[HuggingFace][Neuronx] Training - Optimum Neuron 0.0.25 - Neuron sdk 2.20.0 - Transformers to 4.43.2

dlc_developer_config.toml

@@ -1,6 +1,6 @@
 [dev]
 # Set to "huggingface", for example, if you are a huggingface developer. Default is ""
-partner_developer = ""
+partner_developer = "huggingface"
 # Please only set it to true if you are preparing an EI related PR
 # Do remember to revert it back to false before merging any PR (including EI dedicated PR)
 ei_mode = false
@@ -9,7 +9,7 @@ ei_mode = false
 neuron_mode = false
 # Please only set it to true if you are preparing a NEURONX related PR
 # Do remember to revert it back to false before merging any PR (including NEURONX dedicated PR)
-neuronx_mode = false
+neuronx_mode = true
 # Please only set it to true if you are preparing a GRAVITON related PR
 # Do remember to revert it back to false before merging any PR (including GRAVITON dedicated PR)
 graviton_mode = false
@@ -34,11 +34,11 @@ deep_canary_mode = false
 [build]
 # Add in frameworks you would like to build. By default, builds are disabled unless you specify building an image.
 # available frameworks - ["autogluon", "huggingface_tensorflow", "huggingface_pytorch", "huggingface_tensorflow_trcomp", "huggingface_pytorch_trcomp", "pytorch_trcomp", "tensorflow", "mxnet", "pytorch", "stabilityai_pytorch"]
-build_frameworks = []
+build_frameworks = ["huggingface_pytorch"]
 
 # By default we build both training and inference containers. Set true/false values to determine which to build.
 build_training = true
-build_inference = true
+build_inference = false
 
 # Set do_build to "false" to skip builds and test the latest image built by this PR
 # Note: at least one build is required to set do_build to "false"

huggingface/pytorch/training/buildspec-neuronx.yml

@@ -34,9 +34,9 @@ images:
     device_type: &DEVICE_TYPE neuronx
     python_version: &DOCKER_PYTHON_VERSION py3
     tag_python_version: &TAG_PYTHON_VERSION py310
-    neuron_sdk_version: &NEURON_SDK_VERSION sdk2.19.1
+    neuron_sdk_version: &NEURON_SDK_VERSION sdk2.20.0
     os_version: &OS_VERSION ubuntu20.04
-    transformers_version: &TRANSFORMERS_VERSION 4.41.1
+    transformers_version: &TRANSFORMERS_VERSION 4.43.2
     datasets_version: &DATASETS_VERSION 2.18.0
     tag: !join [ *VERSION, '-', 'transformers', *TRANSFORMERS_VERSION, '-', *DEVICE_TYPE, '-', *TAG_PYTHON_VERSION,"-", *NEURON_SDK_VERSION, '-', *OS_VERSION ]
     docker_file: !join [ docker/, *SHORT_VERSION, /, *DOCKER_PYTHON_VERSION, /, *NEURON_SDK_VERSION, /Dockerfile., *DEVICE_TYPE ]

huggingface/pytorch/training/docker/2.1/py3/sdk2.20.0/Dockerfile.neuronx

@@ -0,0 +1,55 @@
+# https://github.com/aws/deep-learning-containers/blob/master/available_images.md
+# refer to the above page to pull latest PyTorch Neuronx image
+
+# docker image region us-west-2
+FROM 763104351884.dkr.ecr.us-west-2.amazonaws.com/pytorch-training-neuronx:2.1.2-neuronx-py310-sdk2.20.0-ubuntu20.04
+
+
+LABEL maintainer="Amazon AI"
+LABEL dlc_major_version="1"
+
+# version args
+ARG OPTIMUM_NEURON_VERSION=0.0.25
+ARG TRANSFORMERS_VERSION
+ARG DATASETS_VERSION
+ARG GEVENT_VERSION=24.10.3
+ARG PYTHON=python3
+
+ # install Hugging Face libraries and its dependencies
+RUN pip install --no-cache-dir \
+	"sagemaker>=2.214.3,<3" \
+	evaluate \
+	transformers[sklearn,sentencepiece,audio,vision]==${TRANSFORMERS_VERSION} \
+	datasets==${DATASETS_VERSION} \
+	optimum-neuron==${OPTIMUM_NEURON_VERSION} \
+    peft \
+	gevent==${GEVENT_VERSION}
+
+# Pin numpy to version required by neuronx-cc
+# Update Pillow and urllib version to fix high and critical vulnerabilities
+RUN pip install -U \
+	"numpy>=1.24.3,<=1.25.2" \
+	"numba==0.58.1" \
+	"Pillow==10.3.0" \
+	"requests<2.32.0" \
+	"urllib3>=1.26.17,<1.27"
+
+RUN apt-get update \
+ && apt install -y --no-install-recommends \
+    git-lfs \
+	libgssapi-krb5-2 \
+	libexpat1 \
+	expat \
+	libarchive13 \
+ && apt-get upgrade -y apparmor \
+ && apt-get clean \
+ && rm -rf /var/lib/apt/lists/*
+
+RUN HOME_DIR=/root \
+ && curl -o ${HOME_DIR}/oss_compliance.zip https://aws-dlinfra-utilities.s3.amazonaws.com/oss_compliance.zip \
+ && unzip ${HOME_DIR}/oss_compliance.zip -d ${HOME_DIR}/ \
+ && cp ${HOME_DIR}/oss_compliance/test/testOSSCompliance /usr/local/bin/testOSSCompliance \
+ && chmod +x /usr/local/bin/testOSSCompliance \
+ && chmod +x ${HOME_DIR}/oss_compliance/generate_oss_compliance.sh \
+ && ${HOME_DIR}/oss_compliance/generate_oss_compliance.sh ${HOME_DIR} ${PYTHON} \
+ && rm -rf ${HOME_DIR}/oss_compliance*