I used this command ngc registry model list nvidia/iva/* to view all model available.
Used to be working.
But today,view again and see only resnet_18 classification model. What could be wrong?
TLT 2.0 is coming soon. Please stay tuned.
Do I need to update my current tlt in my system? or just api changed only?
No changing if you still use 1.0.1 docker.
But for coming TLT 2.0 docker, you need to download the docker and its corresponding new pretrained models.
I run
ngc registry model list nvidia/tlt*. That is all I see. Not as complete as the previous version.
+------------------+--------------------+----------------+------------------+-------------------+-----------+---------------+------------+
| Name | Repository | Latest Version | Application | Framework | Precision | Last Modified | Permission |
+------------------+--------------------+----------------+------------------+-------------------+-----------+---------------+------------+
| DashCamNet | nvidia/tlt_dashcam | | Object Detection | Transfer Learning | FP32 | Apr 29, 2020 | unlocked |
| | net | | | Toolkit | | | |
| FaceDetectIR | nvidia/tlt_facedet | | Object Detection | Transfer Learning | FP32 | Apr 30, 2020 | unlocked |
| | ectir | | | Toolkit | | | |
| PeopleNet | nvidia/tlt_peoplen | | Object Detection | Transfer Learning | FP32 | Apr 30, 2020 | unlocked |
| | et | | | Toolkit | | | |
| TLT Image | nvidia/tlt_pretrai | | Classification | Transfer Learning | FP32 | Apr 30, 2020 | unlocked |
| Classification | ned_classification | | | Toolkit | | | |
| TLT DetectNet V2 | nvidia/tlt_pretrai | | Object Detection | Transfer Learning | FP32 | Apr 30, 2020 | unlocked |
| Detection | ned_detectnet_v2 | | | Toolkit | | | |
| TLT Object | nvidia/tlt_pretrai | | Object Detection | Transfer Learning | FP32 | Apr 30, 2020 | unlocked |
| Detection | ned_object_detecti | | | Toolkit | | | |
| | on | | | | | | |
| TrafficCamNet | nvidia/tlt_traffic | | Object Detection | Transfer Learning | FP32 | Apr 29, 2020 | unlocked |
| | camnet | | | Toolkit | | | |
| VehicleMakeNet | nvidia/tlt_vehicle | | Classification | Transfer Learning | FP32 | Apr 29, 2020 | unlocked |
| | makenet | | | Toolkit | | | |
| VehicleTypeNet | nvidia/tlt_vehicle | | Classification | Transfer Learning | FP32 | Apr 29, 2020 | unlocked |
| | typenet | | | Toolkit | | | |
+------------------+--------------------+----------------+------------------+-------------------+-----------+---------------+------------+
Then this command
ngc registry model download-version nvidia/tlt_pretrained_detectnet_v2
doesn’t download the pretrained.
Downloaded resnet18.hdf5 and when I train, I have error as
WARNING:tensorflow:From /usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.
Instructions for updating:
Colocations handled automatically by placer.
2020-04-30 07:18:57,369 [WARNING] tensorflow: From /usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.
Instructions for updating:
Colocations handled automatically by placer.
WARNING:tensorflow:From /usr/local/lib/python2.7/dist-packages/horovod/tensorflow/__init__.py:91: div (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.
Instructions for updating:
Deprecated in favor of operator or tf.math.divide.
2020-04-30 07:18:57,386 [WARNING] tensorflow: From /usr/local/lib/python2.7/dist-packages/horovod/tensorflow/__init__.py:91: div (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.
Instructions for updating:
Deprecated in favor of operator or tf.math.divide.
Traceback (most recent call last):
File "/usr/local/bin/tlt-train-g1", line 8, in <module>
sys.exit(main())
File "./common/magnet_train.py", line 37, in main
File "</usr/local/lib/python2.7/dist-packages/decorator.pyc:decorator-gen-2>", line 2, in main
File "./detectnet_v2/utilities/timer.py", line 46, in wrapped_fn
File "./detectnet_v2/scripts/train.py", line 633, in main
File "./detectnet_v2/scripts/train.py", line 557, in run_experiment
File "./detectnet_v2/scripts/train.py", line 443, in train_gridbox
File "./detectnet_v2/scripts/train.py", line 259, in build_gridbox_model
File "./detectnet_v2/model/detectnet_model.py", line 291, in load_model_weights
TypeError: 'NoneType' object has no attribute '__getitem__'
My specs file is
detectnet_v2_retrain_resnet18_kitti.log (4.1 KB)
What could be wrong?
Hi batu,
I want to confirm one thing. Now, you are running at 1.0.1 docker, but your ngc pre-trained model is not downloaded previously,right?
I am not sure if the ngc pretrained models available now are only compatible with 2.0 docker.
I need to check with internal team.
yes I am using 1.0.1 docker and I used pretrained model downloaded using this command ngc registry model list nvidia/iva/*. I used a lot of pretrained models from ngc for classification to object detection, like frcnn, ssd. Used to be fine using this command ngc registry model list nvidia/iva/*. Just today only trying to test detectnet and can’t download using the command. I followed the wget command you suggested, so can download. But can’t train and got that error. Please let me know if I need to use docker 2.0.
Please update to 2.0 docker and retry. TLT 2.0 is released.