Intel Internship
EasyRec
This is a public repository of recommender. It will be replaced by DeepRec.
training on easyrec
pull repo from github
1
2git clone https://github.com/alibaba/EasyRec.git
cd EasyRec/download dataset
1
wget https://easyrec.oss-cn-beijing.aliyuncs.com/data/easyrec_data_20210818.tar.gz
generate .py file based on proto
bash scripts/gen_proto.sh
install dependency
pip install -r requirements/runtime.txt
pip install -r requirements/tests.txt
train existing model
1
python -m easy_rec.python.train_eval --pipeline_config_path samples/model_config/<selected model>
open dashboard
cd EasyRec/experiments/<selected model>
tensorboard --logdir .
DeepRec
This is a public repository of alibaba recommender. It is an upgraded version of EasyRec.
testing on deeprec
build wheel and image
./ci_build bazel_build cpu
enter container
CI_DOCKER_EXTRA_PARAMS=-it ./ci_build bash
install environment in container
pip install wheels/tensorflow/<.whl>
build wheel
./ci_build bazel_build cpu
Alibaba Remote Machine
Connect to alibaba remote machine.
set proxy in MobaXterm.
K83 usage – with kubectl
check nodes
kubectl get nodes
,--all-namespaces
to get all nodes under different namespacescheck existing tfjobs
kubectl get tfjobs -o wide
note: if the tfjob you are intending to create already exists, delete this tfjob
kubectl delete tfjob <tfjob name>
create a new tfjob
kubectl create -f <.yaml>
check pod status
watch -n 1 kubectl get pods -o wide
note: if the status is pending, try change the parameter of
resources
in you yaml filecheck training log
kubectl logs -f <tfjob name>-chief-0