I didn't go through this video till today. But I was sitting on my work machine and I saw this video and cleared up my doubts. Commenting as i can't comment from my work laptop. Great video sir.
Thank you for making me aware of this library. The alternative I used was to name variables with dimension postfix like "letter_w_N_M_e = letter_sne.reshape(w,N,M,512) ". This library can make this even more clear and offers the opportunity to combine multiple lines.
Based on the activity and their packaged release they have not published since 2021 (github.com/facebookresearch/vissl/releases). May not be a good idea to continue with the libraries that have stopped or stale development. I was looking into SSL last year and had kind of liked these two. See if they are helpful github.com/vturrisi/solo-learn github.com/lightly-ai/lightly
@@KapilSachdeva I can't figure out how to do a stack or concatenate operation on two matrices having different shape(having either equal rows or column but not both) using einops.rearrange. Only seems to be working with same shape arrays.
If you can create an example of these matrices and what result you want to see then may be I can suggest how to do it using einops api. You can use Google colab to create the example and then share the public link with me.
I tried it you can't use rearrange for this use case. What you want to use is einops.pack Here is a colab notebook with an example of einops.pack for your tensors. colab.research.google.com/drive/1Tsfqwl52me1pvfn2V6H_zE4q7buZKTRF?usp=sharing See the doc as well einops.rocks/4-pack-and-unpack/#short-summary-so-far
Thank you, the tutorial helped me a lot to get started with Einops.
I didn't go through this video till today. But I was sitting on my work machine and I saw this video and cleared up my doubts.
Commenting as i can't comment from my work laptop. Great video sir.
🙏
Thanks a lot this was very informative.
my man, you've got a killer voice by the way😉
🙏
Thank you for making me aware of this library. The alternative I used was to name variables with dimension postfix like "letter_w_N_M_e = letter_sne.reshape(w,N,M,512) ". This library can make this even more clear and offers the opportunity to combine multiple lines.
This library is very good. There are other functions like repeat, pack and unpack that are also good so check them out as well.
@@KapilSachdeva I will, thank you
Perfect introduction to Einops! Thanks
🙏
this is called perfect teaching.. 😊 enlightened after watching
can you please touch optical flow estimation part as well?
Thanks Sir!
🙏
Einops is OP!
Damn, einops is nice.
Thank you so much sir it is a great video 👌👌🙌🙌
🙏
You are genius please provide an end implementation pytorch of famous models using the available modules.
🙏
Can you make a video on VISSL library?
Based on the activity and their packaged release they have not published since 2021 (github.com/facebookresearch/vissl/releases). May not be a good idea to continue with the libraries that have stopped or stale development.
I was looking into SSL last year and had kind of liked these two. See if they are helpful
github.com/vturrisi/solo-learn
github.com/lightly-ai/lightly
You has really guts to tell the truth, the permute、reshape、view、squeeze、unsqueeze、np.repeat、....all of these forming a mess not readible
🙏 … yeah, code quality and readability is a very important aspect.
what is the difference between einsum and einops?
Many differences.
openreview.net/pdf?id=oapKSVM2bcj
See the appendix B of the einops paper where a detailed comparison is provided
@@KapilSachdeva I can't figure out how to do a stack or concatenate operation on two matrices having different shape(having either equal rows or column but not both) using einops.rearrange. Only seems to be working with same shape arrays.
If you can create an example of these matrices and what result you want to see then may be I can suggest how to do it using einops api. You can use Google colab to create the example and then share the public link with me.
I saw a comment notification with a colab link but the comment has been removed (most likely by youtube). Can you try to again?
I tried it you can't use rearrange for this use case. What you want to use is einops.pack
Here is a colab notebook with an example of einops.pack for your tensors.
colab.research.google.com/drive/1Tsfqwl52me1pvfn2V6H_zE4q7buZKTRF?usp=sharing
See the doc as well
einops.rocks/4-pack-and-unpack/#short-summary-so-far