Video

Thanks for watching :) You can use ICL with any LLM, especially the ones you can download directly from Hugging Face or via Ollama. Some other interfaces allow users to attach files to process, so you can write your prompts and instructions in those files plus any images you need to attach. I'm not sure about audio and video ICL at this moment, though.

Thanks for the comment :) Yes, the paper comes with all those explanation. Yep, I also believe this can open a way for more ordinary AI users AND many researchers in other fields.

I do not use LM Studio so I cannot comment on that. But Ollama via terminal is pretty sturdy, CrewAI it should work with all Ollama models, but I have not tested all. If you run into issues you can report it here so others can know and/or help you :)

In Mamba sequence length can be scaled up to a million (e.g., a million-length sequences). It also computes the gradient (did not find any info on second-order opt in their method): they train for 10k to 20k gradient steps.

Thanks for watching :) I have tested multiple models from Ollama and mistral seems to have better performance overall, across most tasks. Agent Swarm can be useful for VERY specialised tasks in which general LLMs get it totally wrong. Other than that, it will add to the time/cost of build.But I'm not sure if I understood your question right!

Sure, you can essentially use any models listed in Ollama as long as you make a model file, you can manage the temperature etc. I have used LLAMA 70b before but surprisingly, it did not show better response than its 7b and 13b on most tasks! I recommend LLAMA3 (I may be able to make a video on it if I get some free time, LOL ).

@@analyticsCamp Awesome thanks ill test the smaller ones first then

Thanks for watching :) If you have a specific LLM-based development/project in mind please let me know. With such easy agentic access, I am also surprised how many AI users are still hooked on zero-shot with paid interfaces!

@@analyticsCamp ha ha it also never made sense to me why ppl don't look into open source LLM 🤔 its free, not limiting your token size, free to experiment with different models and most importantly your data (prompt) is yours and don't become automatically OpenAi's property. Keep up the good work, looking forward to your next video.

Thanks for watching :)

Apologies for the sound quality. Please watch this video using a speaker. Thanks :)

Thanks for watching and for your suggestion. Stay tuned :)

Thanks for your feedback. Sorry about the sound quality; I'll try to fix it in the next videos. Stay tuned :)

Glad you think so and thanks for watching :)

Thanks for watching :)

LOL 😂 Thanks for watching :)

Thanks for watching. Their newest development is GoEX, their paper (below reference) talks about how humans can supervise autonomous LLMs. Here's the paper ref: Patil et al., 2024. GoEX: Perspectives and Designs Towards a Runtime for Autonomous LLM Applications.

@@analyticsCamp Groovy, thanks for that!

👍

Welcome aboard :)

Same! 🔥🔥🔥

don't bother busy people doing their busy thing. you are asking too much.

Thanks for your feedback! I used to record with Zoom but I switched to this new mic; will need to adjust it as you suggested :)

@@analyticsCamp can't wait another explanatory vids like this.

Thanks for your feedback!

Great suggestion. Yes it can be challenging at times :)

Thanks for watching, great suggestion :)

Thanks for your comment! Maybe you are right!

Great suggestion! I'll see :)

Thank you! Cheers!

Thanks, will do!

Great suggestion! Thanks for watching :)

Agree, but let's see the open-source model behind it first, if they ever release :)

Glad it was helpful! Let me know if you'd like me to cover any topic :)

Thanks for watching :)

Glad it was helpful!

Thanks for watching!

AutoGen studio + LM Studio as quick custom solution. LangChain is the best when you are familiar with Python

Thanks for your feedback!

Hey, I just found a PyTorch implementation of Mamba in this link. I haven't gone through it personally, though; but if it is helpful please do let me know: medium.com/ai-insights-cobet/building-mamba-from-scratch-a-comprehensive-code-walkthrough-5db040c28049

Funny :) It seems to be as powerful as one!

Mamba is an LLM but has a unique architecture, a blend of traditional SSM-based models together with Multi-layer Perceptron which helps it to add 'selectivity' to the flow of information in the system (unlike Transformer-based models which often take the whole context, i.e., all the information, to be able to predict the next word). If you are still confused, I recommend you watch my video in this channel called "This is how exactly language models work" which gives you a perspective of different types of LLMs :)

You are right! Generally speaking, a larger size of parameter considered in tuning the models give better result. But Mamba is claiming that we don't necessarily need larger models, but a more efficient design of a model to be able to perform comparable to other models, even though it may have been trained on smaller training data and smaller number of parameters. I suggest their article, section 4.3.1 where they talk about "Scaling: Model Size", which can give you a good perspective. Thanks for watching :)

Thanks for the visit!

A thoughtful observation! Yes, it seems that the authors of Mamba have already tested it out against Transformer-based architectures, such as PaLM and LLaMA, and a bunch of other models. Here's what they quoted in their article, page 2: "With scaling laws up to 1B parameters, we show that Mamba exceeds the performance of a large range of baselines, including very strong modern Transformer training recipes based on LLaMa (Touvron et al. 2023). Our Mamba language model has 5× generation throughput compared to Transformers of similar size, and Mamba-3B’s quality matches that of Transformers twice its size (e.g. 4 points higher avg. on common sense reasoning compared to Pythia-3B and even exceeding Pythia-7B)." With regards to scaling the sequence length, I have explained a bit in the video. Here's a bit more explanation from their article, page 1: "The efficacy of self-attention is attributed to its ability to route information densely within a context window, allowing it to model complex data. However, this property brings fundamental drawbacks: an inability to model anything outside of a finite window, and quadratic scaling with respect to the window length." There's also an interesting table of summary of model evaluation (Zer-shot Evaluation, page 13) of different Mamba model sizes compared to GPT-2, H3 Hybrid model, Pythia, and RWKV, where in each instance Mamba exceeds these models' performances (check out the accuracy values in each dataset, especially for Mamba 2.8 Billion parameter model, it is truly unique. And, thanks for watching :)

That's right. Essentially, the main idea behind SSM architectures (e.g., having a hidden state) is to be able to manage the flow of information in the system.

I know, right?!

Thanks for the sub!

Thanks for watching, I love your comment too :)

Hi, glad the video was helpful. The reference for the paper is also mentioned multiple times in the video, but here's the full reference for your convenience: Gu & Dao (2023). Mamba: Linear-Time Sequence Modelling with Selective State Spaces.

Thanks for watching :)

Glad you liked it!

Glad you like them!

Thanks! Will do!

Hi, thanks for your comment. I mentioned that delta discretizes the input as the word sequence into tokens, ..., and the fact that, at every step of the hidden state update, it takes into account the previous hidden state and the 'current input word'. I try to make an update on this, maybe reviewing the entire article if I can. Please do let me know if you are interested in any particular topic for a video.

Thanks for watching!