gnu: llama-cpp: Update to 0.0.0-b5013.

* gnu/packages/machine-learning.scm (llama-cpp): Update to 0.0.0-b5013.
[inputs]: Add curl, glslang, and python-gguf-llama-cpp.
[native-inputs]: bash -> bash-minimal.
[source, homepage]: Update URL.
[python-scripts]: Rely on upstream to install them.  Delete phase.
[fix-tests]: Fix an additional test.
* gnu/packages/patches/llama-cpp-vulkan-optional.patch: Delete.
* gnu/local.mk: Unregister patch.

Change-Id: Ic297534cd142cb83e3964eae21b4eb807b74e9bc
Signed-off-by: Danny Milosavljevic <dannym@friendly-machines.com>

gnu/packages/machine-learning.scm

@@ -80,6 +80,7 @@
   #:use-module (gnu packages compression)
   #:use-module (gnu packages cpp)
   #:use-module (gnu packages cran)
+  #:use-module (gnu packages curl)
   #:use-module (gnu packages crates-check)
   #:use-module (gnu packages crates-crypto)
   #:use-module (gnu packages crates-io)
@@ -642,7 +643,7 @@ Performance is achieved by using the LLVM JIT compiler.")
   (deprecated-package "guile-aiscm-next" guile-aiscm))
 
 (define-public llama-cpp
-  (let ((tag "b4549"))
+  (let ((tag "b5013"))
     (package
       (name "llama-cpp")
       (version (string-append "0.0.0-" tag))
@@ -650,19 +651,19 @@ Performance is achieved by using the LLVM JIT compiler.")
        (origin
          (method git-fetch)
          (uri (git-reference
-               (url "https://github.com/ggerganov/llama.cpp")
+               (url "https://github.com/ggml-org/llama.cpp")
                (commit tag)))
          (file-name (git-file-name name tag))
          (sha256
-          (base32 "1xf2579q0r8nv06kj8padi6w9cv30w58vdys65nq8yzm3dy452a1"))
+          (base32 "0s73dz871x53dr366lkzq19f677bwgma2ri8m5vhbfa9p8yp4p3r"))))
-         (patches
-          (search-patches "llama-cpp-vulkan-optional.patch"))))
       (build-system cmake-build-system)
       (arguments
        (list
         #:configure-flags
-        #~(list "-DBUILD_SHARED_LIBS=ON"
+        #~(list #$(string-append "-DGGML_BUILD_NUMBER=" tag)
+                "-DBUILD_SHARED_LIBS=ON"
                 "-DGGML_VULKAN=ON"
+                "-DLLAMA_CURL=ON"
                 "-DGGML_BLAS=ON"
                 "-DGGML_BLAS_VENDOR=OpenBLAS"
                 (string-append "-DBLAS_INCLUDE_DIRS="
@@ -692,32 +693,17 @@ Performance is achieved by using the LLVM JIT compiler.")
                 (substitute* "ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp"
                  (("\"/bin/sh\"")
                   (string-append "\"" (search-input-file inputs "/bin/sh") "\"")))))
-            (add-after 'unpack 'disable-unrunable-tests
+            (add-after 'unpack 'fix-tests
               (lambda _
                 ;; test-eval-callback downloads ML model from network, cannot
                 ;; run in Guix build environment
                 (substitute* '("examples/eval-callback/CMakeLists.txt")
                   (("COMMAND llama-eval-callback")
-                   "COMMAND true llama-eval-callback"))))
+                   "COMMAND true llama-eval-callback"))
-            (add-before 'install 'install-python-scripts
+                ;; Help it find the test files it needs
-              (lambda _
+                (substitute* "tests/test-chat.cpp"
-                (let ((bin (string-append #$output "/bin/")))
+                  (("\"\\.\\./\"") "\"../source/\""))))
-                  (define (make-script script)
+            (add-after 'install 'wrap-python-scripts
-                    (let ((suffix (if (string-suffix? ".py" script) "" ".py")))
-                      (call-with-input-file
-                          (string-append "../source/" script suffix)
-                        (lambda (input)
-                          (call-with-output-file (string-append bin script)
-                            (lambda (output)
-                              (format output "#!~a/bin/python3\n~a"
-                                      #$(this-package-input "python")
-                                      (get-string-all input))))))
-                      (chmod (string-append bin script) #o555)))
-                  (mkdir-p bin)
-                  (make-script "convert_hf_to_gguf")
-                  (make-script "convert_llama_ggml_to_gguf")
-                  (make-script "convert_hf_to_gguf_update.py"))))
-            (add-after 'install-python-scripts 'wrap-python-scripts
               (assoc-ref python:%standard-phases 'wrap))
             (add-after 'install 'remove-tests
               (lambda* (#:key outputs #:allow-other-keys)
@@ -725,12 +711,13 @@ Performance is achieved by using the LLVM JIT compiler.")
                                        (string-append (assoc-ref outputs "out")
                                                       "/bin")
                                        "^test-")))))))
-      (inputs (list python vulkan-headers vulkan-loader))
+      (inputs (list curl glslang python python-gguf-llama-cpp
-      (native-inputs (list pkg-config shaderc bash))
+                    vulkan-headers vulkan-loader))
+      (native-inputs (list pkg-config shaderc bash-minimal))
       (propagated-inputs
        (list python-numpy python-pytorch python-sentencepiece openblas))
       (properties '((tunable? . #true))) ;use AVX512, FMA, etc. when available
-      (home-page "https://github.com/ggerganov/llama.cpp")
+      (home-page "https://github.com/ggml-org/llama.cpp")
       (synopsis "Port of Facebook's LLaMA model in C/C++")
       (description "This package provides a port to Facebook's LLaMA collection
 of foundation language models.  It requires models parameters to be downloaded