for the past 1 year , the best transformers explanation video and fine tune video I have ever seen . You have the talent to make things easier to comprehensive . thank you !!!
thank you, I was involved with NLP back in the dark ages (pre-2017). your vids helped me connect the dots between what I knew and modern practice. thanks for sharing your expertise. cheers!
I like your style of representing information! Thank you for making intake into Community. We'll share this video with our ML/AI community on Discord for sure.
Can you explain the Decoder-only Transformers Training vs Inference, I saw the encoder-decoder but in decoder only we don't have the cross attention so I'm little confuesd. Thanks a lot + can you please share the excalidraw diagram it would really help also for the encoder-decoder vid, pls pls pls
Hi, why attention mask is added to the attn weights instead of multiplied (1h00:11)? if you add the attention weight with zero the weights will not be ignored
I am curious how did you run the gpt2 locally. I cloned the repo, and I added the root of transformers to the path. Then it starts to run the test code but the changes - like print statements in the original gpt2 code do not show up.
Hi, you can do that by doing pip install -e . (the -e flag is short for "editable"). See the details here: huggingface.co/docs/transformers/en/installation#editable-install
for the past 1 year , the best transformers explanation video and fine tune video I have ever seen . You have the talent to make things easier to comprehensive . thank you !!!
thank you, I was involved with NLP back in the dark ages (pre-2017). your vids helped me connect the dots between what I knew and modern practice. thanks for sharing your expertise. cheers!
Same here, did some research in 2011/2012 … wish I stayed with it now. Lol
The single video one needs to watch to understand literally every computation! Thanks a lot.
I like your style of representing information! Thank you for making intake into Community. We'll share this video with our ML/AI community on Discord for sure.
Please do more videos like this. It's amazing. Can't wait to see more🥰
Thank you so much for the in-depth explanation.
Such an amazing video! Thanks for your work! Would you mind sharing your excalidraw ?
Great video! waiting for the benefits of using past_key_values and transformer tools on fine tuning
The best explanatios , your channel is a gem ❤
Can you explain the Decoder-only Transformers Training vs Inference, I saw the encoder-decoder but in decoder only we don't have the cross attention so I'm little confuesd. Thanks a lot
+ can you please share the excalidraw diagram it would really help also for the encoder-decoder vid, pls pls pls
This video is gold standard....If can upload excalidraw diagram it would be great.
Please create more indepth videos like this on LoRA, QLoRA, RAG etc
Hi, why attention mask is added to the attn weights instead of multiplied (1h00:11)? if you add the attention weight with zero the weights will not be ignored
Thanks for your efforts, helped a lot
is there any way to get the whole graph you've drawn?😀
the best, thank u so much
Can you share the template you drew, please?
Thank you for the video. It is awesome.
Thank you so much for this video. It helped me a lot 💓
Thanks for this amazing explanation!! can you please share the draw from excalidraw ?
I am new to this, I am just trying to understand if this during Inference or Training. I guess it is during Inference. please correct me
Thanks so much for the coooool videos. I appreciate the efforts. wondering if you can share the excalidraw notes.
Are Values added to attention weights or the operation is matrix multiplication?
The attention weights are multiplied by the values, in order to produce the attention output.
Thank you for this amazing explanation - is there pr chance a way to share your diagram :)
I am curious how did you run the gpt2 locally. I cloned the repo, and I added the root of transformers to the path. Then it starts to run the test code but the changes - like print statements in the original gpt2 code do not show up.
Hi, you can do that by doing pip install -e . (the -e flag is short for "editable"). See the details here: huggingface.co/docs/transformers/en/installation#editable-install
Awesome thanks @NeilsRogge, I will try it and look at the link
Brilliant
Super Video!!! Haven't seen a better video on explaining transformers...Any chance that you could upload the excali file for us?
another banger
your vids rock
Just woderful. My search ends.