feat: add support for metal and cuda in the build command (#17)

Author: giladgd

.github/workflows/build.yml

@@ -14,7 +14,7 @@ jobs:
         with:
           node-version: "20"
       - name: Install modules
-        run: npm ci --ignore-scripts
+        run: npm ci
       - name: Build
         run: npm run build
       - name: Generate docs
@@ -117,7 +117,7 @@ jobs:
         env:
           ARTIFACT_NAME: ${{ matrix.config.artifact }}
         run: |
-          npm ci --ignore-scripts
+          npm ci
           
           npx zx -y <<'EOF'
           
@@ -221,7 +221,7 @@ jobs:
         with:
           node-version: "20"
       - name: Install modules
-        run: npm ci --ignore-scripts
+        run: npm ci
       - uses: actions/download-artifact@v3
         with:
           path: artifacts
@@ -239,6 +239,8 @@ jobs:
           
           echo "Built binaries:"
           ls llamaBins
+      - name: Add "postinstall" script to package.json
+        run: npm run addPostinstallScript
       - name: Release
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/test.yml

@@ -9,7 +9,7 @@ jobs:
         with:
           node-version: "18"
       - name: Install modules
-        run: npm ci --ignore-scripts
+        run: npm ci
       - name: ESLint
         run: npx eslint --ext .js --ext .ts .
       - name: TypeScript validity

llama/binding.gyp

@@ -18,7 +18,7 @@
       ],
       "cflags": ["-fexceptions"],
       "cflags_cc": ["-fexceptions"],
-      "defines": [ "GGML_USE_K_QUANTS", "NAPI_CPP_EXCEPTIONS" ],
+      "defines": ["GGML_USE_K_QUANTS", "NAPI_CPP_EXCEPTIONS"],
       "msvs_settings": {
         "VCCLCompilerTool": { "AdditionalOptions": [ '/arch:AVX2', '/EHsc' ] }
       }

package-lock.json

@@ -40,10 +40,10 @@
     "node": ">=18.0.0"
   },
   "scripts": {
-    "postinstall": "node ./dist/cli/cli.js postinstall",
     "prepare": "[ $CI = true ] || [ -d '.husky/_' ] || husky install",
     "prebuild": "rm -rf ./dist ./tsconfig.tsbuildinfo",
     "build": "tsc --build tsconfig.json --force",
+    "addPostinstallScript": "npm pkg set scripts.postinstall=\"node ./dist/cli/cli.js postinstall\"",
     "generate-docs": "typedoc",
     "prewatch": "rm -rf ./dist ./tsconfig.tsbuildinfo",
     "watch": "tsc --build tsconfig.json --watch --force",
@@ -123,6 +123,7 @@
     "node-stream-zip": "^1.15.0",
     "octokit": "^3.1.0",
     "ora": "^7.0.1",
+    "simple-git": "^3.19.1",
     "uuid": "^9.0.0",
     "yargs": "^17.7.2"
   }

package.json

@@ -1,12 +1,16 @@
+import process from "process";
 import {CommandModule} from "yargs";
 import chalk from "chalk";
 import {compileLlamaCpp} from "../../utils/compileLLamaCpp.js";
 import withOra from "../../utils/withOra.js";
 import {clearTempFolder} from "../../utils/clearTempFolder.js";
+import {defaultLlamaCppCudaSupport, defaultLlamaCppMetalSupport} from "../../config.js";
 
 type BuildCommand = {
     arch?: string,
-    nodeTarget?: string
+    nodeTarget?: string,
+    metal: boolean,
+    cuda: boolean
 };
 
 export const BuildCommand: CommandModule<object, BuildCommand> = {
@@ -23,12 +27,30 @@ export const BuildCommand: CommandModule<object, BuildCommand> = {
                 alias: "t",
                 type: "string",
                 description: "The Node.js version to compile llama.cpp for. Example: v18.0.0"
+            })
+            .option("metal", {
+                type: "boolean",
+                default: defaultLlamaCppMetalSupport,
+                description: "Compile llama.cpp with Metal support. Can also be set via the NODE_LLAMA_CPP_METAL environment variable"
+            })
+            .option("cuda", {
+                type: "boolean",
+                default: defaultLlamaCppCudaSupport,
+                description: "Compile llama.cpp with CUDA support. Can also be set via the NODE_LLAMA_CPP_CUDA environment variable"
             });
     },
     handler: BuildLlamaCppCommand
 };
 
-export async function BuildLlamaCppCommand({arch, nodeTarget}: BuildCommand) {
+export async function BuildLlamaCppCommand({arch, nodeTarget, metal, cuda}: BuildCommand) {
+    if (metal && process.platform === "darwin") {
+        console.log(`${chalk.yellow("Metal:")} enabled`);
+    }
+
+    if (cuda) {
+        console.log(`${chalk.yellow("CUDA:")} enabled`);
+    }
+
     await withOra({
         loading: chalk.blue("Compiling llama.cpp"),
         success: chalk.blue("Compiled llama.cpp"),
@@ -37,7 +59,9 @@ export async function BuildLlamaCppCommand({arch, nodeTarget}: BuildCommand) {
         await compileLlamaCpp({
             arch: arch ? arch : undefined,
             nodeTarget: nodeTarget ? nodeTarget : undefined,
-            setUsedBingFlag: true
+            setUsedBingFlag: true,
+            metal,
+            cuda
         });
     });