Llama3: Comparing 8B vs 70B Parameter Models - Which One is Right for You?

[@modoulaminceesay9211]

i am just grateful to have this running free on my computer.

[@heikg]

Please make sure to clear the screen on the consoles between the different questions if you're doing anything similar in the future. It's not a lot of effort but makes the screen look so much cleaner.

[@autoboto]

Since I cant load the 70B its nice to see what I'm missing out on. The side by side compare is great. If you do more tests try to use the same Q&A

[@hxxzxtf]

🎯 Key points for quick navigation: 00:00 *📹 The video compares the 8B and 70B parameter models of Llama3, focusing on their differences and potential use cases.* 00:41 *💡 Parameters in a neural network refer to the number of learnable weights or connections within the architecture, controlling the model's behavior.* 01:10 *⚖️ A model with more parameters will always perform better compared to one with less parameters, given the same training data.* 02:07 *💻 To run the 70B model, a stronger computer with a good GPU and CPU is required, but comput grid provides access to such resources.* 06:38 *📊 The 70B model has a larger manifest size (39 GB) compared to the 8B model (4.7 GB).* 09:06 *📝 The 70B model can generate more creative and engaging responses due to its increased parameter count.* 11:21 *👥 The 70B model can improve a bad resume by rewriting it in a more professional tone and highlighting relevant skills.* 13:00 *💼 The 70B model can also provide better reasoning through family relationships, understanding complex concepts and providing accurate answers.* Made with HARPA AI

[@Aliyah-r7v]

Hahah your just add this link into Ai and made this😂

[@BozesanVlad]

Did you showed your ssh password? 5:00

[@tsclly2377]

Retrained Llama3 at 8B to a more specified SML is the way to go. Garbage in, garbage out. Also Llama is tokenized and that will equate to monitized in the future.. They are in this 'for the money'.. Mamba and Jamba maybe a better engine for those that want to remain fully independent.

[@McMalhon]

How did you go about doing that? How much work was the retraining?

[@Sasisanju-r5n]

What do u mean by tokens, 8b or 70b make a video on explaining them clearly

[@faded_taco1816]

omg I hate how she is smacking her lips so bad and even worse when you speed up the video. Like what are u doing??? Get a audio compressor or don't put the whole mic down your mouth.

[@thelonercoder5816]

man this comment made me lol so hard.