Skip to content

Commit f4625a2

Browse files
pribaltapribalta
committed
Update checkpoint MaskRCNN
Signed-off-by: Pablo Ribalta <pribalta@nvidia.com>
1 parent 9ba22d1 commit f4625a2

10 files changed

Lines changed: 19 additions & 25 deletions

‎TensorFlow2/Segmentation/MaskRCNN/download_and_process_pretrained_weights.sh‎

Lines changed: 10 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -14,7 +14,7 @@
1414
# See the License for the specific language governing permissions and
1515
# limitations under the License.
1616

17-
#mkdir -p weights/
17+
mkdir -p /model
1818
cd /model
1919

2020
# DOWNLOAD CHECKPOINTS
@@ -29,7 +29,7 @@ wget -N ${BASE_URL}/variables/variables.data-00000-of-00001 -P ${DEST_DIR}/varia
2929
wget -N ${BASE_URL}/variables/variables.index -P ${DEST_DIR}/variables
3030

3131
## ====================== resnet-nhwc-2018-02-07 ====================== ##
32-
BASE_URL="https://storage.googleapis.com/cloud-tpu-artifacts/resnet/resnet-nhwc-2018-02-07"
32+
BASE_URL="https://storage.googleapis.com/cloud-tpu-checkpoints/retinanet/resnet50-checkpoint-2018-02-07"
3333
DEST_DIR="resnet/resnet-nhwc-2018-02-07"
3434

3535
wget -N ${BASE_URL}/checkpoint -P ${DEST_DIR}
@@ -38,12 +38,12 @@ wget -N ${BASE_URL}/model.ckpt-112603.index -P ${DEST_DIR}
3838
wget -N ${BASE_URL}/model.ckpt-112603.meta -P ${DEST_DIR}
3939

4040
## ====================== resnet-nhwc-2018-10-14 ====================== ##
41-
BASE_URL="https://storage.googleapis.com/cloud-tpu-artifacts/resnet/resnet-nhwc-2018-10-14"
42-
DEST_DIR="resnet/resnet-nhwc-2018-10-14"
43-
44-
wget -N ${BASE_URL}/model.ckpt-112602.data-00000-of-00001 -P ${DEST_DIR}
45-
wget -N ${BASE_URL}/model.ckpt-112602.index -P ${DEST_DIR}
46-
wget -N ${BASE_URL}/model.ckpt-112602.meta -P ${DEST_DIR}
41+
#BASE_URL="https://storage.googleapis.com/cloud-tpu-artifacts/resnet/resnet-nhwc-2018-10-14"
42+
#DEST_DIR="resnet/resnet-nhwc-2018-10-14"
43+
#
44+
#wget -N ${BASE_URL}/model.ckpt-112602.data-00000-of-00001 -P ${DEST_DIR}
45+
#wget -N ${BASE_URL}/model.ckpt-112602.index -P ${DEST_DIR}
46+
#wget -N ${BASE_URL}/model.ckpt-112602.meta -P ${DEST_DIR}
4747

4848
# VERIFY CHECKPOINTS
4949
echo "Verifying and Processing Checkpoints..."
@@ -64,8 +64,8 @@ python inspect_checkpoint.py --file_name=mask-rcnn/1555659850/ckpt/model.ckpt \
6464
python inspect_checkpoint.py --file_name=resnet/resnet-nhwc-2018-02-07/model.ckpt-112603 \
6565
> resnet/resnet-nhwc-2018-02-07/tensors_and_shape.txt
6666

67-
python inspect_checkpoint.py --file_name=resnet/resnet-nhwc-2018-10-14/model.ckpt-112602 \
68-
> resnet/resnet-nhwc-2018-10-14/tensors_and_shape.txt
67+
#python inspect_checkpoint.py --file_name=resnet/resnet-nhwc-2018-10-14/model.ckpt-112602 \
68+
# > resnet/resnet-nhwc-2018-10-14/tensors_and_shape.txt
6969

7070
python inspect_checkpoint.py --file_name=resnet/extracted_from_maskrcnn/resnet50.ckpt \
7171
> resnet/extracted_from_maskrcnn/tensors_and_shape.txt

‎TensorFlow2/Segmentation/MaskRCNN/scripts/benchmark_inference.py‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -36,7 +36,7 @@ def main():
3636
f'python {main_path}'
3737
f' --mode eval'
3838
f' --model_dir "{flags.model_dir}"'
39-
f' --checkpoint "{os.path.join(flags.weights_dir, "resnet/resnet-nhwc-2018-10-14/model.ckpt-112602")}"'
39+
f' --checkpoint "{os.path.join(flags.weights_dir, "resnet/resnet-nhwc-2018-02-07/model.ckpt-112603")}"'
4040
f' --validation_file_pattern "{os.path.join(flags.data_dir, "val*.tfrecord")}"'
4141
f' --val_json_file "{os.path.join(flags.data_dir, "annotations/instances_val2017.json")}"'
4242
f' --num_steps_per_eval 200'

‎TensorFlow2/Segmentation/MaskRCNN/scripts/benchmark_training.py‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -38,7 +38,7 @@ def main():
3838
f'python {main_path}'
3939
f' --mode train'
4040
f' --model_dir "{flags.model_dir}"'
41-
f' --checkpoint "{os.path.join(flags.weights_dir, "resnet/resnet-nhwc-2018-10-14/model.ckpt-112602")}"'
41+
f' --checkpoint "{os.path.join(flags.weights_dir, "resnet/resnet-nhwc-2018-02-07/model.ckpt-112603")}"'
4242
f' --training_file_pattern "{os.path.join(flags.data_dir, "train*.tfrecord")}"'
4343
f' --init_learning_rate 0.04'
4444
f' --total_steps 200'

‎TensorFlow2/Segmentation/MaskRCNN/scripts/train_1GPU.sh‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -22,7 +22,7 @@ export CUDA_VISIBLE_DEVICES=0
2222

2323
python ${BASEDIR}/../mask_rcnn_main.py \
2424
--mode="train_and_eval" \
25-
--checkpoint="/model/resnet/resnet-nhwc-2018-10-14/model.ckpt-112602" \
25+
--checkpoint="/model/resnet/resnet-nhwc-2018-02-07/model.ckpt-112603" \
2626
--eval_samples=5000 \
2727
--init_learning_rate=0.005 \
2828
--learning_rate_steps="240000,320000" \

‎TensorFlow2/Segmentation/MaskRCNN/scripts/train_4GPU.sh‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -32,7 +32,7 @@ mpirun \
3232
--allow-run-as-root \
3333
python ${BASEDIR}/../mask_rcnn_main.py \
3434
--mode="train_and_eval" \
35-
--checkpoint="/model/resnet/resnet-nhwc-2018-10-14/model.ckpt-112602" \
35+
--checkpoint="/model/resnet/resnet-nhwc-2018-02-07/model.ckpt-112603" \
3636
--eval_samples=5000 \
3737
--init_learning_rate=0.02 \
3838
--learning_rate_steps="60000,80000" \

‎TensorFlow2/Segmentation/MaskRCNN/scripts/train_8GPU.sh‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -30,7 +30,7 @@ mpirun \
3030
--allow-run-as-root \
3131
python ${BASEDIR}/../mask_rcnn_main.py \
3232
--mode="train_and_eval" \
33-
--checkpoint="/model/resnet/resnet-nhwc-2018-10-14/model.ckpt-112602" \
33+
--checkpoint="/model/resnet/resnet-nhwc-2018-02-07/model.ckpt-112603" \
3434
--eval_samples=5000 \
3535
--init_learning_rate=0.04 \
3636
--learning_rate_steps="30000,40000" \

‎TensorFlow2/Segmentation/MaskRCNN/scripts/train_AMP_1GPU.sh‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -22,7 +22,7 @@ export CUDA_VISIBLE_DEVICES=0
2222

2323
python ${BASEDIR}/../mask_rcnn_main.py \
2424
--mode="train_and_eval" \
25-
--checkpoint="/model/resnet/resnet-nhwc-2018-10-14/model.ckpt-112602" \
25+
--checkpoint="/model/resnet/resnet-nhwc-2018-02-07/model.ckpt-112603" \
2626
--eval_samples=5000 \
2727
--init_learning_rate=0.005 \
2828
--learning_rate_steps="240000,320000" \

‎TensorFlow2/Segmentation/MaskRCNN/scripts/train_AMP_4GPU.sh‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -32,7 +32,7 @@ mpirun \
3232
--allow-run-as-root \
3333
python ${BASEDIR}/../mask_rcnn_main.py \
3434
--mode="train_and_eval" \
35-
--checkpoint="/model/resnet/resnet-nhwc-2018-10-14/model.ckpt-112602" \
35+
--checkpoint="/model/resnet/resnet-nhwc-2018-02-07/model.ckpt-112603" \
3636
--eval_samples=5000 \
3737
--init_learning_rate=0.02 \
3838
--learning_rate_steps="60000,80000" \

‎TensorFlow2/Segmentation/MaskRCNN/scripts/train_AMP_8GPU.sh‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -30,7 +30,7 @@ mpirun \
3030
--allow-run-as-root \
3131
python ${BASEDIR}/../mask_rcnn_main.py \
3232
--mode="train_and_eval" \
33-
--checkpoint="/model/resnet/resnet-nhwc-2018-10-14/model.ckpt-112602" \
33+
--checkpoint="/model/resnet/resnet-nhwc-2018-02-07/model.ckpt-112603" \
3434
--eval_samples=5000 \
3535
--init_learning_rate=0.04 \
3636
--learning_rate_steps="30000,40000" \

‎TensorFlow2/Segmentation/MaskRCNN/weights/extract_RN50_weights.py‎

Lines changed: 1 addition & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -20,7 +20,6 @@
2020
import getopt
2121
import logging
2222
import tensorflow as tf
23-
from distutils.version import LooseVersion
2423

2524
"""
2625
python weights/extract_RN50_weights.py \
@@ -45,12 +44,7 @@ def rename(checkpoint_dir, save_to, dry_run, verbose):
4544

4645
total_vars_loaded = 0
4746

48-
if LooseVersion(tf.__version__) < LooseVersion("2.0.0"):
49-
file_list = tf.contrib.framework.list_variables(checkpoint_dir)
50-
else:
51-
file_list = tf.train.list_variables(checkpoint_dir)
52-
53-
for var_name, _ in file_list:
47+
for var_name, _ in tf.train.list_variables(checkpoint_dir):
5448

5549
if "resnet50" in var_name:
5650
# Load the variable

0 commit comments

Comments
 (0)