Merge pull request #550 from rhatdan/nvidia

Make more options optional
containers · Jun 12, 2024 · 178f2c1 · 178f2c1
2 parents 7e17de4 + 7f07935
commit 178f2c1
Show file tree

Hide file tree

Showing 7 changed files with 46 additions and 49 deletions.
diff --git a/training/amd-bootc/Containerfile b/training/amd-bootc/Containerfile
@@ -19,7 +19,17 @@ RUN sed -i -e '/additionalimage.*/a "/usr/lib/containers/storage",' \
 	cp /run/.input/ilab /usr/local/bin/ilab
 
 ARG INSTRUCTLAB_IMAGE="quay.io/ai-lab/instructlab-amd:latest"
-ARG VLLM_IMAGE
+ARG VLLM_IMAGE="quay.io/ai-lab/vllm:latest"
+
+ARG SSHPUBKEY
+
+# The --build-arg "SSHPUBKEY=$(cat ~/.ssh/id_rsa.pub)" option inserts your
+# public key into the image, allowing root access via ssh.
+RUN if [ -n "${SSHPUBKEY}" ]; then \
+    set -eu; mkdir -p /usr/ssh && \
+        echo 'AuthorizedKeysFile /usr/ssh/%u.keys .ssh/authorized_keys .ssh/authorized_keys2' >> /etc/ssh/sshd_config.d/30-auth-system.conf && \
+	    echo ${SSHPUBKEY} > /usr/ssh/root.keys && chmod 0600 /usr/ssh/root.keys; \
+fi
 
 RUN sed -i 's/__REPLACE_TRAIN_DEVICE__/cuda/' /usr/local/bin/ilab
 RUN sed -i 's/__REPLACE_CONTAINER_DEVICE__/nvidia.com\/gpu=all/' /usr/local/bin/ilab

diff --git a/training/amd-bootc/Makefile b/training/amd-bootc/Makefile
@@ -11,9 +11,10 @@ bootc: prepare-files growfs
 		$(ARCH:%=--platform linux/%) \
 		$(EXTRA_RPM_PACKAGES:%=--build-arg EXTRA_RPM_PACKAGES=%) \
 		$(FROM:%=--from=%) \
+		$(INSTRUCTLAB_IMAGE:%=--build-arg INSTRUCTLAB_IMAGE=%) \
 		$(SOURCE_DATE_EPOCH:%=--timestamp=%) \
-		--build-arg "INSTRUCTLAB_IMAGE=$(INSTRUCTLAB_IMAGE)" \
-		--build-arg "VLLM_IMAGE=$(VLLM_IMAGE)" \
+		$(VLLM_IMAGE:%=--build-arg VLLM_IMAGE=%) \
+		$(SSH_PUBKEY:%=--build-arg SSHPUBKEY=%) \
 		--cap-add SYS_ADMIN \
 		--file Containerfile \
 		--security-opt label=disable \

diff --git a/training/common/Makefile.common b/training/common/Makefile.common
@@ -32,39 +32,17 @@ KERNEL_VERSION ?=
 INSTRUCTLAB_IMAGE = $(REGISTRY)/$(REGISTRY_ORG)/instructlab-$(VENDOR):$(IMAGE_TAG)
 VLLM_IMAGE = $(REGISTRY)/$(REGISTRY_ORG)/vllm:$(IMAGE_TAG)
 TRAIN_IMAGE = $(REGISTRY)/$(REGISTRY_ORG)/deepspeed-trainer:$(IMAGE_TAG)
-INSTRUCTLAB_IMAGE_ID = $(shell $(CONTAINER_TOOL) image inspect $(INSTRUCTLAB_IMAGE) --format {{.Id}})
-VLLM_IMAGE_ID = $(shell $(CONTAINER_TOOL) image inspect $(VLLM_IMAGE) --format {{.Id}})
-TRAIN_IMAGE_ID = $(shell $(CONTAINER_TOOL) image inspect $(TRAIN_IMAGE) --format {{.Id}})
-WRAPPER = $(CURDIR)/../ilab-wrapper/ilab
-QLORA_WRAPPER = $(CURDIR)/../ilab-wrapper/ilab-qlora
-TRAIN_WRAPPER =  $(CURDIR)/../ilab-wrapper/ilab-training-launcher
 OUTDIR = $(CURDIR)/../build
 
 SSH_PUBKEY ?= $(shell cat ${HOME}/.ssh/id_rsa.pub 2> /dev/null)
 
 .PHONY: prepare-files
-prepare-files: $(OUTDIR)/$(WRAPPER) $(OUTDIR)/$(QLORA_WRAPPER) $(OUTDIR)/$(TRAIN_WRAPPER) $(OUTDIR)/$(INSTRUCTLAB_IMAGE_ID) $(OUTDIR)/$(VLLM_IMAGE_ID) $(OUTDIR)/$(TRAIN_IMAGE_ID) 
+prepare-files: $(OUTDIR)
 
+.PHONY: $(OUTDIR)
 $(OUTDIR):
 	mkdir -p $(OUTDIR)
 
-$(OUTDIR)/$(WRAPPER): $(OUTDIR)
-	cp -pf $(WRAPPER) $(OUTDIR)
-$(OUTDIR)/$(QLORA_WRAPPER): $(OUTDIR)
-	cp -pf $(QLORA_WRAPPER) $(OUTDIR)
-$(OUTDIR)/$(TRAIN_WRAPPER): $(OUTDIR)
-	cp -pf $(TRAIN_WRAPPER) $(OUTDIR)
-
-$(OUTDIR)/$(INSTRUCTLAB_IMAGE_ID):
-	@mkdir -p $(OUTDIR)/$(INSTRUCTLAB_IMAGE_ID)
-	$(CONTAINER_TOOL) push --compress=false $(INSTRUCTLAB_IMAGE) oci:$(OUTDIR)/$(INSTRUCTLAB_IMAGE_ID)/
-$(OUTDIR)/$(VLLM_IMAGE_ID):
-	@mkdir -p $(OUTDIR)/$(VLLM_IMAGE_ID)
-	$(CONTAINER_TOOL) push --compress=false $(VLLM_IMAGE) oci:$(OUTDIR)/$(VLLM_IMAGE_ID)/
-$(OUTDIR)/$(TRAIN_IMAGE_ID):
-	@mkdir -p $(OUTDIR)/$(TRAIN_IMAGE_ID)
-	$(CONTAINER_TOOL) push --compress=false $(TRAIN_IMAGE) oci:$(OUTDIR)/$(TRAIN_IMAGE_ID)/
-
 .PHONY: check-sshkey
 check-sshkey:
 	@test -n "$(SSH_PUBKEY)" || \

diff --git a/training/intel-bootc/Containerfile b/training/intel-bootc/Containerfile
@@ -39,8 +39,19 @@ RUN depmod -a ${KERNEL_VERSION}
 # Include growfs service
 COPY build/usr /usr
 
-ARG INSTRUCTLAB_IMAGE
-ARG VLLM_IMAGE
+ARG INSTRUCTLAB_IMAGE="quay.io/ai-lab/instructlab-intel:latest"
+ARG VLLM_IMAGE="quay.io/ai-lab/vllm:latest"
+
+
+ARG SSHPUBKEY
+
+# The --build-arg "SSHPUBKEY=$(cat ~/.ssh/id_rsa.pub)" option inserts your
+# public key into the image, allowing root access via ssh.
+RUN if [ -n "${SSHPUBKEY}" ]; then \
+    set -eu; mkdir -p /usr/ssh && \
+        echo 'AuthorizedKeysFile /usr/ssh/%u.keys .ssh/authorized_keys .ssh/authorized_keys2' >> /etc/ssh/sshd_config.d/30-auth-system.conf && \
+	    echo ${SSHPUBKEY} > /usr/ssh/root.keys && chmod 0600 /usr/ssh/root.keys; \
+fi
 
 # Prepull the instructlab image
 RUN IID=$(podman --root /usr/lib/containers/storage pull oci:/run/.input/vllm) && \

diff --git a/training/intel-bootc/Makefile b/training/intel-bootc/Makefile
@@ -11,10 +11,11 @@ bootc: growfs prepare-files
 		$(DRIVER_VERSION:%=--build-arg DRIVER_VERSION=%) \
 		$(EXTRA_RPM_PACKAGES:%=--build-arg EXTRA_RPM_PACKAGES=%) \
 		$(FROM:%=--build-arg BASEIMAGE=%) \
+		$(INSTRUCTLAB_IMAGE:%=--build-arg INSTRUCTLAB_IMAGE=%) \
 		$(KERNEL_VERSION:%=--build-arg KERNEL_VERSION=%) \
 		$(SOURCE_DATE_EPOCH:%=--timestamp=%) \
-		--build-arg "INSTRUCTLAB_IMAGE=$(INSTRUCTLAB_IMAGE)" \
-		--build-arg "VLLM_IMAGE=$(VLLM_IMAGE)" \
+		$(VLLM_IMAGE:%=--build-arg VLLM_IMAGE=%) \
+		$(SSH_PUBKEY:%=--build-arg SSHPUBKEY=%) \
 		--cap-add SYS_ADMIN \
 		--file Containerfile \
 		--security-opt label=disable \

diff --git a/training/nvidia-bootc/Containerfile b/training/nvidia-bootc/Containerfile
@@ -140,9 +140,11 @@ ARG SSHPUBKEY
 
 # The --build-arg "SSHPUBKEY=$(cat ~/.ssh/id_rsa.pub)" option inserts your
 # public key into the image, allowing root access via ssh.
-RUN set -eu; mkdir -p /usr/ssh && \
-    echo 'AuthorizedKeysFile /usr/ssh/%u.keys .ssh/authorized_keys .ssh/authorized_keys2' >> /etc/ssh/sshd_config.d/30-auth-system.conf && \
-    echo ${SSHPUBKEY} > /usr/ssh/root.keys && chmod 0600 /usr/ssh/root.keys
+RUN if [ -n "${SSHPUBKEY}" ]; then \
+    set -eu; mkdir -p /usr/ssh && \
+        echo 'AuthorizedKeysFile /usr/ssh/%u.keys .ssh/authorized_keys .ssh/authorized_keys2' >> /etc/ssh/sshd_config.d/30-auth-system.conf && \
+	    echo ${SSHPUBKEY} > /usr/ssh/root.keys && chmod 0600 /usr/ssh/root.keys; \
+fi
 
 # Setup /usr/lib/containers/storage as an additional store for images.
 # Remove once the base images have this set by default.
@@ -153,12 +155,9 @@ RUN grep -q /usr/lib/containers/storage /etc/containers/storage.conf || \
 	cp /run/.input/ilab* /usr/local/bin/
 
 
-ARG INSTRUCTLAB_IMAGE
-ARG INSTRUCTLAB_IMAGE_ID
-ARG VLLM_IMAGE
-ARG VLLM_IMAGE_ID
-ARG TRAIN_IMAGE
-ARG TRAIN_IMAGE_ID
+ARG INSTRUCTLAB_IMAGE="quay.io/ai-lab/instructlab-nvidia:latest"
+ARG VLLM_IMAGE="quay.io/ai-lab/vllm:latest"
+ARG TRAIN_IMAGE="quay.io/ai-lab/deepspeed-trainer:latest"
 ARG GPU_COUNT_COMMAND="nvidia-ctk --quiet cdi list | grep -P nvidia.com/gpu='\\\\d+' | wc -l"
 
 RUN for i in /usr/local/bin/ilab*; do \

diff --git a/training/nvidia-bootc/Makefile b/training/nvidia-bootc/Makefile
@@ -2,7 +2,7 @@ VENDOR ?= nvidia
 IMAGE_NAME ?= $(VENDOR)-bootc
 DTK_IMAGE_NAME ?= $(VENDOR)-builder
 DTK_IMAGE_TAG ?= latest
-DRIVER_TOOLKIT_IMAGE = "${REGISTRY}/${REGISTRY_ORG}/${DTK_IMAGE_NAME}:${DTK_IMAGE_TAG}"
+DRIVER_TOOLKIT_IMAGE = ${REGISTRY}/${REGISTRY_ORG}/${DTK_IMAGE_NAME}:${DTK_IMAGE_TAG}
 
 CUDA_VERSION ?=
 OS_VERSION_MAJOR ?=
@@ -29,21 +29,18 @@ bootc: dtk check-sshkey prepare-files growfs
 	"${CONTAINER_TOOL}" build \
 		$(ARCH:%=--platform linux/%) \
 		$(CUDA_VERSION:%=--build-arg CUDA_VERSION=%) \
+		$(DRIVER_TOOLKIT_IMAGE:%=--build-arg DRIVER_TOOLKIT_IMAGE=%) \
 		$(DRIVER_VERSION:%=--build-arg DRIVER_VERSION=%) \
 		$(DRIVER_VERSION:%=--label driver-version=%) \
 		$(EXTRA_RPM_PACKAGES:%=--build-arg EXTRA_RPM_PACKAGES=%) \
 		$(FROM:%=--build-arg BASEIMAGE=%) \
+		$(INSTRUCTLAB_IMAGE:%=--build-arg INSTRUCTLAB_IMAGE=%) \
 		$(KERNEL_VERSION:%=--build-arg KERNEL_VERSION=%) \
 		$(OS_VERSION_MAJOR:%=--build-arg OS_VERSION_MAJOR=%) \
 		$(SOURCE_DATE_EPOCH:%=--timestamp=%) \
-		--build-arg "INSTRUCTLAB_IMAGE=$(INSTRUCTLAB_IMAGE)" \
-		--build-arg "INSTRUCTLAB_IMAGE_ID=$(INSTRUCTLAB_IMAGE_ID)" \
-		--build-arg "SSHPUBKEY=$(SSH_PUBKEY)" \
-		--build-arg "TRAIN_IMAGE=$(TRAIN_IMAGE)" \
-		--build-arg "TRAIN_IMAGE_ID=$(TRAIN_IMAGE_ID)" \
-		--build-arg "VLLM_IMAGE=$(VLLM_IMAGE)" \
-		--build-arg "VLLM_IMAGE_ID=$(VLLM_IMAGE_ID)" \
-		--build-arg DRIVER_TOOLKIT_IMAGE=${DRIVER_TOOLKIT_IMAGE} \
+		$(TRAIN_IMAGE:%=--build-arg TRAIN_IMAGE=%) \
+		$(VLLM_IMAGE:%=--build-arg VLLM_IMAGE=%) \
+		$(SSH_PUBKEY:%=--build-arg SSHPUBKEY=%) \
 		--cap-add SYS_ADMIN \
 		--file Containerfile \
 		--security-opt label=disable \