intel
diff --git a/‎benchmarks/image_recognition/tensorflow/resnet50v1_5/inference/int8/model_init.py
Lines changed: 8 additions & 7 deletions b/‎benchmarks/image_recognition/tensorflow/resnet50v1_5/inference/int8/model_init.py
Lines changed: 8 additions & 7 deletions
diff --git a/‎benchmarks/language_modeling/tensorflow/bert_large/README.md
Lines changed: 1 addition & 0 deletions b/‎benchmarks/language_modeling/tensorflow/bert_large/README.md
Lines changed: 1 addition & 0 deletions
diff --git a/‎benchmarks/language_modeling/tensorflow/bert_large/inference/int8/.docs/advanced/launch_benchmark_instructions.md
Lines changed: 65 additions & 0 deletions b/‎benchmarks/language_modeling/tensorflow/bert_large/inference/int8/.docs/advanced/launch_benchmark_instructions.md
Lines changed: 65 additions & 0 deletions
diff --git a/‎benchmarks/language_modeling/tensorflow/bert_large/inference/int8/.docs/advanced/model_args.md
Lines changed: 15 additions & 0 deletions b/‎benchmarks/language_modeling/tensorflow/bert_large/inference/int8/.docs/advanced/model_args.md
Lines changed: 15 additions & 0 deletions
diff --git a/‎benchmarks/language_modeling/tensorflow/bert_large/inference/int8/Advanced.md
Lines changed: 110 additions & 0 deletions b/‎benchmarks/language_modeling/tensorflow/bert_large/inference/int8/Advanced.md
Lines changed: 110 additions & 0 deletions
diff --git a/‎benchmarks/language_modeling/tensorflow/bert_large/inference/int8/__init__.py
Lines changed: 19 additions & 0 deletions b/‎benchmarks/language_modeling/tensorflow/bert_large/inference/int8/__init__.py
Lines changed: 19 additions & 0 deletions
diff --git a/‎benchmarks/language_modeling/tensorflow/bert_large/inference/int8/config.json
Lines changed: 8 additions & 0 deletions b/‎benchmarks/language_modeling/tensorflow/bert_large/inference/int8/config.json
Lines changed: 8 additions & 0 deletions
@@ -77,14 +77,15 @@ def run_benchmark_or_accuracy(self):
             cmd = os.path.join(
                 self.args.intelai_models, self.args.mode,
                 "eval_image_classifier_inference_weight_sharing.py")
-        if self.args.gpu:
-            cmd = os.path.join(
-                self.args.intelai_models, self.args.mode, self.args.precision,
-                "eval_image_classifier_inference.py")
         else:
-            cmd = os.path.join(
-                self.args.intelai_models, self.args.mode,
-                "eval_image_classifier_inference.py")
+            if self.args.gpu:
+                cmd = os.path.join(
+                    self.args.intelai_models, self.args.mode, self.args.precision,
+                    "eval_image_classifier_inference.py")
+            else:
+                cmd = os.path.join(
+                    self.args.intelai_models, self.args.mode,
+                    "eval_image_classifier_inference.py")
 
         cmd = self.get_command_prefix(self.args.socket_id) + self.python_exe + " " + cmd
 
 
@@ -3,5 +3,6 @@
 The following documents have instructions for running BERT large:
 * [BFloat16 Inference](/benchmarks/language_modeling/tensorflow/bert_large/inference/bfloat16/README.md)
 * [FP32 Inference](/benchmarks/language_modeling/tensorflow/bert_large/inference/fp32/README.md)
+* [Int8 Inference](/benchmarks/language_modeling/tensorflow/bert_large/inference/int8/README.md)
 * [BFloat16 Training](/benchmarks/language_modeling/tensorflow/bert_large/training/bfloat16/README.md)
 * [FP32 Training](/benchmarks/language_modeling/tensorflow/bert_large/training/fp32/README.md)
@@ -0,0 +1,65 @@
+<!-- 50. Launch benchmark instructions -->
+Once your environment is setup, navigate to the `benchmarks` directory of
+the model zoo and set environment variables for the dataset, checkpoint
+directory, frozen graph, and an output directory where log files will be written.
+```
+cd benchmarks
+
+export DATASET_DIR=<path to the squad dataset>
+export CHECKPOINT_DIR=<path to the pretrained model checkpoints>
+export PRETRAINED_MODEL=<path to the frozen graph .pb file>
+export OUTPUT_DIR=<directory where log files will be saved>
+```
+
+<model name> <mode> can be run in three different modes:
+
+* Benchmark
+  ```
+  python launch_benchmark.py \
+    --model-name=bert_large \
+    --precision=int8 \
+    --mode=inference \
+    --framework=tensorflow \
+    --batch-size=32 \
+    --data-location $DATASET_DIR \
+    --checkpoint $CHECKPOINT_DIR \
+    --in-graph $PRETRAINED_MODEL \
+    --output-dir $OUTPUT_DIR \
+    --docker-image <docker image> \
+    --benchmark-only \
+    -- infer_option=SQuAD
+  ```
+* Profile
+  ```
+  python launch_benchmark.py \
+    --model-name=bert_large \
+    --precision=int8 \
+    --mode=inference \
+    --framework=tensorflow \
+    --batch-size=32 \
+    --data-location $DATASET_DIR \
+    --checkpoint $CHECKPOINT_DIR \
+    --in-graph $PRETRAINED_MODEL \
+    --output-dir $OUTPUT_DIR \
+    --docker-image <docker image> \
+    --accuracy-only \
+    -- infer_option=SQuAD
+  ```
+* Accuracy
+  ```
+  python launch_benchmark.py \
+    --model-name=bert_large \
+    --precision=int8 \
+    --mode=inference \
+    --framework=tensorflow \
+    --batch-size=32 \
+    --data-location $DATASET_DIR \
+    --checkpoint $CHECKPOINT_DIR \
+    --in-graph $PRETRAINED_MODEL \
+    --output-dir $OUTPUT_DIR \
+    --docker-image <docker image> \
+    --accuracy-only \
+    -- infer_option=SQuAD
+  ```
+
+Output files and logs are saved to the ${OUTPUT_DIR} directory.
@@ -0,0 +1,15 @@
+<!-- 70. Model args -->
+Note that args specific to this model are specified after ` -- ` at
+the end of the command (like the `profile=True` arg in the Profile
+command above. Below is a list of all of the model specific args and
+their default values:
+
+| Model arg | Default value |
+|-----------|---------------|
+| doc_stride | `128` |
+| max_seq_length | `384` |
+| profile | `False` |
+| config_file | `bert_config.json` |
+| vocab_file | `vocab.txt` |
+| predict_file | `dev-v1.1.json` |
+| init_checkpoint | `model.ckpt-3649` |
@@ -0,0 +1,110 @@
+<!--- 0. Title -->
+<!-- This document is auto-generated using markdown fragments and the model-builder -->
+<!-- To make changes to this doc, please change the fragments instead of modifying this doc directly -->
+# BERT Large Int8 inference - Advanced Instructions
+
+<!-- 10. Description -->
+This document has advanced instructions for running BERT Large Int8
+inference, which provides more control over the individual parameters that
+are used. For more information on using [`/benchmarks/launch_benchmark.py`](/benchmarks/launch_benchmark.py),
+see the [launch benchmark documentation](/docs/general/tensorflow/LaunchBenchmark.md).
+
+Prior to using these instructions, please follow the setup instructions from
+the model's [README](README.md) and/or the
+[AI Kit documentation](/docs/general/tensorflow/AIKit.md) to get your environment
+setup (if running on bare metal) and download the dataset, pretrained model, etc.
+If you are using AI Kit, please exclude the `--docker-image` flag from the
+commands below, since you will be running the the TensorFlow conda environment
+instead of docker.
+
+<!-- 55. Docker arg -->
+Any of the `launch_benchmark.py` commands below can be run on bare metal by
+removing the `--docker-image` arg. Ensure that you have all of the
+[required prerequisites installed](README.md#run-the-model) in your environment
+before running without the docker container.
+
+If you are new to docker and are running into issues with the container,
+see [this document](/docs/general/docker.md) for troubleshooting tips.
+
+<!-- 50. Launch benchmark instructions -->
+Once your environment is setup, navigate to the `benchmarks` directory of
+the model zoo and set environment variables for the dataset, checkpoint
+directory, frozen graph, and an output directory where log files will be written.
+```
+cd benchmarks
+
+export DATASET_DIR=<path to the squad dataset>
+export CHECKPOINT_DIR=<path to the pretrained model checkpoints>
+export PRETRAINED_MODEL=<path to the frozen graph .pb file>
+export OUTPUT_DIR=<directory where log files will be saved>
+```
+
+BERT Large inference can be run in three different modes:
+
+* Benchmark
+  ```
+  python launch_benchmark.py \
+    --model-name=bert_large \
+    --precision=int8 \
+    --mode=inference \
+    --framework=tensorflow \
+    --batch-size=32 \
+    --data-location $DATASET_DIR \
+    --checkpoint $CHECKPOINT_DIR \
+    --in-graph $PRETRAINED_MODEL \
+    --output-dir $OUTPUT_DIR \
+    --docker-image intel/intel-optimized-tensorflow:tf-r2.5-icx-b631821f \
+    --benchmark-only \
+    -- infer_option=SQuAD
+  ```
+* Profile
+  ```
+  python launch_benchmark.py \
+    --model-name=bert_large \
+    --precision=int8 \
+    --mode=inference \
+    --framework=tensorflow \
+    --batch-size=32 \
+    --data-location $DATASET_DIR \
+    --checkpoint $CHECKPOINT_DIR \
+    --in-graph $PRETRAINED_MODEL \
+    --output-dir $OUTPUT_DIR \
+    --docker-image intel/intel-optimized-tensorflow:tf-r2.5-icx-b631821f \
+    --accuracy-only \
+    -- infer_option=SQuAD
+  ```
+* Accuracy
+  ```
+  python launch_benchmark.py \
+    --model-name=bert_large \
+    --precision=int8 \
+    --mode=inference \
+    --framework=tensorflow \
+    --batch-size=32 \
+    --data-location $DATASET_DIR \
+    --checkpoint $CHECKPOINT_DIR \
+    --in-graph $PRETRAINED_MODEL \
+    --output-dir $OUTPUT_DIR \
+    --docker-image intel/intel-optimized-tensorflow:tf-r2.5-icx-b631821f \
+    --accuracy-only \
+    -- infer_option=SQuAD
+  ```
+
+Output files and logs are saved to the ${OUTPUT_DIR} directory.
+
+<!-- 70. Model args -->
+Note that args specific to this model are specified after ` -- ` at
+the end of the command (like the `profile=True` arg in the Profile
+command above. Below is a list of all of the model specific args and
+their default values:
+
+| Model arg | Default value |
+|-----------|---------------|
+| doc_stride | `128` |
+| max_seq_length | `384` |
+| profile | `False` |
+| config_file | `bert_config.json` |
+| vocab_file | `vocab.txt` |
+| predict_file | `dev-v1.1.json` |
+| init_checkpoint | `model.ckpt-3649` |
+
@@ -0,0 +1,19 @@
+#
+# -*- coding: utf-8 -*-
+#
+# Copyright (c) 2021 Intel Corporation
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+# SPDX-License-Identifier: EPL-2.0
+#
@@ -0,0 +1,8 @@
+{
+    "optimization_parameters": {
+        "KMP_AFFINITY": "fine,verbose,compact,1,0",
+        "KMP_BLOCKTIME": 1,
+        "KMP_SETTINGS": 1
+    }
+}
+