Skip to content

Commit

Permalink
update readme
Browse files Browse the repository at this point in the history
  • Loading branch information
Felix Collins committed Mar 19, 2020
1 parent 784851b commit 183d57b
Showing 1 changed file with 6 additions and 1 deletion.
7 changes: 6 additions & 1 deletion README.md
View file
Original file line number Diff line number Diff line change
@@ -1,5 +1,10 @@
## PORTING to TF2
Current progress is that the ssd300_training.py script seems to be running with TF2.2.0rc, But does not appear to use GPU.
Current progress is that the ssd300_training.py script seems to be running with TF2.2.0rc, on the GPU but something is not right because the GPU usage is very low.

There is this error:
'''WARNING:tensorflow:Gradients do not exist for variables ['conv1_1/bias:0', 'conv1_2/bias:0', 'conv2_1/bias:0', 'conv2_2/bias:0', 'conv3_1/bias:0', 'conv3_2/bias:0', 'conv3_3/bias:0', 'conv4_1/bias:0', 'conv4_2/bias:0', 'conv4_3/bias:0', 'conv5_1/bias:0', 'conv5_2/bias:0', 'conv5_3/bias:0', 'fc6/bias:0', 'fc7/bias:0', 'conv6_1/bias:0', 'conv6_2/bias:0', 'conv7_1/bias:0', 'conv7_2/bias:0', 'conv8_1/bias:0', 'conv8_2/bias:0', 'conv9_1/bias:0', 'conv4_3_norm/conv4_3_norm_gamma:0', 'conv9_2/bias:0', 'conv4_3_norm_mbox_conf/bias:0', 'fc7_mbox_conf/bias:0', 'conv6_2_mbox_conf/bias:0', 'conv7_2_mbox_conf/bias:0', 'conv8_2_mbox_conf/bias:0', 'conv9_2_mbox_conf/bias:0', 'conv4_3_norm_mbox_loc/bias:0', 'fc7_mbox_loc/bias:0', 'conv6_2_mbox_loc/bias:0', 'conv7_2_mbox_loc/bias:0', 'conv8_2_mbox_loc/bias:0', 'conv9_2_mbox_loc/bias:0'] when minimizing the loss.'''

This probably indicates that there is something wrong in the keras_loss_function.py

## SSD: Single-Shot MultiBox Detector implementation in Keras
---
Expand Down

3 comments on commit 183d57b

@Ezzysci
Copy link

@Ezzysci Ezzysci commented on 183d57b Oct 29, 2020 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@felixcollins Thank you for your work. I very much appreciate it. I am new to deep learning. I can't force anything, so I will ask if you could please get the problem fixed. I am using your implementation as the basis of my first deep learning project. I'm using the latest python, and it is not predicting anything. I will implement your updates to the 2018 code but really hope it can work well with the GPU. May you please look into fixing the keras_loss_ssd.

@jessicametzger
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hello, I am running into a similar issue when trying to train ssd300 with tf2. It happens whether I run on gpu or cpu.

Epoch 00001: LearningRateScheduler reducing learning rate to 0.001.
Epoch 1/24
WARNING:tensorflow:Gradients do not exist for variables ['conv4_3/bias:0', 'fc7/bias:0', 'conv6_2/bias:0', 'conv7_2/bias:0', 'conv8_2/bias:0', 'conv4_3_norm/conv4_3_norm_gamma:0', 'conv9_2/bias:0', 'conv4_3_norm_mbox_conf/bias:0', 'fc7_mbox_conf/bias:0', 'conv6_2_mbox_conf/bias:0', 'conv7_2_mbox_conf/bias:0', 'conv8_2_mbox_conf/bias:0', 'conv9_2_mbox_conf/bias:0'] when minimizing the loss.
WARNING:tensorflow:Gradients do not exist for variables ['conv4_3/bias:0', 'fc7/bias:0', 'conv6_2/bias:0', 'conv7_2/bias:0', 'conv8_2/bias:0', 'conv4_3_norm/conv4_3_norm_gamma:0', 'conv9_2/bias:0', 'conv4_3_norm_mbox_conf/bias:0', 'fc7_mbox_conf/bias:0', 'conv6_2_mbox_conf/bias:0', 'conv7_2_mbox_conf/bias:0', 'conv8_2_mbox_conf/bias:0', 'conv9_2_mbox_conf/bias:0'] when minimizing the loss.
WARNING:tensorflow:Gradients do not exist for variables ['conv4_3/bias:0', 'fc7/bias:0', 'conv6_2/bias:0', 'conv7_2/bias:0', 'conv8_2/bias:0', 'conv4_3_norm/conv4_3_norm_gamma:0', 'conv9_2/bias:0', 'conv4_3_norm_mbox_conf/bias:0', 'fc7_mbox_conf/bias:0', 'conv6_2_mbox_conf/bias:0', 'conv7_2_mbox_conf/bias:0', 'conv8_2_mbox_conf/bias:0', 'conv9_2_mbox_conf/bias:0'] when minimizing the loss.
WARNING:tensorflow:Gradients do not exist for variables ['conv4_3/bias:0', 'fc7/bias:0', 'conv6_2/bias:0', 'conv7_2/bias:0', 'conv8_2/bias:0', 'conv4_3_norm/conv4_3_norm_gamma:0', 'conv9_2/bias:0', 'conv4_3_norm_mbox_conf/bias:0', 'fc7_mbox_conf/bias:0', 'conv6_2_mbox_conf/bias:0', 'conv7_2_mbox_conf/bias:0', 'conv8_2_mbox_conf/bias:0', 'conv9_2_mbox_conf/bias:0'] when minimizing the loss.
  4/256 [..............................] - ETA: 4:30 - loss: 1.1389

I am opening an issue in the original ssd_keras github since issues are disabled here. If anyone finds a possible cause for this issue, please let me know.

@jessicametzger
Copy link

@jessicametzger jessicametzger commented on 183d57b Feb 4, 2021 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Here is the issue I opened. Let me know if anyone has run into the same problem training the ssd300 model with tf2 backend, or has found a solution.

Please sign in to comment.