[Benchmark-Py] Adding TF-HUB ALBERT

Author: pavanimajety

tftrt/benchmarking-python/tf_hub/albert/base_run_inference.sh

@@ -0,0 +1,124 @@
+#!/bin/#!/usr/bin/env bash
+
+nvidia-smi
+
+set -x
+
+BASE_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )"
+
+#install packages for the rest of the script
+pip install tensorflow_text tensorflow_hub scipy==1.4.1
+
+
+# Runtime Parameters
+MODEL_NAME=""
+DATASET_NAME=""
+
+# Default Argument Values
+BATCH_SIZE=32
+SEQ_LEN=128
+VOCAB_SIZE=33000
+
+NUM_ITERATIONS=1000
+OUTPUT_TENSOR_NAMES="albert_encoder,albert_encoder_1,albert_encoder_2,albert_encoder_3,albert_encoder_4,albert_encoder_5,albert_encoder_6,albert_encoder_7,albert_encoder_8,albert_encoder_9,albert_encoder_10,albert_encoder_11,albert_encoder_12,albert_encoder_13"
+
+BYPASS_ARGUMENTS=""
+
+# Loop through arguments and process them
+for arg in "$@"
+do
+    case $arg in
+        --model_name=*)
+        MODEL_NAME="${arg#*=}"
+        shift # Remove --model_name from processing
+        ;;
+        --dataset_name=*)
+        DATASET_NAME="${arg#*=}"
+        shift # Remove --dataset_name= from processing
+        ;;
+        --batch_size=*)
+        BATCH_SIZE="${arg#*=}"
+        shift # Remove --batch_size= from processing
+        ;;
+        --sequence_length=*)
+        SEQ_LEN="${arg#*=}"
+        shift # Remove --sequence_length= from processing
+        ;;
+        --num_iterations=*)
+        NUM_ITERATIONS="${arg#*=}"
+        shift # Remove --num_iterations= from processing
+        ;;
+        --vocab_size=*)
+        VOCAB_SIZE="${arg#*=}"
+        shift # Remove --vocab_size= from processing
+        ;;
+        --output_tensors_name=*)
+        OUTPUT_TENSOR_NAMES="${arg#*=}"
+        shift # Remove --output_tensors_name= from processing
+        ;;
+        ######### IGNORE ARGUMENTS BELOW
+        --data_dir=*)
+        shift # Remove --data_dir= from processing
+        ;;
+        --input_saved_model_dir=*)
+        shift # Remove --input_saved_model_dir= from processing
+        ;;
+        --tokenizer_dir=*)
+        shift # Remove --tokenizer_model_dir= from processing
+        ;;
+        --total_max_samples=*)
+        shift # Remove --total_max_samples= from processing
+        ;;
+        *)
+        BYPASS_ARGUMENTS=" ${BYPASS_ARGUMENTS} ${arg}"
+        ;;
+    esac
+done
+
+echo -e "\n********************************************************************"
+echo "[*] MODEL_NAME: ${MODEL_NAME}"
+echo "[*] DATASET_NAME: ${DATASET_NAME}"
+echo ""
+echo "[*] DATA_DIR: ${DATA_DIR}"
+echo "[*] BATCH_SIZE: ${BATCH_SIZE}"
+echo ""
+# Custom T5 Task Flags
+echo "[*] SEQ_LEN: ${SEQ_LEN}"
+echo "[*] OUTPUT_TENSOR_NAMES: ${OUTPUT_TENSOR_NAMES}"
+echo ""
+echo "[*] BYPASS_ARGUMENTS: $(echo \"${BYPASS_ARGUMENTS}\" | tr -s ' ')"
+
+echo -e "********************************************************************\n"
+
+DATA_DIR="/workspace/tftrt/benchmarking-python/tf_hub/albert/data"
+MODEL_DIR="/models/tf_hub/albert/${MODEL_NAME}/"
+TOKENIZER_DIR="/models/tf_hub/albert/tokenizer"
+
+if [[ ! -d ${DATA_DIR} ]]; then
+    echo "ERROR: \`--data_dir=/path/to/directory\` does not exist. [Received: \`${DATA_DIR}\`]"
+    exit 1
+fi
+
+if [[ ! -d ${MODEL_DIR} ]]; then
+    echo "ERROR: \`--input_saved_model_dir=/path/to/directory\` does not exist. [Received: \`${MODEL_DIR}\`]"
+    exit 1
+fi
+
+if [[ ! -d ${TOKENIZER_DIR} ]]; then
+    echo "ERROR: \`--tokenizer_dir=/path/to/directory\` does not exist. [Received: \`${TOKENIZER_DIR}\`]"
+    exit 1
+fi
+
+# Dataset Directory
+
+python ${BASE_DIR}/infer.py \
+    --data_dir=${DATA_DIR} \
+    --calib_data_dir=${DATA_DIR} \
+    --input_saved_model_dir=${MODEL_DIR} \
+    --tokenizer_dir=${TOKENIZER_DIR}\
+    --output_tensors_name=${OUTPUT_TENSOR_NAMES} \
+    `# The following is set because we will be running synthetic benchmarks` \
+    --total_max_samples=1 \
+    --use_synthetic_data  \
+    --num_iterations=${NUM_ITERATIONS} \
+    ${@}