@#define LLAMAFILE_MAJOR 0#define LLAMAFILE_MINOR 8-#define LLAMAFILE_PATCH 9+#define LLAMAFILE_PATCH 10#define LLAMAFILE_VERSION \(100000000 * LLAMAFILE_MAJOR + 1000000 * LLAMAFILE_MINOR + LLAMAFILE_PATCH)