llama: update to commit 71e90e88 (#10192)

943464cc · Jeffrey Morgan · GitHub · 369de832 · 943464cc · 943464cc
Unverified Commit 943464cc authored Apr 16, 2025 by Jeffrey Morgan Committed by GitHub Apr 16, 2025
20 changed files
--- a/llama/llama.cpp/src/llama-model.cpp
+++ b/llama/llama.cpp/src/llama-model.cpp
--- a/llama/llama.cpp/src/llama-model.h
+++ b/llama/llama.cpp/src/llama-model.h
--- a/llama/llama.cpp/src/llama-quant.cpp
+++ b/llama/llama.cpp/src/llama-quant.cpp
--- a/llama/llama.cpp/src/llama-sampling.cpp
+++ b/llama/llama.cpp/src/llama-sampling.cpp
--- a/llama/llama.cpp/src/llama-vocab.cpp
+++ b/llama/llama.cpp/src/llama-vocab.cpp
--- a/llama/llama.cpp/src/llama.cpp
+++ b/llama/llama.cpp/src/llama.cpp
--- a/llama/llama.cpp/src/unicode.cpp
+++ b/llama/llama.cpp/src/unicode.cpp
@@ -220,7 +220,6 @@ static inline std::wstring unicode_wstring_from_utf8(const std::string & s) {
    free(wbuf);
    return ret;
 #else
 #if defined(__clang__)
    // disable C++17 deprecation warning for std::codecvt_utf8
 #    pragma clang diagnostic push

--- a/llama/llama.go
+++ b/llama/llama.go
--- a/llama/patches/0001-ggml-backend-malloc-and-free-using-the-same-compiler.patch
+++ b/llama/patches/0001-ggml-backend-malloc-and-free-using-the-same-compiler.patch
--- a/llama/patches/0002-pretokenizer.patch
+++ b/llama/patches/0002-pretokenizer.patch
--- a/llama/patches/0003-embeddings.patch
+++ b/llama/patches/0003-embeddings.patch
--- a/llama/patches/0004-clip-unicode.patch
+++ b/llama/patches/0004-clip-unicode.patch
--- a/llama/patches/0005-solar-pro.patch
+++ b/llama/patches/0005-solar-pro.patch
--- a/llama/patches/0006-conditional-fattn.patch
+++ b/llama/patches/0006-conditional-fattn.patch
--- a/llama/patches/0007-add-mllama-support.patch
+++ b/llama/patches/0007-add-mllama-support.patch
--- a/llama/patches/0008-add-unpad-operator.patch
+++ b/llama/patches/0008-add-unpad-operator.patch
--- a/llama/patches/0009-fix-deepseek-deseret-regex.patch
+++ b/llama/patches/0009-fix-deepseek-deseret-regex.patch
--- a/llama/patches/0010-Maintain-ordering-for-rules-for-grammar.patch
+++ b/llama/patches/0010-Maintain-ordering-for-rules-for-grammar.patch
--- a/llama/patches/0011-ensure-KV-cache-is-fully-defragmented.patch
+++ b/llama/patches/0011-ensure-KV-cache-is-fully-defragmented.patch
--- a/llama/patches/0011-llama-Ensure-KV-cache-is-fully-defragmented.patch
+++ b/llama/patches/0011-llama-Ensure-KV-cache-is-fully-defragmented.patch