I was inspired and learned the basics of TensorFlow after I completed the TensorFlow specialization on coursera. Personally I think these videos I created give a similar understanding but if you wanna check it out you can. Below you'll find both affiliate and non-affiliate links, the pricing for you is the same but a small commission goes back to the channel if you buy it through the affiliate link which helps me create more future videos. affiliate: bit.ly/3JyvdVK non-affiliate: bit.ly/3qtrK39 Comment and time stamps: 0:00 - Introduction and overview 5:52 - Starter Code 8:40 - TensorBoard Callbacks 14:02 - Plots in Scalars Tab 22:00 - Visualizing Images 33:40 - Confusion Matrix 45:24 - Graphs 49:36 - HParams 1:01:12 - Projector 1:15:59 - TensorFlow Profiler 1:22:30 - Ending & Outro I will add the code to Github soon and add link to it in the description. In the video I showed how to use the projector for visualizing images, which is much more difficult than to use it for text (and quite similar). If you want to see how to do it for text, TensorFlow official tutorial is quite good: www.tensorflow.org/tensorboard/tensorboard_projector_plugin. Mistakes Found In Video: 1:30 - Here I phrased it incorrectly, what I meant to say was that it misclassifies an airplane as a bird in 7% of the cases.
Yeah unfortunately the docs can something be a bit lacking in my opinion, actually sometimes they don't exist. And so you're only option is going through the official tutorials which are amazing most of the time and also what these videos are based off. But it can quickly be overwhelming to go through all of those since information is quite entangled between all of those tutorials. This was my way of going through those and compressing it into something that is watchable for people who prefer that instead of reading and an attempt to make things more clear
Hi Alladin, thanks for a great tutorial I was wondering if we could initialize the model weight with a fixed seed, to ensure the difference in performance is only due to change in hyper parameters and not change in initial weights.
This is the best tutorial on Tensorboard, Thank you so much. One question though, when running the projector its stuck on "Fetching the Sprite Images" (it keeps on loading forever) is there a solution to this? Thank you
Hi all, I am looking for a way to in cooperate the TensorFlow profiler into a custom training loop. At the end of the video he said, he'd link some resources which I am not seeing at the moment. In general it is hard to find any advice about this on the internet. Any help would be greatly appreciated.
this may be a dumb question but what are those distribution how to gain any info from that distributions and where can i learn about those distributiion
Hi! Is it possible to run Tensorboard during prediction? I've been trying to use a Tensorboard callback in a prediction call but when I try to see the logs I just receive a "GRAPH VISUALIZATION FAILED - Malformed GraphDef" message...
The number of hyperparameters used is more than 7 but when trying to visualize through tensorboard, I'm not able to view all the hyperparameters that are used. Please provide a solution to this problem
I was inspired and learned the basics of TensorFlow after I completed the TensorFlow specialization on coursera. Personally I think these videos I created give a similar understanding but if you wanna check it out you can. Below you'll find both affiliate and non-affiliate links, the pricing for you is the same but a small commission goes back to the channel if you buy it through the affiliate link which helps me create more future videos.
affiliate: bit.ly/3JyvdVK
non-affiliate: bit.ly/3qtrK39
Comment and time stamps:
0:00 - Introduction and overview
5:52 - Starter Code
8:40 - TensorBoard Callbacks
14:02 - Plots in Scalars Tab
22:00 - Visualizing Images
33:40 - Confusion Matrix
45:24 - Graphs
49:36 - HParams
1:01:12 - Projector
1:15:59 - TensorFlow Profiler
1:22:30 - Ending & Outro
I will add the code to Github soon and add link to it in the description. In the video I showed how to use the projector for visualizing images, which is much more difficult than to use it for text (and quite similar). If you want to see how to do it for text, TensorFlow official tutorial is quite good: www.tensorflow.org/tensorboard/tensorboard_projector_plugin.
Mistakes Found In Video:
1:30 - Here I phrased it incorrectly, what I meant to say was that it misclassifies an airplane as a bird in 7% of the cases.
Hands down to the best video on tensorboard by far.... even google's own tutorials on tensorboard ain't this good.
There is no best tutorial on tensor board in youtube...
Finally this video came thank you
Thank you :)
Excellent tutorial, I'm leaving a comment so the YT algorithm sees some traffic and hopefully recommends it to more people.
Same.
Alladin, thanks for a great tutorial,Thank you so so soooo much, from Iran
This is exactly what I needed. Thank you so much!
the best tutorial on tensorboard. very good explanations. thank you very much.
Google’s documentation and TF 2.0 rollout has been abysmal but it’s such a good library. Thank you for making these videos!!!
Yeah unfortunately the docs can something be a bit lacking in my opinion, actually sometimes they don't exist. And so you're only option is going through the official tutorials which are amazing most of the time and also what these videos are based off. But it can quickly be overwhelming to go through all of those since information is quite entangled between all of those tutorials. This was my way of going through those and compressing it into something that is watchable for people who prefer that instead of reading and an attempt to make things more clear
This is incredibly useful likes AP.
Great video! I wish your playlist existed when I started working with TF, it'd be even more helpful than it still is!
Hi Alladin, thanks for a great tutorial
I was wondering if we could initialize the model weight with a fixed seed, to ensure the difference in performance is only due to change in hyper parameters and not change in initial weights.
This is the best tutorial on Tensorboard, Thank you so much. One question though, when running the projector its stuck on "Fetching the Sprite Images" (it keeps on loading forever) is there a solution to this?
Thank you
Awesome video, good job and thanks!
nice video, thank you.
Hi Aladdin, Can we use these plots/projectors without writing the custom training loop? E.g. in model.fit() or using Callbacks ?
I tried to look into that but I didn't find any answers to that unfortunately
Hi all, I am looking for a way to in cooperate the TensorFlow profiler into a custom training loop. At the end of the video he said, he'd link some resources which I am not seeing at the moment. In general it is hard to find any advice about this on the internet.
Any help would be greatly appreciated.
can you talk about how to interpret graphs and how to improve model performance?
just thank you
1:19:07 The host is CPU
whereas the device is GPU
this may be a dumb question but what are those distribution how to gain any info from that distributions and where can i learn about those distributiion
i get this error on the tensorboard site
"No dashboards are active for the current data set."
how do i fix this
Hi!
Is it possible to run Tensorboard during prediction?
I've been trying to use a Tensorboard callback in a prediction call but when I try to see the logs I just receive a "GRAPH VISUALIZATION FAILED - Malformed GraphDef" message...
In Visualizing Images, I am getting dll error for importing cv2 while calling img_grid from utils. Can you help me out?
Please how do I delete logs to prevent overwriting of the graphs.
How to get maximum accuracy if we saved as tensorboard file?
How do you use tensorboard if your data is saved in a folder on your computer?
Hello, i have an error ModuleNotFoundError: No module named 'utils'. Сan you please tell me how to solve this problem?
Is there a reason why you haven't upgraded to TF 2.3?
The installation process for windows felt convoluted and unnecessary complicated. I installed it with conda but it only ships with 2.1.
The number of hyperparameters used is more than 7 but when trying to visualize through tensorboard, I'm not able to view all the hyperparameters that are used. Please provide a solution to this problem
how to run Tensorboard on google colaboratory kindly reply.
plz update your code overall
I have a couple videos left then I will spend time making well commented code and update Github page.