Python | AI | Machine Learning | Deep Learning | Data Science | Data Visualization
CoderzColumn Is Designed For Sharing Knowledge On the Topics Like:
* Python * Artificial Intelligence (Deep Learning) * Data Science * Machine Learning * Data Visualization
We focus on explaining various concepts related to these topics in simple terms. Our tutorials are short and simple to understand. Our main aim is to get someone started with technologies faster.
You can support us by donating through BuyMeACoffee (Link Below).
www.buymeacoffee.com/coderzcolumn
CoderzColumn: Developed for Developers by Developers for the Betterment of Development.
So it just run on our local enviroment, no need internet? But what if we share it on streamlit cloud, can it run as we know streamlit cloud doesn't have high resource or even GPU
Yes. It runs on our local computer and does not need internet for it, Now for sharing it on streamlit cloud is not possible with ollama because streamlit does not let u run ollama there and yes it does not have high resource or GPU. You can deploy Ollama on some cloud instance (Gcloud or AWS) and then access it through API. You can then replace code to access ollama through that cloud instance based URL (base_url parameter). That's one solution. Another solution would be to use OpenAI rest API but it is not free. Recently, I have started using Groq REST API which is free. They provide latest Open source models like Llama-3.1, though there is request limit (30 requests per minute). Also, there models are full and not quantized one like Ollama hence more accurate.
@@CoderzColumn aahh i see, so we can build and host our model in server that have high resource, then we use it as rest API for streamlit web app that we share, thx for explain.
Yea. You can use it other platforms than Streamlit because just like openai API, your models are available through API. You can even create models for custom uses. But i feel the cloud instance are bit pricey. You'll need good instance with 16/32/64 GB RAM and one with high GPU RAM. The cost increases fastly with specs. You know if you have some spare system with GPU lying around then you can use it as well.You'll have to keep it on all the time and initially, you'll have to do few firewall settings to make it accessible to outside world on internet. This one can be cheaper hopefully but you'll still have to pay for electricity for keeping this system on (which has GPU hence more electricity usage).
Hi CoderzColumn, you are so generous to share and explain coding clearly. Hope you are okay and keep pursuiting your career successfully in nearby future. <3. Thanks for your free tutorials
Yes, definitely you can use it with Plotly. It'll work fine. Also, I think that with plotly, if you use FigureWidget by wrapping figure object inside it then it'll be better. Check out this article. plotly.com/python/getting-started/#jupyter-notebook-support
Nice video... Just what I'm looking for. This Dash-plotly I used when I need a real time update, at maximum 5 to 5 min refresh. Do you know how this will cost monthly? just to have some idea
If you have set max_instance in config file to 1 then it won't cost you anything. But if you increase count of instances then it'll charge based on usage of instance. I have included info about instance charge pages in video. Please feel free to check them to know more about charges. But with just one instance running, it comes under free tier hence no charges.
I think that when you load transcripts, you can load it in chunks and then create vector store using it. That way you can retrieve timestamps from retrieved relevant documents. Start time of chunk is available as metadata. It'll require bit of coding changes. It won't be accurate but near to answer.
Excellent video, very well explained and easy to follow. I take a look at to your content and is really good and useful for me. Thanks a lot Mr Solanki
Thanks for feedback. Appreciate it. In this tutorial, I have used Groq API for accessing LlaMa-3 hence you should be able to run it even with 2 GB RAM. Because it'll hit Groq API for LLM. You'll need to create api-key on groq though. But if you want to use Local LLMs like the one available from Ollama then you'll need around 8 GB of RAM. It'll be better with more than 8GB but 7-8 B parameters model works with 8 GB as well. Good to have i3 or or above processor. But want to inform you that results were not that good with Llama-3 accessed through Ollama. I have tried it. Maybe because it's quantized version. I would suggest you using Groq API.
'Exception occurred: Error code: 400 - {'error': {'message': "'messages.1' : for 'role:user' the following must be satisfied[('messages.1.content' : one of the following must be satisfied[('messages.1.content' : value must be a string) OR ('messages.1.content.0' : one of the following must be satisfied[('messages.1.content.0.type' : value is not one of the allowed values ['text']) OR ('messages.1.content.0.type' : value is not one of the allowed values ['image_url'])])])]", 'type': 'invalid_request_error'}}' i am not able to fix this error
Can you share which line of code is failing? Is it something failing from Notebook of tutorial or you are trying something new which is failing? I am not able to understand error from this trace only.
Thanks for feedback. I really needed to know this. I'll try to include less bells in intro. I feel i included many this time when things pop up on screen.
Groq provides access to 4 open source LLM (Llama-3 8B & 70B, Gemma 7B & Mistral 8x7B)for free. Groq specializes in hardware creation that speed up token generation for LLMs. These LLMs that they are providing through their API are not theirs but they are open source LLMs which they have deployed on their high performing hardware and provided everyone limited access through API (30 requests per minute). You can login to console.groq.com/settings/limits and check limit for free account. I just now noticed that they added two new models (gemma2 8B and Whisper Large v3). Yes, It's not locally hosted.
Thanks for feedback. I have code present in one my blog. Here's link * coderzcolumn.com/tutorials/data-science/timeline-using-matplotlib I included it in description as well. This is one of my earlier videos where I forgot to include code. Hope this helps.
Thank you Sunny. This has been a great intro and for me, it worked perfectly. I really appreciate your time and effort - you are a good teacher and I have learned much here! Bokeh + Django is a great combination. Regards from Sydney - David
Thanks for the comment. Ideally, if you want to include chart in your Django website, I would recommend to use some Javascript library like Amcharts, d3.js, etc. I have found that it works better with django. But it'll require you to learn javascript though charts are pretty looking. Bokeh is good solution if you want to code in Python for charts and want to include them in django website. At the end of the day, it is javascript based as well. Happy to hear that my tutorial helped you.
@@CoderzColumn Thank you for your advice. I have been considering learning D3, but using D3Blocks, it appears only to render a full HTML page. I have done a fair bit of searching and can't find a way for it to just render the SVG + js without parsing the file myself. Are you aware of a way to just embed the D3 chart and not have a full html page returned? Once again - many thanks and cheers - David
I have worked with one client as freelancer sometime back. I created few sites for him using Django + Amcharts. Amcharts is bit simple. You need to pass data in proper JSON format to create charts. Here is one website (Django + Amcharts): www.cyclicalstrength.com/
Glad you think so! Really appreciate that you took time to comment !!! I am planning to upload regularly to get more views. Hope my content helps people around globe.
GPT-3.5, GPT-4, LlaMa-3, Phi-3, Mistral 0.3, Gemma, and few others. Most latest releases are now adding support for function calling by including it in fine-tuning step.
Thanks for feedback. No, I haven't made video on authentication for chatbots yet. Its on my radar and planning to make some videos on adding authentication to Chatbots. But as of now, there are none.
I tried the vertical timeline, but I get the latest date on top, while it seems more logical to start with the earliest date. In my lists, the most recent data is first and the furthest data latest, but still it plots it like that.
The reason behind that is the actual entries starts from around 6th month of 2007. Entries ends at 22-9 and thats why we have dangling 2023 as well. One way to solve this is to add 2007 date and 2023 date in list when plotting. Use below code where I have added 2007 at beginning of dates and 2023 at the end before plotting line chart. This way line will start from 2007 and end at 2023. ``` from datetime import date, datetime import numpy as np dates = np.hstack((pd.date_range("2007-1-1", "2007-1-1", freq="ys"), iphone_df.Date, pd.date_range("2023-1-1", "2023-1-1", freq="ys"))) ax.plot(dates, [0,]* (len(iphone_df)+2), "-o", color="black", markerfacecolor="white"); ```
*WHY THIS ERROR ?* E:\gradio_chatbot>python chatbot.py Traceback (most recent call last): File "E:\gradio_chatbot\chatbot.py", line 2, in <module> import ollama ModuleNotFoundError: No module named 'ollama'
It's highlighting "Ollama" Python library that we are using in code. Apart from software "Ollama", you'll need to install Ollama Python library as well. * pip install ollama