4 ай бұрын
8 ай бұрын
Жыл бұрын
Пікірлер
@MultiSciGeek 2 күн бұрын
Bro speak slowly plz
@kritisaxena2709 14 күн бұрын
You made it so easy!
@CarstenMaul 22 күн бұрын
kzbin.info/www/bejne/jWqcZmZrZ8Seodk
@Tushii 27 күн бұрын
Hi, can i use this to run phi-3 like so as well?
@ml-explained 27 күн бұрын
Sure if ollama supports phi3 then yes, you just need to swap the model name
@Tushii 26 күн бұрын
@@ml-explained it does, i'll try it out
@GotoRing0 29 күн бұрын
Why didn't you pick openELM 3B to compare? It would be still smaller than PHI-3. Here we saw how 4 times difference in size looks like :-(
@ml-explained 28 күн бұрын
There were some issues in loading 3B model. I'll do some more comparison videos with 3B model as well later.
@ControlProblem Ай бұрын
Good video! But the pronunciation is like "Feye" as in "Fight". Not like "Fee". But overall the content was helpful.
@KarlLew Ай бұрын
Alpha Phi objects to this comment
@erikjohnson9112 Ай бұрын
The bat & ball question results in a range, not a single answer. The bat could be 1.10 and the ball is free (0.0). And the upper range is 1.00 for bat and 0.10 for ball (also valid). Any value in between is also valid, so there is not a single answer without more constraints in the question. UPDATE: I was wrong, I did not maintain the 1.0 difference.
@ml-explained Ай бұрын
I'll check on it, but it's more to see the approach that LLMs take.
@13thxenos 25 күн бұрын
I think you misunderstood the question. There are two values that we know the sum of, and their difference. You can only have 1 valid answer or no answer at all if you consider real numbers, not a range. In your examples, if you put the price of the bat 1.10 and then the ball could not be free, because then their difference is not 1, it would be 1.10. If you put 1 for the bat, you can't have 0.10 for the ball because then their difference is not 1, it is 0.90. So both your answers are not valid ones based on the constraints of the question.
@erikjohnson9112 25 күн бұрын
@@13thxenos You are correct, I forgot to maintain 1.0 difference. A simple system of 2 equations, b-a=1.0 & b+a=1.10 will produce the correct answer when added together to produce b (then substitute for a). My sloppy natural intelligence failed. :)
@erikjohnson9112 25 күн бұрын
I'm leaving my original comment with a correction note at the end. Having comments on a video helps the video author and it shows how even humans can sometimes fail these tests if sampled at the wrong time.
@user-en4ek6xt6w Ай бұрын
Good video thanks you for adding for question and a better sound. You would love to see a comparaison with phi-3 and Llama 3 8B on RAG because i already know that althought they claim to be better than Llama 3 that not true I've tested it
@ml-explained Ай бұрын
Thanks, also please keep up with the feedback, it helps me improve the videos. Next I'll mostly do OpenELM 3B , but tests on RAG will follow :)
@user-en4ek6xt6w Ай бұрын
@@ml-explained Ok thanks you, for now the video was more enjoyable the only think that lack is the editing but all good
@user-en4ek6xt6w Ай бұрын
You should make more test question. Edit out the useless part it make the video long for nothing
@user-en4ek6xt6w Ай бұрын
Your sound is pretty low
@ml-explained Ай бұрын
Thanks for the feedback, I don't have any particular set up for recording but will definitely try to improve it
@MohamedZAARAT Ай бұрын
WILL THIS WORK WITH MY DATA WHICH IS IN EXCEL FORMAT WITH 2 COLUMNS: QUESTIONS & ANSWERS?
@ml-explained Ай бұрын
For any custom requests, please reach out using the contact form at mlexplained.blog
@akentertainments9653 Ай бұрын
will it work for invoices
@ml-explained Ай бұрын
Depends? Are the invoices PDF with text or images ? The problem statement changes on the type of input data. If you want a consultation, feel free to contact by visiting mlexplained.blog
@jonathandegange2450 Ай бұрын
Thanks for your analysis but something seems clearly wrong
@ml-explained Ай бұрын
If you can tell what then it'll help me in future videos
@user-en4ek6xt6w Ай бұрын
Do it with 70B
@ml-explained Ай бұрын
Sure, Next video will be 70B vs Claude/GPT
@user-en4ek6xt6w Ай бұрын
@@ml-explained Ok I will subscribe to not miss it, I really love people testing llm in different scenario like coding, agent communication and rag
@jamesyoungerdds7901 Ай бұрын
Great video, thank you! But looking at the output, unless I'm misunderstanding something, it did show '324' above the word "Awards" near the bottom, as one of the most awarded games in history.
@ml-explained Ай бұрын
That is part of context passed, someone also has pointed that instruct models handle these questions better so I'll be testing that theory out as well. There are a lot of new models out there so will run some interesting tests in future videos so stay subscribed.
@jamesyoungerdds7901 Ай бұрын
@@ml-explained Ok got it, thanks for clarifying, now I see - so it included and printed your prompt/training context above its response 👍
@jeff_65123 Ай бұрын
There's some other issue going on here. If you paste your entire prompt into groq for either of the llama3 models there are zero issues getting the correct answer. As a matter of fact, Gemma and Mixtral 8x7B get it correct as well as does llama 2 70B. Mistral 7B Instruct also gets this correct also. I did take a look at your collab notebook and realized you aren't using the instruct models for mistral or llama 3 but instead treating the base models like instruct models so that's probably the issue. In your ollama tutorial you were using the instruct copy which is the default for ollama.
@ml-explained Ай бұрын
Hmm, Thanks for pointing this out, I'll check and re run the collab by changing the models to instruct. Also I expect larger models to get this correct like 70B Model.
@mickelodiansurname9578 Ай бұрын
you are highlighting a crucial metric ignored at the moment in benchmarks which is the ability of the model to pinpoint data in its context window... far as I see this is probably the most important property for RAG applications. Thanks for the walkthrough, very educational, and subscribed
@mohammadamindadgar7766 Ай бұрын
mistral-7B and llama 3 8B both shouldn't be compared to gpt-4 as gpt-4 model has wayyy more parameters and.
@ml-explained Ай бұрын
It's a fun comparison at the end, we all know small LLMs won't do as well as GPT-4 , but comparison was done in the end just to see the difference.
@muhammadali-jv1kr Ай бұрын
Hi , very good work. I want to use this code for my dataset which comprised of 2000 images. Can you provide the script
@ml-explained Ай бұрын
Hi, for any custom projects or anything, please reach out using the contact form in mlexplained.blog
@RishabhPanchal-cq8tz 2 ай бұрын
Hi bro, openaiembedding function showing error "You exceeded your current quota, please check your plan and billing details." in below function. Can you please tell me any free alternative?? vectordb = Chroma.from_documents(documents, embedding=embedding_func)
@ml-explained Ай бұрын
You can also use ollamaembeddings for the retriever. Also you can use HuggingFace embeddings. I'll soon release a full video on chromadb so stay subscribed for when that comes out
@mahmoudhossamkatr169 2 ай бұрын
I need world chess championship 2023 pdf
@ml-explained 2 ай бұрын
You can easily download it via wikipedia as a PDF
@arigran3055 2 ай бұрын
Can you please also explain in depth for other parameters? Especially min_child_weight.
@kamitp4972 4 ай бұрын
Can it extract or detect tables out of a pdf?
@ml-explained 4 ай бұрын
It can, but this tutorial is just a starter example, extraction of structured data is a bit more complicated and will require a different model and prompt engineering
@kamitp4972 4 ай бұрын
@@ml-explained thanks a lot sir. Could you please make one video on it? It'll be a lot beneficial sir
@andrewndudiefomah248 5 ай бұрын
good morning Please how can I contact you
@ml-explained 5 ай бұрын
You can reach out by visiting mlexplained.blog There is a contact section
@James-hb8qu 5 ай бұрын
This looks interesting but I don't see any training data (/content/grading_gpt_training_data.json) included. Have I missed something?
@ml-explained 5 ай бұрын
Maybe the underlying collab was changed, subscribe for updates, I’ll be doing a video on Mistral models once I get some time off work.
@GihanMarthris 6 ай бұрын
can you do Using Llama2 and LangChain plus huggingface Lama2 in a streamlit App?
@ml-explained 6 ай бұрын
I’ll do some videos on Multi-modal models, will incorporate streamlit in those . Subscribe to stay updated
@CrazyEyezdotpng 6 ай бұрын
You just pasted a bunch of code without explaining how to do it lol? Where did you get the code from?
@ml-explained 6 ай бұрын
I wrote the code, didn’t find it anywhere. But point taken, should’ve gone over the explanation in more detail. You can find the code here - mlexplained.blog/2023/08/22/deploy-machine-learning-model-on-spaces-by-huggingface-using-gradio/
@nirupamadasari8960 6 ай бұрын
Can I use qdrant collection to retrieve top 3 chunks in place of retriever?
@ml-explained 3 ай бұрын
Yes, you’ll need to pass that as a retriever though, can store vectors there . Any vector store can be used as a retriever
@javaria_Ibrahim 6 ай бұрын
Plz make voice bot using streamlit
@ml-explained 6 ай бұрын
Link some articles on what are you looking for and I’ll try
@OfficiallyArtificiallyInformed 7 ай бұрын
Can't you use replicate instead of huggingface?
@ml-explained 7 ай бұрын
Yes, the idea here was to use all open source, but you can plug and play whatever you chose
@parthsonkusre2487 7 ай бұрын
can we make this bot in flowise??
@gummyface 8 ай бұрын
my savior 😭
@sasukeuchiha-ck4hy 8 ай бұрын
Do you have any suggestions on how I can integrate a German-language service chatbot in my website, which has been trained on past customer questions and answers in German, to respond to customer inquiries using this data?
@ml-explained 8 ай бұрын
If you’re interested in exploring custom solutions then you can book a call with me, visit mlexplained.blog in the contact section and book a call. Thank you
@devharal6541 8 ай бұрын
Last response was literally true !! XD
@good-gpt2-chatbot 8 ай бұрын
can you share the pdf that you used in this video?
@ml-explained 8 ай бұрын
You can easily recreate the pdf by downloading the Wikipedia article of the match between Ding and Ian for the world chess championship
@good-gpt2-chatbot 8 ай бұрын
@@ml-explained also, i ran the notebook but in the last cell (the gradio one) i get a 403 error. What might be causing it?
@ml-explained 8 ай бұрын
I’m sorry but I can’t really say just from this info alone
@good-gpt2-chatbot 8 ай бұрын
@@ml-explained ok, i’ll see what I can do. Thank you for the video tho 👍🏻
@twocentswithdeb 9 ай бұрын
Not sure why but for me even if the correlation is =1 it says NOT CORRELATED
@ml-explained 9 ай бұрын
I’ll check the code, probably a bug, but the important thing is the concept, if that’s clear then the purpose of the video is fulfilled
@shivrajnag12 9 ай бұрын
Hi how to create a health chatbot with Llama2, with/wo own Data?
@ml-explained 9 ай бұрын
Without own data you’ll have to look for open source models . With your own data you can use langchain or any icl methods to pass that information to the language model
@liambyliam4 9 ай бұрын
Great tutorial. Have you done any video to make a chatbot where you can ask the question using whisper and then answer it using llama2 and any voice cloning such as resemble ai or eleven labs. Will you mind helping me? I have done it but using chat gpt and it costs. Thank you. How can I can contact you personally?
@ml-explained 9 ай бұрын
Please visit mlexplained.blog to get in touch
@DIY_Foodie 8 ай бұрын
hey please make video of it...and post it here
@ml-explained 8 ай бұрын
I’ll try, with work commitments it’s not easy to make longer videos. But if you like this video do subscribe. Thank you
@LakoMoor 9 ай бұрын
This is greate tutorial ! How run this code with GGML models (use ctransforms) ?
@ml-explained 9 ай бұрын
I haven’t yet experimented with running LLMs on cpu but a nice idea to explore
@LakoMoor 9 ай бұрын
I'm also trying to do it now) So far unsuccessful ((Many artifacts
@tirumaleshtiru5923 9 ай бұрын
well explained. I have one doubt. In this video you are using huggingfaceEmbeddings to create vector db. How LLAMA2 can understand the embeddings of other model which we used to create vector db from pdf ?
@ml-explained 9 ай бұрын
Llama2 Never uses the embeddings. The embeddings are used to do semantic search and the relevant document is passed as context to llama2 model. Hope this answers your question
@tirumaleshtiru5923 9 ай бұрын
@@ml-explained thanks for the response, that answers my doubt. I have one more follow up question. sentense_transformers model is used to create vector db as well as for semantic search. can we use llama2 for this operation? and how can I choose which model is best for semantic search among all the sentese_transformers library models.
@ml-explained 9 ай бұрын
We can’t use llama2 to create embeddings as it’s a decoder only model. Explore other videos on the channel and let me know what topic you’ll like to be covered
@tirumaleshtiru5923 9 ай бұрын
@@ml-explained thanks for clearing the doubts I am having.
@ml-explained 9 ай бұрын
No issues, please consider subscribing and following the blog - mlexplained.blog
@kritisaxena2709 9 ай бұрын
This was really helpful! Looking forward to more of your videos! 😄
@rammohan3401 9 ай бұрын
can i use chat bot retrieve data only from given context without using huggingface
@ml-explained 9 ай бұрын
If you can pass the entire context always then there is no need for a retriever
@mikeportnoy168 9 ай бұрын
Thanks for the great content. I'd like to ask how to modify the code in order to run on the huggingface cloud?
@ml-explained 9 ай бұрын
Video upcoming on how to deploy to spaces
@ml-explained 9 ай бұрын
Here is the video Deploy Your Machine Learning (ML) Model on HuggingFace Spaces using Gradio kzbin.info/www/bejne/o3Wsl2xjo6iigrM
@Insipidityy 9 ай бұрын
Thanks so much for the video! I've been trying to amend your code to include streaming in Gradio but have not been successful. Would you have sample code or a past video I can reference? Thanks!
@ml-explained 9 ай бұрын
Can you specify your exact requirements, I haven’t yet published any video on streaming data using gradio, but lemme look for some resources and publish one in the upcoming week
@Insipidityy 9 ай бұрын
@@ml-explained Exact requirements is to get ConversationalRetrievalChain to stream. From what I understand, there are two ways to code it in Gradio (using gr.Blocks() or gr.ChatInterface). I haven't been successful in either, however.
@ezmode946 9 ай бұрын
i can sit on the sky
@ml-explained 9 ай бұрын
Yes that’s also a nice sentence to check
@nitinab5938 9 ай бұрын
Hey, great tutorial, but I'm getting this error: ValueError: `do_sample` is set to `False`. However, temperature is set to 0.9 -- this flag is only used in sample-based generation modes. Set `do_sample=True` or unset temperature to continue. Any idea how to fix it?
@ml-explained 9 ай бұрын
Hey check out my latest video where I go over how you can set up the pipeline to use temperature > 0. kzbin.info/www/bejne/haaye2h5pb2BgtE Hope this helps
@ml-explained 10 ай бұрын
Full explainer video - kzbin.info/www/bejne/iKSuhpymrLOtfKM
@rjrmatias 10 ай бұрын
thank you very much, excellent video
@ml-explained 10 ай бұрын
Much appreciated
@kritisaxena2709 11 ай бұрын
Really Helpful! I always struggled with this topic