@KGP Talkie what is the purpose of checking if the label is odd or not in align_labels_with_tokens() function? if possible could you give an example? thanks.
hello, i am new to the channel and in machine learning. I ve been watch some of your videos about NLP and read the blog . Thank you so much, it is so complete and organized. So about this videos, i kinda know that we can use spacy pipeline ( ner ) to do a ner task .And we can import the pipeline (ner) from transformers too to do it and if we don't specify the model , it 's gonna take defaulted to dbmdz/bert-large-cased-finetuned-conll03-english in my case . So why do we need to do all of this fine-tunning for NER Tagging . Can you explain me a little bit better, what makes the difference. Thank you.
Hi, Fine tuning is needed if you want to use your own dataset. For example many people uses ner in resume and cv parsing, there they need to do fine tuning. I have shown it with sample data. You can bring your own data and use this same code to fine tune.
this video is the best one that I ever seen for NLP I have never see something like this you are the best you are the one everything is explained carefully and simply to understand I will depend this on my PhD research really thanks from the deep of my heart ♥ but if I have some questions how can I keep in touch with you ... thanks in advanced
Thank you for the tutorial. It helped me a lot, but I still have a few questions after watching the video multiple times for getting better understanding. 1st question: Can I add more entities, other than the predefined ones like "LOC", "MISC", "ORG", and "PER", in my specific use case? I would like the model to be able to recognize a certain entity in my text data. 2nd question: I'm still unclear about the purpose and functionality of the "logits" variable in the compute_metrics(eval_preds) function. Could you please provide more clarification?
ans1: Yes you can add as many entities you want. You need to prepare your own dataset ans2: logits are raw prediction without any probabilities. in my latest video i have explained how logits are converted as final probabilities. Please watch it here: th-cam.com/video/ZYc9za75Chk/w-d-xo.htmlsi=FxAipl_tYqD4mXez
I appreciate your interest! Making these videos allows me to share my passion and knowledge. My motivation comes from the subscribers like you and belief that education can be exciting and transformative. While studying might feel purposeless at times, remember that learning opens doors to opportunities and personal growth. Keep exploring until you find what truly ignites your curiosity. Thanks a lot once again for watching it.
WTF? What’s the point of making videos about one of the biggest products ever created? Jesus Christ. If you have to ask, I don’t think any explanation is going to help 🤦.
@@KGPTalkie I really like different video tutorials on fine-tuning a Hugging Face language model or LLM model for a specific use. However, there are too many tutorials focused on fine-tuning OpenAI GPT-3.5-turbo or GPT-4 LLM. I hope there are more detailed and step-by-step guides on Hugging Face models or open-source LLM.
Both have their own merits. Spacy provides you a lot of additional tools for preprocessing of text data such as lemma stemming pos stopwords etc. Whereas huggingface provides a platform to use transformers for almost any task.
Hey, Great video btw. I was just wondering if it is possible to train with custom data that is not IOB formatted. I have my own dataset which I want to train it on which unfortunately does not have IOB annotations, however it does have basic annotations like brand, color, gender, trying to work on ner project for product listings.
Hahahahaha😂 “if it’s not correct that means it’s wrong.” Um….yeah that’s literally the only way reality works bro. When something isn’t correct, that’s the definition of “incorrect” 😂
It was a great and helpful video . Thank you for you help bro and keep going on .
once again thank you bro ❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤
Teşekkürler.
Thanks 🥰
@KGP Talkie what is the purpose of checking if the label is odd or not in align_labels_with_tokens() function? if possible could you give an example? thanks.
hello, i am new to the channel and in machine learning.
I ve been watch some of your videos about NLP and read the blog . Thank you so much, it is so complete and organized.
So about this videos,
i kinda know that we can use spacy pipeline ( ner ) to do a ner task .And we can import the pipeline (ner) from transformers too to do it and if we don't specify the model , it 's gonna take defaulted to dbmdz/bert-large-cased-finetuned-conll03-english in my case .
So why do we need to do all of this fine-tunning for NER Tagging . Can you explain me a little bit better, what makes the difference.
Thank you.
Hi,
Fine tuning is needed if you want to use your own dataset. For example many people uses ner in resume and cv parsing, there they need to do fine tuning. I have shown it with sample data. You can bring your own data and use this same code to fine tune.
this video is the best one that I ever seen for NLP I have never see something like this you are the best you are the one everything is explained carefully and simply to understand I will depend this on my PhD research really thanks from the deep of my heart ♥ but if I have some questions how can I keep in touch with you ... thanks in advanced
thank you so much for this video!!
You're welcome
thanks a lot. us there a playlist? wheres the part 1 link?
Thank you for the tutorial. It helped me a lot, but I still have a few questions after watching the video multiple times for getting better understanding.
1st question:
Can I add more entities, other than the predefined ones like "LOC", "MISC", "ORG", and "PER", in my specific use case? I would like the model to be able to recognize a certain entity in my text data.
2nd question:
I'm still unclear about the purpose and functionality of the "logits" variable in the compute_metrics(eval_preds) function. Could you please provide more clarification?
ans1: Yes you can add as many entities you want. You need to prepare your own dataset
ans2: logits are raw prediction without any probabilities. in my latest video i have explained how logits are converted as final probabilities. Please watch it here: th-cam.com/video/ZYc9za75Chk/w-d-xo.htmlsi=FxAipl_tYqD4mXez
@@KGPTalkie Thanks!
How can I do with an CSV data which was annotated by me?
what interests you to do all these videos? what makes you so motivated? When I study there is no purpose
++1
I appreciate your interest! Making these videos allows me to share my passion and knowledge. My motivation comes from the subscribers like you and belief that education can be exciting and transformative. While studying might feel purposeless at times, remember that learning opens doors to opportunities and personal growth. Keep exploring until you find what truly ignites your curiosity. Thanks a lot once again for watching it.
there's something called passion, and also he's making a resume for himself, and creating passive income.
WTF? What’s the point of making videos about one of the biggest products ever created? Jesus Christ. If you have to ask, I don’t think any explanation is going to help 🤦.
Thank you so much for the tutorial!!! Meanwhile, I hope u can maybe do a same tutorial by using open-source LLM model like LLAMA2.
Yes sure
@@KGPTalkie I really like different video tutorials on fine-tuning a Hugging Face language model or LLM model for a specific use. However, there are too many tutorials focused on fine-tuning OpenAI GPT-3.5-turbo or GPT-4 LLM. I hope there are more detailed and step-by-step guides on Hugging Face models or open-source LLM.
@limjuroy7078 soon enough I am bringing new huggingface course on udemy for this purpose. It's half done.
@@KGPTalkie great! I bought one of your courses 3 years ago after watching a sentiment analysis tutorial video using DistilBert and Ktrain.
Hi, im trying to annotate a dataset for bert ner, is there a specific tool i can use?
If am not late, you can use label studio
please make a end to end a ner project with streamlit api.
Ok
good
What is the difference between building a model using Huggingface and Spacy transformer?
Both have their own merits. Spacy provides you a lot of additional tools for preprocessing of text data such as lemma stemming pos stopwords etc. Whereas huggingface provides a platform to use transformers for almost any task.
Hey,
Great video btw.
I was just wondering if it is possible to train with custom data that is not IOB formatted. I have my own dataset which I want to train it on which unfortunately does not have IOB annotations, however it does have basic annotations like brand, color, gender, trying to work on ner project for product listings.
You can convert it in said format then train it. We can connect over a call to discuss. Email me at udemy@kgptalkie.com
@@KGPTalkie It would be nice to have a video for the same.
subscribed
Thanks a lot
Cool🎉🎉🎉🎉🎉
Hahahahaha😂 “if it’s not correct that means it’s wrong.” Um….yeah that’s literally the only way reality works bro. When something isn’t correct, that’s the definition of “incorrect” 😂