feat(diffusers): add builds for nvidia-l4t

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

.github/workflows/backend.yml

@@ -90,7 +90,7 @@ jobs:
           - build-type: 'l4t'
             cuda-major-version: "12"
             cuda-minor-version: "0"
-            platforms: 'linux/arm64'
+            platforms: 'linux/amd64'
             tag-latest: 'auto'
             tag-suffix: '-gpu-nvidia-l4t-diffusers'
             runs-on: 'ubuntu-24.04-arm'

.github/workflows/test.yml

@@ -23,20 +23,6 @@ jobs:
       matrix:
         go-version: ['1.21.x']
     steps:
-      - name: Free Disk Space (Ubuntu)
-        uses: jlumbroso/free-disk-space@main
-        with:
-          # this might remove tools that are actually needed,
-          # if set to "true" but frees about 6 GB
-          tool-cache: true
-          # all of these default to true, but feel free to set to
-          # "false" if necessary for your workflow
-          android: true
-          dotnet: true
-          haskell: true
-          large-packages: true
-          docker-images: true
-          swap-storage: true
       - name: Release space from worker
         run: |
           echo "Listing top largest packages"

backend/cpp/llama-cpp/Makefile

@@ -1,5 +1,5 @@
 
-LLAMA_VERSION?=be48528b068111304e4a0bb82c028558b5705f05
+LLAMA_VERSION?=a0552c8beef74e843bb085c8ef0c63f9ed7a2b27
 LLAMA_REPO?=https://github.com/ggerganov/llama.cpp
 
 CMAKE_ARGS?=

backend/go/whisper/Makefile

@@ -6,7 +6,7 @@ CMAKE_ARGS?=
 
 # whisper.cpp version
 WHISPER_REPO?=https://github.com/ggml-org/whisper.cpp
-WHISPER_CPP_VERSION?=b02242d0adb5c6c4896d59ac86d9ec9fe0d0fe33
+WHISPER_CPP_VERSION?=4245c77b654cd384ad9f53a4a302be716b3e5861
 
 export WHISPER_CMAKE_ARGS?=-DBUILD_SHARED_LIBS=OFF
 export WHISPER_DIR=$(abspath ./sources/whisper.cpp)

backend/python/diffusers/requirements-l4t.txt

@@ -6,5 +6,4 @@ accelerate
 compel
 peft
 optimum-quanto
-numpy<2
-sentencepiece
+numpy<2

core/application/startup.go

@@ -59,10 +59,8 @@ func New(opts ...config.AppOption) (*Application, error) {
 		log.Error().Err(err).Msg("error installing models")
 	}
 
-	for _, backend := range options.ExternalBackends {
-		if err := coreStartup.InstallExternalBackends(options.BackendGalleries, options.BackendsPath, nil, backend, "", ""); err != nil {
-			log.Error().Err(err).Msg("error installing external backend")
-		}
+	if err := coreStartup.InstallExternalBackends(options.BackendGalleries, options.BackendsPath, nil, options.ExternalBackends...); err != nil {
+		log.Error().Err(err).Msg("error installing external backends")
 	}
 
 	configLoaderOpts := options.ToConfigLoaderOptions()

core/cli/backends.go

@@ -23,9 +23,7 @@ type BackendsList struct {
 }
 
 type BackendsInstall struct {
-	BackendArgs string `arg:"" optional:"" name:"backend" help:"Backend configuration URL to load"`
-	Name        string `arg:"" optional:"" name:"name" help:"Name of the backend"`
-	Alias       string `arg:"" optional:"" name:"alias" help:"Alias of the backend"`
+	BackendArgs []string `arg:"" optional:"" name:"backends" help:"Backend configuration URLs to load"`
 
 	BackendsCMDFlags `embed:""`
 }
@@ -68,25 +66,27 @@ func (bi *BackendsInstall) Run(ctx *cliContext.Context) error {
 		log.Error().Err(err).Msg("unable to load galleries")
 	}
 
-	progressBar := progressbar.NewOptions(
-		1000,
-		progressbar.OptionSetDescription(fmt.Sprintf("downloading backend %s", bi.BackendArgs)),
-		progressbar.OptionShowBytes(false),
-		progressbar.OptionClearOnFinish(),
-	)
-	progressCallback := func(fileName string, current string, total string, percentage float64) {
-		v := int(percentage * 10)
-		err := progressBar.Set(v)
+	for _, backendName := range bi.BackendArgs {
+
+		progressBar := progressbar.NewOptions(
+			1000,
+			progressbar.OptionSetDescription(fmt.Sprintf("downloading backend %s", backendName)),
+			progressbar.OptionShowBytes(false),
+			progressbar.OptionClearOnFinish(),
+		)
+		progressCallback := func(fileName string, current string, total string, percentage float64) {
+			v := int(percentage * 10)
+			err := progressBar.Set(v)
+			if err != nil {
+				log.Error().Err(err).Str("filename", fileName).Int("value", v).Msg("error while updating progress bar")
+			}
+		}
+
+		err := startup.InstallExternalBackends(galleries, bi.BackendsPath, progressCallback, backendName)
 		if err != nil {
-			log.Error().Err(err).Str("filename", fileName).Int("value", v).Msg("error while updating progress bar")
+			return err
 		}
 	}
-
-	err := startup.InstallExternalBackends(galleries, bi.BackendsPath, progressCallback, bi.BackendArgs, bi.Name, bi.Alias)
-	if err != nil {
-		return err
-	}
-
 	return nil
 }
 

core/startup/backend_preload.go

@@ -1,6 +1,7 @@
 package startup
 
 import (
+	"errors"
 	"fmt"
 	"path/filepath"
 	"strings"
@@ -12,68 +13,49 @@ import (
 	"github.com/rs/zerolog/log"
 )
 
-func InstallExternalBackends(galleries []config.Gallery, backendPath string, downloadStatus func(string, string, string, float64), backend, name, alias string) error {
+func InstallExternalBackends(galleries []config.Gallery, backendPath string, downloadStatus func(string, string, string, float64), backends ...string) error {
+	var errs error
 	systemState, err := system.GetSystemState()
 	if err != nil {
 		return fmt.Errorf("failed to get system state: %w", err)
 	}
-	uri := downloader.URI(backend)
-	switch {
-	case uri.LooksLikeDir():
-		if name == "" { // infer it from the path
-			name = filepath.Base(backend)
-		}
-		log.Info().Str("backend", backend).Str("name", name).Msg("Installing backend from path")
-		if err := gallery.InstallBackend(backendPath, &gallery.GalleryBackend{
-			Metadata: gallery.Metadata{
-				Name: name,
-			},
-			Alias: alias,
-			URI:   backend,
-		}, downloadStatus); err != nil {
-			return fmt.Errorf("error installing backend %s: %w", backend, err)
-		}
-	case uri.LooksLikeOCI() && !uri.LooksLikeOCIFile():
-		if name == "" {
-			return fmt.Errorf("specifying a name is required for OCI images")
-		}
-		log.Info().Str("backend", backend).Str("name", name).Msg("Installing backend from OCI image")
-		if err := gallery.InstallBackend(backendPath, &gallery.GalleryBackend{
-			Metadata: gallery.Metadata{
-				Name: name,
-			},
-			Alias: alias,
-			URI:   backend,
-		}, downloadStatus); err != nil {
-			return fmt.Errorf("error installing backend %s: %w", backend, err)
-		}
-	case uri.LooksLikeOCIFile():
-		name, err := uri.FilenameFromUrl()
-		if err != nil {
-			return fmt.Errorf("failed to get filename from URL: %w", err)
-		}
-		// strip extension if any
-		name = strings.TrimSuffix(name, filepath.Ext(name))
+	for _, backend := range backends {
+		uri := downloader.URI(backend)
+		switch {
+		case uri.LooksLikeDir():
+			name := filepath.Base(backend)
+			log.Info().Str("backend", backend).Str("name", name).Msg("Installing backend from path")
+			if err := gallery.InstallBackend(backendPath, &gallery.GalleryBackend{
+				Metadata: gallery.Metadata{
+					Name: name,
+				},
+				URI: backend,
+			}, downloadStatus); err != nil {
+				errs = errors.Join(err, fmt.Errorf("error installing backend %s", backend))
+			}
+		case uri.LooksLikeOCI():
+			name, err := uri.FilenameFromUrl()
+			if err != nil {
+				return fmt.Errorf("failed to get filename from URL: %w", err)
+			}
+			// strip extension if any
+			name = strings.TrimSuffix(name, filepath.Ext(name))
 
-		log.Info().Str("backend", backend).Str("name", name).Msg("Installing backend from OCI image")
-		if err := gallery.InstallBackend(backendPath, &gallery.GalleryBackend{
-			Metadata: gallery.Metadata{
-				Name: name,
-			},
-			Alias: alias,
-			URI:   backend,
-		}, downloadStatus); err != nil {
-			return fmt.Errorf("error installing backend %s: %w", backend, err)
-		}
-	default:
-		if name != "" || alias != "" {
-			return fmt.Errorf("specifying a name or alias is not supported for this backend")
-		}
-		err := gallery.InstallBackendFromGallery(galleries, systemState, backend, backendPath, downloadStatus, true)
-		if err != nil {
-			return fmt.Errorf("error installing backend %s: %w", backend, err)
+			log.Info().Str("backend", backend).Str("name", name).Msg("Installing backend from OCI image")
+			if err := gallery.InstallBackend(backendPath, &gallery.GalleryBackend{
+				Metadata: gallery.Metadata{
+					Name: name,
+				},
+				URI: backend,
+			}, downloadStatus); err != nil {
+				errs = errors.Join(err, fmt.Errorf("error installing backend %s", backend))
+			}
+		default:
+			err := gallery.InstallBackendFromGallery(galleries, systemState, backend, backendPath, downloadStatus, true)
+			if err != nil {
+				errs = errors.Join(err, fmt.Errorf("error installing backend %s", backend))
+			}
 		}
 	}
-
-	return nil
+	return errs
 }

gallery/index.yaml

@@ -105,8 +105,8 @@
       model: gpt-oss-20b-mxfp4.gguf
   files:
     - filename: gpt-oss-20b-mxfp4.gguf
+      sha256: 52f57ab7d3df3ba9173827c1c6832e73375553a846f3e32b49f1ae2daad688d4
       uri: huggingface://ggml-org/gpt-oss-20b-GGUF/gpt-oss-20b-mxfp4.gguf
-      sha256: be37a636aca0fc1aae0d32325f82f6b4d21495f06823b5fbc1898ae0303e9935
 - !!merge <<: *gptoss
   name: "gpt-oss-120b"
   url: "github:mudler/LocalAI/gallery/harmony.yaml@master"
@@ -119,14 +119,14 @@
       model: gpt-oss-120b-mxfp4-00001-of-00003.gguf
   files:
     - filename: gpt-oss-120b-mxfp4-00001-of-00003.gguf
+      sha256: 40b630223b9fc43820fa0aae5d0ab61020f5858d1719642357753dca9e7df29f
       uri: huggingface://ggml-org/gpt-oss-120b-GGUF/gpt-oss-120b-mxfp4-00001-of-00003.gguf
-      sha256: e2865eb6c1df7b2ffbebf305cd5d9074d5ccc0fe3b862f98d343a46dad1606f9
     - filename: gpt-oss-120b-mxfp4-00002-of-00003.gguf
+      sha256: fbdb8cdec70edb82c53bfc69cc0f54a34759a23d317fa0771a63be6571907b38
       uri: huggingface://ggml-org/gpt-oss-120b-GGUF/gpt-oss-120b-mxfp4-00002-of-00003.gguf
-      sha256: 346492f65891fb27cac5c74a8c07626cbfeb4211cd391ec4de37dbbe3109a93b
     - filename: gpt-oss-120b-mxfp4-00003-of-00003.gguf
+      sha256: b326bfd8ac696c4b9a14e9e84d5529b2bb86847aea0e65443cbf075accba8b71
       uri: huggingface://ggml-org/gpt-oss-120b-GGUF/gpt-oss-120b-mxfp4-00003-of-00003.gguf
-      sha256: 66dca81040933f5a49177e82c479c51319cefb83bd22dad9f06dad45e25f1463
 - !!merge <<: *gptoss
   name: "openai_gpt-oss-20b-neo"
   icon: https://huggingface.co/DavidAU/Openai_gpt-oss-20b-NEO-GGUF/resolve/main/matrix1.gif
@@ -312,7 +312,7 @@
     - https://huggingface.co/Dream-org/Dream-v0-Instruct-7B
     - https://huggingface.co/bartowski/Dream-org_Dream-v0-Instruct-7B-GGUF
   description: |
-    This is the instruct model of Dream 7B, which is an open diffusion large language model with top-tier performance.
+      This is the instruct model of Dream 7B, which is an open diffusion large language model with top-tier performance.
   overrides:
     parameters:
       model: Dream-org_Dream-v0-Instruct-7B-Q4_K_M.gguf
@@ -14446,21 +14446,6 @@
     - filename: entfane_math-genius-7B-Q4_K_M.gguf
       sha256: cd3a3c898a2dfb03d17a66db81b743f2d66981e0ceb92e8669a4af61217feed7
       uri: huggingface://bartowski/entfane_math-genius-7B-GGUF/entfane_math-genius-7B-Q4_K_M.gguf
-- !!merge <<: *mistral03
-  name: "impish_nemo_12b"
-  url: "github:mudler/LocalAI/gallery/chatml.yaml@master"
-  icon: https://huggingface.co/SicariusSicariiStuff/Impish_Nemo_12B/resolve/main/Images/Impish_Nemo_12B.png
-  urls:
-    - https://huggingface.co/SicariusSicariiStuff/Impish_Nemo_12B
-    - https://huggingface.co/SicariusSicariiStuff/Impish_Nemo_12B_GGUF
-  description: "August 2025, Impish_Nemo_12B — my best model yet. And unlike a typical Nemo, this one can take in much higher temperatures (works well with 1+). Oh, and regarding following the character card: It somehow gotten even better, to the point of it being straight up uncanny \U0001F643 (I had to check twice that this model was loaded, and not some 70B!)\n\nI feel like this model could easily replace models much larger than itself for adventure or roleplay, for assistant tasks, obviously not, but the creativity here? Off the charts. Characters have never felt so alive and in the moment before — they’ll use insinuation, manipulation, and, if needed (or provoked) — force. They feel so very present.\n\nThat look on Neo’s face when he opened his eyes and said, “I know Kung Fu”? Well, Impish_Nemo_12B had pretty much the same moment — and it now knows more than just Kung Fu, much, much more. It wasn’t easy, and it’s a niche within a niche, but as promised almost half a year ago — it is now done.\n\nImpish_Nemo_12B is smart, sassy, creative, and got a lot of unhingedness too — these are baked-in deep into every interaction. It took the innate Mistral's relative freedom, and turned it up to 11. It very well maybe too much for many, but after testing and interacting with so many models, I find this 'edge' of sorts, rather fun and refreshing.\n\nAnyway, the dataset used is absolutely massive, tons of new types of data and new domains of knowledge (Morrowind fandom, fighting, etc...). The whole dataset is a very well-balanced mix, and resulted in a model with extremely strong common sense for a 12B. Regarding response length — there's almost no response-length bias here, this one is very much dynamic and will easily adjust reply length based on 1–3 examples of provided dialogue.\n\nOh, and the model comes with 3 new Character Cards, 2 Roleplay and 1 Adventure!\n"
-  overrides:
-    parameters:
-      model: Impish_Nemo_12B-Q6_K.gguf
-  files:
-    - filename: Impish_Nemo_12B-Q6_K.gguf
-      sha256: e0ce3adbed2718e144f477721c2ad68b6e3cccd95fc27dbe8f0135be76c99c72
-      uri: huggingface://SicariusSicariiStuff/Impish_Nemo_12B_GGUF/Impish_Nemo_12B-Q6_K.gguf
 - &mudler
   url: "github:mudler/LocalAI/gallery/mudler.yaml@master" ### START mudler's LocalAI specific-models
   name: "LocalAI-llama3-8b-function-call-v0.2"
@@ -19475,14 +19460,14 @@
   url: "github:mudler/LocalAI/gallery/flux-ggml.yaml@master"
   icon: https://huggingface.co/black-forest-labs/FLUX.1-Kontext-dev/media/main/teaser.png
   description: |
-    FLUX.1 Kontext [dev] is a 12 billion parameter rectified flow transformer capable of editing images based on text instructions. For more information, please read our blog post and our technical report. You can find information about the [pro] version in here.
-    Key Features
-        Change existing images based on an edit instruction.
-        Have character, style and object reference without any finetuning.
-        Robust consistency allows users to refine an image through multiple successive edits with minimal visual drift.
-        Trained using guidance distillation, making FLUX.1 Kontext [dev] more efficient.
-        Open weights to drive new scientific research, and empower artists to develop innovative workflows.
-        Generated outputs can be used for personal, scientific, and commercial purposes, as described in the FLUX.1 [dev] Non-Commercial License.
+      FLUX.1 Kontext [dev] is a 12 billion parameter rectified flow transformer capable of editing images based on text instructions. For more information, please read our blog post and our technical report. You can find information about the [pro] version in here.
+      Key Features
+          Change existing images based on an edit instruction.
+          Have character, style and object reference without any finetuning.
+          Robust consistency allows users to refine an image through multiple successive edits with minimal visual drift.
+          Trained using guidance distillation, making FLUX.1 Kontext [dev] more efficient.
+          Open weights to drive new scientific research, and empower artists to develop innovative workflows.
+          Generated outputs can be used for personal, scientific, and commercial purposes, as described in the FLUX.1 [dev] Non-Commercial License.
   urls:
     - https://huggingface.co/black-forest-labs/FLUX.1-Kontext-dev
     - https://huggingface.co/QuantStack/FLUX.1-Kontext-dev-GGUF

pkg/downloader/uri.go

@@ -98,19 +98,19 @@ func (uri URI) DownloadWithAuthorizationAndCallback(basePath string, authorizati
 }
 
 func (u URI) FilenameFromUrl() (string, error) {
-	if f := filenameFromUrl(string(u)); f != "" {
-		return f, nil
+	f, err := filenameFromUrl(string(u))
+	if err != nil || f == "" {
+		f = utils.MD5(string(u))
+		if strings.HasSuffix(string(u), ".yaml") || strings.HasSuffix(string(u), ".yml") {
+			f = f + ".yaml"
+		}
+		err = nil
 	}
 
-	f := utils.MD5(string(u))
-	if strings.HasSuffix(string(u), ".yaml") || strings.HasSuffix(string(u), ".yml") {
-		f = f + ".yaml"
-	}
-
-	return f, nil
+	return f, err
 }
 
-func filenameFromUrl(urlstr string) string {
+func filenameFromUrl(urlstr string) (string, error) {
 	// strip anything after @
 	if strings.Contains(urlstr, "@") {
 		urlstr = strings.Split(urlstr, "@")[0]
@@ -118,13 +118,13 @@ func filenameFromUrl(urlstr string) string {
 
 	u, err := url.Parse(urlstr)
 	if err != nil {
-		return ""
+		return "", fmt.Errorf("error due to parsing url: %w", err)
 	}
 	x, err := url.QueryUnescape(u.EscapedPath())
 	if err != nil {
-		return ""
+		return "", fmt.Errorf("error due to escaping: %w", err)
 	}
-	return filepath.Base(x)
+	return filepath.Base(x), nil
 }
 
 func (u URI) LooksLikeURL() bool {
@@ -158,10 +158,6 @@ func (s URI) LooksLikeOCI() bool {
 		strings.HasPrefix(string(s), "docker.io")
 }
 
-func (s URI) LooksLikeOCIFile() bool {
-	return strings.HasPrefix(string(s), OCIFilePrefix)
-}
-
 func (s URI) ResolveURL() string {
 	switch {
 	case strings.HasPrefix(string(s), GithubURI2):