feat: Add llama.cpp web assembly support

Primarily for calculating embeddings on the client Signed-off-by: Gordon Smith <[email protected]>
hpcc-systems · Aug 30, 2024 · 5602a88 · 5602a88
1 parent 7c8edd5
commit 5602a88
Show file tree

Hide file tree

Showing 41 changed files with 5,115 additions and 848 deletions.
diff --git a/.gitignore b/.gitignore
@@ -1,6 +1,7 @@
 .nyc_output/
-/.vscode/c_cpp_properties.json
-/.vscode/ipch
+.vscode/c_cpp_properties.json
+.vscode/ipch
+.vscode/settings.json
 .nx/
 bin/
 build/
@@ -18,6 +19,7 @@ dist-test/
 /emsdk*
 /lib-*
 node_modules/
+/packages/llama/docs
 /rust
 /third-party
 /src-ts/*.wasm.ts

diff --git a/.vscode/launch.json b/.vscode/launch.json
@@ -4,6 +4,25 @@
     // For more information, visit: https://go.microsoft.com/fwlink/?linkid=830387
     "version": "0.2.0",
     "configurations": [
+        {
+            "name": "test-node",
+            "type": "node",
+            "request": "launch",
+            "cwd": "${workspaceFolder}",
+            "runtimeArgs": [
+                "run-script",
+                "test-node"
+            ],
+            "runtimeExecutable": "npm",
+            "skipFiles": [
+                "<node_internals>/**"
+            ],
+            "outFiles": [
+                "${workspaceFolder}/packages/**/*.js",
+                "${workspaceFolder}/packages/**/*.c",
+                "!**/node_modules/**"
+            ],
+        },
         {
             "name": "index-browser",
             "type": "msedge",
@@ -75,24 +94,6 @@
                 "webpack:///*": "/*"
             },
         },
-        {
-            "name": "test-node",
-            "type": "node",
-            "request": "launch",
-            "runtimeArgs": [
-                "run-script",
-                "test-node"
-            ],
-            "runtimeExecutable": "npm",
-            "skipFiles": [
-                "<node_internals>/**"
-            ],
-            "outFiles": [
-                "${workspaceFolder}/**/*.js",
-                "${workspaceFolder}/**/*.c",
-                "!**/node_modules/**"
-            ],
-        },
         {
             "name": "esbuild",
             "type": "node",

diff --git a/.vscode/settings.json b/.vscode/settings.json
@@ -1,5 +1,3 @@
 {
-    "cmake.cmakePath": "${workspaceFolder}/scripts/cmake.sh",
-    "ecl.launchConfiguration": "not found",
-    "ecl.targetCluster": {}
+    "cmake.cmakePath": "${workspaceFolder}/scripts/cmake.sh"
 }
diff --git a/.vscode/tasks.json b/.vscode/tasks.json
@@ -1,12 +1,10 @@
 {
-    // See https://go.microsoft.com/fwlink/?LinkId=733558 
-    // for the documentation about the tasks.json format
     "version": "2.0.0",
     "tasks": [
         {
             "type": "npm",
-            "label": "Compile Types Watch",
-            "script": "compile-types-watch",
+            "label": "build-types-watch",
+            "script": "build-types-watch",
             "problemMatcher": [
                 "$tsc-watch"
             ],
@@ -16,9 +14,48 @@
         },
         {
             "type": "npm",
-            "label": "Compile TypeScript Watch",
-            "script": "compile-ts-watch",
-            "problemMatcher": [],
+            "label": "build-ts-watch",
+            "script": "build-ts-watch",
+            "problemMatcher": [
+                "$tsc-watch"
+            ],
+            "presentation": {
+                "group": "group-build"
+            }
+        },
+        {
+            "type": "npm",
+            "label": "build-cpp",
+            "script": "build-cpp",
+            "problemMatcher": [
+                "$gcc"
+            ],
+            "presentation": {
+                "group": "group-build"
+            }
+        },
+        {
+            "type": "npm",
+            "label": "build-cpp-watch",
+            "script": "build-cpp-watch",
+            "dependsOn": [
+                "build-cpp"
+            ],
+            "problemMatcher": {
+                "owner": "cpp",
+                "fileLocation": [
+                    "relative",
+                    "${workspaceFolder}"
+                ],
+                "pattern": {
+                    "regexp": "^(.*):(\\d+):(\\d+):\\s+(warning|error):\\s+(.*)$",
+                    "file": 1,
+                    "line": 2,
+                    "column": 3,
+                    "severity": 4,
+                    "message": 5
+                }
+            },
             "presentation": {
                 "group": "group-build"
             }
@@ -27,23 +64,21 @@
             "type": "npm",
             "label": "Web Server",
             "script": "serve",
-            "problemMatcher": [],
             "presentation": {
                 "group": "group-build"
             }
         },
         {
             "label": "build",
             "dependsOn": [
-                "Compile Types Watch",
-                "Compile TypeScript Watch",
-                "Web Server"
+                "build-types-watch",
+                "build-ts-watch",
+                "build-cpp-watch"
             ],
             "group": {
                 "kind": "build",
                 "isDefault": true
-            },
-            "problemMatcher": []
+            }
         }
     ]
 }
diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -8,7 +8,9 @@ set(CMAKE_INSTALL_PREFIX "..")
 set(VCPKG_INCLUDE_DIR ${VCPKG_INSTALLED_DIR}/${VCPKG_TARGET_TRIPLET}/include)
 
 # See:  https://github.com/emscripten-core/emscripten/blob/main/src/settings.js
-set(EM_CPP_FLAGS "")
+set(EM_CPP_FLAGS 
+    "-flto"
+)
 
 set(EM_LINK_FLAGS 
     "-sASSERTIONS=0"
@@ -35,11 +37,12 @@ set(EM_LINK_FLAGS
     "-sUSE_GLFW=0"
     "-sALLOW_UNIMPLEMENTED_SYSCALLS=1"
     "-sINCOMING_MODULE_JS_API=\"['wasmBinary']\""
+    "--no-entry"
     "--pre-js ${CMAKE_CURRENT_SOURCE_DIR}/src-cpp/src/pre.js"
     "--post-js ${CMAKE_CURRENT_SOURCE_DIR}/src-cpp/src/post.js"
 )
 
-if (CMAKE_BUILD_TYPE STREQUAL "Debug")
+if (CMAKE_BUILD_TYPE STREQUAL "Debug" OR CMAKE_BUILD_TYPE STREQUAL "RelWithDebInfo")
     set(EM_LINK_FLAGS 
         ${EM_LINK_FLAGS}
         "-sUSE_ES6_IMPORT_META=1"
@@ -79,8 +82,8 @@ if (CMAKE_BUILD_TYPE STREQUAL "Debug")
     set(PACK_MODE "-d")
 endif ()
 
-add_subdirectory(src-cpp)
 add_subdirectory(packages/base91/src-cpp)
 add_subdirectory(packages/expat/src-cpp)
 add_subdirectory(packages/graphviz/src-cpp)
+add_subdirectory(packages/llama/src-cpp)
 add_subdirectory(packages/zstd/src-cpp)
diff --git a/CMakePresets.json b/CMakePresets.json
@@ -11,6 +11,7 @@
         },
         {
             "name": "vcpkg-emscripten",
+            "hidden": true,
             "inherits": [
                 "vcpkg"
             ],
@@ -31,6 +32,16 @@
                 "CMAKE_BUILD_TYPE": "Debug"
             }
         },
+        {
+            "name": "vcpkg-emscripten-RelWithDebInfo",
+            "inherits": [
+                "vcpkg-emscripten"
+            ],
+            "binaryDir": "${sourceDir}/build",
+            "cacheVariables": {
+                "CMAKE_BUILD_TYPE": "RelWithDebInfo"
+            }
+        },
         {
             "name": "vcpkg-emscripten-MinSizeRel",
             "inherits": [

diff --git a/docs/.vitepress/config.js b/docs/.vitepress/config.js
@@ -38,6 +38,7 @@ export default {
                     { text: 'DuckDB', link: '/duckdb/src/duckdb/classes/DuckDB' },
                     { text: 'Expat', link: '/expat/src/expat/classes/Expat' },
                     { text: 'Graphviz', link: '/graphviz/src/graphviz/classes/Graphviz' },
+                    { text: 'Llama', link: '/llama/src/llama/classes/Llama' },
                     { text: 'Zstd', link: '/zstd/src/zstd/classes/Zstd' },
                 ]
             }

diff --git a/docs/index.md b/docs/index.md
@@ -35,6 +35,9 @@ features:
   - title: GraphViz
     details: The Graphviz layout algorithms take descriptions of graphs in a simple text language, and make diagrams in useful formats, such as images and SVG for web pages or display in an interactive graph browser.
     link: /graphviz/src/graphviz/classes/Graphviz
+  - title: Llama
+    details: Inference of Meta's LLaMA model (and others) in pure C/C++.
+    link: /llama/src/llama/classes/Llama
   - title: Zstd
     details:  Zstandard is a fast compression algorithm, providing high compression ratios and is backed by an extremely fast decoder.
     link: /zstd/src/zstd/classes/Zstd