add cli

.dockerignore

@@ -29,4 +29,3 @@ web/node_modules/
 web/coverage/
 web/.svelte-kit
 web/build/
-web/.env

.github/DISCUSSION_TEMPLATE/feature-request.yaml

@@ -1,13 +1,11 @@
-title: "[Feature] feature-name-goes-here"
+title: "[Feature] <feature-name-goes-here>"
 labels: ["feature"]
 
 body:
   - type: markdown
     attributes:
       value: |
-        Please use this form to request new feature for Immich.
-        Stick to only a single feature per request. If you list multiple different features at once, 
-        your request will be closed.
+        Please use this form to request new feature for Immich
 
   - type: checkboxes
     attributes:

.github/workflows/cli.yml

@@ -56,10 +56,10 @@ jobs:
         uses: actions/checkout@v4
 
       - name: Set up QEMU
-        uses: docker/setup-qemu-action@v3.2.0
+        uses: docker/setup-qemu-action@v3.1.0
 
       - name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v3.6.1
+        uses: docker/setup-buildx-action@v3.4.0
 
       - name: Login to GitHub Container Registry
         uses: docker/login-action@v3
@@ -88,7 +88,7 @@ jobs:
             type=raw,value=latest,enable=${{ github.event_name == 'release' }}
 
       - name: Build and push image
-        uses: docker/build-push-action@v6.6.1
+        uses: docker/build-push-action@v6.3.0
         with:
           file: cli/Dockerfile
           platforms: linux/amd64,linux/arm64

.github/workflows/docker-cleanup.yml

@@ -35,7 +35,7 @@ jobs:
     steps:
       - name: Clean temporary images
         if: "${{ env.TOKEN != '' }}"
-        uses: stumpylog/image-cleaner-action/ephemeral@v0.8.0
+        uses: stumpylog/image-cleaner-action/ephemeral@v0.7.0
         with:
           token: "${{ env.TOKEN }}"
           owner: "immich-app"
@@ -64,7 +64,7 @@ jobs:
     steps:
       - name: Clean untagged images
         if: "${{ env.TOKEN != '' }}"
-        uses: stumpylog/image-cleaner-action/untagged@v0.8.0
+        uses: stumpylog/image-cleaner-action/untagged@v0.7.0
         with:
           token: "${{ env.TOKEN }}"
           owner: "immich-app"

.github/workflows/docker.yml

@@ -63,10 +63,10 @@ jobs:
         uses: actions/checkout@v4
 
       - name: Set up QEMU
-        uses: docker/setup-qemu-action@v3.2.0
+        uses: docker/setup-qemu-action@v3.1.0
 
       - name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v3.6.1
+        uses: docker/setup-buildx-action@v3.4.0
 
       - name: Login to Docker Hub
         # Only push to Docker Hub when making a release
@@ -115,7 +115,7 @@ jobs:
           fi
 
       - name: Build and push image
-        uses: docker/build-push-action@v6.6.1
+        uses: docker/build-push-action@v6.3.0
         with:
           context: ${{ matrix.context }}
           file: ${{ matrix.file }}

cli/.eslintignore

@@ -0,0 +1 @@
+/dist

cli/.eslintrc.cjs

@@ -0,0 +1,28 @@
+module.exports = {
+  parser: '@typescript-eslint/parser',
+  parserOptions: {
+    project: 'tsconfig.json',
+    sourceType: 'module',
+    tsconfigRootDir: __dirname,
+  },
+  plugins: ['@typescript-eslint/eslint-plugin'],
+  extends: ['plugin:@typescript-eslint/recommended', 'plugin:prettier/recommended', 'plugin:unicorn/recommended'],
+  root: true,
+  env: {
+    node: true,
+  },
+  ignorePatterns: ['.eslintrc.js'],
+  rules: {
+    '@typescript-eslint/interface-name-prefix': 'off',
+    '@typescript-eslint/explicit-function-return-type': 'off',
+    '@typescript-eslint/explicit-module-boundary-types': 'off',
+    '@typescript-eslint/no-explicit-any': 'off',
+    '@typescript-eslint/no-floating-promises': 'error',
+    'unicorn/prefer-module': 'off',
+    'unicorn/prevent-abbreviations': 'off',
+    'unicorn/no-process-exit': 'off',
+    'unicorn/import-style': 'off',
+    curly: 2,
+    'prettier/prettier': 0,
+  },
+};

cli/.nvmrc

@@ -1 +1 @@
-20.16.0
+20.15.1

cli/Dockerfile

@@ -1,4 +1,4 @@
-FROM node:20.16.0-alpine3.20@sha256:eb8101caae9ac02229bd64c024919fe3d4504ff7f329da79ca60a04db08cef52 AS core
+FROM node:20.15.1-alpine3.20@sha256:34b7aa411056c85dbf71d240d26516949b3f72b318d796c26b57caaa1df5639a as core
 
 WORKDIR /usr/src/open-api/typescript-sdk
 COPY open-api/typescript-sdk/package*.json open-api/typescript-sdk/tsconfig*.json ./

cli/eslint.config.mjs

@@ -1,60 +0,0 @@
-import { FlatCompat } from '@eslint/eslintrc';
-import js from '@eslint/js';
-import typescriptEslint from '@typescript-eslint/eslint-plugin';
-import tsParser from '@typescript-eslint/parser';
-import globals from 'globals';
-import path from 'node:path';
-import { fileURLToPath } from 'node:url';
-
-const __filename = fileURLToPath(import.meta.url);
-const __dirname = path.dirname(__filename);
-const compat = new FlatCompat({
-  baseDirectory: __dirname,
-  recommendedConfig: js.configs.recommended,
-  allConfig: js.configs.all,
-});
-
-export default [
-  {
-    ignores: ['eslint.config.mjs', 'dist'],
-  },
-  ...compat.extends(
-    'plugin:@typescript-eslint/recommended',
-    'plugin:prettier/recommended',
-    'plugin:unicorn/recommended',
-  ),
-  {
-    plugins: {
-      '@typescript-eslint': typescriptEslint,
-    },
-
-    languageOptions: {
-      globals: {
-        ...globals.node,
-      },
-
-      parser: tsParser,
-      ecmaVersion: 5,
-      sourceType: 'module',
-
-      parserOptions: {
-        project: 'tsconfig.json',
-        tsconfigRootDir: __dirname,
-      },
-    },
-
-    rules: {
-      '@typescript-eslint/interface-name-prefix': 'off',
-      '@typescript-eslint/explicit-function-return-type': 'off',
-      '@typescript-eslint/explicit-module-boundary-types': 'off',
-      '@typescript-eslint/no-explicit-any': 'off',
-      '@typescript-eslint/no-floating-promises': 'error',
-      'unicorn/prefer-module': 'off',
-      'unicorn/prevent-abbreviations': 'off',
-      'unicorn/no-process-exit': 'off',
-      'unicorn/import-style': 'off',
-      curly: 2,
-      'prettier/prettier': 0,
-    },
-  },
-];

cli/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@immich/cli",
-  "version": "2.2.13",
+  "version": "2.2.8",
   "description": "Command Line Interface (CLI) for Immich",
   "type": "module",
   "exports": "./dist/index.js",
@@ -13,33 +13,30 @@
     "cli"
   ],
   "devDependencies": {
-    "@eslint/eslintrc": "^3.1.0",
-    "@eslint/js": "^9.8.0",
     "@immich/sdk": "file:../open-api/typescript-sdk",
     "@types/byte-size": "^8.1.0",
     "@types/cli-progress": "^3.11.0",
     "@types/lodash-es": "^4.17.12",
     "@types/mock-fs": "^4.13.1",
-    "@types/node": "^20.14.13",
-    "@typescript-eslint/eslint-plugin": "^8.0.0",
-    "@typescript-eslint/parser": "^8.0.0",
-    "@vitest/coverage-v8": "^2.0.5",
-    "byte-size": "^9.0.0",
+    "@types/node": "^20.14.10",
+    "@typescript-eslint/eslint-plugin": "^7.0.0",
+    "@typescript-eslint/parser": "^7.0.0",
+    "@vitest/coverage-v8": "^1.2.2",
+    "byte-size": "^8.1.1",
     "cli-progress": "^3.12.0",
     "commander": "^12.0.0",
-    "eslint": "^9.0.0",
+    "eslint": "^8.56.0",
     "eslint-config-prettier": "^9.1.0",
     "eslint-plugin-prettier": "^5.1.3",
-    "eslint-plugin-unicorn": "^55.0.0",
-    "globals": "^15.9.0",
+    "eslint-plugin-unicorn": "^54.0.0",
     "mock-fs": "^5.2.0",
     "prettier": "^3.2.5",
     "prettier-plugin-organize-imports": "^4.0.0",
     "typescript": "^5.3.3",
     "vite": "^5.0.12",
     "vite-tsconfig-paths": "^4.3.2",
-    "vitest": "^2.0.5",
-    "vitest-fetch-mock": "^0.3.0",
+    "vitest": "^1.2.2",
+    "vitest-fetch-mock": "^0.2.2",
     "yaml": "^2.3.1"
   },
   "scripts": {
@@ -67,6 +64,6 @@
     "lodash-es": "^4.17.21"
   },
   "volta": {
-    "node": "20.16.0"
+    "node": "20.15.1"
   }
 }

deployment/modules/cloudflare/docs-release/.terraform.lock.hcl

@@ -2,37 +2,37 @@
 # Manual edits may be lost in future updates.
 
 provider "registry.opentofu.org/cloudflare/cloudflare" {
-  version     = "4.38.0"
-  constraints = "4.38.0"
+  version     = "4.36.0"
+  constraints = "4.36.0"
   hashes = [
-    "h1:+27KAHKHBDvv3dqyJv5vhtdKQZJzoZXoMqIyronlHNw=",
-    "h1:/uV9RgOUhkxElkHhWs8fs5ZbX9vj6RCBfP0oJO0JF30=",
-    "h1:1DNAdMugJJOAWD/XYiZenYYZLy7fw2ctjT4YZmkRCVQ=",
-    "h1:1wn4PmCLdT7mvd74JkCGmJDJxTQDkcxc+1jNbmwnMHA=",
-    "h1:BIHB4fBxHg2bA9KbL92njhyctxKC8b6hNDp60y5QBss=",
-    "h1:HCQpvKPsMsR4HO5eDqt+Kao7T7CYeEH7KZIO7xMcC6M=",
-    "h1:HTomuzocukpNLwtWzeSF3yteCVsyVKbwKmN66u9iPac=",
-    "h1:YDxsUBhBAwHSXLzVwrSlSBOwv1NvLyry7s5SfCV7VqQ=",
-    "h1:dchVhxo+Acd1l2RuZ88tW9lWj4422QMfgtxKvKCjYrw=",
-    "h1:eypa+P4ZpsEGMPFuCE+6VkRefu0TZRFmVBOpK+PDOPY=",
-    "h1:f3yjse2OsRZj7ZhR7BLintJMlI4fpyt8HyDP/zcEavw=",
-    "h1:mSJ7xj8K+xcnEmGg7lH0jjzyQb157wH94ULTAlIV+HQ=",
-    "h1:tt+2J2Ze8VIdDq2Hr6uHlTJzAMBRpErBwTYx0uD5ilE=",
-    "h1:uQW8SKxmulqrAisO+365mIf2FueINAp5PY28bqCPCug=",
-    "zh:171ab67cccceead4514fafb2d39e4e708a90cce79000aaf3c29aab7ed4457071",
-    "zh:18aa7228447baaaefc49a43e8eff970817a7491a63d8937e796357a3829dd979",
-    "zh:2cbaab6092e81ba6f41fa60a50f14e980c8ec327ee11d0b21f16a478be4b7567",
-    "zh:53b8e49c06f5b31a8c681f8c0669cf43e78abe71657b8182a221d096bb514965",
-    "zh:6037cfc60b4b647aabae155fcb46d649ed7c650e0287f05db52b2068f1e27c8a",
-    "zh:62460982ce1a869eebfca675603fbbd50416cf6b69459fb855bfbe5ae2b97607",
-    "zh:65f6f3a8470917b6398baa5eb4f74b3932b213eac7c0202798bfad6fd1ee17df",
+    "h1:00/Y+l17VV4RquGSfwDnYsGYzyf2ZmdQwUgeIzXC7eg=",
+    "h1:489GpKItA/VRIUA5S4+F8MsnurGVciRvUFyIV81MJTU=",
+    "h1:7cnczyKGj3+gvaJ0r5JIVWLXPbQfkHYejac76MJx+I8=",
+    "h1:8rmr1PjJc14Xmor2eEvo5/WBojylt1eYdx6VbSU3Ulo=",
+    "h1:HjgphNjtgny5tkcUAQoGgBdcuQ+0IyhL8yLsiBqWAP0=",
+    "h1:LH3umxdBnJcAyeVoBLVn+PC0F0CzN6v9UN6lb6CqQPE=",
+    "h1:Xx6WUD/zB8fM9SjkFx06Fgx2K7aGJIVvsJS2pwqALEM=",
+    "h1:YizL5YN9zQ8YkSR6V/G201YrCVdnkF9EUIK4lpROWiA=",
+    "h1:aPcXVGjYcCJdqvWSzc/dEjwj05LnbWZje8IanygVjcI=",
+    "h1:eKCvfashdCqfDcFGXE2gq+XxAURD5SzuaQ9Brs3zLos=",
+    "h1:gpKcBYkBcfn/uF1A8W7MD/OysMZW7EU4QVYvPEEnxGc=",
+    "h1:kCkcxZZnkKAnMz9scUQHb19d9/l9FPOHovAyrvtA618=",
+    "h1:t8mXXnICTeKqoD29uvyLFHVWMfMzTUrJuHje8lpI0zU=",
+    "h1:zjzavjIdLDGRYsWd3v0HJz6ul12Cewj9RW/cqAQ4DxI=",
+    "zh:02665712b3893307596b3caab99cf1f2502d5caca18e22d4b37bb535e628e102",
+    "zh:1514b0d3ef62934484ac471113ee68cddec0c21e56b4f710922741fe9b6e6fdf",
+    "zh:1fab4dfcecbcea13267b42e5ff05ba0692aa2dcb247b8e633fea0daf49feb156",
+    "zh:24d8367295fe1f1b2be37802aecb96edf32f743364663ffe781d1bb92438395d",
+    "zh:34e84e7940c99dcf65663cfd25afac22bf5c8a5ff2cd21900c67180d3a072be9",
+    "zh:3d71d63204a329acf1d1de8638f2c725243cb94cf444d2d7acde54b3d1ac1696",
+    "zh:57831ba88e779a762bcfa224ba9eac8bc22ef9cd70cd541d848b351e0ba6a75c",
+    "zh:6407560f2e548afcb4852c91efc664627a9ee565c31a9c81fc9ea1806fca0567",
+    "zh:738ddbc664d75f4859aa09444a27809bc398795a8ea8f5be8531040690287712",
+    "zh:841ca2b2d78b6f8d33ec3435bc090c5e04a3a7d85c80df11227a7ea00d36f6b1",
     "zh:890df766e9b839623b1f0437355032a3c006226a6c200cd911e15ee1a9014e9f",
-    "zh:8b5cebe64bf04105a49178a165b6a8800a9a33bae6767143a47fe4977755f805",
-    "zh:a5596635db0993ee3c3060fbc2227d91b239466e96d2d82642625a5aa2486988",
-    "zh:b3a9c63038441f13c311fd4b2c7e69e571445e5a7365a20c7cc9046b7e6c8aba",
-    "zh:b585e7e4d7648a540b14b9182819214896ca9337729eeb1f2034833b17db754d",
-    "zh:d2c3c545318ac8542369e9fc8228e29ee585febdf203a450fad3e0eded71ce02",
-    "zh:e95dd2d6c3525073af47d47b763cb81b6a51b20cabf76f789c69328922da9ecf",
-    "zh:eee6e590b36d6c6168a7daae8afa74a8721fd7aa9f62a710f04a311975100722",
+    "zh:8b3d3d63354032ab9b2403c50728e9aa4e83c7367eaad2d18794221addeafc0f",
+    "zh:9e293443fe3127e488f540229983c1b9688268185f87567bb3d18e794697acd2",
+    "zh:b3a22439156e46461213db183e2e89569cd2e8d7cbcfc4b9f90469090e105807",
+    "zh:f430feb5d51891e84028459e57039045dea4f1f5fcf671161d8ac2d8f28763f3",
   ]
 }

deployment/modules/cloudflare/docs-release/config.tf

@@ -5,7 +5,7 @@ terraform {
   required_providers {
     cloudflare = {
       source = "cloudflare/cloudflare"
-      version = "4.38.0"
+      version = "4.36.0"
     }
   }
 }

deployment/modules/cloudflare/docs/.terraform.lock.hcl

@@ -2,37 +2,37 @@
 # Manual edits may be lost in future updates.
 
 provider "registry.opentofu.org/cloudflare/cloudflare" {
-  version     = "4.38.0"
-  constraints = "4.38.0"
+  version     = "4.36.0"
+  constraints = "4.36.0"
   hashes = [
-    "h1:+27KAHKHBDvv3dqyJv5vhtdKQZJzoZXoMqIyronlHNw=",
-    "h1:/uV9RgOUhkxElkHhWs8fs5ZbX9vj6RCBfP0oJO0JF30=",
-    "h1:1DNAdMugJJOAWD/XYiZenYYZLy7fw2ctjT4YZmkRCVQ=",
-    "h1:1wn4PmCLdT7mvd74JkCGmJDJxTQDkcxc+1jNbmwnMHA=",
-    "h1:BIHB4fBxHg2bA9KbL92njhyctxKC8b6hNDp60y5QBss=",
-    "h1:HCQpvKPsMsR4HO5eDqt+Kao7T7CYeEH7KZIO7xMcC6M=",
-    "h1:HTomuzocukpNLwtWzeSF3yteCVsyVKbwKmN66u9iPac=",
-    "h1:YDxsUBhBAwHSXLzVwrSlSBOwv1NvLyry7s5SfCV7VqQ=",
-    "h1:dchVhxo+Acd1l2RuZ88tW9lWj4422QMfgtxKvKCjYrw=",
-    "h1:eypa+P4ZpsEGMPFuCE+6VkRefu0TZRFmVBOpK+PDOPY=",
-    "h1:f3yjse2OsRZj7ZhR7BLintJMlI4fpyt8HyDP/zcEavw=",
-    "h1:mSJ7xj8K+xcnEmGg7lH0jjzyQb157wH94ULTAlIV+HQ=",
-    "h1:tt+2J2Ze8VIdDq2Hr6uHlTJzAMBRpErBwTYx0uD5ilE=",
-    "h1:uQW8SKxmulqrAisO+365mIf2FueINAp5PY28bqCPCug=",
-    "zh:171ab67cccceead4514fafb2d39e4e708a90cce79000aaf3c29aab7ed4457071",
-    "zh:18aa7228447baaaefc49a43e8eff970817a7491a63d8937e796357a3829dd979",
-    "zh:2cbaab6092e81ba6f41fa60a50f14e980c8ec327ee11d0b21f16a478be4b7567",
-    "zh:53b8e49c06f5b31a8c681f8c0669cf43e78abe71657b8182a221d096bb514965",
-    "zh:6037cfc60b4b647aabae155fcb46d649ed7c650e0287f05db52b2068f1e27c8a",
-    "zh:62460982ce1a869eebfca675603fbbd50416cf6b69459fb855bfbe5ae2b97607",
-    "zh:65f6f3a8470917b6398baa5eb4f74b3932b213eac7c0202798bfad6fd1ee17df",
+    "h1:00/Y+l17VV4RquGSfwDnYsGYzyf2ZmdQwUgeIzXC7eg=",
+    "h1:489GpKItA/VRIUA5S4+F8MsnurGVciRvUFyIV81MJTU=",
+    "h1:7cnczyKGj3+gvaJ0r5JIVWLXPbQfkHYejac76MJx+I8=",
+    "h1:8rmr1PjJc14Xmor2eEvo5/WBojylt1eYdx6VbSU3Ulo=",
+    "h1:HjgphNjtgny5tkcUAQoGgBdcuQ+0IyhL8yLsiBqWAP0=",
+    "h1:LH3umxdBnJcAyeVoBLVn+PC0F0CzN6v9UN6lb6CqQPE=",
+    "h1:Xx6WUD/zB8fM9SjkFx06Fgx2K7aGJIVvsJS2pwqALEM=",
+    "h1:YizL5YN9zQ8YkSR6V/G201YrCVdnkF9EUIK4lpROWiA=",
+    "h1:aPcXVGjYcCJdqvWSzc/dEjwj05LnbWZje8IanygVjcI=",
+    "h1:eKCvfashdCqfDcFGXE2gq+XxAURD5SzuaQ9Brs3zLos=",
+    "h1:gpKcBYkBcfn/uF1A8W7MD/OysMZW7EU4QVYvPEEnxGc=",
+    "h1:kCkcxZZnkKAnMz9scUQHb19d9/l9FPOHovAyrvtA618=",
+    "h1:t8mXXnICTeKqoD29uvyLFHVWMfMzTUrJuHje8lpI0zU=",
+    "h1:zjzavjIdLDGRYsWd3v0HJz6ul12Cewj9RW/cqAQ4DxI=",
+    "zh:02665712b3893307596b3caab99cf1f2502d5caca18e22d4b37bb535e628e102",
+    "zh:1514b0d3ef62934484ac471113ee68cddec0c21e56b4f710922741fe9b6e6fdf",
+    "zh:1fab4dfcecbcea13267b42e5ff05ba0692aa2dcb247b8e633fea0daf49feb156",
+    "zh:24d8367295fe1f1b2be37802aecb96edf32f743364663ffe781d1bb92438395d",
+    "zh:34e84e7940c99dcf65663cfd25afac22bf5c8a5ff2cd21900c67180d3a072be9",
+    "zh:3d71d63204a329acf1d1de8638f2c725243cb94cf444d2d7acde54b3d1ac1696",
+    "zh:57831ba88e779a762bcfa224ba9eac8bc22ef9cd70cd541d848b351e0ba6a75c",
+    "zh:6407560f2e548afcb4852c91efc664627a9ee565c31a9c81fc9ea1806fca0567",
+    "zh:738ddbc664d75f4859aa09444a27809bc398795a8ea8f5be8531040690287712",
+    "zh:841ca2b2d78b6f8d33ec3435bc090c5e04a3a7d85c80df11227a7ea00d36f6b1",
     "zh:890df766e9b839623b1f0437355032a3c006226a6c200cd911e15ee1a9014e9f",
-    "zh:8b5cebe64bf04105a49178a165b6a8800a9a33bae6767143a47fe4977755f805",
-    "zh:a5596635db0993ee3c3060fbc2227d91b239466e96d2d82642625a5aa2486988",
-    "zh:b3a9c63038441f13c311fd4b2c7e69e571445e5a7365a20c7cc9046b7e6c8aba",
-    "zh:b585e7e4d7648a540b14b9182819214896ca9337729eeb1f2034833b17db754d",
-    "zh:d2c3c545318ac8542369e9fc8228e29ee585febdf203a450fad3e0eded71ce02",
-    "zh:e95dd2d6c3525073af47d47b763cb81b6a51b20cabf76f789c69328922da9ecf",
-    "zh:eee6e590b36d6c6168a7daae8afa74a8721fd7aa9f62a710f04a311975100722",
+    "zh:8b3d3d63354032ab9b2403c50728e9aa4e83c7367eaad2d18794221addeafc0f",
+    "zh:9e293443fe3127e488f540229983c1b9688268185f87567bb3d18e794697acd2",
+    "zh:b3a22439156e46461213db183e2e89569cd2e8d7cbcfc4b9f90469090e105807",
+    "zh:f430feb5d51891e84028459e57039045dea4f1f5fcf671161d8ac2d8f28763f3",
   ]
 }

deployment/modules/cloudflare/docs/config.tf

@@ -5,7 +5,7 @@ terraform {
   required_providers {
     cloudflare = {
       source = "cloudflare/cloudflare"
-      version = "4.38.0"
+      version = "4.36.0"
     }
   }
 }

docker/docker-compose.dev.yml

@@ -46,8 +46,6 @@ services:
     depends_on:
       - redis
       - database
-    healthcheck:
-      disable: false
 
   immich-web:
     container_name: immich_web
@@ -93,12 +91,10 @@ services:
     depends_on:
       - database
     restart: unless-stopped
-    healthcheck:
-      disable: false
 
   redis:
     container_name: immich_redis
-    image: redis:6.2-alpine@sha256:e3b17ba9479deec4b7d1eeec1548a253acc5374d68d3b27937fcfe4df8d18c7e
+    image: redis:6.2-alpine@sha256:328fe6a5822256d065debb36617a8169dbfbd77b797c525288e465f56c1d392b
     healthcheck:
       test: redis-cli ping || exit 1
 

docker/docker-compose.prod.yml

@@ -21,8 +21,6 @@ services:
       - redis
       - database
     restart: always
-    healthcheck:
-      disable: false
 
   immich-machine-learning:
     container_name: immich_machine_learning
@@ -35,19 +33,15 @@ services:
       dockerfile: Dockerfile
       args:
         - DEVICE=cpu # set to one of [armnn, cuda, openvino, openvino-wsl] for accelerated inference
-    ports:
-      - 3003:3003
     volumes:
       - model-cache:/cache
     env_file:
       - .env
     restart: always
-    healthcheck:
-      disable: false
 
   redis:
     container_name: immich_redis
-    image: redis:6.2-alpine@sha256:e3b17ba9479deec4b7d1eeec1548a253acc5374d68d3b27937fcfe4df8d18c7e
+    image: redis:6.2-alpine@sha256:328fe6a5822256d065debb36617a8169dbfbd77b797c525288e465f56c1d392b
     healthcheck:
       test: redis-cli ping || exit 1
     restart: always
@@ -71,7 +65,7 @@ services:
       interval: 5m
       start_interval: 30s
       start_period: 5m
-    command: ["postgres", "-c", "shared_preload_libraries=vectors.so", "-c", 'search_path="$$user", public, vectors', "-c", "logging_collector=on", "-c", "max_wal_size=2GB", "-c", "shared_buffers=512MB", "-c", "wal_compression=on"]
+    command: ["postgres", "-c" ,"shared_preload_libraries=vectors.so", "-c", 'search_path="$$user", public, vectors', "-c", "logging_collector=on", "-c", "max_wal_size=2GB", "-c", "shared_buffers=512MB", "-c", "wal_compression=on"]
     restart: always
 
   # set IMMICH_METRICS=true in .env to enable metrics
@@ -91,7 +85,7 @@ services:
     command: ['./run.sh', '-disable-reporting']
     ports:
       - 3000:3000
-    image: grafana/grafana:11.1.3-ubuntu@sha256:e10453733015f31103cb530425f32c994816b50102886fa885dafea2c50a711c
+    image: grafana/grafana:11.1.0-ubuntu@sha256:c7fc29ec783d5e7fc1bdfaad6f92345a345cffbc5d21c388ca228175006fc107
     volumes:
       - grafana-data:/var/lib/grafana
 

docker/docker-compose.yml

@@ -16,7 +16,6 @@ services:
     #   file: hwaccel.transcoding.yml
     #   service: cpu # set to one of [nvenc, quicksync, rkmpp, vaapi, vaapi-wsl] for accelerated transcoding
     volumes:
-      # Do not edit the next line. If you want to change the media storage location on your system, edit the value of UPLOAD_LOCATION in the .env file
       - ${UPLOAD_LOCATION}:/usr/src/app/upload
       - /etc/localtime:/etc/localtime:ro
     env_file:
@@ -27,8 +26,6 @@ services:
       - redis
       - database
     restart: always
-    healthcheck:
-      disable: false
 
   immich-machine-learning:
     container_name: immich_machine_learning
@@ -43,12 +40,10 @@ services:
     env_file:
       - .env
     restart: always
-    healthcheck:
-      disable: false
 
   redis:
     container_name: immich_redis
-    image: docker.io/redis:6.2-alpine@sha256:e3b17ba9479deec4b7d1eeec1548a253acc5374d68d3b27937fcfe4df8d18c7e
+    image: docker.io/redis:6.2-alpine@sha256:328fe6a5822256d065debb36617a8169dbfbd77b797c525288e465f56c1d392b
     healthcheck:
       test: redis-cli ping || exit 1
     restart: always
@@ -62,14 +57,13 @@ services:
       POSTGRES_DB: ${DB_DATABASE_NAME}
       POSTGRES_INITDB_ARGS: '--data-checksums'
     volumes:
-      # Do not edit the next line. If you want to change the database storage location on your system, edit the value of DB_DATA_LOCATION in the .env file
       - ${DB_DATA_LOCATION}:/var/lib/postgresql/data
     healthcheck:
       test: pg_isready --dbname='${DB_DATABASE_NAME}' --username='${DB_USERNAME}' || exit 1; Chksum="$$(psql --dbname='${DB_DATABASE_NAME}' --username='${DB_USERNAME}' --tuples-only --no-align --command='SELECT COALESCE(SUM(checksum_failures), 0) FROM pg_stat_database')"; echo "checksum failure count is $$Chksum"; [ "$$Chksum" = '0' ] || exit 1
       interval: 5m
       start_interval: 30s
       start_period: 5m
-    command: ["postgres", "-c", "shared_preload_libraries=vectors.so", "-c", 'search_path="$$user", public, vectors', "-c", "logging_collector=on", "-c", "max_wal_size=2GB", "-c", "shared_buffers=512MB", "-c", "wal_compression=on"]
+    command: ["postgres", "-c" ,"shared_preload_libraries=vectors.so", "-c", 'search_path="$$user", public, vectors', "-c", "logging_collector=on", "-c", "max_wal_size=2GB", "-c", "shared_buffers=512MB", "-c", "wal_compression=on"]
     restart: always
 
 volumes:

docs/.nvmrc

@@ -1 +1 @@
-20.16.0
+20.15.1

docs/blog/2024/immich-core-team-goes-fulltime.mdx

@@ -1,7 +1,7 @@
 ---
 title: The Immich core team goes full-time
 authors: [alextran]
-tags: [update, announcement, FUTO]
+tags: [update, announcement, futo]
 date: 2024-05-01T00:00
 ---
 

docs/blog/2024/immich-licensing.mdx

@@ -1,91 +0,0 @@
----
-title: Licensing announcement - Purchase a license to support Immich
-authors: [alextran]
-tags: [update, announcement, FUTO]
-date: 2024-07-18T00:00
----
-
-Hello everybody,
-
-Firstly, on behalf of the Immich team, I'd like to thank everybody for your continuous support of Immich since the very first day! Your contributions, encouragement, and community engagement have helped bring Immich to its current state. The team and I are forever grateful for that.
-
-Since our [last announcement of the core team joining FUTO to work on Immich full-time](https://immich.app/blog/2024/immich-core-team-goes-fulltime), one of the goals of our new position is to foster a healthy relationship between the developers and the users. We believe that this enables us to create great software, establish transparent policies and build trust.
-
-We want to build a great software application that brings value to you and your loved ones' lives. We are not using you as a product, i.e., selling or tracking your data. We are not putting annoying ads into our software. We respect your privacy. We want to be compensated for the hard work we put in to build Immich for you.
-
-With those notes, we have enabled a way for you to financially support the continued development of Immich, ensuring the software can move forward and will be maintained, by offering a lifetime license of the software. We think if you like and use software, you should pay for it, but _we're never going to force anyone to pay or try to limit Immich for those who don't._
-
-There are two types of license that you can choose to purchase: **Server License** and **Individual License**.
-
-### Server License
-
-This is a lifetime license costing **$99.99**. The license is applied to the whole server. You and all users that use your server are licensed.
-
-### Individual License
-
-This is a lifetime license costing **$24.99**. The license is applied to a single user, and can be used on any server they choose to connect to.
-
-<img
-  width="837"
-  alt="license-social-gh"
-  src="https://github.com/user-attachments/assets/241932ed-ef3b-44ec-a9e2-ee80754e0cca"
-/>
-
-You can purchase the license on [our page - https://buy.immich.app](https://buy.immich.app).
-
-Starting with release `v1.109.0` you can purchase and enter your purchased license key directly in the app.
-
-<img
-  width="1414"
-  alt="license-page-gh"
-  src="https://github.com/user-attachments/assets/364fc32a-f6ef-4594-9fea-28d5a26ad77c"
-/>
-
-## Thank you
-
-Thank you again for your support, this will help create a strong foundation and stability for the Immich team to continue developing and maintaining the project that you love to use.
-
-<p align="center">
-  <img
-    src="https://media.giphy.com/media/v1.Y2lkPTc5MGI3NjExbjY2eWc5Y2F0ZW56MmR4aWE0dDhzZXlidXRmYWZyajl1bWZidXZpcyZlcD12MV9pbnRlcm5hbF9naWZfYnlfaWQmY3Q9Zw/87CKDqErVfMqY/giphy.gif"
-    width="550"
-    title="SUPPORT THE PROJECT!"
-  />
-</p>
-
-<br />
-<br />
-
-Cheers! 🎉
-
-Immich team
-
-# FAQ
-
-### 1. Where can I purchase a license?
-
-There are several places where you can purchase the license from
-
-- [https://buy.immich.app](https://buy.immich.app)
-- [https://pay.futo.org](https://pay.futo.org/)
-- or directly from the app.
-
-### 2. Do I need both _Individual License_ and _Server License_?
-
-No,
-
-If you are the admin and the sole user, or your instance has less than a total of 4 users, you can buy the **Individual License** for each user.
-
-If your instance has more than 4 users, it is more cost-effective to buy the **Server License**, which will license all the users on your instance.
-
-### 3. What do I do if I don't pay?
-
-You can continue using Immich without any restriction.
-
-### 4. Will there be any paywalled features?
-
-No, there will never be any paywalled features.
-
-### 5. Where can I get support regarding payment issues?
-
-You can email us with your `orderId` and your email address `billing@futo.org` or on our Discord server.

docs/blog/2024/update-july-2024.mdx

@@ -1,7 +1,6 @@
 ---
 title: Immich Update - July 2024
 authors: [alextran]
-date: 2024-07-01T00:00
 tags: [update, v1.106.0]
 ---
 

docs/docs/FAQ.mdx

@@ -167,7 +167,7 @@ Immich uses CLIP models. For more information about CLIP and its capabilities, r
 
 ### How does facial recognition work?
 
-See [How Facial Recognition Works](/docs/features/facial-recognition#How-Facial-Recognition-Works) for details.
+For face detection and recognition, Immich uses [InsightFace models](https://github.com/deepinsight/insightface/tree/master/model_zoo).
 
 ### How can I disable machine learning?
 
@@ -181,15 +181,19 @@ However, disabling all jobs will not disable the machine learning service itself
 
 ### I'm getting errors about models being corrupt or failing to download. What do I do?
 
-You can delete the model cache volume, where models are downloaded. This will give the service a clean environment to download the model again. If models are failing to download entirely, you can manually download them from [Hugging Face][huggingface] and place them in the cache folder.
+You can delete the model cache volume, where models are downloaded. This will give the service a clean environment to download the model again. If models are failing to download entirely, you can manually download them from [Huggingface][huggingface] and place them in the cache folder.
 
 ### Can I use a custom CLIP model?
 
-No, this is not supported. Only models listed in the [Hugging Face][huggingface] page are compatible. Feel free to make a feature request if there's a model not listed here that you think should be added.
+No, this is not supported. Only models listed in the [Huggingface][huggingface] page are compatible. Feel free to make a feature request if there's a model not listed here that you think should be added.
 
 ### I want to be able to search in other languages besides English. How can I do that?
 
-You can change to a multilingual CLIP model. See [here](/docs/features/smart-search#CLIP-model) for instructions.
+You can change to a multilingual model listed [here](https://huggingface.co/collections/immich-app/multilingual-clip-654eb08c2382f591eeb8c2a7) by going to Administration > Machine Learning Settings > Smart Search and replacing the name of the model. Be sure to re-run Smart Search on all assets after this change. You can then search in over 100 languages.
+
+:::note
+Feel free to make a feature request if there's a model you want to use that isn't in [Immich Huggingface list][huggingface].
+:::
 
 ### Does Immich support Facial Recognition for videos?
 
@@ -230,7 +234,7 @@ ls clip/ facial-recognition/
 
 ### Why is Immich slow on low-memory systems like the Raspberry Pi?
 
-Immich optionally uses transcoding and machine learning for several features. However, it can be too heavy to run on a Raspberry Pi. You can [mitigate](/docs/FAQ#can-i-lower-cpu-and-ram-usage) this or host Immich's machine-learning container on a [more powerful system](/docs/guides/remote-machine-learning), or [disable](/docs/FAQ#how-can-i-disable-machine-learning) machine learning entirely.
+Immich optionally uses machine learning for several features. However, it can be too heavy to run on a Raspberry Pi. You can [mitigate](/docs/FAQ#can-i-lower-cpu-and-ram-usage) this or host Immich's machine-learning container on a [more powerful system](/docs/guides/remote-machine-learning), or [disable](/docs/FAQ#how-can-i-disable-machine-learning) machine learning entirely.
 
 ### Can I lower CPU and RAM usage?
 
@@ -239,12 +243,10 @@ The initial backup is the most intensive due to the number of jobs running. The
 - Lower the job concurrency for these jobs to 1.
 - Under Settings > Transcoding Settings > Threads, set the number of threads to a low number like 1 or 2.
 - Under Settings > Machine Learning Settings > Facial Recognition > Model Name, you can change the facial recognition model to `buffalo_s` instead of `buffalo_l`. The former is a smaller and faster model, albeit not as good.
-  - For facial recognition on new images to work properly, You must re-run the Face Detection job for all images after this.
-- At the container level, you can [set resource constraints](/docs/FAQ#can-i-limit-cpu-and-ram-usage) to lower usage further.
-  - It's recommended to only apply these constraints _after_ taking some of the measures here for best performance.
+- For facial recognition on new images to work properly, You must re-run the Face Detection job for all images after this.
 - If these changes are not enough, see [below](/docs/FAQ#how-can-i-disable-machine-learning) for instructions on how to disable machine learning.
 
-### Can I limit CPU and RAM usage?
+### Can I limit the amount of CPU and RAM usage?
 
 By default, a container has no resource constraints and can use as much of a given resource as the host's kernel scheduler allows. To limit this, you can add the following to the `docker-compose.yml` block of any containers that you want to have limited resources.
 
@@ -264,8 +266,6 @@ deploy:
 </details>
 For more details, you can look at the [original docker docs](https://docs.docker.com/config/containers/resource_constraints/) or use this [guide](https://www.baeldung.com/ops/docker-memory-limit).
 
-Note that memory constraints work by terminating the container, so this can introduce instability if set too low.
-
 ### How can I boost machine learning speed?
 
 :::note
@@ -275,16 +275,21 @@ This advice improves throughput, not latency. This is to say that it will make S
 You can increase throughput by increasing the job concurrency for machine learning jobs (Smart Search, Face Detection). With higher concurrency, the host will work on more assets in parallel. You can do this by navigating to Administration > Settings > Job Settings and increasing concurrency as needed.
 
 :::danger
-On a normal machine, 2 or 3 concurrent jobs can probably max the CPU. Storage speed and latency can quickly become the limiting factor beyond this, particularly when using HDDs.
+On a normal machine, 2 or 3 concurrent jobs can probably max the CPU. Beyond this, note that storage speed and latency may quickly become the limiting factor; particularly when using HDDs.
 
-The concurrency can be increased more comfortably with a GPU, but should still not be above 16 in most cases.
+Do not exaggerate with the amount of jobs because you're probably thoroughly overloading the server.
 
-Do not exaggerate with the job concurrency because you're probably thoroughly overloading the server.
+More details can be found [here](https://discord.com/channels/979116623879368755/994044917355663450/1174711719994605708)
 :::
 
-### My server shows Server Status Offline | Version Unknown. What can I do?
+### Why is Immich using so much of my CPU?
 
-You need to enable WebSockets on your reverse proxy.
+When a large number of assets are uploaded to Immich, it makes sense that the CPU and RAM will be heavily used for machine learning work and creating image thumbnails.
+Once this process is completed, the percentage of CPU usage will drop to around 3-5% usage
+
+### My server shows Server Status Offline | Version Unknown what can I do?
+
+You need to enable Websocket on your reverse proxy.
 
 ---
 
@@ -294,12 +299,6 @@ You need to enable WebSockets on your reverse proxy.
 
 Immich components are typically deployed using docker. To see logs for deployed docker containers, you can use the [Docker CLI](https://docs.docker.com/engine/reference/commandline/cli/), specifically the `docker logs` command. For examples, see [Docker Help](/docs/guides/docker-help.md).
 
-### How can I reduce the log verbosity of Redis?
-
-To decrease Redis logs, you can add the following line to the `redis:` section of the `docker-compose.yml`:
-
-`    command: redis-server --loglevel warning`
-
 ### How can I run Immich as a non-root user?
 
 You can change the user in the container by setting the `user` argument in `docker-compose.yml` for each service.

docs/docs/administration/backup-and-restore.md

@@ -149,21 +149,9 @@ for more info read the [release notes](https://github.com/immich-app/immich/rele
   - Preview images (small thumbnails and large previews) for each asset and thumbnails for recognized faces.
   - Stored in `UPLOAD_LOCATION/thumbs/<userID>`.
 - **Encoded Assets:**
-
   - Videos that have been re-encoded from the original for wider compatibility. The original is not removed.
   - Stored in `UPLOAD_LOCATION/encoded-video/<userID>`.
 
-- **Postgres**
-
-  - The Immich database containing all the information to allow the system to function properly.  
-    **Note:** This folder will only appear to users who have made the changes mentioned in [v1.102.0](https://github.com/immich-app/immich/discussions/8930) (an optional, non-mandatory change) or who started with this version.
-  - Stored in `UPLOAD_LOCATION/postgres`.
-
-  :::danger
-  A backup of this folder does not constitute a backup of your database!
-  Follow the instructions listed [here](/docs/administration/backup-and-restore#database) to learn how to perform a proper backup.
-  :::
-
 </TabItem>
   <TabItem value="Storage Template On" label="Storage Template On">
 
@@ -199,19 +187,8 @@ When you turn off the storage template engine, it will leave the assets in `UPLO
   - Files uploaded through mobile apps.
   - Temporarily located in `UPLOAD_LOCATION/upload/<userID>`.
   - Transferred to `UPLOAD_LOCATION/library/<userID>` upon successful upload.
-- **Postgres**
-
-  - The Immich database containing all the information to allow the system to function properly.  
-    **Note:** This folder will only appear to users who have made the changes mentioned in [v1.102.0](https://github.com/immich-app/immich/discussions/8930) (an optional, non-mandatory change) or who started with this version.
-  - Stored in `UPLOAD_LOCATION/postgres`.
-
-  :::danger
-  A backup of this folder does not constitute a backup of your database!
-  Follow the instructions listed [here](/docs/administration/backup-and-restore#database) to learn how to perform a proper backup.
-  :::
 
 </TabItem>
-
 </Tabs>
 
 :::danger

docs/docs/developer/setup.md

@@ -24,7 +24,7 @@ This environment includes the services below. Additional details are available i
 - Web app - [`/web`](https://github.com/immich-app/immich/tree/main/web)
 - Machine learning - [`/machine-learning`](https://github.com/immich-app/immich/tree/main/machine-learning)
 - Redis
-- PostgreSQL development database with exposed port `5432` so you can use any database client to access it
+- PostgreSQL development database with exposed port `5432` so you can use any database client to acess it
 
 All the services are packaged to run as with single Docker Compose command.
 

docs/docs/features/facial-recognition.md

@@ -2,7 +2,7 @@
 
 ## Overview
 
-Immich recognizes faces in your photos and videos and groups them together into people. You can then assign names to these people and search for them.
+Immich recognizes faces in your photos and videos and groups them together. You can then assign names to the faces and search for them.
 
 The list of people is shown in the Explore page.
 
@@ -18,75 +18,13 @@ The asset detail view will also show the faces that are recognized in the asset.
 
 ## Actions
 
-Additional actions you can do include:
+Additional actions you can do with a detected person are:
 
-- Changing the feature photo of the person
-- Setting a person's date of birth
-- Merging two or more detected faces into one person
-- Hiding the faces of a person from the Explore page and detail view
-- Assigning an unrecognized face to a person
+- Change the feature face photo of the person
+- Set date of birth
+- Merge two or more detected faces into one person
+- Hide face
 
 It can be found from the app bar when you access the detail view of a person.
 
 <img src={require('./img/facial-recognition-4.png').default} title='Facial Recognition 4' width="70%"/>
-
-## How Face Detection Works
-
-Face detection sends the generated preview image to the machine learning service for processing. The service checks if it has the relevant model downloaded and downloads it if not. The image is decoded, pre-processed and passed to the face detection model (with hardware acceleration if configured). The bounding boxes and scores outputted from this model are used to crop and preprocess the image once again to be passed to a facial recognition model (also accelerated if configured). The embeddings from the recognition model, together with the bounding boxes and scores from the face detection model, are then sent back to the server to be added to the database. The embeddings in particular are indexed so they can be searched quickly during facial recognition clustering.
-
-## How Facial Recognition Works
-
-The facial recognition algorithm we use is derived from [DBSCAN](https://www.youtube.com/watch?v=RDZUdRSDOok), a popular clustering algorithm. It essentially treats each detected face as a point in a graph and aims to group points that are close to each other.
-
-:::note
-An important concept is whether something is a _core point_. A core point has a minimum number of points around it within a certain distance. A non-core point can only be assigned to a cluster if it can reach a core point; a non-core point can't be used to extend a cluster even if it's part of one. In Immich, the _Minimum Recognized Faces_ setting controls the threshold to be considered a core point.
-:::
-
-For each face, it looks around it to find other faces within a certain distance. Faces within this distance are considered similar, so it then checks if any of these faces are associated with a person.
-
-If there is an existing person, it assigns the person of the most similar face to the face being processed.
-
-If there is none, then it has to determine something from the DBSCAN algorithm: whether the face is a _core point_. If there are a certain number of similar faces (by default 3, including the face being considered), then this face is a core point. A new person is created for this face and the face is assigned to it. When other faces are processed, if they're similar to this face, they'll see that it has an associated person and can be assigned to that person.
-
-However, if there aren't enough similar faces, no new person will be created. Instead, the face will wait for all the other faces to be processed to see if any matches that previously didn't have an associated person now do. If they do, then the face will be assigned to that person. If not, this face will be considered an outlier, such as a stranger in the background of an image.
-
-The algorithm has some subtle differences compared to DBSCAN:
-
-- DBSCAN doesn't have a concept of incremental clustering: it clusters all points at once. In contrast, facial recognition has to evolve as more assets are added without re-clustering everything each time.
-  - The algorithm described above works within a set of queued assets. Once these faces are processed and a new round of faces are detected, the behavior will not be the same as traditional DBSCAN since it preserves the clusters (people) generated from the previous round.
-    - Facial recognition tries to wait for face detection and thumbnail generation to complete before starting for this reason: the larger the set of faces in the queue, the better the results will be.
-    - Re-running facial recognition on all assets afterwards does behave like DBSCAN, however.
-- DBSCAN is designed for range-based searches (i.e. points within a distance), but high-dimensional vector indices are generally optimized for getting the closest K results. The recognition algorithm doesn't try to get _all_ similar faces within a distance for performance reasons. Instead, it searches for a small number of matches for each face. The end result should be very similar if not identical, but with possibly different performance characteristics.
-  - Because of this, part of the recognition process is handled during a nightly job to ensure that unassigned faces with potential matches can be recognized.
-
-:::tip
-If you didn't import your assets at once or if the server was able to process jobs faster than you could upload them, it's possible that the clustering was suboptimal. If you haven't put effort into the current results, it may be worth re-running facial recognition on all assets for the best starting point. If it's too late for that, you can also manually assign a selection of unassigned faces and queue _Missing_ for Facial Recognition to help it learn and assign more faces automatically.
-:::
-
-## Configuration
-
-Navigating to Administration > Settings > Machine Learning Settings > Facial Recognition will show the options available.
-
-:::tip
-It's better to only tweak the parameters here than to set them to something very different unless you're ready to test a variety of options. If you do need to set a parameter to a strict setting, relaxing other settings can be a good option to compensate, and vice versa.
-:::
-
-### Facial recognition model
-
-There are a few different models available; the default is typically considered the best. On more constrained systems where the default is too intensive, you can choose a smaller model instead.
-
-### Minimum detection score
-
-This setting affects whether a result from the face detecton model is filtered out as a false positive. It may seem tempting to set this low to detect more faces, but it can lead to false positives that are difficult to deal with and can harm facial recognition. It is strongly recommended not to go below 0.5 for this setting. Setting it to a very high number like 0.9 is also not recommended: the default is already biased toward precision, so a threshold that high leads to many undetected faces.
-
-After changing this setting, it will only apply to new face detection jobs. To apply the new setting to all assets, you need to re-run face detection for all assets.
-
-### Maximum recognition distance
-
-The distance threshold described in How Facial Recognition Works. The default works well for most people, but it may be worth lowering it if the library has twins or otherwise very similar looking people. A threshold that's too low just means needing to merge duplicate people after facial recognition, whereas a threshold too high can produce unsalvageable results. It is strongly recommended not to go below 0.3 or above 0.7.
-
-### Minimum recognized faces
-
-The core point threshold described in How Facial Recognition Works. This setting has a few implications. First, it takes effect immediately in that people with fewer faces than this are hidden from view. Secondly, it makes clustering more robust as it prevents loosely-related faces from being linked to each other by requiring a certain level of density.
-
-Increasing this setting is a good idea if you increase the recognition distance or reduce the minimum detection score. Setting it to 1 effectively disables the concept of core points, but can be an option if you prefer a more hands-on approach.

docs/docs/features/hardware-transcoding.md

@@ -123,7 +123,6 @@ Once this is done, you can continue to step 3 of "Basic Setup".
 
 - You may want to choose a slower preset than for software transcoding to maintain quality and efficiency
 - While you can use VAAPI with NVIDIA and Intel devices, prefer the more specific APIs since they're more optimized for their respective devices
-- You can confirm the device is being recognized and used by checking its utilization (via `nvtop` for NVIDIA, `intel_gpu_top` for Intel, etc.) when transcoding. A lack of error logs when transcoding also indicates that it's being used.
 
 [hw-file]: https://github.com/immich-app/immich/releases/latest/download/hwaccel.transcoding.yml
 [nvct]: https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html

docs/docs/features/libraries.md

@@ -104,19 +104,18 @@ The `immich-server` container will need access to the gallery. Modify your docke
   immich-server:
     volumes:
       - ${UPLOAD_LOCATION}:/usr/src/app/upload
-+     - /mnt/nas/christmas-trip:/mnt/nas/christmas-trip:ro
-+     - /home/user/old-pics:/home/user/old-pics:ro
++     - /mnt/nas/christmas-trip:/mnt/media/christmas-trip:ro
++     - /home/user/old-pics:/mnt/media/old-pics:ro
 +     - /mnt/media/videos:/mnt/media/videos:ro
-+     - /mnt/media/videos2:/mnt/media/videos2 # the files in this folder can be deleted, as it does not end with :ro
 +     - "C:/Users/user_name/Desktop/my media:/mnt/media/my-media:ro" # import path in Windows system.
 ```
 
 :::tip
-The `ro` flag at the end only gives read-only access to the volumes. This will disallow the images from being deleted in the web UI.
+The `ro` flag at the end only gives read-only access to the volumes. While Immich does not modify files, it's a good practice to mount read-only.
 :::
 
 :::info
-_Remember to run `docker compose up -d` to register the changes. Make sure you can see the mounted path in the container._
+_Remember to bring the container `docker compose down/up` to register the changes. Make sure you can see the mounted path in the container._
 :::
 
 ### Create External Libraries

docs/docs/features/ml-hardware-acceleration.md

@@ -32,13 +32,12 @@ You do not need to redo any machine learning jobs after enabling hardware accele
   - Where and how you can get this file depends on device and vendor, but typically, the device vendor also supplies these
   - The `hwaccel.ml.yml` file assumes the path to it is `/usr/lib/libmali.so`, so update accordingly if it is elsewhere
   - The `hwaccel.ml.yml` file assumes an additional file `/lib/firmware/mali_csffw.bin`, so update accordingly if your device's driver does not require this file
-- Optional: Configure your `.env` file, see [environment variables](/docs/install/environment-variables) for ARM NN specific settings
 
 #### CUDA
 
 - The GPU must have compute capability 5.2 or greater.
 - The server must have the official NVIDIA driver installed.
-- The installed driver must be >= 545 (it must support CUDA 12.3.2).
+- The installed driver must be >= 535 (it must support CUDA 12.2).
 - On Linux (except for WSL2), you also need to have [NVIDIA Container Toolkit][nvct] installed.
 
 #### OpenVINO

docs/docs/features/monitoring.md

@@ -66,7 +66,7 @@ The provided file is just a starting point. There are a ton of ways to configure
 After bringing down the containers with `docker compose down` and back up with `docker compose up -d`, a Prometheus instance will now collect metrics from the immich server and microservices containers. Note that we didn't need to expose any new ports for these containers - the communication is handled in the internal Docker network.
 
 :::note
-To see exactly what metrics are made available, you can additionally add `8081:8081` to the server container's ports and `8082:8082` to the microservices container's ports. Visiting the `/metrics` endpoint for these services will show the same raw data that Prometheus collects.
+To see exactly what metrics are made available, you can additionally add `8081:8081` to the server container's ports and `8082:8081` to the microservices container's ports. Visiting the `/metrics` endpoint for these services will show the same raw data that Prometheus collects.
 :::
 
 ### Usage

docs/docs/features/smart-search.md

@@ -7,30 +7,29 @@ Immich uses Postgres as its search database for both metadata and smart search.
 
 Smart search is powered by the [pgvecto.rs](https://github.com/tensorchord/pgvecto.rs) extension, utilizing machine learning models like [CLIP](https://openai.com/research/clip) to provide relevant search results. This allows for freeform searches without requiring specific keywords in the image or video metadata.
 
+Archived photos are not included in search results by default. To include them, mark the checkbox in [advanced search filters](/docs/features/smart-search#advanced-search-filters).
+
+:::tip Alternative CLIP Models
+More powerful models can be used for more accurate search results. For more information, see the related [FAQ](/docs/FAQ#can-i-use-a-custom-clip-model).
+:::
+
+:::info
+Smart Search is currently limited to 5,000 results for a single search on the web.
+:::
+
 ## Advanced Search Filters
 
 In addition, Immich offers advanced search functionality, allowing you to find specific content using customizable search filters. These filters include location, one or more faces, specific albums, and more. You can try out the search filters on the [Demo site](https://demo.immich.app).
 
-The filters smart search allows you to search by include:
+Smart search features include:
 
-- People
-- Location
-  - Country
-  - State
-  - City
-- Camera
-  - Make
-  - Model
-- Date range
-- File name or extension
-- Media type
-  - Image (including live/motion photos)
-  - Video
-  - All
-- Condition
-  - Not in any album
-  - Archived
-  - Favorited
+- Search for one or more faces (with or without context search).
+- Search by Country or State or City or by all three.
+- Search by camera make and model.
+- Search by date range.
+- Search by file name.
+- Search by media types: image, video or all (**Note:** Image includes live images).
+- Search by condition: not in any album or archive or Favorite or all conditions.
 
 <Tabs>
   <TabItem value="Computer" label="Computer" default>
@@ -48,27 +47,3 @@ Some search examples:
 
 </TabItem>
 </Tabs>
-
-## Configuration
-
-Navigating to `Administration > Settings > Machine Learning Settings > Smart Search` will show the options available.
-
-### CLIP model
-
-More powerful models can be used for more accurate search results, but are slower and can require more server resources. Check out the models [here][huggingface-clip] for more options!
-
-[Multilingual models][huggingface-multilingual-clip] are also available so users can search in their native language. These models support over 100 languages; the `nllb` models in particular support 200.
-:::note
-Multilingual models are much slower and larger and perform slightly worse for English than English-only models. For this reason, only use them if you actually intend to search in a language besides English.
-
-As a special case, the `ViT-H-14-quickgelu__dfn5b` and `ViT-H-14-378-quickgelu__dfn5b` models are excellent at many European languages despite not specifically being multilingual. They're very intensive regardless, however - especially the latter.
-:::
-
-Once you've chosen a model, change this setting to the name of the model you chose. Be sure to re-run Smart Search on all assets after this change.
-
-:::note
-Feel free to make a feature request if there's a model you want to use that we don't currently support.
-:::
-
-[huggingface-clip]: https://huggingface.co/collections/immich-app/clip-654eaefb077425890874cd07
-[huggingface-multilingual-clip]: https://huggingface.co/collections/immich-app/multilingual-clip-654eb08c2382f591eeb8c2a7

docs/docs/guides/custom-locations.md

@@ -13,14 +13,14 @@ In our `.env` file, we will define variables that will help us in the future whe
 
 # Custom location where your uploaded, thumbnails, and transcoded video files are stored
 - UPLOAD_LOCATION=./library
-+ UPLOAD_LOCATION=/custom/path/immich/immich_files
-+ THUMB_LOCATION=/custom/path/immich/thumbs
-+ ENCODED_VIDEO_LOCATION=/custom/path/immich/encoded-video
-+ PROFILE_LOCATION=/custom/path/immich/profile
++ UPLOAD_LOCATION=/custom/location/on/your/system/immich/immich_files
++ THUMB_LOCATION=/custom/location/on/your/system/immich/thumbs
++ ENCODED_VIDEO_LOCATION=/custom/location/on/your/system/immich/encoded-video
++ PROFILE_LOCATION=/custom/location/on/your/system/immich/profile
 ...
 ```
 
-After defining the locations for these files, we will edit the `docker-compose.yml` file accordingly and add the new variables to the `immich-server` container.
+After defining the locations for these files, we will edit the `docker-compose.yml` file accordingly and add the new variables to the `immich-server` and `immich-microservices` containers.
 
 ```diff title="docker-compose.yml"
 services:
@@ -29,6 +29,16 @@ services:
       - ${UPLOAD_LOCATION}:/usr/src/app/upload
 +     - ${THUMB_LOCATION}:/usr/src/app/upload/thumbs
 +     - ${ENCODED_VIDEO_LOCATION}:/usr/src/app/upload/encoded-video
++     - ${PROFILE_LOCATION}:/usr/src/app/upload/profile
+      - /etc/localtime:/etc/localtime:ro
+
+...
+
+  immich-microservices:
+      volumes:
+      - ${UPLOAD_LOCATION}:/usr/src/app/upload
++     - ${THUMB_LOCATION}:/usr/src/app/upload/thumbs
++     - ${ENCODED_VIDEO_LOCATION}:/usr/src/app/upload/encoded-video
 +     - ${PROFILE_LOCATION}:/usr/src/app/upload/profile
       - /etc/localtime:/etc/localtime:ro
 ```
@@ -36,6 +46,7 @@ services:
 Restart Immich to register the changes.
 
 ```
+docker compose down
 docker compose up -d
 ```
 

docs/docs/guides/database-queries.md

@@ -5,7 +5,7 @@ Keep in mind that mucking around in the database might set the moon on fire. Avo
 :::
 
 :::tip
-Run `docker exec -it immich_postgres psql --dbname=immich --username=<DB_USERNAME>` to connect to the database via the container directly.
+Run `docker exec -it immich_postgres psql immich <DB_USERNAME>` to connect to the database via the container directly.
 
 (Replace `<DB_USERNAME>` with the value from your [`.env` file](/docs/install/environment-variables#database)).
 :::
@@ -106,9 +106,3 @@ SELECT "key", "value" FROM "system_metadata" WHERE "key" = 'system-config';
 ```sql title="Delete person and unset it for the faces it was associated with"
 DELETE FROM "person" WHERE "name" = 'PersonNameHere';
 ```
-
-## Postgres internal
-
-```sql title="Change DB_PASSWORD"
-ALTER USER <DB_USERNAME> WITH ENCRYPTED PASSWORD 'newpasswordhere';
-```

docs/docs/guides/external-library.md

@@ -6,18 +6,36 @@ in a directory on the same machine.
 
 # Mount the directory into the containers.
 
-Edit `docker-compose.yml` to add one or more new mount points in the section `immich-server:` under `volumes:`.
-If you want Immich to be able to delete the images in the external library, remove `:ro` from the end of the mount point.
+Edit `docker-compose.yml` to add two new mount points in the section `immich-server:` under `volumes:`
 
 ```diff
 immich-server:
     volumes:
-        - ${UPLOAD_LOCATION}:/usr/src/app/upload
-+       - /home/user/photos1:/home/user/photos1:ro
-+       - /mnt/photos2:/mnt/photos2:ro # you can delete this line if you only have one mount point, or you can add more lines if you have more than two
++       - ${EXTERNAL_PATH}:/usr/src/app/external
 ```
 
-Restart Immich by running `docker compose up -d`.
+Edit `.env` to define `EXTERNAL_PATH`, substituting in the correct path for your computer:
+
+```
+EXTERNAL_PATH=<your-path-here>
+```
+
+On my computer, for example, I use this path:
+
+```
+EXTERNAL_PATH=/home/tenino/photos
+```
+
+:::info EXTERNAL_PATH design
+The design choice to put the EXTERNAL_PATH into .env rather than put two copies of the absolute path in the yml file in order to make everything easier, so if you have two copies of the same path that have to be kept in sync, then someday later when you move the data, update only one of the paths, without everything will break mysteriously.
+:::
+
+Restart Immich.
+
+```
+docker compose down
+docker compose up -d
+```
 
 # Create the library
 

docs/docs/guides/remote-machine-learning.md

@@ -4,11 +4,14 @@ To alleviate [performance issues on low-memory systems](/docs/FAQ.mdx#why-is-imm
 
 - Set the URL in Machine Learning Settings on the Admin Settings page to point to the designated ML system, e.g. `http://workstation:3003`.
 - Copy the following `docker-compose.yml` to your ML system.
-  - If using [hardware acceleration](/docs/features/ml-hardware-acceleration), the [hwaccel.ml.yml](https://github.com/immich-app/immich/releases/latest/download/hwaccel.ml.yml) file also needs to be added
 - Start the container by running `docker compose up -d`.
 
 :::info
-Smart Search and Face Detection will use this feature, but Facial Recognition is handled in the server.
+Starting with version v1.93.0 face detection work and face recognize were split. From now on face detection is done in the immich_machine_learning container, but facial recognition is done in the `microservices` worker.
+:::
+
+:::note
+The [hwaccel.ml.yml](https://github.com/immich-app/immich/releases/latest/download/hwaccel.ml.yml) file also needs to be in the same folder if trying to use [hardware acceleration](/docs/features/ml-hardware-acceleration).
 :::
 
 ```yaml
@@ -34,7 +37,3 @@ volumes:
 ```
 
 Please note that version mismatches between both hosts may cause instabilities and bugs, so make sure to always perform updates together.
-
-:::caution
-As an internal service, the machine learning container has no security measures whatsoever. Please be mindful of where it's deployed and who can access it.
-:::

docs/docs/install/environment-variables.md

@@ -38,19 +38,17 @@ Regardless of filesystem, it is not recommended to use a network share for your
 
 ## General
 
-| Variable                            | Description                                         |           Default            | Containers               | Workers            |
-| :---------------------------------- | :-------------------------------------------------- | :--------------------------: | :----------------------- | :----------------- |
-| `TZ`                                | Timezone                                            |                              | server                   | microservices      |
-| `IMMICH_ENV`                        | Environment (production, development)               |         `production`         | server, machine learning | api, microservices |
-| `IMMICH_LOG_LEVEL`                  | Log Level (verbose, debug, log, warn, error)        |            `log`             | server, machine learning | api, microservices |
-| `IMMICH_MEDIA_LOCATION`             | Media Location                                      |   `./upload`<sup>\*1</sup>   | server                   | api, microservices |
-| `IMMICH_CONFIG_FILE`                | Path to config file                                 |                              | server                   | api, microservices |
-| `NO_COLOR`                          | Set to `true` to disable color-coded log output     |           `false`            | server, machine learning |                    |
-| `CPU_CORES`                         | Amount of cores available to the immich server      | auto-detected cpu core count | server                   |                    |
-| `IMMICH_API_METRICS_PORT`           | Port for the OTEL metrics                           |            `8081`            | server                   | api                |
-| `IMMICH_MICROSERVICES_METRICS_PORT` | Port for the OTEL metrics                           |            `8082`            | server                   | microservices      |
-| `IMMICH_PROCESS_INVALID_IMAGES`     | When `true`, generate thumbnails for invalid images |                              | server                   | microservices      |
-| `IMMICH_TRUSTED_PROXIES`            | List of comma separated IPs set as trusted proxies  |                              | server                   | api                |
+| Variable                            | Description                                     |           Default            | Containers               | Workers            |
+| :---------------------------------- | :---------------------------------------------- | :--------------------------: | :----------------------- | :----------------- |
+| `TZ`                                | Timezone                                        |                              | server                   | microservices      |
+| `IMMICH_ENV`                        | Environment (production, development)           |         `production`         | server, machine learning | api, microservices |
+| `IMMICH_LOG_LEVEL`                  | Log Level (verbose, debug, log, warn, error)    |            `log`             | server, machine learning | api, microservices |
+| `IMMICH_MEDIA_LOCATION`             | Media Location                                  |   `./upload`<sup>\*1</sup>   | server                   | api, microservices |
+| `IMMICH_CONFIG_FILE`                | Path to config file                             |                              | server                   | api, microservices |
+| `NO_COLOR`                          | Set to `true` to disable color-coded log output |           `false`            | server, machine learning |                    |
+| `CPU_CORES`                         | Amount of cores available to the immich server  | auto-detected cpu core count | server                   |                    |
+| `IMMICH_API_METRICS_PORT`           | Port for the OTEL metrics                       |            `8081`            | server                   | api                |
+| `IMMICH_MICROSERVICES_METRICS_PORT` | Port for the OTEL metrics                       |            `8082`            | server                   | microservices      |
 
 \*1: With the default `WORKDIR` of `/usr/src/app`, this path will resolve to `/usr/src/app/upload`.
 It only need to be set if the Immich deployment method is changing.
@@ -157,21 +155,18 @@ Redis (Sentinel) URL example JSON before encoding:
 
 ## Machine Learning
 
-| Variable                                         | Description                                                                                         |                Default                | Containers       |
-| :----------------------------------------------- | :-------------------------------------------------------------------------------------------------- | :-----------------------------------: | :--------------- |
-| `MACHINE_LEARNING_MODEL_TTL`                     | Inactivity time (s) before a model is unloaded (disabled if \<= 0)                                  |                 `300`                 | machine learning |
-| `MACHINE_LEARNING_MODEL_TTL_POLL_S`              | Interval (s) between checks for the model TTL (disabled if \<= 0)                                   |                 `10`                  | machine learning |
-| `MACHINE_LEARNING_CACHE_FOLDER`                  | Directory where models are downloaded                                                               |               `/cache`                | machine learning |
-| `MACHINE_LEARNING_REQUEST_THREADS`<sup>\*1</sup> | Thread count of the request thread pool (disabled if \<= 0)                                         |          number of CPU cores          | machine learning |
-| `MACHINE_LEARNING_MODEL_INTER_OP_THREADS`        | Number of parallel model operations                                                                 |                  `1`                  | machine learning |
-| `MACHINE_LEARNING_MODEL_INTRA_OP_THREADS`        | Number of threads for each model operation                                                          |                  `2`                  | machine learning |
-| `MACHINE_LEARNING_WORKERS`<sup>\*2</sup>         | Number of worker processes to spawn                                                                 |                  `1`                  | machine learning |
-| `MACHINE_LEARNING_WORKER_TIMEOUT`                | Maximum time (s) of unresponsiveness before a worker is killed                                      | `120` (`300` if using OpenVINO image) | machine learning |
-| `MACHINE_LEARNING_PRELOAD__CLIP`                 | Name of a CLIP model to be preloaded and kept in cache                                              |                                       | machine learning |
-| `MACHINE_LEARNING_PRELOAD__FACIAL_RECOGNITION`   | Name of a facial recognition model to be preloaded and kept in cache                                |                                       | machine learning |
-| `MACHINE_LEARNING_ANN`                           | Enable ARM-NN hardware acceleration if supported                                                    |                `True`                 | machine learning |
-| `MACHINE_LEARNING_ANN_FP16_TURBO`                | Execute operations in FP16 precision: increasing speed, reducing precision (applies only to ARM-NN) |                `False`                | machine learning |
-| `MACHINE_LEARNING_ANN_TUNING_LEVEL`              | ARM-NN GPU tuning level (1: rapid, 2: normal, 3: exhaustive)                                        |                  `2`                  | machine learning |
+| Variable                                         | Description                                                          |       Default       | Containers       |
+| :----------------------------------------------- | :------------------------------------------------------------------- | :-----------------: | :--------------- |
+| `MACHINE_LEARNING_MODEL_TTL`                     | Inactivity time (s) before a model is unloaded (disabled if \<= 0)   |        `300`        | machine learning |
+| `MACHINE_LEARNING_MODEL_TTL_POLL_S`              | Interval (s) between checks for the model TTL (disabled if \<= 0)    |        `10`         | machine learning |
+| `MACHINE_LEARNING_CACHE_FOLDER`                  | Directory where models are downloaded                                |      `/cache`       | machine learning |
+| `MACHINE_LEARNING_REQUEST_THREADS`<sup>\*1</sup> | Thread count of the request thread pool (disabled if \<= 0)          | number of CPU cores | machine learning |
+| `MACHINE_LEARNING_MODEL_INTER_OP_THREADS`        | Number of parallel model operations                                  |         `1`         | machine learning |
+| `MACHINE_LEARNING_MODEL_INTRA_OP_THREADS`        | Number of threads for each model operation                           |         `2`         | machine learning |
+| `MACHINE_LEARNING_WORKERS`<sup>\*2</sup>         | Number of worker processes to spawn                                  |         `1`         | machine learning |
+| `MACHINE_LEARNING_WORKER_TIMEOUT`                | Maximum time (s) of unresponsiveness before a worker is killed       |        `120`        | machine learning |
+| `MACHINE_LEARNING_PRELOAD__CLIP`                 | Name of a CLIP model to be preloaded and kept in cache               |                     | machine learning |
+| `MACHINE_LEARNING_PRELOAD__FACIAL_RECOGNITION`   | Name of a facial recognition model to be preloaded and kept in cache |                     | machine learning |
 
 \*1: It is recommended to begin with this parameter when changing the concurrency levels of the machine learning service and then tune the other ones.
 

docs/docs/install/script.md

@@ -8,10 +8,6 @@ sidebar_position: 20
 This method is experimental and not currently recommended for production use. For production, please refer to installing with [Docker Compose](/docs/install/docker-compose.mdx).
 :::
 
-:::note
-The install script only supports Linux operating systems and requires Docker to be already installed on the system.
-:::
-
 In the shell, from a directory of your choice, run the following command:
 
 ```bash

docs/docusaurus.config.js

@@ -145,44 +145,11 @@ const config = {
                 label: 'Installation',
                 to: '/docs/install/requirements',
               },
-              {
-                label: 'Contributing',
-                to: '/docs/overview/support-the-project',
-              },
-              {
-                label: 'Privacy Policy',
-                to: '/privacy-policy',
-              },
             ],
           },
           {
-            title: 'Documentation',
+            title: 'Community',
             items: [
-              {
-                label: 'Roadmap',
-                to: '/roadmap',
-              },
-              {
-                label: 'API',
-                to: '/docs/api',
-              },
-              {
-                label: 'Cursed Knowledge',
-                to: '/cursed-knowledge',
-              },
-            ],
-          },
-          {
-            title: 'Links',
-            items: [
-              {
-                label: 'GitHub',
-                href: 'https://github.com/immich-app/immich',
-              },
-              {
-                label: 'YouTube',
-                href: 'https://www.youtube.com/@immich-app',
-              },
               {
                 label: 'Discord',
                 href: 'https://discord.immich.app',
@@ -193,6 +160,23 @@ const config = {
               },
             ],
           },
+          {
+            title: 'Links',
+            items: [
+              // {
+              //   label: "Blog",
+              //   to: "/blog",
+              // },
+              {
+                label: 'GitHub',
+                href: 'https://github.com/immich-app/immich',
+              },
+              {
+                label: 'YouTube',
+                href: 'https://www.youtube.com/@immich-app',
+              },
+            ],
+          },
         ],
         copyright: `Immich is available as open source under the terms of the GNU AGPL v3 License.`,
       },

docs/package-lock.json

@@ -12754,9 +12754,9 @@
       }
     },
     "node_modules/postcss": {
-      "version": "8.4.40",
-      "resolved": "https://registry.npmjs.org/postcss/-/postcss-8.4.40.tgz",
-      "integrity": "sha512-YF2kKIUzAofPMpfH6hOi2cGnv/HrUlfucspc7pDyvv7kGdqXrfj8SCl/t8owkEgKEuu8ZcRjSOxFxVLqwChZ2Q==",
+      "version": "8.4.39",
+      "resolved": "https://registry.npmjs.org/postcss/-/postcss-8.4.39.tgz",
+      "integrity": "sha512-0vzE+lAiG7hZl1/9I8yzKLx3aR9Xbof3fBHKunvMfOCYAtMhrsnccJY2iTURb9EZd5+pLuiNV9/c/GZJOHsgIw==",
       "funding": [
         {
           "type": "opencollective",
@@ -13600,9 +13600,9 @@
       }
     },
     "node_modules/prettier": {
-      "version": "3.3.3",
-      "resolved": "https://registry.npmjs.org/prettier/-/prettier-3.3.3.tgz",
-      "integrity": "sha512-i2tDNA0O5IrMO757lfrdQZCc2jPNDVntV0m/+4whiDfWaTKfMNgR7Qz0NAeGz/nRqF4m5/6CLzbP4/liHt12Ew==",
+      "version": "3.3.2",
+      "resolved": "https://registry.npmjs.org/prettier/-/prettier-3.3.2.tgz",
+      "integrity": "sha512-rAVeHYMcv8ATV5d508CFdn+8/pHPpXeIid1DdrPwXnaAdH7cqjVbpJaT5eq4yRAFU/lsbwYwSF/n5iNrdJHPQA==",
       "dev": true,
       "license": "MIT",
       "bin": {
@@ -13747,10 +13747,9 @@
       }
     },
     "node_modules/qs": {
-      "version": "6.13.0",
-      "resolved": "https://registry.npmjs.org/qs/-/qs-6.13.0.tgz",
-      "integrity": "sha512-+38qI9SOr8tfZ4QmJNplMUxqjbe7LKvvZgWdExBOmd+egZTtjLB67Gu0HRX3u/XOq7UU2Nx6nsjvS16Z9uwfpg==",
-      "license": "BSD-3-Clause",
+      "version": "6.12.1",
+      "resolved": "https://registry.npmjs.org/qs/-/qs-6.12.1.tgz",
+      "integrity": "sha512-zWmv4RSuB9r2mYQw3zxQuHWeU+42aKi1wWig/j4ele4ygELZ7PEO6MM7rim9oAQH2A5MWfsAVf/jPvTPgCbvUQ==",
       "dependencies": {
         "side-channel": "^1.0.6"
       },
@@ -16015,9 +16014,9 @@
       }
     },
     "node_modules/tailwindcss": {
-      "version": "3.4.7",
-      "resolved": "https://registry.npmjs.org/tailwindcss/-/tailwindcss-3.4.7.tgz",
-      "integrity": "sha512-rxWZbe87YJb4OcSopb7up2Ba4U82BoiSGUdoDr3Ydrg9ckxFS/YWsvhN323GMcddgU65QRy7JndC7ahhInhvlQ==",
+      "version": "3.4.4",
+      "resolved": "https://registry.npmjs.org/tailwindcss/-/tailwindcss-3.4.4.tgz",
+      "integrity": "sha512-ZoyXOdJjISB7/BcLTR6SEsLgKtDStYyYZVLsUtWChO4Ps20CBad7lfJKVDiejocV4ME1hLmyY0WJE3hSDcmQ2A==",
       "license": "MIT",
       "dependencies": {
         "@alloc/quick-lru": "^5.2.0",
@@ -16377,9 +16376,9 @@
       }
     },
     "node_modules/typescript": {
-      "version": "5.5.4",
-      "resolved": "https://registry.npmjs.org/typescript/-/typescript-5.5.4.tgz",
-      "integrity": "sha512-Mtq29sKDAEYP7aljRgtPOpTvOfbwRWlS6dPRzwjdE+C0R4brX/GUyhHSecbHMFLNBLcJIPt9nl9yG5TZ1weH+Q==",
+      "version": "5.5.3",
+      "resolved": "https://registry.npmjs.org/typescript/-/typescript-5.5.3.tgz",
+      "integrity": "sha512-/hreyEujaB0w76zKo6717l3L0o/qEUtRgdvUBvlkhoWeOVMjMuHNHk0BRBzikzuGDqNmPQbg5ifMEqsHLiIUcQ==",
       "license": "Apache-2.0",
       "bin": {
         "tsc": "bin/tsc",
@@ -16715,16 +16714,12 @@
       }
     },
     "node_modules/url": {
-      "version": "0.11.4",
-      "resolved": "https://registry.npmjs.org/url/-/url-0.11.4.tgz",
-      "integrity": "sha512-oCwdVC7mTuWiPyjLUz/COz5TLk6wgp0RCsN+wHZ2Ekneac9w8uuV0njcbbie2ME+Vs+d6duwmYuR3HgQXs1fOg==",
-      "license": "MIT",
+      "version": "0.11.3",
+      "resolved": "https://registry.npmjs.org/url/-/url-0.11.3.tgz",
+      "integrity": "sha512-6hxOLGfZASQK/cijlZnZJTq8OXAkt/3YGfQX45vvMYXpZoo8NdWZcY73K108Jf759lS1Bv/8wXnHDTSz17dSRw==",
       "dependencies": {
         "punycode": "^1.4.1",
-        "qs": "^6.12.3"
-      },
-      "engines": {
-        "node": ">= 0.4"
+        "qs": "^6.11.2"
       }
     },
     "node_modules/url-loader": {
@@ -16788,8 +16783,7 @@
     "node_modules/url/node_modules/punycode": {
       "version": "1.4.1",
       "resolved": "https://registry.npmjs.org/punycode/-/punycode-1.4.1.tgz",
-      "integrity": "sha512-jmYNElW7yvO7TV33CjSmvSiE2yco3bV2czu/OzDKdMNVZQWfxCblURLhf+47syQRBntjfLdd/H0egrzIG+oaFQ==",
-      "license": "MIT"
+      "integrity": "sha512-jmYNElW7yvO7TV33CjSmvSiE2yco3bV2czu/OzDKdMNVZQWfxCblURLhf+47syQRBntjfLdd/H0egrzIG+oaFQ=="
     },
     "node_modules/util": {
       "version": "0.10.4",

docs/package.json

@@ -56,6 +56,6 @@
     "node": ">=20"
   },
   "volta": {
-    "node": "20.16.0"
+    "node": "20.15.1"
   }
 }

docs/src/components/community-projects.tsx

@@ -63,11 +63,6 @@ const projects: CommunityProjectProps[] = [
     description: 'Powershell Module for the Immich API',
     url: 'https://github.com/hanpq/PSImmich',
   },
-  {
-    title: 'Immich Distribution',
-    description: 'Snap package for easy install and zero-care auto updates of Immich. Self-hosted photo management.',
-    url: 'https://immich-distribution.nsg.cc',
-  },
 ];
 
 function CommunityProject({ title, description, url }: CommunityProjectProps): JSX.Element {

docs/src/pages/cursed-knowledge.tsx

@@ -1,13 +1,4 @@
-import {
-  mdiCalendarToday,
-  mdiCrosshairsOff,
-  mdiLeadPencil,
-  mdiLockOff,
-  mdiLockOutline,
-  mdiSpeedometerSlow,
-  mdiWeb,
-  mdiWrap,
-} from '@mdi/js';
+import { mdiCalendarToday, mdiLeadPencil, mdiLockOutline, mdiSpeedometerSlow, mdiWeb } from '@mdi/js';
 import Layout from '@theme/Layout';
 import React from 'react';
 import { Item as TimelineItem, Timeline } from '../components/timeline';
@@ -17,41 +8,6 @@ const withLanguage = (date: Date) => (language: string) => date.toLocaleDateStri
 type Item = Omit<TimelineItem, 'done' | 'getDateLabel'> & { date: Date };
 
 const items: Item[] = [
-  {
-    icon: mdiWrap,
-    iconColor: 'gray',
-    title: 'Carriage returns in bash scripts are cursed',
-    description: 'Git can be configured to automatically convert LF to CRLF on checkout and CRLF breaks bash scripts.',
-    link: {
-      url: 'https://github.com/immich-app/immich/pull/11613',
-      text: '#11613',
-    },
-    date: new Date(2024, 7, 7),
-  },
-  {
-    icon: mdiLockOff,
-    iconColor: 'red',
-    title: 'Fetch inside Cloudflare Workers is cursed',
-    description:
-      'Fetch requests in Cloudflare Workers use http by default, even if you explicitly specify https, which can often cause redirect loops.',
-    link: {
-      url: 'https://community.cloudflare.com/t/does-cloudflare-worker-allow-secure-https-connection-to-fetch-even-on-flexible-ssl/68051/5',
-      text: 'Cloudflare',
-    },
-    date: new Date(2024, 7, 7),
-  },
-  {
-    icon: mdiCrosshairsOff,
-    iconColor: 'gray',
-    title: 'GPS sharing on mobile is cursed',
-    description:
-      'Some phones will silently strip GPS data from images when apps without location permission try to access them.',
-    link: {
-      url: 'https://github.com/immich-app/immich/discussions/11268',
-      text: '#11268',
-    },
-    date: new Date(2024, 6, 21),
-  },
   {
     icon: mdiLeadPencil,
     iconColor: 'gold',

docs/src/pages/index.tsx

@@ -41,7 +41,7 @@ function HomepageHeader() {
             Discord
           </Link>
         </div>
-        <img src="/img/immich-screenshots.webp" alt="screenshots" width={'70%'} />
+        <img src="/img/immich-screenshots.png" alt="screenshots" width={'70%'} />
         <div className="flex flex-col sm:flex-row place-items-center place-content-center mt-4 gap-1">
           <div className="h-24">
             <a href="https://play.google.com/store/apps/details?id=app.alextran.immich">

docs/src/pages/privacy-policy.tsx

@@ -1,114 +0,0 @@
-import React from 'react';
-import Link from '@docusaurus/Link';
-import Layout from '@theme/Layout';
-import { useColorMode } from '@docusaurus/theme-common';
-function HomepageHeader() {
-  const { isDarkTheme } = useColorMode();
-
-  return (
-    <header>
-      <section className="max-w-[900px] m-4 p-4 md:p-6 md:m-auto md:my-12 border border-red-400 rounded-2xl  bg-slate-200 dark:bg-immich-dark-gray">
-        <section>
-          <h1>Privacy Policy</h1>
-          <p>Last updated: July 31st 2024</p>
-          <p>
-            Welcome to Immich. We are committed to respecting your privacy. This Privacy Policy sets out how we collect,
-            use, and share information when you use our Immich app.
-          </p>
-        </section>
-
-        {/* 1. Scope of This Policy */}
-        <section>
-          <h2>1. Scope of This Policy</h2>
-          <p>
-            This Privacy Policy applies to the Immich app ("we", "our", or "us") and covers our collection, use, and
-            disclosure of your information. This Policy does not cover any third-party websites, services, or
-            applications that can be accessed through our app, or third-party services you may access through Immich.
-          </p>
-        </section>
-
-        {/* 2. Information We Collect */}
-        <section>
-          <h2>2. Information We Collect</h2>
-          <div>
-            <p>
-              <strong>Locally Stored Data</strong>: Immich stores all your photos, albums, settings, and locally on your
-              device. We do not have access to this data, nor do we transmit or store it on any of our servers.
-            </p>
-          </div>
-
-          <div>
-            <p>
-              <strong>Purchase Information:</strong> When you make a purchase within the{' '}
-              <a href="https://buy.immich.app">https://buy.immich.app</a>, we collect the following information for tax
-              calculation purposes:
-            </p>
-            <ul>
-              <li>Country of origin</li>
-              <li>Postal code (if the user is from Canada or the United States)</li>
-            </ul>
-          </div>
-        </section>
-
-        {/* 3. Use of Your Information */}
-        <section>
-          <h2>3. Use of Your Information</h2>
-          <p>
-            <strong>Tax Calculation:</strong> The country of origin and postal code (for users from Canada or the United
-            States) are collected solely for determining the applicable tax rates on your purchase.
-          </p>
-        </section>
-
-        {/* 4. Sharing of Your Information */}
-        <section>
-          <h2>4. Sharing of Your Information</h2>
-          <ul>
-            <li>
-              <strong>Tax Authorities:</strong> The purchase information may be shared with tax authorities as required
-              by law.
-            </li>
-            <li>
-              <strong>Payment Providers:</strong> The purchase information may be shared with payment providers where
-              required.
-            </li>
-          </ul>
-        </section>
-
-        {/* 5. Changes to This Policy */}
-        <section>
-          <h2>5. Changes to This Policy</h2>
-          <p>
-            We may update our Privacy Policy from time to time. If we make any changes, we will notify you by revising
-            the "Last updated" date at the top of this policy. It's encouraged that users frequently check this page for
-            any changes to stay informed about how we are helping to protect the personal information we collect.
-          </p>
-        </section>
-
-        {/* 6. Contact Us */}
-        <section>
-          <h2>6. Contact Us</h2>
-          <p>
-            If you have any questions about this Privacy Policy, please contact us at{' '}
-            <a href="mailto:immich@futo.org">immich@futo.org</a>
-          </p>
-        </section>
-      </section>
-    </header>
-  );
-}
-
-export default function Home(): JSX.Element {
-  return (
-    <Layout
-      title="Home"
-      description="immich Self-hosted photo and video backup solution directly from your mobile phone "
-      noFooter={true}
-    >
-      <HomepageHeader />
-      <div className="flex flex-col place-items-center place-content-center">
-        <p>This project is available under GNU AGPL v3 license.</p>
-        <p className="text-xs">Privacy should not be a luxury</p>
-      </div>
-    </Layout>
-  );
-}

docs/src/pages/roadmap.tsx

@@ -66,16 +66,12 @@ import {
   mdiVectorCombine,
   mdiVideo,
   mdiWeb,
-  mdiLicense,
 } from '@mdi/js';
 import Layout from '@theme/Layout';
 import React from 'react';
 import { Item, Timeline } from '../components/timeline';
 
 const releases = {
-  // TODO
-  'v1.110.0': new Date(2024, 5, 11),
-  'v1.109.0': new Date(2024, 6, 18),
   'v1.106.1': new Date(2024, 5, 11),
   'v1.104.0': new Date(2024, 4, 13),
   'v1.103.0': new Date(2024, 3, 29),
@@ -224,20 +220,6 @@ const roadmap: Item[] = [
 ];
 
 const milestones: Item[] = [
-  {
-    icon: mdiStar,
-    iconColor: 'gold',
-    title: '40,000 Stars',
-    description: 'Reached 40K Stars on GitHub!',
-    getDateLabel: withLanguage(new Date(2024, 6, 21)),
-  },
-  withRelease({
-    icon: mdiLicense,
-    iconColor: 'gold',
-    title: 'Supporter Badge',
-    description: 'The option to buy Immich to support its development!',
-    release: 'v1.109.0',
-  }),
   withRelease({
     icon: mdiHistory,
     title: 'Versioned documentation',
@@ -254,7 +236,7 @@ const milestones: Item[] = [
   withRelease({
     icon: mdiContentDuplicate,
     title: 'Similar image detection',
-    description: "Detect duplicate assets that aren't exactly identical",
+    description: 'Detect duplicate assets that aren’t exactly identical',
     release: 'v1.106.1',
   }),
   withRelease({

docs/static/archived-versions.json

@@ -1,24 +1,4 @@
 [
-  {
-    "label": "v1.111.0",
-    "url": "https://v1.111.0.archive.immich.app"
-  },
-  {
-    "label": "v1.110.0",
-    "url": "https://v1.110.0.archive.immich.app"
-  },
-  {
-    "label": "v1.109.2",
-    "url": "https://v1.109.2.archive.immich.app"
-  },
-  {
-    "label": "v1.109.1",
-    "url": "https://v1.109.1.archive.immich.app"
-  },
-  {
-    "label": "v1.109.0",
-    "url": "https://v1.109.0.archive.immich.app"
-  },
   {
     "label": "v1.108.0",
     "url": "https://v1.108.0.archive.immich.app"

e2e/.eslintrc.cjs

@@ -0,0 +1,32 @@
+module.exports = {
+  parser: '@typescript-eslint/parser',
+  parserOptions: {
+    project: 'tsconfig.json',
+    sourceType: 'module',
+    tsconfigRootDir: __dirname,
+  },
+  plugins: ['@typescript-eslint/eslint-plugin'],
+  extends: ['plugin:@typescript-eslint/recommended', 'plugin:prettier/recommended', 'plugin:unicorn/recommended'],
+  root: true,
+  env: {
+    node: true,
+  },
+  ignorePatterns: ['.eslintrc.js'],
+  rules: {
+    '@typescript-eslint/interface-name-prefix': 'off',
+    '@typescript-eslint/explicit-function-return-type': 'off',
+    '@typescript-eslint/explicit-module-boundary-types': 'off',
+    '@typescript-eslint/no-explicit-any': 'off',
+    '@typescript-eslint/no-floating-promises': 'error',
+    'unicorn/prefer-module': 'off',
+    'unicorn/import-style': 'off',
+    curly: 2,
+    'prettier/prettier': 0,
+    'unicorn/prevent-abbreviations': 'off',
+    'unicorn/filename-case': 'off',
+    'unicorn/no-null': 'off',
+    'unicorn/prefer-top-level-await': 'off',
+    'unicorn/prefer-event-target': 'off',
+    'unicorn/no-thenable': 'off',
+  },
+};

e2e/.nvmrc

@@ -1 +1 @@
-20.16.0
+20.15.1

e2e/docker-compose.yml

@@ -35,7 +35,7 @@ services:
       - 2283:3001
 
   redis:
-    image: redis:6.2-alpine@sha256:e3b17ba9479deec4b7d1eeec1548a253acc5374d68d3b27937fcfe4df8d18c7e
+    image: redis:6.2-alpine@sha256:328fe6a5822256d065debb36617a8169dbfbd77b797c525288e465f56c1d392b
 
   database:
     image: tensorchord/pgvecto-rs:pg14-v0.2.0@sha256:90724186f0a3517cf6914295b5ab410db9ce23190a2d9d0b9dd6463e3fa298f0

e2e/eslint.config.mjs

@@ -1,64 +0,0 @@
-import { FlatCompat } from '@eslint/eslintrc';
-import js from '@eslint/js';
-import typescriptEslint from '@typescript-eslint/eslint-plugin';
-import tsParser from '@typescript-eslint/parser';
-import globals from 'globals';
-import path from 'node:path';
-import { fileURLToPath } from 'node:url';
-
-const __filename = fileURLToPath(import.meta.url);
-const __dirname = path.dirname(__filename);
-const compat = new FlatCompat({
-  baseDirectory: __dirname,
-  recommendedConfig: js.configs.recommended,
-  allConfig: js.configs.all,
-});
-
-export default [
-  {
-    ignores: ['eslint.config.mjs'],
-  },
-  ...compat.extends(
-    'plugin:@typescript-eslint/recommended',
-    'plugin:prettier/recommended',
-    'plugin:unicorn/recommended',
-  ),
-  {
-    plugins: {
-      '@typescript-eslint': typescriptEslint,
-    },
-
-    languageOptions: {
-      globals: {
-        ...globals.node,
-      },
-
-      parser: tsParser,
-      ecmaVersion: 5,
-      sourceType: 'module',
-
-      parserOptions: {
-        project: 'tsconfig.json',
-        tsconfigRootDir: __dirname,
-      },
-    },
-
-    rules: {
-      '@typescript-eslint/interface-name-prefix': 'off',
-      '@typescript-eslint/explicit-function-return-type': 'off',
-      '@typescript-eslint/explicit-module-boundary-types': 'off',
-      '@typescript-eslint/no-explicit-any': 'off',
-      '@typescript-eslint/no-floating-promises': 'error',
-      'unicorn/prefer-module': 'off',
-      'unicorn/import-style': 'off',
-      curly: 2,
-      'prettier/prettier': 0,
-      'unicorn/prevent-abbreviations': 'off',
-      'unicorn/filename-case': 'off',
-      'unicorn/no-null': 'off',
-      'unicorn/prefer-top-level-await': 'off',
-      'unicorn/prefer-event-target': 'off',
-      'unicorn/no-thenable': 'off',
-    },
-  },
-];

e2e/package.json

@@ -1,6 +1,6 @@
 {
   "name": "immich-e2e",
-  "version": "1.111.0",
+  "version": "1.108.0",
   "description": "",
   "main": "index.js",
   "type": "module",
@@ -19,26 +19,23 @@
   "author": "",
   "license": "GNU Affero General Public License version 3",
   "devDependencies": {
-    "@eslint/eslintrc": "^3.1.0",
-    "@eslint/js": "^9.8.0",
     "@immich/cli": "file:../cli",
     "@immich/sdk": "file:../open-api/typescript-sdk",
     "@playwright/test": "^1.44.1",
     "@types/luxon": "^3.4.2",
-    "@types/node": "^20.14.13",
+    "@types/node": "^20.14.10",
     "@types/oidc-provider": "^8.5.1",
     "@types/pg": "^8.11.0",
     "@types/pngjs": "^6.0.4",
     "@types/supertest": "^6.0.2",
-    "@typescript-eslint/eslint-plugin": "^8.0.0",
-    "@typescript-eslint/parser": "^8.0.0",
-    "@vitest/coverage-v8": "^2.0.5",
-    "eslint": "^9.0.0",
+    "@typescript-eslint/eslint-plugin": "^7.1.0",
+    "@typescript-eslint/parser": "^7.1.0",
+    "@vitest/coverage-v8": "^1.3.0",
+    "eslint": "^8.57.0",
     "eslint-config-prettier": "^9.1.0",
     "eslint-plugin-prettier": "^5.1.3",
-    "eslint-plugin-unicorn": "^55.0.0",
-    "exiftool-vendored": "^28.0.0",
-    "globals": "^15.9.0",
+    "eslint-plugin-unicorn": "^54.0.0",
+    "exiftool-vendored": "^27.0.0",
     "jose": "^5.6.3",
     "luxon": "^3.4.4",
     "oidc-provider": "^8.5.1",
@@ -50,9 +47,9 @@
     "supertest": "^7.0.0",
     "typescript": "^5.3.3",
     "utimes": "^5.2.1",
-    "vitest": "^2.0.5"
+    "vitest": "^1.3.0"
   },
   "volta": {
-    "node": "20.16.0"
+    "node": "20.15.1"
   }
 }

e2e/src/api/specs/asset.e2e-spec.ts

@@ -472,44 +472,6 @@ describe('/asset', () => {
       expect(status).toEqual(200);
     });
 
-    it('should update date time original when sidecar file contains DateTimeOriginal', async () => {
-      const sidecarData = `<?xpacket begin='?' id='W5M0MpCehiHzreSzNTczkc9d'?>
-<x:xmpmeta xmlns:x='adobe:ns:meta/' x:xmptk='Image::ExifTool 12.40'>
-<rdf:RDF xmlns:rdf='http://www.w3.org/1999/02/22-rdf-syntax-ns#'>
- <rdf:Description rdf:about=''
-  xmlns:exif='http://ns.adobe.com/exif/1.0/'>
-  <exif:ExifVersion>0220</exif:ExifVersion>  <exif:DateTimeOriginal>2024-07-11T10:32:52Z</exif:DateTimeOriginal>
-  <exif:GPSVersionID>2.3.0.0</exif:GPSVersionID>
- </rdf:Description>
-</rdf:RDF>
-</x:xmpmeta>
-<?xpacket end='w'?>`;
-
-      const { id } = await utils.createAsset(user1.accessToken, {
-        sidecarData: {
-          bytes: Buffer.from(sidecarData),
-          filename: 'example.xmp',
-        },
-      });
-      await utils.waitForQueueFinish(admin.accessToken, 'metadataExtraction');
-
-      const assetInfo = await utils.getAssetInfo(user1.accessToken, id);
-      expect(assetInfo.exifInfo?.dateTimeOriginal).toBe('2024-07-11T10:32:52.000Z');
-
-      const { status, body } = await request(app)
-        .put(`/assets/${id}`)
-        .set('Authorization', `Bearer ${user1.accessToken}`)
-        .send({ dateTimeOriginal: '2023-11-19T18:11:00.000-07:00' });
-
-      expect(body).toMatchObject({
-        id,
-        exifInfo: expect.objectContaining({
-          dateTimeOriginal: '2023-11-20T01:11:00.000Z',
-        }),
-      });
-      expect(status).toEqual(200);
-    });
-
     it('should reject invalid gps coordinates', async () => {
       for (const test of [
         { latitude: 12 },
@@ -545,22 +507,6 @@ describe('/asset', () => {
       expect(status).toEqual(200);
     });
 
-    it.skip('should geocode country from gps data in the middle of nowhere', async () => {
-      const { status } = await request(app)
-        .put(`/assets/${user1Assets[0].id}`)
-        .set('Authorization', `Bearer ${user1.accessToken}`)
-        .send({ latitude: 42, longitude: 69 });
-      expect(status).toEqual(200);
-
-      await utils.waitForQueueFinish(admin.accessToken, 'metadataExtraction');
-
-      const asset = await getAssetInfo({ id: user1Assets[0].id }, { headers: asBearerAuth(user1.accessToken) });
-      expect(asset).toMatchObject({
-        id: user1Assets[0].id,
-        exifInfo: expect.objectContaining({ city: null, country: 'Kazakhstan' }),
-      });
-    });
-
     it('should set the description', async () => {
       const { status, body } = await request(app)
         .put(`/assets/${user1Assets[0].id}`)
@@ -1137,7 +1083,7 @@ describe('/asset', () => {
           type: AssetTypeEnum.Image,
           originalFileName: '14bit-uncompressed-(3_2).arw',
           resized: true,
-          fileCreatedAt: '2016-01-08T14:08:01.000Z',
+          fileCreatedAt: '2016-01-08T15:08:01.000Z',
           exifInfo: {
             make: 'SONY',
             model: 'ILCE-7M2',
@@ -1149,7 +1095,7 @@ describe('/asset', () => {
             iso: 100,
             lensModel: 'E 25mm F2',
             fileSizeInByte: 49_512_448,
-            dateTimeOriginal: '2016-01-08T14:08:01.000Z',
+            dateTimeOriginal: '2016-01-08T15:08:01.000Z',
             latitude: null,
             longitude: null,
             orientation: '1',
@@ -1224,25 +1170,17 @@ describe('/asset', () => {
     // into the test here.
     it.each([
       {
-        filepath: 'formats/motionphoto/samsung-one-ui-5.jpg',
+        filepath: 'formats/motionphoto/Samsung One UI 5.jpg',
         checksum: 'fr14niqCq6N20HB8rJYEvpsUVtI=',
       },
       {
-        filepath: 'formats/motionphoto/samsung-one-ui-6.jpg',
+        filepath: 'formats/motionphoto/Samsung One UI 6.jpg',
         checksum: 'lT9Uviw/FFJYCjfIxAGPTjzAmmw=',
       },
       {
-        filepath: 'formats/motionphoto/samsung-one-ui-6.heic',
+        filepath: 'formats/motionphoto/Samsung One UI 6.heic',
         checksum: '/ejgzywvgvzvVhUYVfvkLzFBAF0=',
       },
-      {
-        filepath: 'formats/motionphoto/pixel-6-pro.jpg',
-        checksum: 'bFhLGbdK058PSk4FTfrSnoKWykc=',
-      },
-      {
-        filepath: 'formats/motionphoto/pixel-8a.jpg',
-        checksum: '7YdY+WF0h+CXHbiXpi0HiCMTTjs=',
-      },
     ])(`should extract motionphoto video from $filepath`, async ({ filepath, checksum }) => {
       const response = await utils.createAsset(admin.accessToken, {
         assetData: {

e2e/src/api/specs/map.e2e-spec.ts

@@ -159,75 +159,4 @@ describe('/map', () => {
       expect(body).toEqual(expect.objectContaining({ id: 'immich-map-dark' }));
     });
   });
-
-  describe('GET /map/reverse-geocode', () => {
-    it('should require authentication', async () => {
-      const { status, body } = await request(app).get('/map/reverse-geocode');
-      expect(status).toBe(401);
-      expect(body).toEqual(errorDto.unauthorized);
-    });
-
-    it('should throw an error if a lat is not provided', async () => {
-      const { status, body } = await request(app)
-        .get('/map/reverse-geocode?lon=123')
-        .set('Authorization', `Bearer ${admin.accessToken}`);
-      expect(status).toBe(400);
-      expect(body).toEqual(errorDto.badRequest(['lat must be a number between -90 and 90']));
-    });
-
-    it('should throw an error if a lat is not a number', async () => {
-      const { status, body } = await request(app)
-        .get('/map/reverse-geocode?lat=abc&lon=123.456')
-        .set('Authorization', `Bearer ${admin.accessToken}`);
-      expect(status).toBe(400);
-      expect(body).toEqual(errorDto.badRequest(['lat must be a number between -90 and 90']));
-    });
-
-    it('should throw an error if a lat is out of range', async () => {
-      const { status, body } = await request(app)
-        .get('/map/reverse-geocode?lat=91&lon=123.456')
-        .set('Authorization', `Bearer ${admin.accessToken}`);
-      expect(status).toBe(400);
-      expect(body).toEqual(errorDto.badRequest(['lat must be a number between -90 and 90']));
-    });
-
-    it('should throw an error if a lon is not provided', async () => {
-      const { status, body } = await request(app)
-        .get('/map/reverse-geocode?lat=75')
-        .set('Authorization', `Bearer ${admin.accessToken}`);
-      expect(status).toBe(400);
-      expect(body).toEqual(errorDto.badRequest(['lon must be a number between -180 and 180']));
-    });
-
-    const reverseGeocodeTestCases = [
-      {
-        name: 'Vaucluse',
-        lat: -33.858_977_058_663_13,
-        lon: 151.278_490_730_270_48,
-        results: [{ city: 'Vaucluse', state: 'New South Wales', country: 'Australia' }],
-      },
-      {
-        name: 'Ravenhall',
-        lat: -37.765_732_399_174_75,
-        lon: 144.752_453_164_883_3,
-        results: [{ city: 'Ravenhall', state: 'Victoria', country: 'Australia' }],
-      },
-      {
-        name: 'Scarborough',
-        lat: -31.894_346_156_789_997,
-        lon: 115.757_617_103_904_64,
-        results: [{ city: 'Scarborough', state: 'Western Australia', country: 'Australia' }],
-      },
-    ];
-
-    it.each(reverseGeocodeTestCases)(`should resolve to $name`, async ({ lat, lon, results }) => {
-      const { status, body } = await request(app)
-        .get(`/map/reverse-geocode?lat=${lat}&lon=${lon}`)
-        .set('Authorization', `Bearer ${admin.accessToken}`);
-      expect(status).toBe(200);
-      expect(Array.isArray(body)).toBe(true);
-      expect(body.length).toBe(results.length);
-      expect(body).toEqual(results);
-    });
-  });
 });

e2e/src/api/specs/person.e2e-spec.ts

@@ -6,19 +6,10 @@ import request from 'supertest';
 import { beforeAll, beforeEach, describe, expect, it } from 'vitest';
 
 const invalidBirthday = [
-  {
-    birthDate: 'false',
-    response: ['birthDate must be a string in the format yyyy-MM-dd', 'Birth date cannot be in the future'],
-  },
-  {
-    birthDate: '123567',
-    response: ['birthDate must be a string in the format yyyy-MM-dd', 'Birth date cannot be in the future'],
-  },
-  {
-    birthDate: 123_567,
-    response: ['birthDate must be a string in the format yyyy-MM-dd', 'Birth date cannot be in the future'],
-  },
-  { birthDate: '9999-01-01', response: ['Birth date cannot be in the future'] },
+  { birthDate: 'false', response: 'birthDate must be a date string' },
+  { birthDate: '123567', response: 'birthDate must be a date string' },
+  { birthDate: 123_567, response: 'birthDate must be a date string' },
+  { birthDate: new Date(9999, 0, 0).toISOString(), response: ['Birth date cannot be in the future'] },
 ];
 
 describe('/people', () => {
@@ -74,7 +65,6 @@ describe('/people', () => {
 
       expect(status).toBe(200);
       expect(body).toEqual({
-        hasNextPage: false,
         total: 3,
         hidden: 1,
         people: [
@@ -90,7 +80,6 @@ describe('/people', () => {
 
       expect(status).toBe(200);
       expect(body).toEqual({
-        hasNextPage: false,
         total: 3,
         hidden: 1,
         people: [
@@ -99,21 +88,6 @@ describe('/people', () => {
         ],
       });
     });
-
-    it('should support pagination', async () => {
-      const { status, body } = await request(app)
-        .get('/people')
-        .set('Authorization', `Bearer ${admin.accessToken}`)
-        .query({ withHidden: true, page: 2, size: 1 });
-
-      expect(status).toBe(200);
-      expect(body).toEqual({
-        hasNextPage: true,
-        total: 3,
-        hidden: 1,
-        people: [expect.objectContaining({ name: 'visible_person' })],
-      });
-    });
   });
 
   describe('GET /people/:id', () => {
@@ -194,13 +168,13 @@ describe('/people', () => {
         .set('Authorization', `Bearer ${admin.accessToken}`)
         .send({
           name: 'New Person',
-          birthDate: '1990-01-01',
+          birthDate: '1990-01-01T05:00:00.000Z',
         });
       expect(status).toBe(201);
       expect(body).toMatchObject({
         id: expect.any(String),
         name: 'New Person',
-        birthDate: '1990-01-01',
+        birthDate: '1990-01-01T05:00:00.000Z',
       });
     });
   });
@@ -242,7 +216,7 @@ describe('/people', () => {
       const { status, body } = await request(app)
         .put(`/people/${visiblePerson.id}`)
         .set('Authorization', `Bearer ${admin.accessToken}`)
-        .send({ birthDate: '1990-01-01' });
+        .send({ birthDate: '1990-01-01T05:00:00.000Z' });
       expect(status).toBe(200);
       expect(body).toMatchObject({ birthDate: '1990-01-01' });
     });

e2e/src/api/specs/search.e2e-spec.ts

@@ -1,4 +1,4 @@
-import { AssetMediaResponseDto, LoginResponseDto, deleteAssets, updateAsset } from '@immich/sdk';
+import { AssetMediaResponseDto, LoginResponseDto, deleteAssets, getMapMarkers, updateAsset } from '@immich/sdk';
 import { DateTime } from 'luxon';
 import { readFile } from 'node:fs/promises';
 import { join } from 'node:path';
@@ -32,6 +32,9 @@ describe('/search', () => {
   let assetOneJpg5: AssetMediaResponseDto;
   let assetSprings: AssetMediaResponseDto;
   let assetLast: AssetMediaResponseDto;
+  let cities: string[];
+  let states: string[];
+  let countries: string[];
 
   beforeAll(async () => {
     await utils.resetDatabase();
@@ -46,9 +49,9 @@ describe('/search', () => {
       { filename: '/albums/nature/silver_fir.jpg' },
       { filename: '/formats/heic/IMG_2682.heic' },
       { filename: '/formats/jpg/el_torcal_rocks.jpg' },
-      { filename: '/formats/motionphoto/samsung-one-ui-6.jpg' },
-      { filename: '/formats/motionphoto/samsung-one-ui-6.heic' },
-      { filename: '/formats/motionphoto/samsung-one-ui-5.jpg' },
+      { filename: '/formats/motionphoto/Samsung One UI 6.jpg' },
+      { filename: '/formats/motionphoto/Samsung One UI 6.heic' },
+      { filename: '/formats/motionphoto/Samsung One UI 5.jpg' },
 
       { filename: '/metadata/gps-position/thompson-springs.jpg', dto: { isArchived: true } },
 
@@ -82,7 +85,7 @@ describe('/search', () => {
     // note: the coordinates here are not the actual coordinates of the images and are random for most of them
     const coordinates = [
       { latitude: 48.853_41, longitude: 2.3488 }, // paris
-      { latitude: 35.6895, longitude: 139.691_71 }, // tokyo
+      { latitude: 63.0695, longitude: -151.0074 }, // denali
       { latitude: 52.524_37, longitude: 13.410_53 }, // berlin
       { latitude: 1.314_663_1, longitude: 103.845_409_3 }, // singapore
       { latitude: 41.013_84, longitude: 28.949_66 }, // istanbul
@@ -98,15 +101,16 @@ describe('/search', () => {
       { latitude: 31.634_16, longitude: -7.999_94 }, // marrakesh
       { latitude: 38.523_735_4, longitude: -78.488_619_4 }, // tanners ridge
       { latitude: 59.938_63, longitude: 30.314_13 }, // st. petersburg
+      { latitude: 35.6895, longitude: 139.691_71 }, // tokyo
     ];
 
-    const updates = coordinates.map((dto, i) =>
-      updateAsset({ id: assets[i].id, updateAssetDto: dto }, { headers: asBearerAuth(admin.accessToken) }),
+    const updates = assets.map((asset, i) =>
+      updateAsset({ id: asset.id, updateAssetDto: coordinates[i] }, { headers: asBearerAuth(admin.accessToken) }),
     );
 
     await Promise.all(updates);
-    for (const [i] of coordinates.entries()) {
-      await utils.waitForWebsocketEvent({ event: 'assetUpdate', id: assets[i].id });
+    for (const asset of assets) {
+      await utils.waitForWebsocketEvent({ event: 'assetUpdate', id: asset.id });
     }
 
     [
@@ -133,6 +137,12 @@ describe('/search', () => {
     assetLast = assets.at(-1) as AssetMediaResponseDto;
 
     await deleteAssets({ assetBulkDeleteDto: { ids: [assetSilver.id] } }, { headers: asBearerAuth(admin.accessToken) });
+
+    const mapMarkers = await getMapMarkers({}, { headers: asBearerAuth(admin.accessToken) });
+    const nonTrashed = mapMarkers.filter((mark) => mark.id !== assetSilver.id);
+    cities = [...new Set(nonTrashed.map((mark) => mark.city).filter((entry): entry is string => !!entry))].sort();
+    states = [...new Set(nonTrashed.map((mark) => mark.state).filter((entry): entry is string => !!entry))].sort();
+    countries = [...new Set(nonTrashed.map((mark) => mark.country).filter((entry): entry is string => !!entry))].sort();
   }, 30_000);
 
   afterAll(async () => {
@@ -305,126 +315,29 @@ describe('/search', () => {
       {
         should: 'should search by originalFilename with spaces',
         deferred: () => ({
-          dto: { originalFileName: 'samsung-one', type: 'IMAGE' },
+          dto: { originalFileName: 'Samsung One', type: 'IMAGE' },
           assets: [assetOneJpg5, assetOneJpg6, assetOneHeic6],
         }),
       },
       {
         should: 'should search by city',
-        deferred: () => ({
-          dto: {
-            city: 'Accra',
-            includeNull: true,
-          },
-          assets: [assetHeic],
-        }),
-      },
-      {
-        should: "should search city ('')",
-        deferred: () => ({
-          dto: {
-            city: '',
-            isVisible: true,
-            includeNull: true,
-          },
-          assets: [assetLast],
-        }),
-      },
-      {
-        should: 'should search city (null)',
-        deferred: () => ({
-          dto: {
-            city: null,
-            isVisible: true,
-            includeNull: true,
-          },
-          assets: [assetLast],
-        }),
+        deferred: () => ({ dto: { city: 'Accra' }, assets: [assetHeic] }),
       },
       {
         should: 'should search by state',
-        deferred: () => ({
-          dto: {
-            state: 'New York',
-            includeNull: true,
-          },
-          assets: [assetDensity],
-        }),
-      },
-      {
-        should: "should search state ('')",
-        deferred: () => ({
-          dto: {
-            state: '',
-            isVisible: true,
-            withExif: true,
-            includeNull: true,
-          },
-          assets: [assetLast, assetNotocactus],
-        }),
-      },
-      {
-        should: 'should search state (null)',
-        deferred: () => ({
-          dto: {
-            state: null,
-            isVisible: true,
-            includeNull: true,
-          },
-          assets: [assetLast, assetNotocactus],
-        }),
+        deferred: () => ({ dto: { state: 'New York' }, assets: [assetDensity] }),
       },
       {
         should: 'should search by country',
-        deferred: () => ({
-          dto: {
-            country: 'France',
-            includeNull: true,
-          },
-          assets: [assetFalcon],
-        }),
-      },
-      {
-        should: "should search country ('')",
-        deferred: () => ({
-          dto: {
-            country: '',
-            isVisible: true,
-            includeNull: true,
-          },
-          assets: [assetLast],
-        }),
-      },
-      {
-        should: 'should search country (null)',
-        deferred: () => ({
-          dto: {
-            country: null,
-            isVisible: true,
-            includeNull: true,
-          },
-          assets: [assetLast],
-        }),
+        deferred: () => ({ dto: { country: 'France' }, assets: [assetFalcon] }),
       },
       {
         should: 'should search by make',
-        deferred: () => ({
-          dto: {
-            make: 'Canon',
-            includeNull: true,
-          },
-          assets: [assetFalcon, assetDenali],
-        }),
+        deferred: () => ({ dto: { make: 'Canon' }, assets: [assetFalcon, assetDenali] }),
       },
       {
         should: 'should search by model',
-        deferred: () => ({
-          dto: {
-            model: 'Canon EOS 7D',
-            includeNull: true,
-          },
-          assets: [assetDenali],
-        }),
+        deferred: () => ({ dto: { model: 'Canon EOS 7D' }, assets: [assetDenali] }),
       },
       {
         should: 'should allow searching the upload library (libraryId: null)',
@@ -537,79 +450,32 @@ describe('/search', () => {
 
     it('should get suggestions for country', async () => {
       const { status, body } = await request(app)
-        .get('/search/suggestions?type=country&includeNull=true')
+        .get('/search/suggestions?type=country')
         .set('Authorization', `Bearer ${admin.accessToken}`);
-      expect(body).toEqual([
-        'Cuba',
-        'France',
-        'Georgia',
-        'Germany',
-        'Ghana',
-        'Japan',
-        'Morocco',
-        "People's Republic of China",
-        'Russian Federation',
-        'Singapore',
-        'Spain',
-        'Switzerland',
-        'United States of America',
-        null,
-      ]);
+      expect(body).toEqual(countries);
       expect(status).toBe(200);
     });
 
     it('should get suggestions for state', async () => {
       const { status, body } = await request(app)
-        .get('/search/suggestions?type=state&includeNull=true')
+        .get('/search/suggestions?type=state')
         .set('Authorization', `Bearer ${admin.accessToken}`);
-      expect(body).toEqual([
-        'Andalusia',
-        'Berlin',
-        'Glarus',
-        'Greater Accra',
-        'Havana',
-        'Île-de-France',
-        'Marrakesh-Safi',
-        'Mississippi',
-        'New York',
-        'Shanghai',
-        'St.-Petersburg',
-        'Tbilisi',
-        'Tokyo',
-        'Virginia',
-        null,
-      ]);
+      expect(body).toHaveLength(states.length);
+      expect(body).toEqual(expect.arrayContaining(states));
       expect(status).toBe(200);
     });
 
     it('should get suggestions for city', async () => {
       const { status, body } = await request(app)
-        .get('/search/suggestions?type=city&includeNull=true')
+        .get('/search/suggestions?type=city')
         .set('Authorization', `Bearer ${admin.accessToken}`);
-      expect(body).toEqual([
-        'Accra',
-        'Berlin',
-        'Glarus',
-        'Havana',
-        'Marrakesh',
-        'Montalbán de Córdoba',
-        'New York City',
-        'Novena',
-        'Paris',
-        'Philadelphia',
-        'Saint Petersburg',
-        'Shanghai',
-        'Stanley',
-        'Tbilisi',
-        'Tokyo',
-        null,
-      ]);
+      expect(body).toEqual(cities);
       expect(status).toBe(200);
     });
 
     it('should get suggestions for camera make', async () => {
       const { status, body } = await request(app)
-        .get('/search/suggestions?type=camera-make&includeNull=true')
+        .get('/search/suggestions?type=camera-make')
         .set('Authorization', `Bearer ${admin.accessToken}`);
       expect(body).toEqual([
         'Apple',
@@ -619,14 +485,13 @@ describe('/search', () => {
         'PENTAX Corporation',
         'samsung',
         'SONY',
-        null,
       ]);
       expect(status).toBe(200);
     });
 
     it('should get suggestions for camera model', async () => {
       const { status, body } = await request(app)
-        .get('/search/suggestions?type=camera-model&includeNull=true')
+        .get('/search/suggestions?type=camera-model')
         .set('Authorization', `Bearer ${admin.accessToken}`);
       expect(body).toEqual([
         'Canon EOS 7D',
@@ -641,7 +506,6 @@ describe('/search', () => {
         'SM-F711N',
         'SM-S906U',
         'SM-T970',
-        null,
       ]);
       expect(status).toBe(200);
     });

e2e/src/api/specs/server.e2e-spec.ts

@@ -254,7 +254,7 @@ describe('/server', () => {
         .set('Authorization', `Bearer ${admin.accessToken}`)
         .send(serverLicense);
       const { status } = await request(app).get('/server/license').set('Authorization', `Bearer ${admin.accessToken}`);
-      expect(status).toBe(404);
+      expect(status).toBe(200);
     });
   });
 

e2e/src/api/specs/shared-link.e2e-spec.ts

@@ -112,13 +112,6 @@ describe('/shared-links', () => {
       expect(resp.header['content-type']).toContain('text/html');
       expect(resp.text).toContain(`<meta name="description" content="1 shared photos & videos" />`);
     });
-
-    it('should have fqdn og:image meta tag for shared asset', async () => {
-      const resp = await request(shareUrl).get(`/${linkWithAssets.key}`);
-      expect(resp.status).toBe(200);
-      expect(resp.header['content-type']).toContain('text/html');
-      expect(resp.text).toContain(`<meta property="og:image" content="http://`);
-    });
   });
 
   describe('GET /shared-links', () => {

e2e/src/utils.ts

@@ -50,7 +50,7 @@ type CommandResponse = { stdout: string; stderr: string; exitCode: number | null
 type EventType = 'assetUpload' | 'assetUpdate' | 'assetDelete' | 'userDelete' | 'assetHidden';
 type WaitOptions = { event: EventType; id?: string; total?: number; timeout?: number };
 type AdminSetupOptions = { onboarding?: boolean };
-type FileData = { bytes?: Buffer; filename: string };
+type AssetData = { bytes?: Buffer; filename: string };
 
 const dbUrl = 'postgres://postgres:postgres@127.0.0.1:5433/immich';
 export const baseUrl = 'http://127.0.0.1:2283';
@@ -291,10 +291,7 @@ export const utils = {
 
   createAsset: async (
     accessToken: string,
-    dto?: Partial<Omit<AssetMediaCreateDto, 'assetData' | 'sidecarData'>> & {
-      assetData?: FileData;
-      sidecarData?: FileData;
-    },
+    dto?: Partial<Omit<AssetMediaCreateDto, 'assetData'>> & { assetData?: AssetData },
   ) => {
     const _dto = {
       deviceAssetId: 'test-1',
@@ -316,10 +313,6 @@ export const utils = {
       .attach('assetData', assetData, filename)
       .set('Authorization', `Bearer ${accessToken}`);
 
-    if (dto?.sidecarData?.bytes) {
-      void builder.attach('sidecarData', dto.sidecarData.bytes, dto.sidecarData.filename);
-    }
-
     for (const [key, value] of Object.entries(_dto)) {
       void builder.field(key, String(value));
     }
@@ -332,7 +325,7 @@ export const utils = {
   replaceAsset: async (
     accessToken: string,
     assetId: string,
-    dto?: Partial<Omit<AssetMediaCreateDto, 'assetData'>> & { assetData?: FileData },
+    dto?: Partial<Omit<AssetMediaCreateDto, 'assetData'>> & { assetData?: AssetData },
   ) => {
     const _dto = {
       deviceAssetId: 'test-1',
@@ -424,12 +417,12 @@ export const utils = {
 
   createPartner: (accessToken: string, id: string) => createPartner({ id }, { headers: asBearerAuth(accessToken) }),
 
-  setAuthCookies: async (context: BrowserContext, accessToken: string, domain = '127.0.0.1') =>
+  setAuthCookies: async (context: BrowserContext, accessToken: string) =>
     await context.addCookies([
       {
         name: 'immich_access_token',
         value: accessToken,
-        domain,
+        domain: '127.0.0.1',
         path: '/',
         expires: 1_742_402_728,
         httpOnly: true,
@@ -439,7 +432,7 @@ export const utils = {
       {
         name: 'immich_auth_type',
         value: 'password',
-        domain,
+        domain: '127.0.0.1',
         path: '/',
         expires: 1_742_402_728,
         httpOnly: true,
@@ -449,7 +442,7 @@ export const utils = {
       {
         name: 'immich_is_authenticated',
         value: 'true',
-        domain,
+        domain: '127.0.0.1',
         path: '/',
         expires: 1_742_402_728,
         httpOnly: false,

e2e/src/web/specs/asset-viewer/navbar.e2e-spec.ts

@@ -10,9 +10,6 @@ test.describe('Asset Viewer Navbar', () => {
     utils.initSdk();
     await utils.resetDatabase();
     admin = await utils.adminSetup();
-  });
-
-  test.beforeEach(async () => {
     asset = await utils.createAsset(admin.accessToken);
   });
 
@@ -52,14 +49,4 @@ test.describe('Asset Viewer Navbar', () => {
       }
     });
   });
-
-  test.describe('actions', () => {
-    test('favorite asset with shortcut', async ({ context, page }) => {
-      await utils.setAuthCookies(context, admin.accessToken);
-      await page.goto(`/photos/${asset.id}`);
-      await page.waitForSelector('#immich-asset-viewer');
-      await page.keyboard.press('f');
-      await expect(page.locator('#notification-list').getByTestId('message')).toHaveText('Added to favorites');
-    });
-  });
 });

e2e/src/web/specs/asset-viewer/slideshow.e2e-spec.ts

@@ -1,56 +0,0 @@
-import { AssetMediaResponseDto, LoginResponseDto } from '@immich/sdk';
-import { expect, type Page, test } from '@playwright/test';
-import { utils } from 'src/utils';
-
-test.describe('Slideshow', () => {
-  let admin: LoginResponseDto;
-  let asset: AssetMediaResponseDto;
-
-  test.beforeAll(async () => {
-    utils.initSdk();
-    await utils.resetDatabase();
-    admin = await utils.adminSetup();
-    asset = await utils.createAsset(admin.accessToken);
-  });
-
-  const openSlideshow = async (page: Page) => {
-    await page.goto(`/photos/${asset.id}`);
-    await page.waitForSelector('#immich-asset-viewer');
-    await page.getByRole('button', { name: 'More' }).click();
-    await page.getByRole('menuitem', { name: 'Slideshow' }).click();
-  };
-
-  test('open slideshow', async ({ context, page }) => {
-    await utils.setAuthCookies(context, admin.accessToken);
-    await openSlideshow(page);
-    await expect(page.getByRole('button', { name: 'Exit Slideshow' })).toBeVisible();
-  });
-
-  test('exit slideshow with button', async ({ context, page }) => {
-    await utils.setAuthCookies(context, admin.accessToken);
-    await openSlideshow(page);
-
-    const exitButton = page.getByRole('button', { name: 'Exit Slideshow' });
-    await exitButton.click();
-    await expect(exitButton).not.toBeVisible();
-  });
-
-  test('exit slideshow with shortcut', async ({ context, page }) => {
-    await utils.setAuthCookies(context, admin.accessToken);
-    await openSlideshow(page);
-
-    const exitButton = page.getByRole('button', { name: 'Exit Slideshow' });
-    await expect(exitButton).toBeVisible();
-    await page.keyboard.press('Escape');
-    await expect(exitButton).not.toBeVisible();
-  });
-
-  test('favorite shortcut is disabled', async ({ context, page }) => {
-    await utils.setAuthCookies(context, admin.accessToken);
-    await openSlideshow(page);
-
-    await expect(page.getByRole('button', { name: 'Exit Slideshow' })).toBeVisible();
-    await page.keyboard.press('f');
-    await expect(page.locator('#notification-list')).not.toBeVisible();
-  });
-});

e2e/src/web/specs/auth.e2e-spec.ts

@@ -13,7 +13,7 @@ test.describe('Registration', () => {
   test('admin registration', async ({ page }) => {
     // welcome
     await page.goto('/');
-    await page.getByRole('link', { name: 'Getting Started' }).click();
+    await page.getByRole('button', { name: 'Getting Started' }).click();
 
     // register
     await expect(page).toHaveTitle(/Admin Registration/);

e2e/src/web/specs/photo-viewer.e2e-spec.ts

@@ -25,7 +25,7 @@ test.describe('Photo Viewer', () => {
 
   test('initially shows a loading spinner', async ({ page }) => {
     await page.route(`/api/assets/${asset.id}/thumbnail**`, async (route) => {
-      // slow down the request for thumbnail, so spinner has chance to show up
+      // slow down the request for thumbnail, so spiner has chance to show up
       await new Promise((f) => setTimeout(f, 2000));
       await route.continue();
     });
@@ -40,7 +40,7 @@ test.describe('Photo Viewer', () => {
     await page.goto(`/photos/${asset.id}`);
     await expect.poll(async () => await imageLocator(page).getAttribute('src')).toContain('thumbnail');
     const box = await imageLocator(page).boundingBox();
-    expect(box).toBeTruthy();
+    expect(box).toBeTruthy;
     const { x, y, width, height } = box!;
     await page.mouse.move(x + width / 2, y + height / 2);
     await page.mouse.wheel(0, -1);

e2e/src/web/specs/websocket.e2e-spec.ts

@@ -1,25 +0,0 @@
-import { LoginResponseDto } from '@immich/sdk';
-import { expect, test } from '@playwright/test';
-import { utils } from 'src/utils';
-
-test.describe('Websocket', () => {
-  let admin: LoginResponseDto;
-
-  test.beforeAll(async () => {
-    utils.initSdk();
-    await utils.resetDatabase();
-    admin = await utils.adminSetup();
-  });
-
-  test('connects using ipv4', async ({ page, context }) => {
-    await utils.setAuthCookies(context, admin.accessToken);
-    await page.goto('http://127.0.0.1:2283/');
-    await expect(page.locator('#sidebar')).toContainText('Server Online');
-  });
-
-  test('connects using ipv6', async ({ page, context }) => {
-    await utils.setAuthCookies(context, admin.accessToken, '[::1]');
-    await page.goto('http://[::1]:2283/');
-    await expect(page.locator('#sidebar')).toContainText('Server Online');
-  });
-});

e2e/vitest.config.ts

@@ -13,7 +13,6 @@ export default defineConfig({
     include: ['src/{api,cli,immich-admin}/specs/*.e2e-spec.ts'],
     globalSetup,
     testTimeout: 15_000,
-    pool: 'threads',
     poolOptions: {
       threads: {
         singleThread: true,

machine-learning/Dockerfile

@@ -1,21 +1,23 @@
 ARG DEVICE=cpu
 
-FROM python:3.11-bookworm@sha256:d0131ce0ff4bdb5e9eae6bc86ebde891c207d5cac1f3f582b5de0f903cc68384 AS builder-cpu
+FROM python:3.11-bookworm@sha256:7bec1574675e7fd9e3a540a03cd7d6811c59ca261bd300cd665369d8f435298a as builder-cpu
 
-FROM builder-cpu AS builder-openvino
+FROM openvino/ubuntu22_runtime:2023.3.0@sha256:176646df619032ea6c10faf842867119c393e7497b7f88b5e307e932a0fd5aa8 as builder-openvino
+USER root
+RUN apt-get update && apt-get install -y --no-install-recommends python3-dev
 
-FROM builder-cpu AS builder-cuda
+FROM builder-cpu as builder-cuda
 
-FROM builder-cpu AS builder-armnn
+FROM builder-cpu as builder-armnn
 
 ENV ARMNN_PATH=/opt/armnn
 COPY ann /opt/ann
 RUN mkdir /opt/armnn && \
-    curl -SL "https://github.com/ARM-software/armnn/releases/download/v24.05/ArmNN-linux-aarch64.tar.gz" | tar -zx -C /opt/armnn && \
+    curl -SL "https://github.com/ARM-software/armnn/releases/download/v23.11/ArmNN-linux-aarch64.tar.gz" | tar -zx -C /opt/armnn && \
     cd /opt/ann && \
     sh build.sh
 
-FROM builder-${DEVICE} AS builder
+FROM builder-${DEVICE} as builder
 
 ARG DEVICE
 ENV PYTHONDONTWRITEBYTECODE=1 \
@@ -34,27 +36,25 @@ RUN python3 -m venv /opt/venv
 COPY poetry.lock pyproject.toml ./
 RUN poetry install --sync --no-interaction --no-ansi --no-root --with ${DEVICE} --without dev
 
-FROM python:3.11-slim-bookworm@sha256:a90e299af8a9cd6b59c4aaed2b024c78561476978244a1ab89742a4a5ac8c974 AS prod-cpu
+FROM python:3.11-slim-bookworm@sha256:17ec9dc2367aa748559d0212f34665ec4df801129de32db705ea34654b5bc77a as prod-cpu
 
-FROM prod-cpu AS prod-openvino
+FROM openvino/ubuntu22_runtime:2023.3.0@sha256:176646df619032ea6c10faf842867119c393e7497b7f88b5e307e932a0fd5aa8 as prod-openvino
+USER root
+# TODO: remove this once the image has the fix for https://github.com/intel/compute-runtime/issues/710
+ENV NEOReadDebugKeys=1 \
+    OverrideGpuAddressSpace=48
 
-COPY scripts/configure-apt.sh ./
-RUN ./configure-apt.sh && \
-    apt-get update && \
-    apt-get install -t unstable --no-install-recommends -yqq intel-opencl-icd && \
-    rm configure-apt.sh
-
-FROM nvidia/cuda:12.3.2-cudnn9-runtime-ubuntu22.04@sha256:fa44193567d1908f7ca1f3abf8623ce9c63bc8cba7bcfdb32702eb04d326f7a8 AS prod-cuda
+FROM nvidia/cuda:12.2.2-cudnn8-runtime-ubuntu22.04@sha256:2d913b09e6be8387e1a10976933642c73c840c0b735f0bf3c28d97fc9bc422e0 as prod-cuda
 
 COPY --from=builder-cuda /usr/local/bin/python3 /usr/local/bin/python3
 COPY --from=builder-cuda /usr/local/lib/python3.11 /usr/local/lib/python3.11
 COPY --from=builder-cuda /usr/local/lib/libpython3.11.so /usr/local/lib/libpython3.11.so
 
-FROM prod-cpu AS prod-armnn
+FROM prod-cpu as prod-armnn
 
 ENV LD_LIBRARY_PATH=/opt/armnn
 
-RUN apt-get update && apt-get install -y --no-install-recommends ocl-icd-libopencl1 mesa-opencl-icd libgomp1 && \
+RUN apt-get update && apt-get install -y --no-install-recommends ocl-icd-libopencl1 mesa-opencl-icd && \
     rm -rf /var/lib/apt/lists/* && \
     mkdir --parents /etc/OpenCL/vendors && \
     echo "/usr/lib/libmali.so" > /etc/OpenCL/vendors/mali.icd && \
@@ -70,13 +70,10 @@ COPY --from=builder-armnn \
     /opt/ann/build.sh \
     /opt/armnn/
 
-FROM prod-${DEVICE} AS prod
-ARG DEVICE
+FROM prod-${DEVICE} as prod
 
 RUN apt-get update && \
-    apt-get install -y --no-install-recommends tini $(if ! [ "$DEVICE" = "openvino" ]; then echo "libmimalloc2.0"; fi) && \
-    apt-get autoremove -yqq && \
-    apt-get clean && \
+    apt-get install -y --no-install-recommends tini libmimalloc2.0 && \
     rm -rf /var/lib/apt/lists/*
 
 WORKDIR /usr/src/app

machine-learning/ann/ann.cpp

@@ -48,22 +48,21 @@ public:
              bool saveCachedNetwork,
              const char *cachedNetworkPath)
     {
-        NetworkId netId = -2;
-        while (netId == -2)
+        INetworkPtr network = loadModel(modelPath);
+        IOptimizedNetworkPtr optNet = OptimizeNetwork(network.get(), fastMath, fp16, saveCachedNetwork, cachedNetworkPath);
+        const IOInfos infos = getIOInfos(optNet.get());
+        NetworkId netId;
+        mutex.lock();
+        Status status = runtime->LoadNetwork(netId, std::move(optNet));
+        mutex.unlock();
+        if (status != Status::Success)
         {
-            try
-            {
-                netId = loadInternal(modelPath, fastMath, fp16, saveCachedNetwork, cachedNetworkPath);
-            }
-            catch (InvalidArgumentException e)
-            {
-                // fp16 models do not support the forced fp16-turbo (runtime fp32->fp16 conversion)
-                if (fp16)
-                    fp16 = false;
-                else
-                    netId = -1;
-            }
+            return -1;
         }
+        spinLock.lock();
+        ioInfos[netId] = infos;
+        mutexes.emplace(netId, std::make_unique<std::mutex>());
+        spinLock.unlock();
         return netId;
     }
 
@@ -118,8 +117,6 @@ public:
     Ann(int tuningLevel, const char *tuningFile)
     {
         IRuntime::CreationOptions runtimeOptions;
-        runtimeOptions.m_ProfilingOptions.m_EnableProfiling = false;
-        runtimeOptions.m_ProfilingOptions.m_TimelineEnabled = false;
         BackendOptions backendOptions{"GpuAcc",
                                       {
                                           {"TuningLevel", tuningLevel},
@@ -136,30 +133,6 @@ public:
     };
 
 private:
-    int loadInternal(const char *modelPath,
-                     bool fastMath,
-                     bool fp16,
-                     bool saveCachedNetwork,
-                     const char *cachedNetworkPath)
-    {
-        NetworkId netId = -1;
-        INetworkPtr network = loadModel(modelPath);
-        IOptimizedNetworkPtr optNet = OptimizeNetwork(network.get(), fastMath, fp16, saveCachedNetwork, cachedNetworkPath);
-        const IOInfos infos = getIOInfos(optNet.get());
-        mutex.lock();
-        Status status = runtime->LoadNetwork(netId, std::move(optNet));
-        mutex.unlock();
-        if (status != Status::Success)
-        {
-            return -1;
-        }
-        spinLock.lock();
-        ioInfos[netId] = infos;
-        mutexes.emplace(netId, std::make_unique<std::mutex>());
-        spinLock.unlock();
-        return netId;
-    }
-
     INetworkPtr loadModel(const char *modelPath)
     {
         const auto path = std::string(modelPath);
@@ -199,8 +172,6 @@ private:
         options.SetReduceFp32ToFp16(fp16);
         options.SetShapeInferenceMethod(shapeInferenceMethod);
         options.SetAllowExpandedDims(allowExpandedDims);
-        options.SetDebugToFileEnabled(false);
-        options.SetProfilingEnabled(false);
 
         BackendOptions gpuAcc("GpuAcc", {{"FastMathEnabled", fastMath}});
         if (cachedNetworkPath)
@@ -261,8 +232,8 @@ private:
     IRuntime *runtime;
     std::map<NetworkId, IOInfos> ioInfos;
     std::map<NetworkId, std::unique_ptr<std::mutex>> mutexes; // mutex per network to not execute the same the same network concurrently
-    std::mutex mutex;                                         // global mutex for load/unload calls to the runtime
-    SpinLock spinLock;                                        // fast spin lock to guard access to the ioInfos and mutexes maps
+    std::mutex mutex; // global mutex for load/unload calls to the runtime
+    SpinLock spinLock; // fast spin lock to guard access to the ioInfos and mutexes maps
 };
 
 extern "C" void *init(int logLevel, int tuningLevel, const char *tuningFile)

machine-learning/ann/ann.py

@@ -65,7 +65,7 @@ class Ann(metaclass=_Singleton):
         self.input_shapes: dict[int, tuple[tuple[int], ...]] = {}
         self.ann: int | None = None
         self.new()
-
+        
         if self.tuning_file is not None:
             # make sure tuning file exists (without clearing contents)
             # once filled, the tuning file reduces the cost/time of the first
@@ -105,7 +105,7 @@ class Ann(metaclass=_Singleton):
             raise ValueError("model_path must be a file with extension .armnn, .tflite or .onnx")
         if not exists(model_path):
             raise ValueError("model_path must point to an existing file!")
-
+        
         save_cached_network = False
         if cached_network_path is not None and not exists(cached_network_path):
             save_cached_network = True
@@ -120,8 +120,6 @@ class Ann(metaclass=_Singleton):
             save_cached_network,
             cached_network_path.encode() if cached_network_path is not None else None,
         )
-        if net_id < 0:
-            raise ValueError("Cannot load model!")
 
         self.input_shapes[net_id] = tuple(
             self.shape(net_id, input=True, index=i) for i in range(self.tensors(net_id, input=True))

machine-learning/ann/build.sh

@@ -1,3 +0,0 @@
-#!/usr/bin/env sh
-
-g++ -shared -O3 -o libann.so -fuse-ld=gold -std=c++17 -I"$ARMNN_PATH"/include -larmnn -larmnnDeserializer -larmnnTfLiteParser -larmnnOnnxParser -L"$ARMNN_PATH" ann.cpp

machine-learning/ann/export/build-converter.sh

@@ -1,4 +0,0 @@
-#!/usr/bin/env sh
-
-cd armnn-23.11/ || exit
-g++ -o ../armnnconverter -O1 -DARMNN_ONNX_PARSER -DARMNN_SERIALIZER -DARMNN_TF_LITE_PARSER -fuse-ld=gold -std=c++17 -Iinclude -Isrc/armnnUtils -Ithird-party -larmnn -larmnnDeserializer -larmnnTfLiteParser -larmnnOnnxParser -larmnnSerializer -L../armnn src/armnnConverter/ArmnnConverter.cpp

machine-learning/ann/export/env.yaml

@@ -1,201 +0,0 @@
-name: annexport
-channels:
-  - pytorch
-  - nvidia
-  - conda-forge
-dependencies:
-  - _libgcc_mutex=0.1=conda_forge
-  - _openmp_mutex=4.5=2_kmp_llvm
-  - aiohttp=3.9.1=py310h2372a71_0
-  - aiosignal=1.3.1=pyhd8ed1ab_0
-  - arpack=3.8.0=nompi_h0baa96a_101
-  - async-timeout=4.0.3=pyhd8ed1ab_0
-  - attrs=23.1.0=pyh71513ae_1
-  - aws-c-auth=0.7.3=h28f7589_1
-  - aws-c-cal=0.6.1=hc309b26_1
-  - aws-c-common=0.9.0=hd590300_0
-  - aws-c-compression=0.2.17=h4d4d85c_2
-  - aws-c-event-stream=0.3.1=h2e3709c_4
-  - aws-c-http=0.7.11=h00aa349_4
-  - aws-c-io=0.13.32=he9a53bd_1
-  - aws-c-mqtt=0.9.3=hb447be9_1
-  - aws-c-s3=0.3.14=hf3aad02_1
-  - aws-c-sdkutils=0.1.12=h4d4d85c_1
-  - aws-checksums=0.1.17=h4d4d85c_1
-  - aws-crt-cpp=0.21.0=hb942446_5
-  - aws-sdk-cpp=1.10.57=h85b1a90_19
-  - blas=2.120=openblas
-  - blas-devel=3.9.0=20_linux64_openblas
-  - brotli-python=1.0.9=py310hd8f1fbe_9
-  - bzip2=1.0.8=hd590300_5
-  - c-ares=1.23.0=hd590300_0
-  - ca-certificates=2023.11.17=hbcca054_0
-  - certifi=2023.11.17=pyhd8ed1ab_0
-  - charset-normalizer=3.3.2=pyhd8ed1ab_0
-  - click=8.1.7=unix_pyh707e725_0
-  - colorama=0.4.6=pyhd8ed1ab_0
-  - coloredlogs=15.0.1=pyhd8ed1ab_3
-  - cuda-cudart=11.7.99=0
-  - cuda-cupti=11.7.101=0
-  - cuda-libraries=11.7.1=0
-  - cuda-nvrtc=11.7.99=0
-  - cuda-nvtx=11.7.91=0
-  - cuda-runtime=11.7.1=0
-  - dataclasses=0.8=pyhc8e2a94_3
-  - datasets=2.14.7=pyhd8ed1ab_0
-  - dill=0.3.7=pyhd8ed1ab_0
-  - filelock=3.13.1=pyhd8ed1ab_0
-  - flatbuffers=23.5.26=h59595ed_1
-  - freetype=2.12.1=h267a509_2
-  - frozenlist=1.4.0=py310h2372a71_1
-  - fsspec=2023.10.0=pyhca7485f_0
-  - ftfy=6.1.3=pyhd8ed1ab_0
-  - gflags=2.2.2=he1b5a44_1004
-  - glog=0.6.0=h6f12383_0
-  - glpk=5.0=h445213a_0
-  - gmp=6.3.0=h59595ed_0
-  - gmpy2=2.1.2=py310h3ec546c_1
-  - huggingface_hub=0.17.3=pyhd8ed1ab_0
-  - humanfriendly=10.0=pyhd8ed1ab_6
-  - icu=73.2=h59595ed_0
-  - idna=3.6=pyhd8ed1ab_0
-  - importlib-metadata=7.0.0=pyha770c72_0
-  - importlib_metadata=7.0.0=hd8ed1ab_0
-  - joblib=1.3.2=pyhd8ed1ab_0
-  - keyutils=1.6.1=h166bdaf_0
-  - krb5=1.21.2=h659d440_0
-  - lcms2=2.15=h7f713cb_2
-  - ld_impl_linux-64=2.40=h41732ed_0
-  - lerc=4.0.0=h27087fc_0
-  - libabseil=20230125.3=cxx17_h59595ed_0
-  - libarrow=12.0.1=hb87d912_8_cpu
-  - libblas=3.9.0=20_linux64_openblas
-  - libbrotlicommon=1.0.9=h166bdaf_9
-  - libbrotlidec=1.0.9=h166bdaf_9
-  - libbrotlienc=1.0.9=h166bdaf_9
-  - libcblas=3.9.0=20_linux64_openblas
-  - libcrc32c=1.1.2=h9c3ff4c_0
-  - libcublas=11.10.3.66=0
-  - libcufft=10.7.2.124=h4fbf590_0
-  - libcufile=1.8.1.2=0
-  - libcurand=10.3.4.101=0
-  - libcurl=8.5.0=hca28451_0
-  - libcusolver=11.4.0.1=0
-  - libcusparse=11.7.4.91=0
-  - libdeflate=1.19=hd590300_0
-  - libedit=3.1.20191231=he28a2e2_2
-  - libev=4.33=hd590300_2
-  - libevent=2.1.12=hf998b51_1
-  - libffi=3.4.2=h7f98852_5
-  - libgcc-ng=13.2.0=h807b86a_3
-  - libgfortran-ng=13.2.0=h69a702a_3
-  - libgfortran5=13.2.0=ha4646dd_3
-  - libgoogle-cloud=2.12.0=hac9eb74_1
-  - libgrpc=1.54.3=hb20ce57_0
-  - libhwloc=2.9.3=default_h554bfaf_1009
-  - libiconv=1.17=hd590300_1
-  - libjpeg-turbo=2.1.5.1=hd590300_1
-  - liblapack=3.9.0=20_linux64_openblas
-  - liblapacke=3.9.0=20_linux64_openblas
-  - libnghttp2=1.58.0=h47da74e_1
-  - libnpp=11.7.4.75=0
-  - libnsl=2.0.1=hd590300_0
-  - libnuma=2.0.16=h0b41bf4_1
-  - libnvjpeg=11.8.0.2=0
-  - libopenblas=0.3.25=pthreads_h413a1c8_0
-  - libpng=1.6.39=h753d276_0
-  - libprotobuf=3.21.12=hfc55251_2
-  - libsentencepiece=0.1.99=h180e1df_0
-  - libsqlite=3.44.2=h2797004_0
-  - libssh2=1.11.0=h0841786_0
-  - libstdcxx-ng=13.2.0=h7e041cc_3
-  - libthrift=0.18.1=h8fd135c_2
-  - libtiff=4.6.0=h29866fb_1
-  - libutf8proc=2.8.0=h166bdaf_0
-  - libuuid=2.38.1=h0b41bf4_0
-  - libwebp-base=1.3.2=hd590300_0
-  - libxcb=1.15=h0b41bf4_0
-  - libxml2=2.11.6=h232c23b_0
-  - libzlib=1.2.13=hd590300_5
-  - llvm-openmp=17.0.6=h4dfa4b3_0
-  - lz4-c=1.9.4=hcb278e6_0
-  - mkl=2022.2.1=h84fe81f_16997
-  - mkl-devel=2022.2.1=ha770c72_16998
-  - mkl-include=2022.2.1=h84fe81f_16997
-  - mpc=1.3.1=hfe3b2da_0
-  - mpfr=4.2.1=h9458935_0
-  - mpmath=1.3.0=pyhd8ed1ab_0
-  - multidict=6.0.4=py310h2372a71_1
-  - multiprocess=0.70.15=py310h2372a71_1
-  - ncurses=6.4=h59595ed_2
-  - numpy=1.26.2=py310hb13e2d6_0
-  - onnx=1.14.0=py310ha3deec4_1
-  - onnx2torch=1.5.13=pyhd8ed1ab_0
-  - onnxruntime=1.16.3=py310hd4b7fbc_1_cpu
-  - open-clip-torch=2.23.0=pyhd8ed1ab_1
-  - openblas=0.3.25=pthreads_h7a3da1a_0
-  - openjpeg=2.5.0=h488ebb8_3
-  - openssl=3.2.0=hd590300_1
-  - orc=1.9.0=h2f23424_1
-  - packaging=23.2=pyhd8ed1ab_0
-  - pandas=2.1.4=py310hcc13569_0
-  - pillow=10.0.1=py310h29da1c1_1
-  - pip=23.3.1=pyhd8ed1ab_0
-  - protobuf=4.21.12=py310heca2aa9_0
-  - pthread-stubs=0.4=h36c2ea0_1001
-  - pyarrow=12.0.1=py310h0576679_8_cpu
-  - pyarrow-hotfix=0.6=pyhd8ed1ab_0
-  - pysocks=1.7.1=pyha2e5f31_6
-  - python=3.10.13=hd12c33a_0_cpython
-  - python-dateutil=2.8.2=pyhd8ed1ab_0
-  - python-flatbuffers=23.5.26=pyhd8ed1ab_0
-  - python-tzdata=2023.3=pyhd8ed1ab_0
-  - python-xxhash=3.4.1=py310h2372a71_0
-  - python_abi=3.10=4_cp310
-  - pytorch=1.13.1=cpu_py310hd11e9c7_1
-  - pytorch-cuda=11.7=h778d358_5
-  - pytorch-mutex=1.0=cuda
-  - pytz=2023.3.post1=pyhd8ed1ab_0
-  - pyyaml=6.0.1=py310h2372a71_1
-  - rdma-core=28.9=h59595ed_1
-  - re2=2023.03.02=h8c504da_0
-  - readline=8.2=h8228510_1
-  - regex=2023.10.3=py310h2372a71_0
-  - requests=2.31.0=pyhd8ed1ab_0
-  - s2n=1.3.49=h06160fa_0
-  - sacremoses=0.0.53=pyhd8ed1ab_0
-  - safetensors=0.3.3=py310hcb5633a_1
-  - sentencepiece=0.1.99=hff52083_0
-  - sentencepiece-python=0.1.99=py310hebdb9f0_0
-  - sentencepiece-spm=0.1.99=h180e1df_0
-  - setuptools=68.2.2=pyhd8ed1ab_0
-  - six=1.16.0=pyh6c4a22f_0
-  - sleef=3.5.1=h9b69904_2
-  - snappy=1.1.10=h9fff704_0
-  - sympy=1.12=pypyh9d50eac_103
-  - tbb=2021.11.0=h00ab1b0_0
-  - texttable=1.7.0=pyhd8ed1ab_0
-  - timm=0.9.12=pyhd8ed1ab_0
-  - tk=8.6.13=noxft_h4845f30_101
-  - tokenizers=0.14.1=py310h320607d_2
-  - torchvision=0.14.1=cpu_py310hd3d2ac3_1
-  - tqdm=4.66.1=pyhd8ed1ab_0
-  - transformers=4.35.2=pyhd8ed1ab_0
-  - typing-extensions=4.9.0=hd8ed1ab_0
-  - typing_extensions=4.9.0=pyha770c72_0
-  - tzdata=2023c=h71feb2d_0
-  - ucx=1.14.1=h64cca9d_5
-  - urllib3=2.1.0=pyhd8ed1ab_0
-  - wcwidth=0.2.12=pyhd8ed1ab_0
-  - wheel=0.42.0=pyhd8ed1ab_0
-  - xorg-libxau=1.0.11=hd590300_0
-  - xorg-libxdmcp=1.1.3=h7f98852_0
-  - xxhash=0.8.2=hd590300_0
-  - xz=5.2.6=h166bdaf_0
-  - yaml=0.2.5=h7f98852_2
-  - yarl=1.9.3=py310h2372a71_0
-  - zipp=3.17.0=pyhd8ed1ab_0
-  - zlib=1.2.13=hd590300_5
-  - zstd=1.5.5=hfc55251_0
-  - pip:
-      - git+https://github.com/fyfrey/TinyNeuralNetwork.git

machine-learning/ann/export/run.py

@@ -1,157 +0,0 @@
-import logging
-import os
-import platform
-import subprocess
-from abc import abstractmethod
-
-import onnx
-import open_clip
-import torch
-from onnx2torch import convert
-from onnxruntime.tools.onnx_model_utils import fix_output_shapes, make_input_shape_fixed
-from tinynn.converter import TFLiteConverter
-
-
-class ExportBase(torch.nn.Module):
-    input_shape: tuple[int, ...]
-
-    def __init__(self, device: torch.device, name: str):
-        super().__init__()
-        self.device = device
-        self.name = name
-        self.optimize = 5
-        self.nchw_transpose = False
-
-    @abstractmethod
-    def forward(self, input_tensor: torch.Tensor) -> torch.Tensor | tuple[torch.Tensor]:
-        pass
-
-    def dummy_input(self) -> torch.FloatTensor:
-        return torch.rand((1, 3, 224, 224), device=self.device)
-
-
-class ArcFace(ExportBase):
-    input_shape = (1, 3, 112, 112)
-
-    def __init__(self, onnx_model_path: str, device: torch.device):
-        name, _ = os.path.splitext(os.path.basename(onnx_model_path))
-        super().__init__(device, name)
-        onnx_model = onnx.load_model(onnx_model_path)
-        make_input_shape_fixed(onnx_model.graph, onnx_model.graph.input[0].name, self.input_shape)
-        fix_output_shapes(onnx_model)
-        self.model = convert(onnx_model).to(device)
-        if self.device.type == "cuda":
-            self.model = self.model.half()
-
-    def forward(self, input_tensor: torch.Tensor) -> torch.FloatTensor:
-        embedding: torch.FloatTensor = self.model(
-            input_tensor.half() if self.device.type == "cuda" else input_tensor
-        ).float()
-        assert isinstance(embedding, torch.FloatTensor)
-        return embedding
-
-    def dummy_input(self) -> torch.FloatTensor:
-        return torch.rand(self.input_shape, device=self.device)
-
-
-class RetinaFace(ExportBase):
-    input_shape = (1, 3, 640, 640)
-
-    def __init__(self, onnx_model_path: str, device: torch.device):
-        name, _ = os.path.splitext(os.path.basename(onnx_model_path))
-        super().__init__(device, name)
-        self.optimize = 3
-        self.model = convert(onnx_model_path).eval().to(device)
-        if self.device.type == "cuda":
-            self.model = self.model.half()
-
-    def forward(self, input_tensor: torch.Tensor) -> tuple[torch.FloatTensor]:
-        out: torch.Tensor = self.model(input_tensor.half() if self.device.type == "cuda" else input_tensor)
-        return tuple(o.float() for o in out)
-
-    def dummy_input(self) -> torch.FloatTensor:
-        return torch.rand(self.input_shape, device=self.device)
-
-
-class ClipVision(ExportBase):
-    input_shape = (1, 3, 224, 224)
-
-    def __init__(self, model_name: str, weights: str, device: torch.device):
-        super().__init__(device, model_name + "__" + weights)
-        self.model = open_clip.create_model(
-            model_name,
-            weights,
-            precision="fp16" if device.type == "cuda" else "fp32",
-            jit=False,
-            require_pretrained=True,
-            device=device,
-        )
-
-    def forward(self, input_tensor: torch.Tensor) -> torch.FloatTensor:
-        embedding: torch.Tensor = self.model.encode_image(
-            input_tensor.half() if self.device.type == "cuda" else input_tensor,
-            normalize=True,
-        ).float()
-        return embedding
-
-
-def export(model: ExportBase) -> None:
-    model.eval()
-    for param in model.parameters():
-        param.requires_grad = False
-    dummy_input = model.dummy_input()
-    model(dummy_input)
-    jit = torch.jit.trace(model, dummy_input)  # type: ignore[no-untyped-call,attr-defined]
-    tflite_model_path = f"output/{model.name}.tflite"
-    os.makedirs("output", exist_ok=True)
-
-    converter = TFLiteConverter(
-        jit,
-        dummy_input,
-        tflite_model_path,
-        optimize=model.optimize,
-        nchw_transpose=model.nchw_transpose,
-    )
-    # segfaults on ARM, must run on x86_64 / AMD64
-    converter.convert()
-
-    armnn_model_path = f"output/{model.name}.armnn"
-    os.environ["LD_LIBRARY_PATH"] = "armnn"
-    subprocess.run(
-        [
-            "./armnnconverter",
-            "-f",
-            "tflite-binary",
-            "-m",
-            tflite_model_path,
-            "-i",
-            "input_tensor",
-            "-o",
-            "output_tensor",
-            "-p",
-            armnn_model_path,
-        ]
-    )
-
-
-def main() -> None:
-    if platform.machine() not in ("x86_64", "AMD64"):
-        raise RuntimeError(f"Can only run on x86_64 / AMD64, not {platform.machine()}")
-
-    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-    if device.type != "cuda":
-        logging.warning(
-            "No CUDA available, cannot create fp16 model! proceeding to create a fp32 model (use only for testing)"
-        )
-    models = [
-        ClipVision("ViT-B-32", "openai", device),
-        ArcFace("buffalo_l_rec.onnx", device),
-        RetinaFace("buffalo_l_det.onnx", device),
-    ]
-    for model in models:
-        export(model)
-
-
-if __name__ == "__main__":
-    with torch.no_grad():
-        main()

machine-learning/app/config.py

@@ -30,8 +30,6 @@ class Settings(BaseSettings):
     model_inter_op_threads: int = 0
     model_intra_op_threads: int = 0
     ann: bool = True
-    ann_fp16_turbo: bool = False
-    ann_tuning_level: int = 2
     preload: PreloadModelData | None = None
 
     class Config:

machine-learning/app/models/constants.py

@@ -2,64 +2,53 @@ from app.config import clean_name
 from app.schemas import ModelSource
 
 _OPENCLIP_MODELS = {
-    "RN101__openai",
-    "RN101__yfcc15m",
-    "RN50__cc12m",
     "RN50__openai",
     "RN50__yfcc15m",
-    "RN50x16__openai",
+    "RN50__cc12m",
+    "RN101__openai",
+    "RN101__yfcc15m",
     "RN50x4__openai",
+    "RN50x16__openai",
     "RN50x64__openai",
-    "ViT-B-16-SigLIP-256__webli",
-    "ViT-B-16-SigLIP-384__webli",
-    "ViT-B-16-SigLIP-512__webli",
-    "ViT-B-16-SigLIP-i18n-256__webli",
-    "ViT-B-16-SigLIP__webli",
-    "ViT-B-16-plus-240__laion400m_e31",
-    "ViT-B-16-plus-240__laion400m_e32",
-    "ViT-B-16__laion400m_e31",
-    "ViT-B-16__laion400m_e32",
-    "ViT-B-16__openai",
-    "ViT-B-32__laion2b-s34b-b79k",
+    "ViT-B-32__openai",
     "ViT-B-32__laion2b_e16",
     "ViT-B-32__laion400m_e31",
     "ViT-B-32__laion400m_e32",
-    "ViT-B-32__openai",
-    "ViT-H-14-378-quickgelu__dfn5b",
-    "ViT-H-14-quickgelu__dfn5b",
-    "ViT-H-14__laion2b-s32b-b79k",
-    "ViT-L-14-336__openai",
-    "ViT-L-14-quickgelu__dfn2b",
-    "ViT-L-14__laion2b-s32b-b82k",
+    "ViT-B-32__laion2b-s34b-b79k",
+    "ViT-B-16__openai",
+    "ViT-B-16__laion400m_e31",
+    "ViT-B-16__laion400m_e32",
+    "ViT-B-16-plus-240__laion400m_e31",
+    "ViT-B-16-plus-240__laion400m_e32",
+    "ViT-L-14__openai",
     "ViT-L-14__laion400m_e31",
     "ViT-L-14__laion400m_e32",
-    "ViT-L-14__openai",
-    "ViT-L-16-SigLIP-256__webli",
-    "ViT-L-16-SigLIP-384__webli",
-    "ViT-SO400M-14-SigLIP-384__webli",
+    "ViT-L-14__laion2b-s32b-b82k",
+    "ViT-L-14-336__openai",
+    "ViT-H-14__laion2b-s32b-b79k",
     "ViT-g-14__laion2b-s12b-b42k",
-    "XLM-Roberta-Base-ViT-B-32__laion5b_s13b_b90k",
+    "ViT-L-14-quickgelu__dfn2b",
+    "ViT-H-14-quickgelu__dfn5b",
+    "ViT-H-14-378-quickgelu__dfn5b",
     "XLM-Roberta-Large-ViT-H-14__frozen_laion5b_s13b_b90k",
-    "nllb-clip-base-siglip__mrl",
     "nllb-clip-base-siglip__v1",
-    "nllb-clip-large-siglip__mrl",
     "nllb-clip-large-siglip__v1",
 }
 
 
 _MCLIP_MODELS = {
     "LABSE-Vit-L-14",
-    "XLM-Roberta-Large-Vit-B-16Plus",
     "XLM-Roberta-Large-Vit-B-32",
+    "XLM-Roberta-Large-Vit-B-16Plus",
     "XLM-Roberta-Large-Vit-L-14",
 }
 
 
 _INSIGHTFACE_MODELS = {
     "antelopev2",
-    "buffalo_s",
-    "buffalo_m",
     "buffalo_l",
+    "buffalo_m",
+    "buffalo_s",
 }
 
 

machine-learning/app/models/transforms.py

@@ -47,8 +47,8 @@ def pil_to_cv2(image: Image.Image) -> NDArray[np.uint8]:
 def decode_pil(image_bytes: bytes | IO[bytes] | Image.Image) -> Image.Image:
     if isinstance(image_bytes, Image.Image):
         return image_bytes
-    image: Image.Image = Image.open(BytesIO(image_bytes) if isinstance(image_bytes, bytes) else image_bytes)
-    image.load()
+    image = Image.open(BytesIO(image_bytes) if isinstance(image_bytes, bytes) else image_bytes)
+    image.load()  # type: ignore
     if not image.mode == "RGB":
         image = image.convert("RGB")
     return image

machine-learning/app/sessions/ann.py

@@ -20,13 +20,12 @@ class AnnSession:
     def __init__(self, model_path: Path, cache_dir: Path = settings.cache_folder) -> None:
         self.model_path = model_path
         self.cache_dir = cache_dir
-        self.ann = Ann(tuning_level=settings.ann_tuning_level, tuning_file=(cache_dir / "gpu-tuning.ann").as_posix())
+        self.ann = Ann(tuning_level=3, tuning_file=(cache_dir / "gpu-tuning.ann").as_posix())
 
         log.info("Loading ANN model %s ...", model_path)
         self.model = self.ann.load(
             model_path.as_posix(),
             cached_network_path=model_path.with_suffix(".anncache").as_posix(),
-            fp16=settings.ann_fp16_turbo,
         )
         log.info("Loaded ANN model with ID %d", self.model)
 

machine-learning/app/sessions/ort.py

@@ -83,21 +83,17 @@ class OrtSession:
 
     @property
     def _provider_options_default(self) -> list[dict[str, Any]]:
-        provider_options = []
+        options = []
         for provider in self.providers:
             match provider:
                 case "CPUExecutionProvider" | "CUDAExecutionProvider":
-                    options = {"arena_extend_strategy": "kSameAsRequested"}
+                    option = {"arena_extend_strategy": "kSameAsRequested"}
                 case "OpenVINOExecutionProvider":
-                    options = {
-                        "device_type": "GPU",
-                        "precision": "FP32",
-                        "cache_dir": (self.model_path.parent / "openvino").as_posix(),
-                    }
+                    option = {"device_type": "GPU_FP32", "cache_dir": (self.model_path.parent / "openvino").as_posix()}
                 case _:
-                    options = {}
-            provider_options.append(options)
-        return provider_options
+                    option = {}
+            options.append(option)
+        return options
 
     @property
     def sess_options(self) -> ort.SessionOptions:

machine-learning/app/test_main.py

@@ -212,7 +212,7 @@ class TestOrtSession:
         session = OrtSession(model_path, providers=["OpenVINOExecutionProvider", "CPUExecutionProvider"])
 
         assert session.provider_options == [
-            {"device_type": "GPU", "precision": "FP32", "cache_dir": "/cache/ViT-B-32__openai/openvino"},
+            {"device_type": "GPU_FP32", "cache_dir": "/cache/ViT-B-32__openai/openvino"},
             {"arena_extend_strategy": "kSameAsRequested"},
         ]
 
@@ -268,9 +268,9 @@ class TestAnnSession:
 
         AnnSession(model_path, cache_dir)
 
-        ann_session.assert_called_once_with(tuning_level=2, tuning_file=(cache_dir / "gpu-tuning.ann").as_posix())
+        ann_session.assert_called_once_with(tuning_level=3, tuning_file=(cache_dir / "gpu-tuning.ann").as_posix())
         ann_session.return_value.load.assert_called_once_with(
-            model_path.as_posix(), cached_network_path=model_path.with_suffix(".anncache").as_posix(), fp16=False
+            model_path.as_posix(), cached_network_path=model_path.with_suffix(".anncache").as_posix()
         )
         info.assert_has_calls(
             [

machine-learning/export/ann/Dockerfile

@@ -0,0 +1,35 @@
+FROM mambaorg/micromamba:bookworm-slim@sha256:333f7598ff2c2400fb10bfe057709c68b7daab5d847143af85abcf224a07271a as builder
+
+USER root
+RUN apt-get update && apt-get install -y --no-install-recommends \
+  build-essential \
+  cmake \
+  curl \
+  git
+USER $MAMBA_USER
+
+WORKDIR /home/mambauser
+ENV ARMNN_PATH=armnn
+COPY --chown=$MAMBA_USER:$MAMBA_USER scripts/* .
+RUN ./download-armnn.sh && \
+  ./build-converter.sh && \
+  ./build.sh
+
+COPY --chown=$MAMBA_USER:$MAMBA_USER conda-lock.yml .
+RUN micromamba create -y -p /home/mambauser/venv -f conda-lock.yml && \
+  micromamba clean --all --yes
+ENV PATH="/home/mambauser/venv/bin:${PATH}"
+
+FROM gcr.io/distroless/base-debian12
+# FROM mambaorg/micromamba:bookworm-slim@sha256:333f7598ff2c2400fb10bfe057709c68b7daab5d847143af85abcf224a07271a
+
+WORKDIR /export/ann
+ENV PYTHONDONTWRITEBYTECODE=1 \
+  LD_LIBRARY_PATH=/export/ann/armnn \
+  PATH="/opt/venv/bin:${PATH}"
+
+COPY --from=builder /home/mambauser/armnnconverter /home/mambauser/armnn ./
+COPY --from=builder /home/mambauser/venv /opt/venv
+COPY --chown=$MAMBA_USER:$MAMBA_USER onnx2ann onnx2ann
+
+ENTRYPOINT ["python", "-m", "onnx2ann"]

machine-learning/export/ann/env.yaml

@@ -0,0 +1,21 @@
+name: onnx2ann
+channels:
+  - conda-forge
+dependencies:
+  - python>=3.11,<4.0
+  - onnx>=1.16.1
+  # - onnxruntime>=1.18.1 # conda only has gpu version
+  - psutil>=6.0.0
+  - flatbuffers>=24.3.25
+  - ml_dtypes>=0.3.1
+  - typer-slim>=0.12.3
+  - huggingface_hub>=0.23.4
+  - pip
+  - pip:
+    - onnxruntime>=1.18.1 # conda only has gpu version
+    - onnxsim>=0.4.36
+    - onnx2tf>=1.24.1
+    - onnx_graphsurgeon>=0.5.2
+    - simple_onnx_processing_tools>=1.1.32
+    - tf_keras>=2.16.0
+    - git+https://github.com/microsoft/onnxconverter-common.git

machine-learning/export/ann/onnx2ann/__main__.py

@@ -0,0 +1,99 @@
+import os
+import platform
+from typing import Annotated, Optional
+
+import typer
+
+from onnx2ann.export import Exporter, ModelType, Precision
+
+app = typer.Typer(add_completion=False, pretty_exceptions_show_locals=False)
+
+
+@app.command()
+def export(
+    model_name: Annotated[
+        str, typer.Argument(..., help="The name of the model to be exported as it exists in Hugging Face.")
+    ],
+    model_type: Annotated[ModelType, typer.Option(..., "--type", "-t", help="The type of model to be exported.")],
+    input_shapes: Annotated[
+        list[str],
+        typer.Option(
+            ...,
+            "--input-shape",
+            "-s",
+            help="The shape of an input tensor to the model, each dimension separated by commas. "
+            "Multiple shapes can be provided for multiple inputs.",
+        ),
+    ],
+    precision: Annotated[
+        Precision,
+        typer.Option(
+            ...,
+            "--precision",
+            "-p",
+            help="The precision of the exported model. `float16` requires a GPU.",
+        ),
+    ] = Precision.FLOAT32,
+    cache_dir: Annotated[
+        str,
+        typer.Option(
+            ...,
+            "--cache-dir",
+            "-c",
+            help="Directory where pre-export models will be stored.",
+            envvar="CACHE_DIR",
+            show_envvar=True,
+        ),
+    ] = "~/.cache/huggingface",
+    output_dir: Annotated[
+        str,
+        typer.Option(
+            ...,
+            "--output-dir",
+            "-o",
+            help="Directory where exported models will be stored.",
+        ),
+    ] = "output",
+    auth_token: Annotated[
+        Optional[str],
+        typer.Option(
+            ...,
+            "--auth-token",
+            "-t",
+            help="If uploading models to Hugging Face, the auth token of the user or organisation.",
+            envvar="HF_AUTH_TOKEN",
+            show_envvar=True,
+        ),
+    ] = None,
+    force_export: Annotated[
+        bool,
+        typer.Option(
+            ...,
+            "--force-export",
+            "-f",
+            help="Export the model even if an exported model already exists in the output directory.",
+        ),
+    ] = False,
+) -> None:
+    if platform.machine() not in ("x86_64", "AMD64"):
+        msg = f"Can only run on x86_64 / AMD64, not {platform.machine()}"
+        raise RuntimeError(msg)
+    os.environ.setdefault("LD_LIBRARY_PATH", "armnn")
+    parsed_input_shapes = [tuple(map(int, shape.split(","))) for shape in input_shapes]
+    model = Exporter(
+        model_name, model_type, input_shapes=parsed_input_shapes, cache_dir=cache_dir, force_export=force_export
+    )
+    model_dir = os.path.join("output", model_name)
+    output_dir = os.path.join(model_dir, model_type)
+    armnn_model = model.to_armnn(output_dir, precision)
+
+    if not auth_token:
+        return
+
+    from huggingface_hub import upload_file
+
+    relative_path = os.path.relpath(armnn_model, start=model_dir)
+    upload_file(path_or_fileobj=armnn_model, path_in_repo=relative_path, repo_id=model.repo_name, token=auth_token)
+
+
+app()

machine-learning/export/ann/onnx2ann/export.py

@@ -0,0 +1,129 @@
+import os
+import subprocess
+from enum import StrEnum
+
+from onnx2ann.helpers import onnx_make_armnn_compatible, onnx_make_inputs_fixed
+
+
+class ModelType(StrEnum):
+    VISUAL = "visual"
+    TEXTUAL = "textual"
+    RECOGNITION = "recognition"
+    DETECTION = "detection"
+
+
+class Precision(StrEnum):
+    FLOAT16 = "float16"
+    FLOAT32 = "float32"
+
+
+class Exporter:
+    def __init__(
+        self,
+        model_name: str,
+        model_type: str,
+        input_shapes: list[tuple[int, ...]],
+        optimization_level: int = 5,
+        cache_dir: str = os.environ.get("CACHE_DIR", "~/.cache/huggingface"),
+        force_export: bool = False,
+    ):
+        self.model_name = model_name.split("/")[-1]
+        self.model_type = model_type
+        self.optimize = optimization_level
+        self.input_shapes = input_shapes
+        self.cache_dir = os.path.join(cache_dir, self.repo_name)
+        self.force_export = force_export
+
+    def download(self) -> str:
+        model_path = os.path.join(self.cache_dir, self.model_type, "model.onnx")
+        if os.path.isfile(model_path):
+            print(f"Model is already downloaded at {model_path}")
+            return model_path
+        from huggingface_hub import snapshot_download
+
+        snapshot_download(
+            self.repo_name, cache_dir=self.cache_dir, local_dir=self.cache_dir, local_dir_use_symlinks=False
+        )
+        return model_path
+
+    def to_onnx_static(self, precision: Precision) -> str:
+        import onnx
+        from onnxconverter_common import float16
+        onnx_path_original = self.download()
+        static_dir = os.path.join(self.cache_dir, self.model_type, "static")
+
+        static_path = os.path.join(static_dir, f"model.onnx")
+        if self.force_export and not os.path.isfile(static_path):
+            print(f"Making {self} static")
+            os.makedirs(static_dir, exist_ok=True)
+            onnx_make_inputs_fixed(onnx_path_original, static_path, self.input_shapes)
+            onnx_make_armnn_compatible(static_path)
+            print(f"Finished making {self} static")
+
+        model = onnx.load(static_path)
+        self.inputs = [input_.name for input_ in model.graph.input]
+        self.outputs = [output_.name for output_ in model.graph.output]
+        if precision == Precision.FLOAT16:
+            static_path = os.path.join(static_dir, f"model_{precision}.onnx")
+            print(f"Converting {self} to {precision} precision")
+            model = float16.convert_float_to_float16(model, keep_io_types=True, disable_shape_infer=True)
+            onnx.save(model, static_path)
+            print(f"Finished converting {self} to {precision} precision")
+        # self.inputs, self.outputs = onnx_get_inputs_outputs(static_path)
+        return static_path
+
+    def to_tflite(self, output_dir: str, precision: Precision) -> str:
+        onnx_model = self.to_onnx_static(precision)
+        tflite_dir = os.path.join(output_dir, precision)
+        tflite_model = os.path.join(tflite_dir, f"model_{precision}.tflite")
+        if self.force_export or not os.path.isfile(tflite_model):
+            import onnx2tf
+
+            print(f"Exporting {self} to TFLite with {precision} precision (this might take a few minutes)")
+            onnx2tf.convert(
+                input_onnx_file_path=onnx_model,
+                output_folder_path=tflite_dir,
+                keep_shape_absolutely_input_names=self.inputs,
+                # verbosity="warn",
+                copy_onnx_input_output_names_to_tflite=True,
+                output_signaturedefs=True,
+                not_use_onnxsim=True,
+            )
+            print(f"Finished exporting {self} to TFLite with {precision} precision")
+
+        return tflite_model
+
+    def to_armnn(self, output_dir: str, precision: Precision) -> tuple[str, str]:
+        armnn_model = os.path.join(output_dir, "model.armnn")
+        if not self.force_export and os.path.isfile(armnn_model):
+            return armnn_model
+
+        tflite_model_dir = os.path.join(output_dir, "tflite")
+        tflite_model = self.to_tflite(tflite_model_dir, precision)
+
+        args = ["./armnnconverter", "-f", "tflite-binary", "-m", tflite_model, "-p", armnn_model]
+        args.append("-i")
+        args.extend(self.inputs)
+        args.append("-o")
+        args.extend(self.outputs)
+
+        print(f"Exporting {self} to ARM NN with {precision} precision")
+        try:
+            if (stdout := subprocess.check_output(args, stderr=subprocess.STDOUT).decode()):
+                print(stdout)
+            print(f"Finished exporting {self} to ARM NN with {precision} precision")
+        except subprocess.CalledProcessError as e:
+            print(e.output.decode())
+            try:
+                from shutil import rmtree
+
+                rmtree(tflite_model_dir, ignore_errors=True)
+            finally:
+                raise e
+
+    @property
+    def repo_name(self) -> str:
+        return f"immich-app/{self.model_name}"
+
+    def __repr__(self) -> str:
+        return f"{self.model_name} ({self.model_type})"

machine-learning/export/ann/onnx2ann/helpers.py

@@ -0,0 +1,260 @@
+from typing import Any
+
+
+def onnx_make_armnn_compatible(model_path: str) -> None:
+    """
+    i can explain
+    armnn only supports up to 4d tranposes, but the model has a 5d transpose due to a redundant unsqueeze
+    this function folds the unsqueeze+transpose+squeeze into a single 4d transpose
+    it also switches from gather ops to slices since armnn has different dimension semantics for gathers
+    also fixes batch normalization being in training mode
+    """
+
+    import numpy as np
+    import onnx
+    from onnx_graphsurgeon import Constant, Node, Variable, export_onnx, import_onnx
+
+    proto = onnx.load(model_path)
+    graph = import_onnx(proto)
+
+    gather_idx = 1
+    squeeze_idx = 1
+    for node in graph.nodes:
+        for link1 in node.outputs:
+            if "Unsqueeze" in link1.name:
+                for node1 in link1.outputs:
+                    for link2 in node1.outputs:
+                        if "Transpose" in link2.name:
+                            for node2 in link2.outputs:
+                                if node2.attrs.get("perm") == [3, 1, 2, 0, 4]:
+                                    node2.attrs["perm"] = [2, 0, 1, 3]
+                                    link2.shape = link1.shape
+                                    for link3 in node2.outputs:
+                                        if "Squeeze" in link3.name:
+                                            link3.shape = [link3.shape[x] for x in [0, 1, 2, 4]]
+                                            for node3 in link3.outputs:
+                                                for link4 in node3.outputs:
+                                                    link4.shape = link3.shape
+                                                    try:
+                                                        idx = link2.inputs.index(node1)
+                                                        link2.inputs[idx] = node
+                                                    except ValueError:
+                                                        pass
+
+                                                    node.outputs = [link2]
+                                                    if "Gather" in link4.name:
+                                                        for node4 in link4.outputs:
+                                                            axis = node1.attrs.get("axis", 0)
+                                                            index = node4.inputs[1].values
+                                                            slice_link = Variable(
+                                                                f"onnx::Slice_123{gather_idx}",
+                                                                dtype=link4.dtype,
+                                                                shape=[1] + link3.shape[1:],
+                                                            )
+                                                            slice_node = Node(
+                                                                op="Slice",
+                                                                inputs=[
+                                                                    link3,
+                                                                    Constant(
+                                                                        f"SliceStart_123{gather_idx}",
+                                                                        np.array([index]),
+                                                                    ),
+                                                                    Constant(
+                                                                        f"SliceEnd_123{gather_idx}",
+                                                                        np.array([index + 1]),
+                                                                    ),
+                                                                    Constant(
+                                                                        f"SliceAxis_123{gather_idx}",
+                                                                        np.array([axis]),
+                                                                    ),
+                                                                ],
+                                                                outputs=[slice_link],
+                                                                name=f"Slice_123{gather_idx}",
+                                                            )
+                                                            graph.nodes.append(slice_node)
+                                                            gather_idx += 1
+
+                                                            for link5 in node4.outputs:
+                                                                for node5 in link5.outputs:
+                                                                    try:
+                                                                        idx = node5.inputs.index(link5)
+                                                                        node5.inputs[idx] = slice_link
+                                                                    except ValueError:
+                                                                        pass
+            elif node.op == "LayerNormalization":
+                for node1 in link1.outputs:
+                    if node1.op == "Gather":
+                        for link2 in node1.outputs:
+                            for node2 in link2.outputs:
+                                axis = node1.attrs.get("axis", 0)
+                                index = node1.inputs[1].values
+                                slice_link = Variable(
+                                    f"onnx::Slice_123{gather_idx}",
+                                    dtype=link2.dtype,
+                                    shape=[1, *link2.shape],
+                                )
+                                slice_node = Node(
+                                    op="Slice",
+                                    inputs=[
+                                        node1.inputs[0],
+                                        Constant(
+                                            f"SliceStart_123{gather_idx}",
+                                            np.array([index]),
+                                        ),
+                                        Constant(
+                                            f"SliceEnd_123{gather_idx}",
+                                            np.array([index + 1]),
+                                        ),
+                                        Constant(
+                                            f"SliceAxis_123{gather_idx}",
+                                            np.array([axis]),
+                                        ),
+                                    ],
+                                    outputs=[slice_link],
+                                    name=f"Slice_123{gather_idx}",
+                                )
+                                graph.nodes.append(slice_node)
+                                gather_idx += 1
+
+                                squeeze_link = Variable(
+                                    f"onnx::Squeeze_123{squeeze_idx}",
+                                    dtype=link2.dtype,
+                                    shape=link2.shape,
+                                )
+                                squeeze_node = Node(
+                                    op="Squeeze",
+                                    inputs=[
+                                        slice_link,
+                                        Constant(
+                                            f"SqueezeAxis_123{squeeze_idx}",
+                                            np.array([0]),
+                                        ),
+                                    ],
+                                    outputs=[squeeze_link],
+                                    name=f"Squeeze_123{squeeze_idx}",
+                                )
+                                graph.nodes.append(squeeze_node)
+                                squeeze_idx += 1
+                                try:
+                                    idx = node2.inputs.index(link2)
+                                    node2.inputs[idx] = squeeze_link
+                                except ValueError:
+                                    pass
+            elif node.op == "Reshape":
+                for node1 in link1.outputs:
+                    if node1.op == "Gather":
+                        node2s = [n for link in node1.outputs for n in link.outputs]
+                        if any(n.op == "Abs" for n in node2s):
+                            axis = node1.attrs.get("axis", 0)
+                            index = node1.inputs[1].values
+                            slice_link = Variable(
+                                f"onnx::Slice_123{gather_idx}",
+                                dtype=node1.outputs[0].dtype,
+                                shape=[1, *node1.outputs[0].shape],
+                            )
+                            slice_node = Node(
+                                op="Slice",
+                                inputs=[
+                                    node1.inputs[0],
+                                    Constant(
+                                        f"SliceStart_123{gather_idx}",
+                                        np.array([index]),
+                                    ),
+                                    Constant(
+                                        f"SliceEnd_123{gather_idx}",
+                                        np.array([index + 1]),
+                                    ),
+                                    Constant(
+                                        f"SliceAxis_123{gather_idx}",
+                                        np.array([axis]),
+                                    ),
+                                ],
+                                outputs=[slice_link],
+                                name=f"Slice_123{gather_idx}",
+                            )
+                            graph.nodes.append(slice_node)
+                            gather_idx += 1
+
+                            squeeze_link = Variable(
+                                f"onnx::Squeeze_123{squeeze_idx}",
+                                dtype=node1.outputs[0].dtype,
+                                shape=node1.outputs[0].shape,
+                            )
+                            squeeze_node = Node(
+                                op="Squeeze",
+                                inputs=[
+                                    slice_link,
+                                    Constant(
+                                        f"SqueezeAxis_123{squeeze_idx}",
+                                        np.array([0]),
+                                    ),
+                                ],
+                                outputs=[squeeze_link],
+                                name=f"Squeeze_123{squeeze_idx}",
+                            )
+                            graph.nodes.append(squeeze_node)
+                            squeeze_idx += 1
+                            for node2 in node2s:
+                                node2.inputs[0] = squeeze_link
+            elif node.op == "BatchNormalization" and node.attrs.get("training_mode") == 1:
+                node.attrs["training_mode"] = 0
+                node.outputs = node.outputs[:1]
+
+    graph.cleanup(remove_unused_node_outputs=True, recurse_subgraphs=True, recurse_functions=True)
+    graph.toposort()
+    graph.fold_constants()
+    updated = export_onnx(graph)
+    onnx_save(updated, model_path)
+
+    # for some reason, reloading the model is necessary to apply the correct shape
+    proto = onnx.load(model_path)
+    graph = import_onnx(proto)
+    for node in graph.nodes:
+        if node.op == "Slice":
+            for link in node.outputs:
+                if "Slice_123" in link.name and link.shape[0] == 3:  # noqa: PLR2004
+                    link.shape[0] = 1
+
+    graph.cleanup(remove_unused_node_outputs=True, recurse_subgraphs=True, recurse_functions=True)
+    graph.toposort()
+    graph.fold_constants()
+    updated = export_onnx(graph)
+    onnx_save(updated, model_path)
+    onnx.shape_inference.infer_shapes_path(model_path, check_type=True, strict_mode=True, data_prop=True)
+
+
+def onnx_make_inputs_fixed(input_path: str, output_path: str, input_shapes: list[tuple[int, ...]]) -> None:
+    import onnx
+    import onnxsim
+    from onnxruntime.tools.onnx_model_utils import fix_output_shapes, make_input_shape_fixed
+
+    model, success = onnxsim.simplify(input_path)
+    if not success:
+        msg = f"Failed to simplify {input_path}"
+        raise RuntimeError(msg)
+    onnx_save(model, output_path)
+    onnx.shape_inference.infer_shapes_path(output_path, check_type=True, strict_mode=True, data_prop=True)
+    model = onnx.load_model(output_path)
+    for input_node, shape in zip(model.graph.input, input_shapes, strict=False):
+        make_input_shape_fixed(model.graph, input_node.name, shape)
+    fix_output_shapes(model)
+    onnx_save(model, output_path)
+    onnx.shape_inference.infer_shapes_path(output_path, check_type=True, strict_mode=True, data_prop=True)
+
+
+def onnx_get_inputs_outputs(model_path: str) -> tuple[list[str], list[str]]:
+    import onnx
+
+    model = onnx.load(model_path)
+    inputs = [input_.name for input_ in model.graph.input]
+    outputs = [output_.name for output_ in model.graph.output]
+    return inputs, outputs
+
+
+def onnx_save(model: Any, output_path: str) -> None:
+    import onnx
+
+    try:
+        onnx.save(model, output_path)
+    except:
+        onnx.save(model, output_path, save_as_external_data=True, all_tensors_to_one_file=False, size_threshold=1_000_000)

machine-learning/export/ann/pyproject.toml

@@ -0,0 +1,56 @@
+[project]
+name = "onnx2ann"
+version = "1.107.2"
+dependencies = [
+    "onnx>=1.16.1",
+    "psutil>=6.0.0",
+    "flatbuffers>=24.3.25",
+    "ml_dtypes>=0.3.1,<1.0.0",
+    "typer-slim>=0.12.3,<1.0.0",
+    "huggingface_hub>=0.23.4,<1.0.0",
+    "onnxruntime>=1.18.1",
+    "onnxsim>=0.4.36,<1.0.0",
+    "onnx2tf>=1.24.0",
+    "onnx_graphsurgeon>=0.5.2,<1.0.0",
+    "simple_onnx_processing_tools>=1.1.32",
+    "tf_keras>=2.16.0",
+    "onnxconverter-common @ git+https://github.com/microsoft/onnxconverter-common"
+]
+requires-python = ">=3.11"
+
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"
+
+[tool.hatch.build.targets.sdist]
+only-include = ["onnx2ann"]
+
+[tool.hatch.metadata]
+allow-direct-references = true
+
+[tool.mypy]
+python_version = "3.12"
+follow_imports = "silent"
+warn_redundant_casts = true
+disallow_any_generics = true
+check_untyped_defs = true
+disallow_untyped_defs = true
+ignore_missing_imports = true
+
+[tool.pydantic-mypy]
+init_forbid_extra = true
+init_typed = true
+warn_required_dynamic_aliases = true
+warn_untyped_fields = true
+
+[tool.ruff]
+line-length = 120
+target-version = "py312"
+
+[tool.ruff.lint]
+extend-select = ["E", "F", "I"]
+extend-ignore = ["FBT001", "FBT002"]
+
+[tool.black]
+line-length = 120
+target-version = ['py312']

machine-learning/export/ann/scripts/ann.cpp

@@ -0,0 +1,281 @@
+#include <fstream>
+#include <mutex>
+#include <atomic>
+
+#include "armnn/IRuntime.hpp"
+#include "armnn/INetwork.hpp"
+#include "armnn/Types.hpp"
+#include "armnnDeserializer/IDeserializer.hpp"
+#include "armnnTfLiteParser/ITfLiteParser.hpp"
+#include "armnnOnnxParser/IOnnxParser.hpp"
+
+using namespace armnn;
+
+struct IOInfos
+{
+    std::vector<BindingPointInfo> inputInfos;
+    std::vector<BindingPointInfo> outputInfos;
+};
+
+// from https://rigtorp.se/spinlock/
+struct SpinLock
+{
+    std::atomic<bool> lock_ = {false};
+
+    void lock()
+    {
+        for (;;)
+        {
+            if (!lock_.exchange(true, std::memory_order_acquire))
+            {
+                break;
+            }
+            while (lock_.load(std::memory_order_relaxed))
+                ;
+        }
+    }
+
+    void unlock() { lock_.store(false, std::memory_order_release); }
+};
+
+class Ann
+{
+
+public:
+    int load(const char *modelPath,
+             bool fastMath,
+             bool fp16,
+             bool saveCachedNetwork,
+             const char *cachedNetworkPath)
+    {
+        INetworkPtr network = loadModel(modelPath);
+        IOptimizedNetworkPtr optNet = OptimizeNetwork(network.get(), fastMath, fp16, saveCachedNetwork, cachedNetworkPath);
+        const IOInfos infos = getIOInfos(optNet.get());
+        NetworkId netId;
+        mutex.lock();
+        Status status = runtime->LoadNetwork(netId, std::move(optNet));
+        mutex.unlock();
+        if (status != Status::Success)
+        {
+            return -1;
+        }
+        spinLock.lock();
+        ioInfos[netId] = infos;
+        mutexes.emplace(netId, std::make_unique<std::mutex>());
+        spinLock.unlock();
+        return netId;
+    }
+
+    void execute(NetworkId netId, const void **inputData, void **outputData)
+    {
+        spinLock.lock();
+        const IOInfos *infos = &ioInfos[netId];
+        auto m = mutexes[netId].get();
+        spinLock.unlock();
+        InputTensors inputTensors;
+        inputTensors.reserve(infos->inputInfos.size());
+        size_t i = 0;
+        for (const BindingPointInfo &info : infos->inputInfos)
+            inputTensors.emplace_back(info.first, ConstTensor(info.second, inputData[i++]));
+        OutputTensors outputTensors;
+        outputTensors.reserve(infos->outputInfos.size());
+        i = 0;
+        for (const BindingPointInfo &info : infos->outputInfos)
+            outputTensors.emplace_back(info.first, Tensor(info.second, outputData[i++]));
+        m->lock();
+        runtime->EnqueueWorkload(netId, inputTensors, outputTensors);
+        m->unlock();
+    }
+
+    void unload(NetworkId netId)
+    {
+        mutex.lock();
+        runtime->UnloadNetwork(netId);
+        mutex.unlock();
+    }
+
+    int tensors(NetworkId netId, bool isInput = false)
+    {
+        spinLock.lock();
+        const IOInfos *infos = &ioInfos[netId];
+        spinLock.unlock();
+        return (int)(isInput ? infos->inputInfos.size() : infos->outputInfos.size());
+    }
+
+    unsigned long shape(NetworkId netId, bool isInput = false, int index = 0)
+    {
+        spinLock.lock();
+        const IOInfos *infos = &ioInfos[netId];
+        spinLock.unlock();
+        const TensorShape shape = (isInput ? infos->inputInfos : infos->outputInfos)[index].second.GetShape();
+        unsigned long s = 0;
+        for (unsigned int d = 0; d < shape.GetNumDimensions(); d++)
+            s |= ((unsigned long)shape[d]) << (d * 16); // stores up to 4 16-bit values in a 64-bit value
+        return s;
+    }
+
+    Ann(int tuningLevel, const char *tuningFile)
+    {
+        IRuntime::CreationOptions runtimeOptions;
+        BackendOptions backendOptions{"GpuAcc",
+                                      {
+                                          {"TuningLevel", tuningLevel},
+                                          {"MemoryOptimizerStrategy", "ConstantMemoryStrategy"}, // SingleAxisPriorityList or ConstantMemoryStrategy
+                                      }};
+        if (tuningFile)
+            backendOptions.AddOption({"TuningFile", tuningFile});
+        runtimeOptions.m_BackendOptions.emplace_back(backendOptions);
+        runtime = IRuntime::CreateRaw(runtimeOptions);
+    };
+    ~Ann()
+    {
+        IRuntime::Destroy(runtime);
+    };
+
+private:
+    INetworkPtr loadModel(const char *modelPath)
+    {
+        const auto path = std::string(modelPath);
+        if (path.rfind(".tflite") == path.length() - 7) // endsWith()
+        {
+            auto parser = armnnTfLiteParser::ITfLiteParser::CreateRaw();
+            return parser->CreateNetworkFromBinaryFile(modelPath);
+        }
+        else if (path.rfind(".onnx") == path.length() - 5) // endsWith()
+        {
+            auto parser = armnnOnnxParser::IOnnxParser::CreateRaw();
+            return parser->CreateNetworkFromBinaryFile(modelPath);
+        }
+        else
+        {
+            std::ifstream ifs(path, std::ifstream::in | std::ifstream::binary);
+            auto parser = armnnDeserializer::IDeserializer::CreateRaw();
+            return parser->CreateNetworkFromBinary(ifs);
+        }
+    }
+
+    static BindingPointInfo getInputTensorInfo(LayerBindingId inputBindingId, TensorInfo info)
+    {
+        const auto newInfo = TensorInfo{info.GetShape(), info.GetDataType(),
+                                        info.GetQuantizationScale(),
+                                        info.GetQuantizationOffset(),
+                                        true};
+        return {inputBindingId, newInfo};
+    }
+
+    IOptimizedNetworkPtr OptimizeNetwork(INetwork *network, bool fastMath, bool fp16, bool saveCachedNetwork, const char *cachedNetworkPath)
+    {
+        const bool allowExpandedDims = false;
+        const ShapeInferenceMethod shapeInferenceMethod = ShapeInferenceMethod::ValidateOnly;
+
+        OptimizerOptionsOpaque options;
+        options.SetReduceFp32ToFp16(fp16);
+        options.SetShapeInferenceMethod(shapeInferenceMethod);
+        options.SetAllowExpandedDims(allowExpandedDims);
+
+        BackendOptions gpuAcc("GpuAcc", {{"FastMathEnabled", fastMath}});
+        if (cachedNetworkPath)
+        {
+            gpuAcc.AddOption({"SaveCachedNetwork", saveCachedNetwork});
+            gpuAcc.AddOption({"CachedNetworkFilePath", cachedNetworkPath});
+        }
+        options.AddModelOption(gpuAcc);
+
+        // No point in using ARMNN for CPU, use ONNX (quantized) instead.
+        // BackendOptions cpuAcc("CpuAcc",
+        //                       {
+        //                           {"FastMathEnabled", fastMath},
+        //                           {"NumberOfThreads", 0},
+        //                       });
+        // options.AddModelOption(cpuAcc);
+
+        BackendOptions allowExDimOpt("AllowExpandedDims",
+                                     {{"AllowExpandedDims", allowExpandedDims}});
+        options.AddModelOption(allowExDimOpt);
+        BackendOptions shapeInferOpt("ShapeInferenceMethod",
+                                     {{"InferAndValidate", shapeInferenceMethod == ShapeInferenceMethod::InferAndValidate}});
+        options.AddModelOption(shapeInferOpt);
+
+        std::vector<BackendId> backends = {
+            BackendId("GpuAcc"),
+            // BackendId("CpuAcc"),
+            // BackendId("CpuRef"),
+        };
+        return Optimize(*network, backends, runtime->GetDeviceSpec(), options);
+    }
+
+    IOInfos getIOInfos(IOptimizedNetwork *optNet)
+    {
+        struct InfoStrategy : IStrategy
+        {
+            void ExecuteStrategy(const IConnectableLayer *layer,
+                                 const BaseDescriptor &descriptor,
+                                 const std::vector<ConstTensor> &constants,
+                                 const char *name,
+                                 const LayerBindingId id = 0) override
+            {
+                IgnoreUnused(descriptor, constants, id);
+                const LayerType lt = layer->GetType();
+                if (lt == LayerType::Input)
+                    ioInfos.inputInfos.push_back(getInputTensorInfo(id, layer->GetOutputSlot(0).GetTensorInfo()));
+                else if (lt == LayerType::Output)
+                    ioInfos.outputInfos.push_back({id, layer->GetInputSlot(0).GetTensorInfo()});
+            }
+            IOInfos ioInfos;
+        };
+
+        InfoStrategy infoStrategy;
+        optNet->ExecuteStrategy(infoStrategy);
+        return infoStrategy.ioInfos;
+    }
+
+    IRuntime *runtime;
+    std::map<NetworkId, IOInfos> ioInfos;
+    std::map<NetworkId, std::unique_ptr<std::mutex>> mutexes; // mutex per network to not execute the same the same network concurrently
+    std::mutex mutex; // global mutex for load/unload calls to the runtime
+    SpinLock spinLock; // fast spin lock to guard access to the ioInfos and mutexes maps
+};
+
+extern "C" void *init(int logLevel, int tuningLevel, const char *tuningFile)
+{
+    LogSeverity level = static_cast<LogSeverity>(logLevel);
+    ConfigureLogging(true, true, level);
+
+    Ann *ann = new Ann(tuningLevel, tuningFile);
+    return ann;
+}
+
+extern "C" void destroy(void *ann)
+{
+    delete ((Ann *)ann);
+}
+
+extern "C" int load(void *ann,
+                    const char *path,
+                    bool fastMath,
+                    bool fp16,
+                    bool saveCachedNetwork,
+                    const char *cachedNetworkPath)
+{
+    return ((Ann *)ann)->load(path, fastMath, fp16, saveCachedNetwork, cachedNetworkPath);
+}
+
+extern "C" void unload(void *ann, NetworkId netId)
+{
+    ((Ann *)ann)->unload(netId);
+}
+
+extern "C" void execute(void *ann, NetworkId netId, const void **inputData, void **outputData)
+{
+    ((Ann *)ann)->execute(netId, inputData, outputData);
+}
+
+extern "C" unsigned long shape(void *ann, NetworkId netId, bool isInput, int index)
+{
+    return ((Ann *)ann)->shape(netId, isInput, index);
+}
+
+extern "C" int tensors(void *ann, NetworkId netId, bool isInput)
+{
+    return ((Ann *)ann)->tensors(netId, isInput);
+}

machine-learning/export/ann/scripts/build-converter.sh

@@ -0,0 +1,4 @@
+#!/usr/bin/env sh
+
+cd armnn-23.11/ || exit
+g++ -o ../armnnconverter -fPIC -O1 -DARMNN_ONNX_PARSER -DARMNN_SERIALIZER -DARMNN_TF_LITE_PARSER -fuse-ld=gold -std=c++17 -Iinclude -Isrc/armnnUtils -Ithird-party -larmnn -larmnnDeserializer -larmnnTfLiteParser -larmnnOnnxParser -larmnnSerializer -L../armnn src/armnnConverter/ArmnnConverter.cpp

machine-learning/export/ann/scripts/build.sh

@@ -0,0 +1,3 @@
+#!/usr/bin/env sh
+
+g++ -shared -O3 -fPIC -o libann.so -fuse-ld=gold -std=c++17 -I"$ARMNN_PATH"/include -larmnn -larmnnDeserializer -larmnnTfLiteParser -larmnnOnnxParser -L"$ARMNN_PATH" ann.cpp

machine-learning/export/ort/Dockerfile

@@ -1,4 +1,4 @@
-FROM mambaorg/micromamba:bookworm-slim@sha256:954e438daab0ad0835430ea84acb27dd47d1ea35a7120c3c9dd9d1a5578f4b13 AS builder
+FROM mambaorg/micromamba:bookworm-slim@sha256:333f7598ff2c2400fb10bfe057709c68b7daab5d847143af85abcf224a07271a as builder
 
 ENV TRANSFORMERS_CACHE=/cache \
   PYTHONDONTWRITEBYTECODE=1 \

machine-learning/export/ort/env.yaml

@@ -2,7 +2,7 @@ name: base
 channels:
   - conda-forge
   - nvidia
-  - pytorch
+  - pytorch-nightly
 platforms:
   - linux-64
 dependencies:
@@ -13,7 +13,7 @@ dependencies:
   - orjson==3.*
   - pip
   - python==3.11.*
-  - pytorch>=2.3
+  - pytorch
   - rich==13.*
   - safetensors==0.*
   - setuptools==68.*
@@ -21,5 +21,5 @@ dependencies:
   - transformers==4.*
   - pip:
     - multilingual-clip
-    - onnxsim
+    - onnx-simplifier
 category: main

machine-learning/export/ort/models/mclip.py

@@ -1,4 +1,3 @@
-import os
 import tempfile
 import warnings
 from pathlib import Path
@@ -9,6 +8,7 @@ from transformers import AutoTokenizer
 
 from .openclip import OpenCLIPModelConfig
 from .openclip import to_onnx as openclip_to_onnx
+from .optimize import optimize
 from .util import get_model_path
 
 _MCLIP_TO_OPENCLIP = {
@@ -19,39 +19,44 @@ _MCLIP_TO_OPENCLIP = {
 }
 
 
-def to_onnx(
-    model_name: str,
-    output_dir_visual: Path | str,
-    output_dir_textual: Path | str,
-) -> tuple[Path, Path]:
-    textual_path = get_model_path(output_dir_textual)
-    with tempfile.TemporaryDirectory() as tmpdir:
-        model = MultilingualCLIP.from_pretrained(model_name, cache_dir=os.environ.get("CACHE_DIR", tmpdir))
-        AutoTokenizer.from_pretrained(model_name).save_pretrained(output_dir_textual)
-
-        model.eval()
-        for param in model.parameters():
-            param.requires_grad_(False)
-
-        export_text_encoder(model, textual_path)
-        visual_path, _ = openclip_to_onnx(_MCLIP_TO_OPENCLIP[model_name], output_dir_visual)
-        assert visual_path is not None, "Visual model export failed"
-    return visual_path, textual_path
-
-
-def export_text_encoder(model: MultilingualCLIP, output_path: Path | str) -> None:
-    output_path = Path(output_path)
-
-    def forward(self: MultilingualCLIP, input_ids: torch.Tensor, attention_mask: torch.Tensor) -> torch.Tensor:
+def forward(self: MultilingualCLIP, input_ids: torch.Tensor, attention_mask: torch.Tensor) -> torch.Tensor:
         embs = self.transformer(input_ids, attention_mask)[0]
         embs = (embs * attention_mask.unsqueeze(2)).sum(dim=1) / attention_mask.sum(dim=1)[:, None]
         embs = self.LinearTransformation(embs)
         return torch.nn.functional.normalize(embs, dim=-1)
 
-    # unfortunately need to monkeypatch for tracing to work here
-    # otherwise it hits the 2GiB protobuf serialization limit
-    MultilingualCLIP.forward = forward
+# unfortunately need to monkeypatch for tracing to work here
+# otherwise it hits the 2GiB protobuf serialization limit
+MultilingualCLIP.forward = forward
 
+
+def to_torchscript(model_name: str) -> torch.jit.ScriptModule:
+    with tempfile.TemporaryDirectory() as tmpdir:
+        model = MultilingualCLIP.from_pretrained(model_name, cache_dir=tmpdir)
+
+        model.eval()
+        for param in model.parameters():
+            param.requires_grad_(False)
+        
+        return model
+
+
+def to_onnx(
+    model_name: str,
+    output_dir_visual: Path | str,
+    output_dir_textual: Path | str,
+) -> None:
+    textual_path = get_model_path(output_dir_textual)
+    model = to_torchscript(model_name)
+    AutoTokenizer.from_pretrained(model_name).save_pretrained(output_dir_textual)
+
+    _text_encoder_to_onnx(model, textual_path)
+    openclip_to_onnx(_MCLIP_TO_OPENCLIP[model_name], output_dir_visual)
+    optimize(textual_path)
+
+
+def _text_encoder_to_onnx(model: MultilingualCLIP, output_path: Path | str) -> None:
+    output_path = Path(output_path)
     args = (torch.ones(1, 77, dtype=torch.int32), torch.ones(1, 77, dtype=torch.int32))
     with warnings.catch_warnings():
         warnings.simplefilter("ignore", UserWarning)
@@ -60,10 +65,10 @@ def export_text_encoder(model: MultilingualCLIP, output_path: Path | str) -> Non
             args,
             output_path.as_posix(),
             input_names=["input_ids", "attention_mask"],
-            output_names=["embedding"],
+            output_names=["text_embedding"],
             opset_version=17,
-            # dynamic_axes={
-            #     "input_ids": {0: "batch_size", 1: "sequence_length"},
-            #     "attention_mask": {0: "batch_size", 1: "sequence_length"},
-            # },
+            dynamic_axes={
+                "input_ids": {0: "batch_size", 1: "sequence_length"},
+                "attention_mask": {0: "batch_size", 1: "sequence_length"},
+            },
         )