Add small demo

language_model/tensorflow/bert/cleanup_scripts/download_demo.sh

@@ -0,0 +1,49 @@
+#!/bin/bash
+
+# pip install --user gdown
+
+DATA_DIR="./wiki"
+
+# Capture MLCube parameter
+while [ "$1" != "" ]; do
+    case $1 in
+    --data_dir=*)
+        DATA_DIR="${1#*=}"
+        ;;
+    esac
+    shift
+done
+
+mkdir -p $DATA_DIR
+
+cd $DATA_DIR
+
+# Downloading files from Google Drive location: https://drive.google.com/drive/folders/1oQF4diVHNPCclykwdvQJw8n_VIWwV0PT
+
+# bert_config.json
+gdown https://drive.google.com/uc?id=1fbGClQMi2CoMv7fwrwTC5YYPooQBdcFW
+
+#License.txt
+gdown https://drive.google.com/uc?id=1SYfj3zsFPvXwo4nUVkAS54JVczBFLCWI
+
+# vocab.txt
+gdown https://drive.google.com/uc?id=1USK108J6hMM_d27xCHi738qBL8_BT1u1
+
+# Download TF-2 checkpoints
+mkdir tf2_ckpt
+
+cd tf2_ckpt
+
+gdown https://drive.google.com/uc?id=1pJhVkACK3p_7Uc-1pAzRaOXodNeeHZ7F
+
+gdown https://drive.google.com/uc?id=1oVBgtSxkXC9rH2SXJv85RXR9-WrMPy-Q
+
+cd ..
+
+# Download dummy data in TFRecord format
+wget https://storage.googleapis.com/mlperf_training_demo/bert/sample_data/tf_data.zip
+unzip tf_data.zip
+rm tf_data.zip
+wget https://storage.googleapis.com/mlperf_training_demo/bert/sample_data/tf_eval_data.zip
+unzip tf_eval_data.zip
+rm tf_eval_data.zip

language_model/tensorflow/bert/mlcube/README.md

@@ -20,6 +20,22 @@ cd ./mlcube
 mlcube describe
 ```
 
+### Demo execution
+
+These tasks will use a demo dataset to execute a faster training workload for a quick demo (~8 min):
+
+```bash
+mlcube run --task=download_demo -Pdocker.build_strategy=always
+
+mlcube run --task=demo -Pdocker.build_strategy=always
+```
+
+It's also possible to execute the two tasks in one single instruction:
+
+```bash
+mlcube run --task=download_demo,demo -Pdocker.build_strategy=always
+```
+
 ### MLCube tasks
 
 Download dataset.

language_model/tensorflow/bert/mlcube/mlcube.yaml

@@ -22,6 +22,11 @@ tasks:
     parameters:
       outputs:
         data_dir: data/
+  download_demo:
+    entrypoint: ./cleanup_scripts/download_demo.sh -a
+    parameters:
+      outputs:
+        data_dir: demo_data/
   process_data:
     entrypoint: ./cleanup_scripts/create_pretraining_data.sh -a
     parameters:
@@ -36,6 +41,23 @@ tasks:
       outputs:
         output_path: tf_data/
         output_eval_path: tf_eval_data/
+  demo:
+    entrypoint: ./run_demo.sh -a
+    parameters:
+      inputs:
+        tfdata_path: demo_data/tf_data/
+        init_checkpoint:
+          type: file
+          default: demo_data/tf2_ckpt/model.ckpt-28252
+        eval_file:
+          type: file
+          default: demo_data/tf_eval_data/eval_10k
+        config_path:
+          type: file
+          default: demo_data/bert_config.json
+      outputs:
+        log_dir: demo_logs/
+        output_dir: demo_output/
   train:
     entrypoint: ./run_and_time.sh -a
     parameters:

language_model/tensorflow/bert/requirements.txt

@@ -1,3 +1,3 @@
-gdown==4.6.4
+gdown==4.7.3
 git+https://github.com/mlcommons/[email protected]
 cloud-tpu-client

language_model/tensorflow/bert/run_demo.sh

@@ -0,0 +1,115 @@
+#!/bin/bash
+
+set +x
+set -e
+
+# start timing
+start=$(date +%s)
+start_fmt=$(date +%Y-%m-%d\ %r)
+echo "STARTING TIMING RUN AT $start_fmt"
+
+# Set variables
+: "${TFDATA_PATH:=./workspace/output_data}"
+: "${INIT_CHECKPOINT:=./workspace/data/tf2_ckpt}"
+: "${EVAL_FILE:=./workspace/tf_eval_data/eval_10k}"
+: "${CONFIG_PATH:=./workspace/data/bert_config.json}"
+: "${LOG_DIR:=./workspace/logs}"
+: "${OUTPUT_DIR:=./workspace/final_output}"
+
+# Handle MLCube parameters
+while [ $# -gt 0 ]; do
+  case "$1" in
+  --tfdata_path=*)
+    TFDATA_PATH="${1#*=}"
+    ;;
+  --config_path=*)
+    CONFIG_PATH="${1#*=}"
+    ;;
+  --init_checkpoint=*)
+    INIT_CHECKPOINT="${1#*=}"
+    ;;
+  --log_dir=*)
+    LOG_DIR="${1#*=}"
+    ;;
+  --output_dir=*)
+    OUTPUT_DIR="${1#*=}"
+    ;;
+  --eval_file=*)
+    EVAL_FILE="${1#*=}"
+    ;;
+  *) ;;
+  esac
+  shift
+done
+
+# run benchmark
+echo "running benchmark"
+
+TF_XLA_FLAGS='--tf_xla_auto_jit=2' \
+  python run_pretraining.py \
+  --bert_config_file=$CONFIG_PATH \
+  --output_dir=$OUTPUT_DIR \
+  --input_file="$TFDATA_PATH/part*" \
+  --init_checkpoint=$INIT_CHECKPOINT \
+  --nodo_eval \
+  --do_train \
+  --eval_batch_size=1 \
+  --learning_rate=0.0001 \
+  --iterations_per_loop=1000 \
+  --max_predictions_per_seq=76 \
+  --max_seq_length=512 \
+  --num_train_steps=10 \
+  --num_warmup_steps=1 \
+  --optimizer=lamb \
+  --save_checkpoints_steps=5 \
+  --start_warmup_step=0 \
+  --num_gpus=1 \
+  --train_batch_size=1 |& tee "$LOG_DIR/train_console.log"
+
+# Copy log file to MLCube log folder
+if [ "$LOG_DIR" != "" ]; then
+  timestamp=$(date +%Y%m%d_%H%M%S)
+  cp bert.log "$LOG_DIR/bert_train_$timestamp.log"
+fi
+
+TF_XLA_FLAGS='--tf_xla_auto_jit=2' \
+  python3 run_pretraining.py \
+  --bert_config_file=$CONFIG_PATH \
+  --output_dir=$OUTPUT_DIR \
+  --input_file=$EVAL_FILE \
+  --do_eval \
+  --nodo_train \
+  --eval_batch_size=8 \
+  --init_checkpoint=$OUTPUT_DIR/model.ckpt-5 \
+  --iterations_per_loop=10 \
+  --learning_rate=0.0001 \
+  --max_eval_steps=10 \
+  --max_predictions_per_seq=76 \
+  --max_seq_length=512 \
+  --num_gpus=1 \
+  --num_train_steps=10 \
+  --num_warmup_steps=1 \
+  --optimizer=lamb \
+  --save_checkpoints_steps=5 \
+  --start_warmup_step=1 \
+  --train_batch_size=24 \
+  --nouse_tpu |& tee "$LOG_DIR/eval_console.log"
+ret_code=$?
+
+# Copy log file to MLCube log folder
+if [ "$LOG_DIR" != "" ]; then
+  timestamp=$(date +%Y%m%d_%H%M%S)
+  cp bert.log "$LOG_DIR/bert_eval_$timestamp.log"
+fi
+
+set +x
+
+sleep 3
+if [[ $ret_code != 0 ]]; then exit $ret_code; fi
+
+# end timing
+end=$(date +%s)
+end_fmt=$(date +%Y-%m-%d\ %r)
+echo "ENDING TIMING RUN AT $end_fmt"
+secs=$((end - start))
+printf 'TOTAL EXECUTION TIME: %dh:%dm:%ds\n' $((secs / 3600)) $((secs % 3600 / 60)) $((secs % 60))