diff --git a/recipes-fsl/packagegroups/packagegroup-fsl-ml.bb b/recipes-fsl/packagegroups/packagegroup-fsl-ml.bb
index 744b4de..10c8fa2 100644
--- a/recipes-fsl/packagegroups/packagegroup-fsl-ml.bb
+++ b/recipes-fsl/packagegroups/packagegroup-fsl-ml.bb
@@ -1,35 +1,75 @@
+# Copyright 2019-2021 NXP
+# Released under the MIT license (see COPYING.MIT for the terms)
+
 DESCRIPTION = "Add packages for AI/ML build"
+LICENSE = "MIT"
+LIC_FILES_CHKSUM = "file://${COMMON_LICENSE_DIR}/MIT;md5=0835ade698e0bcf8506ecda2f7b4f302"
 
-PACKAGE_ARCH = "${MACHINE_SOCARCH}"
+PACKAGE_ARCH = "${TUNE_PKGARCH}"
 
 inherit packagegroup
 
 ML_NNSTREAMER_PKGS_LIST = " \
     nnstreamer \
+    nnstreamer-deepview-rt \
     nnstreamer-protobuf \
     nnstreamer-python3 \
+    nnstreamer-query \
     nnstreamer-tensorflow-lite \
 "
+
 ML_NNSTREAMER_PKGS = ""
 ML_NNSTREAMER_PKGS:mx8-nxp-bsp:imxgpu = "${ML_NNSTREAMER_PKGS_LIST}"
+ML_NNSTREAMER_PKGS:mx8mp-nxp-bsp      = "${ML_NNSTREAMER_PKGS_LIST} nnstreamer-tvm nnstreamer-unittest"
+ML_NNSTREAMER_PKGS:mx9-nxp-bsp        = "${ML_NNSTREAMER_PKGS_LIST}"
 
-ML_PKGS    ?= ""
+# i.MX8MP uses nnshark - other i.MX8/9 use raw gstshark
+ML_GST_PROFILER = ""
+#ML_GST_PROFILER:mx8-nxp-bsp   = "gst-shark"
+#ML_GST_PROFILER:mx8mp-nxp-bsp = "nnshark"
+#ML_GST_PROFILER:mx9-nxp-bsp   = "gst-shark"
+
+# These packages don't require any acceleration
+ML_PKGS            = ""
 ML_PKGS:mx8-nxp-bsp = " \
-    onnxruntime \
-    pytorch \
+    deepview-rt-examples \
+    onnxruntime-tests \
     tensorflow-lite \
     torchvision \
 "
-ML_PKGS:mx8mq-nxp-bsp = " \
-    deepview-rt \
-    tvm \
+ML_PKGS:mx9-nxp-bsp = " \
+    deepview-rt-examples \
+    onnxruntime-tests \
+    tensorflow-lite \
+    torchvision \
 "
-ML_PKGS:mx8mp-nxp-bsp = " \
-    deepview-rt \
+# These packages require acceleration, which currently maps to
+# all SOCs with GPU except for 8M Mini
+ML_ACCELERATED_PKGS                    = ""
+ML_ACCELERATED_PKGS:mx8-nxp-bsp:imxgpu = " \
+    tensorflow-lite-vx-delegate \
+"
+ML_ACCELERATED_PKGS:mx8mp-nxp-bsp      = " \
     tensorflow-lite-vx-delegate \
     tvm \
 "
+ML_ACCELERATED_PKGS:mx8mm-nxp-bsp      = ""
+
+
+# ARM ethos-u package
+ETHOS_U_PKGS = ""
+ETHOS_U_PKGS:mx93-nxp-bsp = " \
+    ethos-u-vela \
+    ethos-u-driver-stack \
+    tensorflow-lite-ethosu-delegate \
+    eiq-examples \
+"
+
+
 RDEPENDS:${PN} = " \
     ${ML_PKGS} \
+    ${ML_ACCELERATED_PKGS} \
     ${ML_NNSTREAMER_PKGS} \
+    ${ML_GST_PROFILER} \
+    ${ETHOS_U_PKGS} \
 "