dlrm

This benchmark seems to require tensorflow.contrib which has been dropped in 2.6.x+ so is no longer functional.

Ensure the virtual environment is prepared as described in MLPerf Inference

Activate the virtual environment first and then

cd ~/src/inference/recommendation/dlrm/tf
python3 dlrm_main.py --batch_size=65536 --nobfloat16_grads_all_reduce --data_dir=gs://dmchen-data/dlrm_tf1_data/criteo --decay_start_step=38000 --decay_steps=40000 --dim_embed=128 --noenable_profiling --noenable_summary --eval_batch_size=65536 --eval_steps=1362 --gcp_project=[ID of your GCP project] --learning_rate=0.7 --lr_warmup_steps=2000 --master=[Name of your cloud TPU] --mlp_bottom=512,256,128 --mlp_top=1024,1024,512,256,1 --model_dir=gs://dmchen-data/dlrm_tf1_data/model_dir_0 --num_dense_features=13 --num_tables_in_ec=13 --num_tpu_shards=128 --optimizer=sgd --nopipeline_execution --replicas_per_host=8 --restore_checkpoint --nosave_checkpoint --sleep_after_init=60 --steps_between_evals=0 --tpu_zone=europe-west4-a --train_steps=0 --use_batched_tfrecords --nouse_cached_data --nouse_synthetic_data --vocab_sizes_embed=39884406,39043,17289,7420,20263,3,7120,1543,63,38532951,2953546,403346,10,2208,11938,155,4,976,14,39979771,25641295,39664984,585935,12972,108,36

TensorFlow

dlrm

Analytics

Related content