I've been searching through countless tutorials on fine-tuning for almost a month now, and I finally found yours. Thank you so much! Please keep making more content.
This instructional vid is something I was looking for. I am looking to do more of a rating type of app, but the sentiment analysis is a good start for that. I would like to get a copy of your free data science guide. Thanks.
Glad it was what you were looking for. For rating type of app consider using Text Classification models. And here you can get my free guide: www.maryammiradi.com/free-guide
Hi Maryam, I'm brand new to learning data science. I really enjoy your step by step implementation and explanations. I would love your free data science guide as a reference, as I continue to explore these topics. Many thanks and please keep up the great work!
@@maryammiradi Thank you Maryam! Would love if you could continue to explore Hugging Face - Time Series Analysis and compare setup/performance vs your previous tutorial on the 36 models.
@jdub8204 Do you mean building Time Series with Hugging Face or Do you mean as separate things because Hugging Face is more for LLMs NLP and Computer Vision where ScikitLearn 36!Models are suitable for Machine Learning problems. Then for Time Series you have got so many different options from Statistics, Machine Learning to Deep Learning up to the latest LLMs that can be used for Time series
@@maryammiradi Wow thanks for the quick replies Maryam. There's so many models/technologies - that I'm confusing myself and their application. In particular, I'm interested in time series forecasting - and predicting behavior based on historical data. Since I'm new, would you recommend focusing on one particular model (eg XGBoost?) Thank you
hello I'm from India, Thanks for the clear explanation could you please make more videos on finetuning as many youtuber have not covered it (clear explanation) , try to make it for language translation it would be great if you make it , Thanks in advance I loved the way you explained the things. little suggestion- Try to reduce the size of the face camera
Thank you! Absolutely! Have you seen my other video from my LLM series. I am explaining LLM ladder & Prompt tuning & Prefix tuning 👋 ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-HkZOGGvZzg4.htmlsi=kxHtTkCfM51eUEGE
Hi Maryam. Really loving your videos. I am seen your last video on time series and really found it helpful. I am exploring the opportunity of Temporal Fusion Transfers on Time Series Forecasting.
@asifurrahman2111 TFTs can forecast multiple time steps into the future, making them great for various applications. In which industry are you applying Time Series?
@@maryammiradi Well I am planning to use TFTs in the field of hydrology. As the data comes from multiple sources and exhibits significant variation, I thought that the TFT model will be ideal.
When I run the trainer.train() syntax at Step #6 (12:20') the following error popped out, what is the problem? All the codes before that line works good. Here is the error: NameError Traceback (most recent call last) in () 1 # Evaluate the model ----> 2 result = trainer.evaluate() 3 print(result) NameError: name 'trainer' is not defined
The Trainer should be applied to your specific model and then you can run the evaluation. trainer = Trainer( model, args, train_dataset=encoded_dataset[“train”], eval_dataset=encoded_dataset[“validation”], tokenizer=tokenizer, compute_metrics=compute_metrics ) trainer.evaluate() trainer.predict(encoded_dataset[“test”])
If you're using Hugging Face's hosted API services, ensure that data is transmitted over secure channels. Hugging Face models are pre-trained and may not be aware of the sensitive nature of your data. Using models locally rather than via API can mitigate this risk.
@@julienduchesneau9747 Would you let me know which part is not clear tp you? To learn NLP and LLMs you need to have some foundation in AI and ML. What is your field of study? Hopefully I can help you with more resources
@@maryammiradi I never used hunging face yet because I don't have a suitable GPU (to try local LLM) so I am lost right at the beggining, I dont even know what do you use to type on. I am just learning for myself, from scratch. With the help of ChatGPT I did a chatbot on my twitch account using Python in VScode. It is super basic because also super cheap. So I watch several youtube video about AI to continu learning and improve my chatBot. So yeah thanks for the reply but you don't have to, I have sub to your channel so I will come back later when I feel I need it.
@julienduchesneau9747 The most important thing that Hugging Face offers is not their local LLMs but their Python library transforms. This library allows you to so any kind of NLP task that you want just with importing this library and then just use their pipeline function. My code is in Google colab. It is notebook with link in the description. What you would learn from this video is not only Hugging face but also what kind of NLP tasks are possible and for which purposes you can use LLMs, like summarization and Sentiment Analysis. Let me know if you have other questions. 👋
@@maryammiradi oh wow really??!!! ok I will definetly watch it again and also search about notebook. Sometime there is so much stuff to learn, especially for a begginer, github ananconda langchain streamlit... AND everthing goes so fast in AI, it is crazy how each month there is new cool stuff. Thanks for the reply also it is really helpfull
@@maryammiradi oh wow really??!!! I will definetly watch it again then and take a look about notebook, another thing that I dont use already but probably a must. So much thing to look into when we begin, Github, anaconda, langchain, streamlit, RAG. AI is so cool tho I am not tired learning about it and it is so crazy to see how things advance month after month. Thanks for the reply I am pretty sure it will help my chatbot a lot