error First step cannot be zero when running train.py #51
Comments
|
I have the same error.Do you find how to solve it? |
|
yes, edit this in your config file in ...\models\research\object_detection\training |
|
Thank you. It work here :)
…On Mon, 4 Jun 2561 at 16:52 blockhunts ***@***.***> wrote:
yes, edit this in your config file in
...\models\research\object_detection\training
train_config: {
batch_size: 1
optimizer {
momentum_optimizer: {
learning_rate: {
manual_step_learning_rate {
initial_learning_rate: 0.0002
schedule {
step: 900000
learning_rate: .00002
}
schedule {
step: 1200000
learning_rate: .000002
}
}
}
momentum_optimizer_value: 0.9
}
use_moving_average: false
}
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#51 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AmB6BPsiQVe1w3fV6oSj5jUW-ATlE7pEks5t5QNIgaJpZM4UPjDR>
.
|
|
if you download model from the github repository files are up to date |
|
I ran into this same error while using the AWS DL AMI (Deep Learning AMI (Ubuntu) Version 10.0 (ami-23c4fb46)) and following, as far as I can tell, the same steps I used on Windows with obvious substitutions since this AMI is Ubuntu. Both Ubuntu and Windows are using TF 1.8. But when I use the train_config that blockhunts mentioned I get: Any ideas? |
|
I see that epratheeban has the solution to my problem mentioned here #11: It's easy. Go to the utils folder. Find the learning_schedules.py file. Go to the line 167. And replace the line 167 with below rate_index = tf.reduce_max(tf.where(tf.greater_equal(global_step, boundaries), |
|
Hi @jim-meyer WARNING:tensorflow:From C:\Users\sadegh\Anaconda3\envs\tensorflow1\lib\site-pack Future major versions of TensorFlow will allow gradients to flow See @{tf.nn.softmax_cross_entropy_with_logits_v2}. Traceback (most recent call last): |
|
TypeError: Cannot convert a list containing a tensor of dtype <dtype: 'int32'> to <dtype: 'float32'> (Tensor is: <tf.Tensor 'Preprocessor/stack_1:0' shape=(1, 3) dtype=int32>) |
|
@tamizharasank what file ? this kind of error copy it in google you will find the fix easily |
|
@tamizharasank did you solve this error? I got the same error, any suggesstions? |
|
After making changes in configure file in training folder I got this error: (tensorflow1) C:\tensorflow1\models\research\object_detection>python train.py --logtostderr --train_dir=training/ --pipeline_config_path=training/faster_rcnn_inception_v2_pets.config Future major versions of TensorFlow will allow gradients to flow See @{tf.nn.softmax_cross_entropy_with_logits_v2}. C:\Users\kayka\Anaconda3\envs\tensorflow1\lib\site-packages\tensorflow\python\ops\gradients_impl.py:108: UserWarning: Converting sparse IndexedSlices to a dense Tensor of unknown shape. This may consume a large amount of memory. |
|
Looks like you probably did not follow all of the steps in 2a, "Download TensorFlow Object Detection API repository from GitHub" and/or 2b, "Download the Faster-RCNN-Inception-V2-COCO model from TensorFlow's model zoo". Try following those steps again exactly and that should fix your problem. |
|
File "C:\tensorflow1\models\research\object_detection\utils\learning_schedules.py", line 160, in manual_stepping i edit the file and save it and when i train it again it's return to it's original value |
|
I'm getting below error while i was trying to run: WARNING: The TensorFlow contrib module will not be included in TensorFlow 2.0.
WARNING:tensorflow:From C:\Users\Asus\Miniconda3\lib\site-packages\tensorflow\python\platform\app.py:125: main (from main) is deprecated and will be removed in a future version. Traceback (most recent call last): |
|
@ShubhranshuMaurya that error seems to indicate that there is something wrong with C:\Tensorflow\workspace raining_demonnotations/label_map.pbtxt. Have you opened that file in a text editor to see if it looks right? That file file should look something like this: item { IIRC this file could also be a binary protobuf file in which case viewing it in a text editor won't tell you much. But if it appears to be binary perhaps you could try creating a text version with your training labels and see if that works. |
|
#tessor flow custom training ERROR:raise ValueError('First step cannot be zero.') ValueError: First step cannot be zero. SOLUTION: object_detection\training\ .config train_config: { |
|
For me it worked with 'step: 1' |
Did you find a solution? |
can you explain what is happening in learning rate?, what does the both step size signify in manual learning rate and also what is initial learning rate? |
|
python train.py --logtostderr -train_dir=training/ --pipeline_config_path=training/ssd_mobilenet_v2_quantized_300x300_coco.config Current thread 0x00005734 (most recent call first): help |
i tried to use the same images (card) provided, i just delete all the processed file (csv,dll) and follow all the step.
And when i tried to issue python train.py
I got this error
Any clues why this happen?
The text was updated successfully, but these errors were encountered: