Great video as always Nick! I just finished my BS in Electrical Engineering and used a couple of your videos to help me with my senior project. Watching your videos makes me want to get into Data Science. Thanks for everything and keep up the awesome work! - Scott
Great stuff really appreciate it, 1 request, could you please make a tutorial on how we can further fine tune these gpt4all models on our own supervised data for a particular domain... it would be really great!!! thanks looking forward to it!
@@NicholasRenotte 100 Days of Coding in Python course :) We'll see where that takes me. I love your succinct energetic delivery, looking forward to more local llm magic videos :)
Love it! Only thing that would make it even better would be some type of summary or chart at the end. I found myself going back and pausing to take notes to get that info ;-). Btw, it’s really cool that you can so easily work with so many models because of GPT4ALL and langchain. Is there a generic way to work with a random model, x? I want to setup guanaco and falcon and play around with fine tuning them. I expect tomorrow there will be something new. I *think* langchain can likely just pull these in regardless, but with one of your models not working right it had me wonder. Thanks for spending the time putting this together. It is very much appreciated!
How many computer resources you require to run that moxel? Last time i checked you require about 16gb of ram to load it in your PC but i would love to hear you out on this
You are so awesome Nick! Thank you for your content it is always so great and useful. With the knowledge you share through your videos I was able to land my first job as an ML Developer in the industry and not in the academics field. You are really awesome!!!!!!!!! - Jorge :)
Hi! It would be awesome if you could make a video about creating a chat-bot that uses free HuggingFace models and huge custom data (like PDF's or TXT file custom knowledge bases) together. I couldn't find a comprehensive guide about this, all I see is the usage of OpenAI API everywhere. Thanks for the video, it has helped me a lot on my journey!
Bro, you are THE MAN. I might try your courses too... I have like twenty other disciplines I do as a small business owner, but damn, you're a good teacher and it would be totally worth it. I just need to make the time, then I'm in. I have a startup that manufactures a small 2-axis CNC machine, and I want to plug in some LLMs to generate roughing passing on my cutters. Problem is, the LLMs seem to be fundamentally incompatible with mathematics and linear algebra. But I think there's still enough ways to hack them and get something working.
I wonder if you could use the wolfram alpha API to help with the tool routes. Haven't done any CNC modelling but I'd imagine there's some path to doing it with a math heavy api or even building an agent environment to CAD software!
@@NicholasRenotte I'll check out the Wolfram Alpha API. I tried to probe GPT3 with some 3D solid situations, such as "Imagine you are drawing a 2D path in the Z-X plane, then revolving around Z. Your origin is (Z0,X0). Create a 10mm diameter cylinder 20mm long. Provide the ordered pairs around the path, starting from the origin and ending at the other end". I've always gotten gibberish with this sort of thing, and I think it's fundamental to a LLM. The LLM would need to understand that I'm describing a cylinder, and that a cylinder is comprised of a path revolved around an axis. What we all need is sort of a "Large Geometrical Language Model" that deeply understands geometry and can visualize 3D solids parametrically. This would be the holy grail of 3D modeling / machining and AI. Also the LLMs can't generate STEP file formats for 3D solids, last I checked (would be sweet). An AI should be able to imagine a solid and propose a working STEP file that you could then open with Fusion 360, etc. When we get there, things will get wild in physical reality. Thoughts?
Yeah, this is where agents would help. e.g. like having access to CAD sftware to visually display the tool paths as it's generating. You would then be able to provide feedback to the LLM based on error/success metrics from the software.
First, thank you for the video ! You said that "if you're already a student, you'll get immediate access" about the fullstackml course. But I don't quite figure out how and where I can enter my student credentials in order to get access to the Full Stack ML course? Or should I make a new account using my university email?
Yep, they can be spun up in HF for fine tuning. You could probs do it with documents, although I found most of them to be pretty slow when working with agents.
@@NicholasRenotte Just for fun/training I was trying to implement something like privategpt without looking at their code. I also wanted to use mpt instruct as I had good results with it in another test. All went ok, it works as intended but very slow, at least on my hardware. Will investigate running on gpu. Thank you for your videos, I have learned a lot from you.
Can you do a video on building an LLM that can do statistical analysis like Monte Carlo simulation from scratch? Bard is capable of doing Monte Carlo simulations and regression analysis.
Great job on this video, I was wondering if you have spent any time benchmarking what kind of hardware resources (CPU, GPU, Memory, Storage, etc) are needed for a LLMs on-prem development and production environment.
Sure can, I built one, I have no idea where I put the code though. If i get some time I'll rebuild it in a vid Henk! Lewis showing promising pace last night 😉, is this the comeback?!
Enjoy these videos and do a great deal of python with GIS. I'm curious if these localized LLMs can reference internal datasources and documents (if we tokenize them) to respond to my prompts?
Sure can, the documents vid used OpenAI. My only concern with using local models is inference speed. These were already pretty slow, I don't think doing vector search and using massive context will help performance....but going to be doing a speed comparison this week (and trying to get my GPU working).
thanks Nick. Very nice tutorial. I have doubt, which model I can conclude is the best one considering the possibility to fine tune and add more knowledge to it? I mean, vicuna, mtp, gpt4all, falcon and many others open source which one I can finetune and use it freely ?
If it needs to run locally then your best bet is one of these. Haven’t tested falcon yet though. For big enterprise apps we’re mainly doing stuff with massive hosted private models!
Great video, but i am keep getting error: illegal hardware instruction /Users/priyanshz/Desktop/NewProj/Nopenai/nonopenai/bin/python I have checked all of the documentation still cant figure out why its not working with m1 chip.
Hi there Nicholas. Thank you for this great content! I am planning on building an llm for our online therapy platfrom to answer specific questions regarding sexual health and relationship issues. Would you recommend we build this chatbot via a proprietary model like chatgpt api or via an open source model? Our goal is to be able to teach it our own data as well and increase it’s capabilities and knowledge in the long run. Thank you!
thank you a lot for these videos Nick you are awesome!! - based on your tests, which model is the best to use in your opinion? - i did not very well understood the difference between using the python agent VS the llm chain and why should we use one or the other ?
Depends on the use case tbh and whether you need a commercially licensable model. For chat MPTs seemed pretty good. For most of the others I liked Vicuna and Snoozy!
thanks for amazing tutorial Nick 😄 but I'm encountering the following error: llm = GPT4All(model=PATH, verbose=False) 'No corresponding model for provided filename C:\Users\user\Desktop onopenai\ggml-gpt4all-j-v1.3-groovy.bin. If this is a custom model, make sure to specify a valid model_type.' Then I modified and a new error appeared llm = GPT4All(model=PATH, backend='gptj', verbose=False) File "c:\users\user\desktop\llm onopenai\lib\site-packages\langchain\llms\gpt4all.py", line 155, in validate_environment values["client"].model.set_thread_count(values["n_threads"]) 'AttributeError: 'GPTJModel' object has no attribute 'set_thread_count'' Appreciate if you can share solution for the errors, thanks a lot
Hey Nic, I'm a little bit confuse. By running this short code, do you actually do an API call to get the answer from the internet, or are you using your machine to perfom some local training. Because I do not understand the 17min that your CPU took perform the computation at the end.
No stress Ange! It’s all done locally for the GPT4All models. This means inference is limited by the specs of the machine. When swapping over to the OpenAI LLM it’s done using their SaaS service.
Quick Questions regarding Tensorflow js: When I use Tensorflow Js. cameraWithTensors with expo to create an object detection app, my camera keeps getting stuck everytime it makes a prediction, could you please explain why this is happening.
can you make a video on how to make a stable diffusion api in node js and upload it online ,create a website for it ,let users create there own image and download it.
What awesome contents you have in your channel :D. I just want to ask if you can also teach how to do facial recognition using an FPGA? I am currently studying HDL/VHDL programming for implementing facial recognition for my college project. However, there's not much resources I can find to learn about the programming and coding aspects of facial recognition so far.
Heya, it really depends on the machine you're using. These models run locally on your computer so the speed will be relative to the compute power of the device. For the basic responses it took around 1 minute to get anything back on my machine.
Hi, I can see it slowly writing words in the terminal but nothings coming up on the streamlit hosted page, is this normal or is there something wrong? I’m using a laptop with GTX 1650 GPU
Hi Nick, i tried running the code for the app-chain on my laptop and i have downloaded the snoozy model. I gave a basic query(what is deep learning?) on streamlit app and it's been 20min now and it's still running. Can you help me with how to get faster results?
20 minutes is steep, is there anything else running on your machine taking up CPU compute? If so shut it down. Otherwise, im working on a GPU accelerated version of the app next week. It took a minute and 30 seconds to get most responses. Summarisation took a little longer.
@@NicholasRenotte Apart from VSCode, chrome(your TH-cam video, GitHub) nothing else is working on the background. I tried changing the the prompt template to the one you mentioned in app-comparison. But it's still running, it's been 5 minutes now since i ran it.
@@NicholasRenotte hi just an update, after changing the prompt from the simple prompt to the better prompt ("the prompt below..." ) It took 10 min to execute this. Maybe the simple prompt and cpu processing is making the response time slower?
@@akashramdev weird, I was getting concurrent non-terminated threads but only when I was force shutting. I’ll play around with the app a little more this w/e to see if I can replicate on my MacBook
Hi , your are great teacher . I have learnt many things from you , specially Longchine. Kindly make tutorial for flowise ai tutorial along with agent like with no code tolls like bubble
Hey Nick, amazing tutorial! For some reason though I'm getting an error everytime I run the code. I have installed all the necessary dependencies, but it still seems to be giving me an error. I'm using python 3.11 on an M1 macbook air - could that be the issue?
Hey man, thank you for the great videos as always. I am planning on getting a laptop with an rtx 4060 for mainly CV, NLP and gpt tasks. is it good enough? I can't do PCs as I move a lot and this is the best I can get with my budget.
Hi can I ask how may RAMs your device has ? I integrated this model with langchain (pdf chat), the answer is pretty descent but took ages . Are there any extra parameters for GPU acceleration ?
Hello, Nicholas. Great tutorials! I am writing a program about face recognition using dlib, face_recognition and opencv. I am wondering how to implement liveness detection as on samsung or other phone without additional sensors (as on iphone). Would be grate to see tutorial on this theme on your channel.
I am novice and I need to get into generative ai. I need to build a product for legal research and drafting. Please suggest. As of now I have started learning python.
hello, very interesting video. can you do a follow up video but with very specific data that are stored in pdf, word, HTML? i would like to see you query but on the very specific topic
Hey nicholas hope you’re doing good Im currently pursuing bsc in analytics and data science in Melbourne from deakin I wanted to know what are job prospects of Data science in australia generally and do employers compare us with masters student or will I have upper hand if I possess more skills ?
Can you make a video on generating any kind of document ..... Let's say a petition letter ( 15-20 page ) based on the user uploaded documents using langchain 🙂
@@NicholasRenotte if possible, in the video please create an app of asking questions with pdf using huggingface model without downloading the model. Also add the functionality of next empty prompt/question space once a question is answered. Thanks a lot , i really appreciate your time and effort
I am interested in the GPU part... is it just a "cuda" flag or is there more to it? Even on my "new rig", this is kinda slow... You're so smart that I suspect you might be an extraterrestrial life form in disguise.
if u use windows and use BASE_PATH, ValidationError: ... GPT4All__root .... change rule 31: to PATH = f'{BASE_PATH}\{model}' (extra slash between variables)
I recently began writing a neural network and in python and your videos have been a great help 😀 I wish to join your course but PayPal isn't an option where I live any ideas?
Straight to point ! Finally a great video
about open-source models! Congrats!
Glad you liked it Anderson!! Wanted to see what the hype was about but more so see how they work in a range of non-chat use cases!!
Great video as always Nick!
I just finished my BS in Electrical Engineering and used a couple of your videos to help me with my senior project. Watching your videos makes me want to get into Data Science. Thanks for everything and keep up the awesome work!
- Scott
Thanks a mil Scott! Nice, definitely a hyper interesting field to be in!!
Just pausing the video to say "YES PLEASE DO HUGGING FACE HUB VIDEO".
DAVID!! You got it.
Please please please 😂
Also @David, just checking to make sure you came back and pressed play again? 😂
@@NicholasRenotte haha yes I did. Great vid as always. Gonna try it out myself when I get the chance.
@@dacoda85 🙏 all the code (inc the big app) is in the GH repo! Finally started adding good READMEs as well.
Great stuff really appreciate it, 1 request, could you please make a tutorial on how we can further fine tune these gpt4all models on our own supervised data for a particular domain... it would be really great!!! thanks looking forward to it!
You got it!! Might take a little while.
@@NicholasRenotte Yea I am looking for this as well! I tried but it seems the ChromaDB expects the openAI embeddings?
@@mitchross2852 updated the code a few days ago on GitHub, take a look and lmk how you go!
The more python I learn, the more fascinating your videos are :) Thanks for sharing!
Yes!! I’ve been brushing up on some topics myself this week. Whatcha been learning?
@@NicholasRenotte 100 Days of Coding in Python course :) We'll see where that takes me. I love your succinct energetic delivery, looking forward to more local llm magic videos :)
Oooooh, nice! You got it, I'm getting back into flow for creating vids. Plenty more to come.
10:50 when using triple quotes, lines should start at column 0, here the format is not llama's expectations
Other question: Have you used QLora to fine tuning any model?
Nup, not yet. Next vid?
@@NicholasRenotte yes, please!
@@NicholasRenotte yes
Love it! Only thing that would make it even better would be some type of summary or chart at the end. I found myself going back and pausing to take notes to get that info ;-).
Btw, it’s really cool that you can so easily work with so many models because of GPT4ALL and langchain. Is there a generic way to work with a random model, x? I want to setup guanaco and falcon and play around with fine tuning them. I expect tomorrow there will be something new. I *think* langchain can likely just pull these in regardless, but with one of your models not working right it had me wonder.
Thanks for spending the time putting this together. It is very much appreciated!
OHHHH mannn, I wanted to do this. I had even prepped the data, will defs do it for the next one.
How many computer resources you require to run that moxel? Last time i checked you require about 16gb of ram to load it in your PC but i would love to hear you out on this
That's so good that someone did it. I think some more extensive tests wpuld be great ro see. And ofcourse more open source model in the feature:-)
Thanks a mil @Marcin, definitely, will see what I can get up to!
You are so awesome Nick!
Thank you for your content it is always so great and useful.
With the knowledge you share through your videos I was able to land my first job as an ML Developer in the industry and not in the academics field.
You are really awesome!!!!!!!!!
- Jorge :)
Hi! It would be awesome if you could make a video about creating a chat-bot that uses free HuggingFace models and huge custom data (like PDF's or TXT file custom knowledge bases) together. I couldn't find a comprehensive guide about this, all I see is the usage of OpenAI API everywhere. Thanks for the video, it has helped me a lot on my journey!
Yes, good suggestion, it will be first and only content on YT.
Bro, you are THE MAN. I might try your courses too... I have like twenty other disciplines I do as a small business owner, but damn, you're a good teacher and it would be totally worth it. I just need to make the time, then I'm in. I have a startup that manufactures a small 2-axis CNC machine, and I want to plug in some LLMs to generate roughing passing on my cutters. Problem is, the LLMs seem to be fundamentally incompatible with mathematics and linear algebra. But I think there's still enough ways to hack them and get something working.
I wonder if you could use the wolfram alpha API to help with the tool routes. Haven't done any CNC modelling but I'd imagine there's some path to doing it with a math heavy api or even building an agent environment to CAD software!
@@NicholasRenotte I'll check out the Wolfram Alpha API. I tried to probe GPT3 with some 3D solid situations, such as "Imagine you are drawing a 2D path in the Z-X plane, then revolving around Z. Your origin is (Z0,X0). Create a 10mm diameter cylinder 20mm long. Provide the ordered pairs around the path, starting from the origin and ending at the other end". I've always gotten gibberish with this sort of thing, and I think it's fundamental to a LLM. The LLM would need to understand that I'm describing a cylinder, and that a cylinder is comprised of a path revolved around an axis. What we all need is sort of a "Large Geometrical Language Model" that deeply understands geometry and can visualize 3D solids parametrically. This would be the holy grail of 3D modeling / machining and AI. Also the LLMs can't generate STEP file formats for 3D solids, last I checked (would be sweet). An AI should be able to imagine a solid and propose a working STEP file that you could then open with Fusion 360, etc. When we get there, things will get wild in physical reality.
Thoughts?
Yeah, this is where agents would help. e.g. like having access to CAD sftware to visually display the tool paths as it's generating. You would then be able to provide feedback to the LLM based on error/success metrics from the software.
Thanks for the knowledge transfer Nick, Awesome work. Keep it up.
Anytime Sarit! Try to share where I can!
Amazing tutorial, im brazilian guy who tried apply this for tech solucions in my country and you will help me that! thanks
shoot, what’s up?
First, thank you for the video ! You said that "if you're already a student, you'll get immediate access" about the fullstackml course. But I don't quite figure out how and where I can enter my student credentials in order to get access to the Full Stack ML course? Or should I make a new account using my university email?
Oh, it’s if you’re already a student in the Full Stack ML course.
How can i download model ? PlZ send a links
Can we fine-tune this implementation or give it access to information from a PDF just like the previous videos?
Yep, they can be spun up in HF for fine tuning. You could probs do it with documents, although I found most of them to be pretty slow when working with agents.
Literally just did my own implementation a few days ago.
I also just went and picked the models from the download dir.
Good content as always!
Nice!! Yeah, I realised a little later on that you could just grab all the GGML files directly from the site. What did you end up building?
@@NicholasRenotte Just for fun/training I was trying to implement something like privategpt without looking at their code.
I also wanted to use mpt instruct as I had good results with it in another test.
All went ok, it works as intended but very slow, at least on my hardware.
Will investigate running on gpu.
Thank you for your videos, I have learned a lot from you.
Great job. Thanks for sharing. Looking forward to more.
Cheers @Lucifer Hesperus!! Plenty more to come.
You mentioned GPT4all class compatibility and hugging face, two topics I’d love to see have videos in the future
Can you do a video on building an LLM that can do statistical analysis like Monte Carlo simulation from scratch? Bard is capable of doing Monte Carlo simulations and regression analysis.
Great job on this video, I was wondering if you have spent any time benchmarking what kind of hardware resources (CPU, GPU, Memory, Storage, etc) are needed for a LLMs on-prem development and production environment.
I haven't...but I plan to. I'm going to be doing this a lot more at work so I'll get a lot more info I can share.
Great video comparising different llm's. Can u do with stramlit like the chat ui where u can see the chathistory and with streaming?
Sure can, I built one, I have no idea where I put the code though. If i get some time I'll rebuild it in a vid Henk! Lewis showing promising pace last night 😉, is this the comeback?!
Enjoy these videos and do a great deal of python with GIS. I'm curious if these localized LLMs can reference internal datasources and documents (if we tokenize them) to respond to my prompts?
Sure can, the documents vid used OpenAI. My only concern with using local models is inference speed. These were already pretty slow, I don't think doing vector search and using massive context will help performance....but going to be doing a speed comparison this week (and trying to get my GPU working).
thanks Nick. Very nice tutorial. I have doubt, which model I can conclude is the best one considering the possibility to fine tune and add more knowledge to it? I mean, vicuna, mtp, gpt4all, falcon and many others open source which one I can finetune and use it freely ?
If it needs to run locally then your best bet is one of these. Haven’t tested falcon yet though. For big enterprise apps we’re mainly doing stuff with massive hosted private models!
@@NicholasRenotte which model can be used to enterprise causes? I know t5 but the main stream models I don't know
A bunch, been using Bloomz and flan-t5-UL
Hi @Nicholas,
Where can I get your wallpaper image?
Great video, but i am keep getting error: illegal hardware instruction /Users/priyanshz/Desktop/NewProj/Nopenai/nonopenai/bin/python
I have checked all of the documentation still cant figure out why its not working with m1 chip.
Hi there Nicholas. Thank you for this great content! I am planning on building an llm for our online therapy platfrom to answer specific questions regarding sexual health and relationship issues. Would you recommend we build this chatbot via a proprietary model like chatgpt api or via an open source model? Our goal is to be able to teach it our own data as well and increase it’s capabilities and knowledge in the long run. Thank you!
thank you a lot for these videos Nick you are awesome!!
- based on your tests, which model is the best to use in your opinion?
- i did not very well understood the difference between using the python agent VS the llm chain and why should we use one or the other ?
Depends on the use case tbh and whether you need a commercially licensable model. For chat MPTs seemed pretty good. For most of the others I liked Vicuna and Snoozy!
Oh the Python agent has access to write and run code which means it can self-debug and solve problems using code.
thanks for amazing tutorial Nick 😄 but I'm encountering the following error:
llm = GPT4All(model=PATH, verbose=False)
'No corresponding model for provided filename C:\Users\user\Desktop
onopenai\ggml-gpt4all-j-v1.3-groovy.bin. If this is a custom model, make sure to specify a valid model_type.'
Then I modified and a new error appeared
llm = GPT4All(model=PATH, backend='gptj', verbose=False)
File "c:\users\user\desktop\llm
onopenai\lib\site-packages\langchain\llms\gpt4all.py", line 155, in validate_environment
values["client"].model.set_thread_count(values["n_threads"])
'AttributeError: 'GPTJModel' object has no attribute 'set_thread_count''
Appreciate if you can share solution for the errors, thanks a lot
Hmmm, can't see any doco on that issue. What version of LC, Python and what type of OS are you using?
@@NicholasRenotte
langchain==0.0.189
Python 3.8.10
Windows 11
Is it necessary to make a project with integration , because I am making projects in python jupyter , colab but not integration any suggestions?
Hey Nic,
I'm a little bit confuse. By running this short code, do you actually do an API call to get the answer from the internet, or are you using your machine to perfom some local training.
Because I do not understand the 17min that your CPU took perform the computation at the end.
No stress Ange! It’s all done locally for the GPT4All models. This means inference is limited by the specs of the machine. When swapping over to the OpenAI LLM it’s done using their SaaS service.
Why does using models from HF take forever to get an O/P? Is there any way to optimize the inference time?
I've got a feeling it wasn't picking up my GPU, will try again next week!
@@NicholasRenotte I guess it was the same in my case. Also, great video. Can't wait for more cool videos on LLM.👍🏻
Quick Questions regarding Tensorflow js: When I use Tensorflow Js. cameraWithTensors with expo to create an object detection app, my camera keeps getting stuck everytime it makes a prediction, could you please explain why this is happening.
can you make a video on how to make a stable diffusion api in node js and upload it online ,create a website for it ,let users create there own image and download it.
Hi Nick, no available models are showing up for me after clicking on GPT4All.
Hi Nick, i am stuck on the project that I am working on which is to measure the objects on picture in millimeters. How can I do in efficient way?
What awesome contents you have in your channel :D. I just want to ask if you can also teach how to do facial recognition using an FPGA? I am currently studying HDL/VHDL programming for implementing facial recognition for my college project. However, there's not much resources I can find to learn about the programming and coding aspects of facial recognition so far.
excelent video! i wanna use one of this models just for answers question or search from a intranet database, what you recommend ?
Also Chae4ek langchain not found.
Damn, just realised he took the repo down. Will update when I'm back at my computer, I'm out of the office atm, try using 0.0.173
Nick why streamlit runs so slow with ml models??
Heya, it really depends on the machine you're using. These models run locally on your computer so the speed will be relative to the compute power of the device. For the basic responses it took around 1 minute to get anything back on my machine.
Are there other open source models that are not llm
Hi, I can see it slowly writing words in the terminal but nothings coming up on the streamlit hosted page, is this normal or is there something wrong? I’m using a laptop with GTX 1650 GPU
Hi Nick, i tried running the code for the app-chain on my laptop and i have downloaded the snoozy model. I gave a basic query(what is deep learning?) on streamlit app and it's been 20min now and it's still running. Can you help me with how to get faster results?
20 minutes is steep, is there anything else running on your machine taking up CPU compute? If so shut it down. Otherwise, im working on a GPU accelerated version of the app next week. It took a minute and 30 seconds to get most responses. Summarisation took a little longer.
@@NicholasRenotte Apart from VSCode, chrome(your TH-cam video, GitHub) nothing else is working on the background. I tried changing the the prompt template to the one you mentioned in app-comparison. But it's still running, it's been 5 minutes now since i ran it.
@@NicholasRenotte hi just an update, after changing the prompt from the simple prompt to the better prompt ("the prompt below..." ) It took 10 min to execute this. Maybe the simple prompt and cpu processing is making the response time slower?
@@akashramdev weird, I was getting concurrent non-terminated threads but only when I was force shutting. I’ll play around with the app a little more this w/e to see if I can replicate on my MacBook
where can i have that folder that have models?
Heya, install instructions to do it are in the GitHub repo. When you install GPT4All it’ll do it.
@@NicholasRenotte thank you for all
How do you continue training the LLM with new or updated information?
Haven't covered it here but you can fine tune certain models. I didn't have a vid planned for it but I'm going to give it a crack soon.
Do that, and you got yourself a new subscriber mate
Hi , your are great teacher . I have learnt many things from you , specially Longchine.
Kindly make tutorial for flowise ai tutorial along with agent like with no code tolls like bubble
thanks for the awesome content, youve made data science so fun to learn :)
Yesssss!! That was always the goal!
Thanks, it really helps a lot!!
Glad you liked it!!
Hey Nick, amazing tutorial! For some reason though I'm getting an error everytime I run the code. I have installed all the necessary dependencies, but it still seems to be giving me an error. I'm using python 3.11 on an M1 macbook air - could that be the issue?
Using my library? I had to use a different branch of LangChain to get it to work. I put it in the requirements.txt
I got the same issue, the github problem :(
What’s the error?
Hey man, thank you for the great videos as always.
I am planning on getting a laptop with an rtx 4060 for mainly CV, NLP and gpt tasks.
is it good enough?
I can't do PCs as I move a lot and this is the best I can get with my budget.
Hey man, does your machine learning class use OpenAI or are you using open source?
Right now all open source, the project will do both but focus mostly on open source!
It is not working on latest langchain version but working on 0.0.173
nice, I had to use a separate branch from GitHub as well. Specific version is in the requirements.txt file!
Hi can I ask how may RAMs your device has ?
I integrated this model with langchain (pdf chat), the answer is pretty descent but took ages .
Are there any extra parameters for GPU acceleration ?
Hi, Please make a video about LLM app deployment
Np, will line something up soonish!
Hello, Nicholas. Great tutorials! I am writing a program about face recognition using dlib, face_recognition and opencv. I am wondering how to implement liveness detection as on samsung or other phone without additional sensors (as on iphone). Would be grate to see tutorial on this theme on your channel.
can you make a video about how to work with hugging face ?
You got it!
I am novice and I need to get into generative ai. I need to build a product for legal research and drafting. Please suggest. As of now I have started learning python.
Take a look at the documents vid and build from there. You would probs want context from existing legal docs and use that to generate.
Great video _ I wonder if it works with new Llama 2 models.
Amazing video. Can you let me know how you can create a chat bot just like this with our own data?
Thanks for this Nick
Anytime ✌️
Hi Nick!
Can this app deployed on web version of Streamlit ?
Hey, what are your machine specs ? In my laptop it's taking hours to return response for a simple short prompt
A little older now, it’s a Ryzen 3700x with a 2070 super and a ton of RAM.
@@NicholasRenotte 😟 how do you run all that on that?
@@BEEETRUS lol I have no idea, legit it sounds like a jet plane when I record these videos. It took me two days just to record all the demo segments.
@@NicholasRenotte is there any way I can get in contact with you I have some questions ?
Can it be modified to accept pdf files?
May I know which python version are you using? Also pip install asks for user/password but doesn;t work even when credentials are correct
Run pip install --user if you're bumping up against credential issues. Ahhh, I think I'm using 3.9.x?
Great job!!! Your vídeos is amazing
hello, very interesting video. can you do a follow up video but with very specific data that are stored in pdf, word, HTML? i would like to see you query but on the very specific topic
Where is model or weight please provide link
Bless you brother. i been waiting for this video. Please make another one with huggingface hub llms . Can we use them for autogpt?
You got it!!
Thanks so much for this video!
Anytime, cheers Michael!
Hi sir , please make a series Algo trading with different strategies platform through machine learning ?
Hey nicholas hope you’re doing good
Im currently pursuing bsc in analytics and data science in Melbourne from deakin I wanted to know what are job prospects of Data science in australia generally and do employers compare us with masters student or will I have upper hand if I possess more skills ?
PLEASE do a tutorial on running GPT4All on a gpu,
I’m so tired of waiting 5 minutes for a response every time I do a test
Can you make a video on generating any kind of document ..... Let's say a petition letter ( 15-20 page ) based on the user uploaded documents using langchain 🙂
I am screaming for past couple of videos and post of yours. PLZ DO A VIDEO ON HUGGINGFACE HUB MODELS
Whipping it up soon!!
@@NicholasRenotte if possible, in the video please create an app of asking questions with pdf using huggingface model without downloading the model. Also add the functionality of next empty prompt/question space once a question is answered. Thanks a lot , i really appreciate your time and effort
a question that cannot be ignored!! can this be used to make money?? type implemented a real chat as an assistant?
Nick, we have bard and gpt so which one is good for what? A video would be nice
Gotta review next week, had my head down trying to get this done!
I am interested in the GPU part... is it just a "cuda" flag or is there more to it?
Even on my "new rig", this is kinda slow...
You're so smart that I suspect you might be an extraterrestrial life form in disguise.
The github link is not working why ?
My bad, was private! Updated
🎉 amazing 💪🏻💪🏻💪🏻 Great tuto thanks 🙏🏻
Cheers Kev!!
Can you teach how to use AI for extract video (mark point of joint ) into 3D animation plsssssssss😊😊😊
can you male a video on chainlit ?
I am really confused. He is saying open source llm, but time and again saying openai. Is this openai's model?
if u use windows and use BASE_PATH,
ValidationError: ... GPT4All__root ....
change rule 31: to PATH = f'{BASE_PATH}\{model}' (extra slash between variables)
Does it support many languages?
Haven't tried tbh, although my guess is it's purely english.
Another high quality video ❤
Thanks a mil @Sayan!
You are my idol, I love your tutorials
Oh thanks so much Jack, you put a huge smile on my face man! I went around my house telling my fam about your comment ✌🏽
Love your channel
Thanks for working on this nick much appreciated❤ also love the precap and editing👏👏
Anytime! Stoked you liked the updated style!
Can you please do a tutorial on GPU
Pleaaase the "run on GPU" version of it! ❤❤❤
I recently began writing a neural network and in python and your videos have been a great help 😀 I wish to join your course but PayPal isn't an option where I live any ideas?
Heya, shoot me an email at nick@coursesfromnick.com and I’ll work something out for you!!
When Nicholas enters into God mode
Next week: Super Saiyan 5
@@NicholasRenotte ❤🔥
How to use GPU with the code
Thanks nice Job👍
With the help of Python and machine learning, we can imagine the impossible and make it happen. Amazing video, Nick🙂..
Waiting for you to do the same for Huggingface Hub as well
Next week I’m on it @Vasanth!!
@@NicholasRenotte Cool
why your head covers the code you write!?
More vids on open source llms pls 🤧🤧
Thx for the tutorial
Anytime!! Glad you liked it!!