update changelog and goMLX to stable version

Author: RJKeevil

CHANGELOG.md

@@ -4,6 +4,26 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](http://keepachangelog.com/)
 and this project adheres to [Semantic Versioning](http://semver.org/).
 
+## [0.6.0] 🌲❄️🎄🎁 - 2025-12-18
+
+### Changed
+
+- Integrated [ONNX Runtime GenAI](https://github.com/microsoft/onnxruntime-genai) backend for significantly faster generative inference and broad model support.
+- Added ObjectDetection pipeline
+- Added WithTPU option to NewXLASession
+- FeatureExtractionPipeline now supports image inputs, enabling vision models like CLIP
+- Updated Onnx Runtime to 1.23.2, and GoMLX to 0.26.0
+
+### Breaking changes
+
+- ORT Gen AI has strong requirements on the name of the base ORT library. It should not be renamed from the release zip (e.g. libonnxruntime.so)
+- WithOnnxLibraryPath should now be the folder contining the ORT library. The library name is now inferred from the current operating system.
+- XLA now uses go-xla to manage PJRT dependencies, see our [Dockerfile](./Dockerfile) for details 
+
+### Fixed
+
+- Model loading path could potentially duplicate paths (thanks @ajroetker)
+
 ## [0.5.10] - 2025-12-08
 
 ### Breaking changes
@@ -17,7 +37,7 @@ and this project adheres to [Semantic Versioning](http://semver.org/).
 - breaking: GetStatistics now returns a Statistics struct for the pipelines rather than a list of strings
 - breaking: pipelineBackends has been renamed to backends
 
-### Changes
+### Changed
 
 - update of onnxruntime_go, goMLX, gopjrt
 

Dockerfile

@@ -3,7 +3,7 @@
 ARG GO_VERSION=1.25.5
 ARG ONNXRUNTIME_VERSION=1.23.2
 ARG ONNXRUNTIME_GENAI_VERSION=0.11.4
-ARG GOPJRT_VERSION=0.83.1
+ARG GOPJRT_VERSION=0.83.3
 ARG BUILD_PLATFORM=linux/amd64
 
 #--- runtime layer with all hugot dependencies for cpu 
@@ -23,6 +23,9 @@ COPY ./scripts/download-onnxruntime-genai.sh /download-onnxruntime-genai.sh
 RUN --mount=src=./go.mod,dst=/go.mod \
     dnf --allowerasing -y install gcc jq bash tar xz gzip glibc-static libstdc++ wget zip git dirmngr sudo which && \
     ln -s /usr/lib64/libstdc++.so.6 /usr/lib64/libstdc++.so && \
+    curl -LO https://download.fedoraproject.org/pub/fedora/linux/releases/43/Everything/x86_64/os/Packages/g/glibc-2.42-4.fc43.x86_64.rpm && \
+    rpm -Uvh --nodeps --force glibc-2.42-4.fc43.x86_64.rpm && \
+    rm glibc-2.42-4.fc43.x86_64.rpm && \
     dnf clean all && \
     # go
     curl -LO https://golang.org/dl/go${GO_VERSION}.linux-amd64.tar.gz && \
@@ -42,7 +45,7 @@ RUN --mount=src=./go.mod,dst=/go.mod \
     sed -i 's/\r//g' /download-onnxruntime-genai.sh && chmod +x /download-onnxruntime-genai.sh && \
     /download-onnxruntime-genai.sh ${ONNXRUNTIME_GENAI_VERSION} && \
     # XLA/goMLX
-    GOPROXY=direct go run github.com/gomlx/go-xla/cmd/pjrt_installer@latest -plugin=amazonlinux -version=v${GOPJRT_VERSION} -path=/usr/local/lib/go-xla && \
+    GOPROXY=direct go run github.com/gomlx/go-xla/cmd/pjrt_installer@latest -plugin=linux -version=v${GOPJRT_VERSION} -path=/usr/local/lib/go-xla && \
     # NON-PRIVILEGED USER
     # create non-privileged testuser with id: 1000
     useradd -u 1000 -m testuser && usermod -a -G wheel testuser && \

README.md

@@ -3,7 +3,7 @@
 ARG GO_VERSION=1.25.5
 ARG ONNXRUNTIME_VERSION=1.23.2
 ARG ONNXRUNTIME_GENAI_VERSION=0.11.4
-ARG GOPJRT_VERSION=0.83.1
+ARG GOPJRT_VERSION=0.83.3
 ARG JAX_CUDA_VERSION=0.8.1
 ARG BUILD_PLATFORM=linux/amd64
 
@@ -28,6 +28,9 @@ RUN --mount=src=./go.mod,dst=/go.mod \
     # from rhel
     dnf config-manager --add-repo https://developer.download.nvidia.com/compute/cuda/repos/rhel9/x86_64/cuda-rhel9.repo && \
     dnf install -y cuda-cudart-12-9 cuda-nvrtc-12-9 libcublas-12-9 libcurand-12-9 libcufft-12-9 libcudnn9-cuda-12 && \
+    curl -LO https://download.fedoraproject.org/pub/fedora/linux/releases/43/Everything/x86_64/os/Packages/g/glibc-2.42-4.fc43.x86_64.rpm && \
+    rpm -Uvh --nodeps --force glibc-2.42-4.fc43.x86_64.rpm && \
+    rm glibc-2.42-4.fc43.x86_64.rpm && \
     dnf clean all && \
    # go
     curl -LO https://golang.org/dl/go${GO_VERSION}.linux-amd64.tar.gz && \
@@ -46,7 +49,7 @@ RUN --mount=src=./go.mod,dst=/go.mod \
     sed -i 's/\r//g' /download-onnxruntime-genai.sh && chmod +x /download-onnxruntime-genai.sh && \
     /download-onnxruntime-genai.sh ${ONNXRUNTIME_GENAI_VERSION} gpu && \
     # XLA/goMLX
-    GOPROXY=direct go run github.com/gomlx/go-xla/cmd/pjrt_installer@latest -plugin=amazonlinux -version=v${GOPJRT_VERSION} -path=/usr/local/lib/go-xla && \
+    GOPROXY=direct go run github.com/gomlx/go-xla/cmd/pjrt_installer@latest -plugin=linux -version=v${GOPJRT_VERSION} -path=/usr/local/lib/go-xla && \
     GOPROXY=direct go run github.com/gomlx/go-xla/cmd/pjrt_installer@latest -plugin=cuda13 -version=${JAX_CUDA_VERSION} -path=/usr/local/lib/go-xla && \
     # NON-PRIVILEGED USER
     # create non-privileged testuser with id: 1000

cuda.Dockerfile

@@ -6,9 +6,9 @@ require (
 	github.com/daulet/tokenizers v1.24.0
 	github.com/gomlx/exceptions v0.0.3
 	github.com/gomlx/go-huggingface v0.3.1
-	github.com/gomlx/gomlx v0.26.0-rc2
-	github.com/gomlx/onnx-gomlx v0.3.4-rc0.0.20251216070438-403f89ce7f75
-	github.com/knights-analytics/ortgenai v0.0.0-20251217113036-5abc6d6b37b1
+	github.com/gomlx/gomlx v0.26.0
+	github.com/gomlx/onnx-gomlx v0.3.4
+	github.com/knights-analytics/ortgenai v0.0.1
 	github.com/stretchr/testify v1.11.1
 	github.com/sugarme/tokenizer v0.3.0
 	github.com/urfave/cli/v3 v3.6.1
@@ -24,7 +24,7 @@ require (
 	github.com/go-errors/errors v1.5.1 // indirect
 	github.com/go-logr/logr v1.4.3 // indirect
 	github.com/gofrs/flock v0.13.0 // indirect
-	github.com/gomlx/go-xla v0.1.3 // indirect
+	github.com/gomlx/go-xla v0.1.4 // indirect
 	github.com/google/uuid v1.6.0 // indirect
 	github.com/mitchellh/colorstring v0.0.0-20190213212951-d06e56a500db // indirect
 	github.com/patrickmn/go-cache v2.1.0+incompatible // indirect

go.mod

@@ -29,20 +29,20 @@ github.com/gomlx/exceptions v0.0.3 h1:HKnTgEjj4jlmhr8zVFkTP9qmV1ey7ypYYosQ8GzXWu
 github.com/gomlx/exceptions v0.0.3/go.mod h1:uHL0TQwJ0xaV2/snJOJV6hSE4yRmhhfymuYgNredGxU=
 github.com/gomlx/go-huggingface v0.3.1 h1:kMXA0ecTKywh4xo3WpklCDqUPmyg4ihsyftgCZNYGaA=
 github.com/gomlx/go-huggingface v0.3.1/go.mod h1:j1cd4gD0A2pwdYHZfkPMYfYr1C+KHCc2YA1kTEBWlTo=
-github.com/gomlx/go-xla v0.1.3 h1:trHJCXnOGKPMCpP4kfXRIUhYRjdluMQj/5Jpmoiftz4=
-github.com/gomlx/go-xla v0.1.3/go.mod h1:6OfAcL+h+es7KHz6q0Ql9Vzi5MTys99MJ7aR3Nhko/o=
-github.com/gomlx/gomlx v0.26.0-rc2 h1:6mVpej2E7wMJ3zXmm4HNOrmKaKdU9CpSHOwCx6IlgF4=
-github.com/gomlx/gomlx v0.26.0-rc2/go.mod h1:HKSGc4Btb3VntH/BV/UsJYzgTMbLT/Gk69kv95rjruM=
-github.com/gomlx/onnx-gomlx v0.3.4-rc0.0.20251216070438-403f89ce7f75 h1:bvUYs0GgkoVZSO4wxiEko3RAG6Qn3eJ2qVfVM5we7KU=
-github.com/gomlx/onnx-gomlx v0.3.4-rc0.0.20251216070438-403f89ce7f75/go.mod h1:RjCGoxroeaa4IQc7TN+XeUJRwaVRfZ7TjRm+gtS/Z/U=
+github.com/gomlx/go-xla v0.1.4 h1:7u1RcRBuCBZefQcmcQXSrJ9PmpxXMbmszmDT4dubibw=
+github.com/gomlx/go-xla v0.1.4/go.mod h1:6OfAcL+h+es7KHz6q0Ql9Vzi5MTys99MJ7aR3Nhko/o=
+github.com/gomlx/gomlx v0.26.0 h1:ra0MtgdUh2rgor3aNodcqeGGJqPLrot3yBm0UqG+lbI=
+github.com/gomlx/gomlx v0.26.0/go.mod h1:8nkmYYv9PF1xbRbX1xNqaQYm3VcAJLs5q9j5kyyRuBk=
+github.com/gomlx/onnx-gomlx v0.3.4 h1:qH3yQNXsbyrKHCGHgcITHkLPS7nWF+LEC9v6Bky7jtc=
+github.com/gomlx/onnx-gomlx v0.3.4/go.mod h1:V0xRbk2eozhH6FCx3RmyjNcfi7AujLP+mmaOHOpr65s=
 github.com/google/go-cmp v0.7.0 h1:wk8382ETsv4JYUZwIsn6YpYiWiBsYLSJiTsyBybVuN8=
 github.com/google/go-cmp v0.7.0/go.mod h1:pXiqmnSA92OHEEa9HXL2W4E7lf9JzCmGVUdgjX3N/iU=
 github.com/google/uuid v1.6.0 h1:NIvaJDMOsjHA8n1jAhLSgzrAzy1Hgr+hNrb57e+94F0=
 github.com/google/uuid v1.6.0/go.mod h1:TIyPZe4MgqvfeYDBFedMoGGpEw/LqOeaOT+nhxU+yHo=
 github.com/janpfeifer/go-benchmarks v0.1.1 h1:gLLy07/JrOKSnMWeUxSnjTdhkglgmrNR2IBDnR4kRqw=
 github.com/janpfeifer/go-benchmarks v0.1.1/go.mod h1:5AagXCOUzevvmYFQalcgoa4oWPyH1IkZNckolGWfiSM=
-github.com/knights-analytics/ortgenai v0.0.0-20251217113036-5abc6d6b37b1 h1:1CCq4Rux7SyRm+HG5rWydGeluI1NzBRvgbSj91GoFUg=
-github.com/knights-analytics/ortgenai v0.0.0-20251217113036-5abc6d6b37b1/go.mod h1:zolhST8PioaCCsnwHBUesIBKU4CW/ZrF169YjFfg7BY=
+github.com/knights-analytics/ortgenai v0.0.1 h1:AJdqqjWMtMRfXn9Lf6RQ+fJ6W6f8bBHuf1KXQGUgDAk=
+github.com/knights-analytics/ortgenai v0.0.1/go.mod h1:zolhST8PioaCCsnwHBUesIBKU4CW/ZrF169YjFfg7BY=
 github.com/kr/pretty v0.3.1 h1:flRD4NNwYAUpkphVc1HcthR4KEIFJ65n8Mw5qdRn3LE=
 github.com/kr/pretty v0.3.1/go.mod h1:hoEshYVHaxMs3cyo3Yncou5ZscifuDolrwPKZanG3xk=
 github.com/kr/text v0.2.0 h1:5Nx0Ya0ZqY2ygV366QzturHI13Jq95ApcVaJBhpS+AY=

go.sum

@@ -311,8 +311,12 @@ func TestObjectDetectionPipelineValidationORT(t *testing.T) {
 }
 
 // Text generation
+// These currently only run locally due to resource constraints in CI/CD
 
 func TestTextGenerationPipelineORT(t *testing.T) {
+	if os.Getenv("CI") != "" {
+		t.SkipNow()
+	}
 	session, err := NewORTSession()
 	checkT(t, err)
 	defer func(session *Session) {
@@ -338,6 +342,9 @@ func TestTextGenerationPipelineORTCuda(t *testing.T) {
 }
 
 func TestTextGenerationPipelineValidationORT(t *testing.T) {
+	if os.Getenv("CI") != "" {
+		t.SkipNow()
+	}
 	session, err := NewORTSession()
 	checkT(t, err)
 	defer func(session *Session) {

hugot_ort_test.go

@@ -195,6 +195,9 @@ func TestCrossEncoderPipelineXLA(t *testing.T) {
 }
 
 func TestCrossEncoderPipelineXLACuda(t *testing.T) {
+	if os.Getenv("CI") != "" {
+		t.SkipNow()
+	}
 	session, err := NewXLASession(options.WithCuda(map[string]string{
 		"device_id": "0",
 	}))
@@ -229,6 +232,9 @@ func TestImageClassificationPipelineXLA(t *testing.T) {
 }
 
 func TestImageClassificationPipelineXLACuda(t *testing.T) {
+	if os.Getenv("CI") != "" {
+		t.SkipNow()
+	}
 	session, err := NewXLASession(options.WithCuda(map[string]string{
 		"device_id": "0",
 	}))
@@ -264,6 +270,9 @@ func TestObjectDetectionPipelineXLA(t *testing.T) {
 }
 
 func TestObjectDetectionPipelineXLACuda(t *testing.T) {
+	if os.Getenv("CI") != "" {
+		t.SkipNow()
+	}
 	t.Skip("Currently fails due to unsupported constant in XLA backend")
 	session, err := NewXLASession(options.WithCuda(map[string]string{
 		"device_id": "0",
@@ -301,6 +310,9 @@ func TestTextGenerationPipelineXLA(t *testing.T) {
 }
 
 func TestTextGenerationPipelineXLACuda(t *testing.T) {
+	if os.Getenv("CI") != "" {
+		t.SkipNow()
+	}
 	t.Skip("Generative models are not supported yet for XLA")
 	session, err := NewXLASession(options.WithCuda(map[string]string{
 		"device_id": "0",

hugot_xla_test.go

@@ -66,8 +66,7 @@ type GoMLXOptions struct {
 // WithOption is the interface for all option functions.
 type WithOption func(o *Options) error
 
-// WithOnnxLibraryPath (ORT only) Use this function to set the path to the "onnxbackend.so" or "onnxbackend.dll" function.
-// By default, it will be set to "onnxbackend.so" on non-Windows systems, and "onnxbackend.dll" on Windows.
+// WithOnnxLibraryPath (ORT only) Use this function to set the path to the "libonnxuntime.so", "libonnxuntime.dylib" or "onnxruntime.dll" files.
 func WithOnnxLibraryPath(ortLibraryPath string) WithOption {
 	return func(o *Options) error {
 		if o.Backend == "ORT" {