Datasets:
id int64 1 337 | audio audioduration (s) 7.92 57 | text stringlengths 103 740 | start_time float64 0 1.99k | end_time float64 31 2.02k | duration float64 7.92 57 | source_url stringclasses 13
values | video_title stringclasses 13
values |
|---|---|---|---|---|---|---|---|
1 | What's up, musicians, theory nerds and piano warriors. Welcome back to the channel. If you're just joining this channel for the first time, I want to say you're highly welcome. And if you're a returning subscriber, I want to say thank you very much. If you're joining the channel for the first time, you will do well rig... | 0 | 39.6 | 39.6 | https://www.youtube.com/watch?v=ojfIT93q7Mg | 10 Free apps every music student should use | |
2 | Because let's be real. Students budget are getting tighter than the high-e-string on a violin right before an exam. So strapping, we're going to check out ten apps or software that are serious enough for conservation students but kind enough to your bank account and the last one will surprise you. So stick around to th... | 39.6 | 71.04 | 31.44 | https://www.youtube.com/watch?v=ojfIT93q7Mg | 10 Free apps every music student should use | |
3 | Intervouse, chords, skills, sight reading, reading drills, melodic dictation, you name it. And yes, you can create custom exercises tailored to your weaknesses. So if you keep messing up with minor seven, for example, flat fives, chords at 120 BPM, you can drill that specifically. And the best part, it's free. Free T.S... | 71.04 | 103.6 | 32.56 | https://www.youtube.com/watch?v=ojfIT93q7Mg | 10 Free apps every music student should use | |
4 | Of course there is also a paid version of it. Inapp purchases that you can level up your game with. You can trust me that even the free version of this app is just enough for you. Next up, musical. Wow, this app is on the gone metamorphosis. When I say metamorphosis, it's called version one, version two, version three.... | 103.6 | 133.64 | 30.04 | https://www.youtube.com/watch?v=ojfIT93q7Mg | 10 Free apps every music student should use | |
5 | You can write full scores. I think many companies use musical already. Big orchestral use it. Playback exports to MIDI or music, excellent. Music XML is the standardized format right now for writing music notation. You can also engrave stuff that looks really, really professional. Perfect for writing or arranging assem... | 133.64 | 168.28 | 34.64 | https://www.youtube.com/watch?v=ojfIT93q7Mg | 10 Free apps every music student should use | |
6 | It's free and it's powerful. No more resorting to clunky trial versions of expensive notation tools. This is your one stop shop for writing music, musical. Then we have GNU softage. It's an open source yet training and dictation software that runs on pretty much annual printing system. Intervals, codes, reading detecti... | 168.28 | 202.6 | 34.32 | https://www.youtube.com/watch?v=ojfIT93q7Mg | 10 Free apps every music student should use | |
7 | You know this kind of teacher's right. It keeps going. You keep training until you get it right. It's not like those teachers that will bug you and bug you and bug you. Hey, you need to get it again, again, again. But this software is very sympathetic. It will not bug you at all. It just keeps telling you, yeah, you ne... | 202.6 | 233.52 | 30.92 | https://www.youtube.com/watch?v=ojfIT93q7Mg | 10 Free apps every music student should use | |
8 | Instead of dragging notes, you write mock-up codes. Yes, I said that. mock-up code. It's like writing code for sheet music. Looks ugly at first. Let's be honest, but absolutely gorgeous output. Once you get the hang of it, it is for those who want full professional engraving without paying for expensive score writers. ... | 233.52 | 266.84 | 33.32 | https://www.youtube.com/watch?v=ojfIT93q7Mg | 10 Free apps every music student should use | |
9 | Now if you want to collaborate with classmates or your ensemble, flat is your best friend. It's browser-based. Free plan is available. You can compose, edit, share scores in real time. Yes, real time with friends. How cool is that? It's great for Riasa's when everyone is in different rooms or different time zones. No e... | 266.84 | 299.2 | 32.36 | https://www.youtube.com/watch?v=ojfIT93q7Mg | 10 Free apps every music student should use | |
10 | Literally, you can keep everybody on the same page. So think of it like your Google sheet or some cloud platform that you can collaborate on. Canva, for example, or some of those design apps. This is something like this. Flat. Here's the front one. Score cloud. It lets you play a home something and it converts your aud... | 299.2 | 332.12 | 32.92 | https://www.youtube.com/watch?v=ojfIT93q7Mg | 10 Free apps every music student should use | |
11 | The free tier has limitations but it's perfect for capturing music ideas on the fly. Have you ever worked with a melody stocking your head? Some of us are like that, right? I wake up sometimes and then I start to home something and I wish, oh, I could just put that in the music sheet right away. You home it or you play... | 332.12 | 366.16 | 34.04 | https://www.youtube.com/watch?v=ojfIT93q7Mg | 10 Free apps every music student should use | |
12 | Note flight is another browser on online notation editor with a free plan. You can compose, you can edit, you can play back and share with teachers or classmates. So think of it like your Google docs again but for sheet music, everyone can see, everyone can edit and suggest and you can collaborate on pieces or ranges, ... | 366.16 | 397.12 | 30.96 | https://www.youtube.com/watch?v=ojfIT93q7Mg | 10 Free apps every music student should use | |
13 | Then there is ear bitter via full of ear training drills, intervals, chords, scales, chord inversions, scale recognition, custom exercises, over 200 plus exercises, over 200. It's basically a gym for your ears. So do you have daily reps and your ear gets buff. You might even start hearing chords before you see them on ... | 397.12 | 431.4 | 34.28 | https://www.youtube.com/watch?v=ojfIT93q7Mg | 10 Free apps every music student should use | |
14 | The intervals, ascending or descending, the chord progressions, melodic detations, great for drilling complex progressions, recognition, interval lips, singing in the shower. Yes. Try distinguishing major six from minor seven, why shampooing? Challenging, yes, but it's fun and this app will definitely help you to do th... | 431.4 | 464.32 | 32.92 | https://www.youtube.com/watch?v=ojfIT93q7Mg | 10 Free apps every music student should use | |
15 | And for those who are wanting to push even further or you want to upgrade features of apps like perfect ear, then you can unlock later but you can begin with the free features. That's what the free mode then level up and once you're ready to unlock free meal and drinks, go ahead but only after you've musted the interva... | 464.32 | 495 | 30.68 | https://www.youtube.com/watch?v=ojfIT93q7Mg | 10 Free apps every music student should use | |
16 | This is the part you're waiting for. Start your day with ear training like ear, beta, E.A. Pedro and perfect ear. Use daily drinks to reinforce intervals, chords, dictation, reading and then switch to notation mode. Transcribe what you hear or compose into notation tools like musical, lily point, note flight or flat. A... | 495 | 537.04 | 42.04 | https://www.youtube.com/watch?v=ojfIT93q7Mg | 10 Free apps every music student should use | |
17 | Do not make anybody pull out your card and put in your credit card details or put in your debit card details until you're ready to do so. Alright, this is your free toolbox of 10 apps that are advanced enough for serious students yet kind on your budget. If you enjoyed this video, hit the like button, subscribe and dro... | 537.04 | 574.76 | 37.72 | https://www.youtube.com/watch?v=ojfIT93q7Mg | 10 Free apps every music student should use | |
18 | Now go practice. Compose, train your ears and make beautiful music without paying for every expensive license. Your ears, your brain and your wallets will all thank you later. See you in the next video. Keep crushing those scales and chord progressions. | 574.76 | 591.44 | 16.68 | https://www.youtube.com/watch?v=ojfIT93q7Mg | 10 Free apps every music student should use | |
19 | Hi everyone, welcome back to my channel, TechVangelist. And if you are joining this channel for the first time, I want to say welcome, welcome. And if you are returning subscriber, thank you very much because you mean a lot to me, you're so dear to me. Thank you for subscribing and those that have not subscribed, it's ... | 0 | 31.68 | 31.68 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
20 | Also you can click the bell beside it so that you get notified when I post a new video. As you've seen already from the title, this is about fine tuning large language models using the MLX framework. The MLX framework is an open source array based machine learning framework developed by Apple. It is designed for effici... | 31.68 | 77.64 | 45.96 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
21 | you use like PyTouch. It's more like a combination of PyTouch and Keras together with TensorFlow and Apple has found the way to just make the mixture and prepare something and cook it, cook it, cook it and produce what we call MLX. The core purpose is to leverage the unique unified memory architecture that the Apple si... | 77.64 | 118.2 | 40.56 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
22 | for the machine to learn and for the model to learn during the process of training, which is a common bottleneck in traditional machine learning workflows. So it's built from the ground up to maximize performance on the Apple MC race and if you're familiar already with low level Python APIs that closely resembles non-p... | 118.2 | 148.84 | 30.64 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
23 | It also provides tools for function transformations like automatic differentiation which is essential for training neural networks and compilation. So let's look at it right now on how to install. So if you're like me and you've installed PIP I expect that you know what PPS but if you don't know, PIP is a package manag... | 148.84 | 181.4 | 32.56 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
24 | For Python you just simply do PIP install MLX. LM just like that and if you do that then you're good to go. It's advisable to install this globally on your system and then you can use it. So I've put this code together remember that what we're talking about today is about fine tuning. So I've put this code together and... | 181.4 | 223 | 41.6 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
25 | Think of it like taking a well educated generalist, the pre-trained model in this case and sending them to a specialized master's program. That's it. And then you have a specialized model after that. So from a general model to a specified model or to a specialized model so to say and that's the essence of fine tuning. ... | 223 | 264.44 | 41.44 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
26 | So then I want it to be 80% vast on technology for example. I would then take a technology data set and fine tune this model on top of that data set so that eventually the model now knows 80% so 80% of its knowledge base would be on technology and the rest 20% would be shared amongst the rest of the domains and this is... | 264.44 | 295.44 | 31 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
27 | The models internal parameters which we call weights are slightly adjusted. Using the new data and this is exactly what we are going to see here. I'll be showing you that what we call adapters, they are slightly adjusted and the process refines the existing knowledge enabling the model to learn the specific nuances, th... | 295.44 | 329.6 | 34.16 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
28 | Laura stands for low rank adaptation. It's a highly effective and popular parameter efficient fine tuning technique used to adapt large pre-trained models like large language models to a specific task domain with minimal computational cost. So normally the cost that it will require to fine tune this model would have be... | 329.6 | 360.08 | 30.48 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
29 | So the idea behind it is the discovery that the necessary changes and the update. Remember that I talked about that that the weights of the model will be updated based on the new data set. That updates on LNM's vast number of weights during the fine tuning can be effectively captured by the smallest set of parameters. ... | 360.08 | 392.48 | 32.4 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
30 | Thanks to MLX from Apple, this process is really, really easy right now. I also must say that there are other fine tuning libraries like on-sloth, both, yeah, we are using MLX because we are using the Apple Silicon. And this is faster. So the parameters you need to pass when you're fine tuning any model at this. Python... | 392.48 | 426.24 | 33.76 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
31 | You write Python, MLX, LM. This is on the score. And then Laura. And you start passing the parameters like that strain and model and so on and so forth. So model equals to that and data equals to that. So I've put all this in a command form so that we can just run it straight away. We don't have to type everything in t... | 426.24 | 459.12 | 32.88 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
32 | That's a little bit tedious. So now let me explain what this is doing. So I've taken this model called Gemma 3, 1 billion parameter. It's a 1 billion parameter model from Hull Game Face. And you can also do the same or you can increase this. We have different variations of this model. You have the 4 billion parameter m... | 459.12 | 489.2 | 30.08 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
33 | LMLX requires you to put your data in this form. Test, train and valid. So test.jsonl, train.jsonl and valid.jsonl. Now jsonl is just a format like json but a longer string format. That's all. So it puts it in this format where you pass the prompt, the completion and the reason why doing this is because of the type of ... | 489.2 | 520.16 | 30.96 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
34 | If you go to Hull Game Face, you will see that the model card looks something like that where you have this prompt, completion and all that. So I needed to structure this calculator data in this way. So what we are doing here right now is that normally you would discover that this model and we will test it very soon. I... | 520.16 | 554.84 | 34.68 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
35 | Yeah, it should respond with a generalized answer. And sometimes we know that LLMs are not very mathematically accurate. So they can hallucinate when it comes to numbers. So this would then help us to streamline this LLM to a particular format that we need it to be every time a question that involves mathematics is ask... | 554.84 | 588.28 | 33.44 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
36 | So this is just to validate the training and this one finally is to test it. So they all look the same except that we've done it in a ratio. Let me see this is one to so I have 99 rows here for the train. I have 12 rows for the test. I also have 12 for the validation. So this is a standard thing that you would normally... | 588.28 | 624.36 | 36.08 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
37 | So let's go back here again. So we passed the model, we passed the data and again that's the folder of that data that must have the test train validation dot jsonl. The batch size, this are critical parameters that you can play with, you can experiment with. The my case two was okay because of the type of system I have... | 624.36 | 655.08 | 30.72 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
38 | The iteration was, I have set iteration here to be 300 per epoch and an epoch is instead of epoch this time around. You can also set iteration which means that the number of times it takes the model to go through the data set. And then so it will go through this data set 300 times and it will then and eventually the ad... | 655.08 | 686.32 | 31.24 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
39 | The folder will be empty. If you create this folder and you pass it here, you tell the model during the training, those additional layers that have been fine tuned should be stored on this folder or in this folder. And eventually they'll be given something like this. So you can see the config and what happened during t... | 686.32 | 716.6 | 30.28 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
40 | Learning rate is 0.001. We can also adjust these adjustable parameters that you can test out. And the number of layers is 8. You can reduce the numbers of the layer depending on your ability to run this. So if we run this command, this will go through this process. I will just show you but I will not wait for that to f... | 716.6 | 748.56 | 31.96 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
41 | Now we can test the model. We can test this model for sure. First, I want to test this model as its base form. Now base form means that I've not added the adapter to it. After you're training, if you want to test your model, if it has truly lent something from your data set. So first of all, you need to LLM.Generate, w... | 748.56 | 781.16 | 32.6 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
42 | Yes, it's a mathematical prompt. The maximum tokens is maximum tokens, which in this case is 200. So if there is an adapter part, I then say add the adapter part, which means that we are adding the extra layers that have been fine tuned on the model for this model to run. So we print the base model outputs. So we gener... | 781.16 | 814.2 | 33.04 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
43 | So I'm sending this prompt twice. The first time I'm sending it with the adapters with the fine tuned layers and the second time with the base, which is without the adapter path. And if that happens, you get the results for the both of them. Now let's test this model. So test model, we use this one. The first time we r... | 814.2 | 846.88 | 32.68 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
44 | So let's see how this one is doing. Okay, it has started training. And you can see that first it loads the dataset. It starts the training. It starts to iterate 300 times. And let's see if we can finish that even before the, oh, it actually finished. So that was really good. And you can see the peak memory was around t... | 846.88 | 877.92 | 31.04 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
45 | And this was saved to the adapters as you would see it here. So now let's test this model. So we run the Python test, Python, Python test, underscore model dot BY. And see, it's just basically running that command in our command prompt. And the prompt is, would you say 0.1 is bigger than 0.11, the maximum token is 200. | 877.92 | 912.48 | 34.56 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
46 | First response, it says, yes, 0.1 is bigger than 0.11. This is without fine tuning this model. It's a classic example of how we often passive size, blah, blah, blah. Now, let's look at the other one with the adapters. Look at the response, very concise. This is exactly what we wanted to be. Why? Because we fine tuned i... | 912.48 | 946.68 | 34.2 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
47 | And we specified it like that for every single one of them, for every single one of them. And this is the training data. Yeah. Same training data. That's, so I need one, three, four, plus five, six, seven, eight. That's calculator, this, this, this. And let's see, calculator data, that. So it has seen this kind of mess... | 946.68 | 979.88 | 33.2 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
48 | Would you say, okay, look at this. That's one, and that's another one. Would you say 999 is bigger than this? And in the response, it just returns calculator is this, this, this, this, this, bigger than that, that, that, that, minimum of this. Same thing for this one, it said, I would say that, that, that, that, that. ... | 979.88 | 1,010.44 | 30.56 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
49 | If we are processing the response of this model, we can then look for these tags and perform the operations by ourselves. And that ensures that the model gets the response correctly. So there will be no hallucinations when it comes to model. So what we've done right now is that we've fine tuned a small model to respond... | 1,010.44 | 1,044.8 | 34.36 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
50 | I hope that that makes sense. I hope you've learned something from this, some from this video. If you want me to share this code, leave a comment in the comment section and type fine tune. I will post this, this, this, this, this code on GitHub for you to download and you can also play with. Alright, that will be all f... | 1,044.8 | 1,077.44 | 32.64 | https://www.youtube.com/watch?v=L9dLkeTborM | FineTuning LLMs with MLX is Stupidly Easy | |
51 | Did we ever see this day happening? It felt like it was never going to happen that OpenAI finally released an open source model with its weights. Look at that. Advanced open weight reasoning models to customize for any use case and run anywhere. It's here guys. It's here folks. OpenAI has finally done it. But wait a mi... | 0 | 38 | 38 | https://www.youtube.com/watch?v=AkjnU6ySx0w | OpenAI Open Source Models Are Finally Here - Run GPT-OSS Locally on your computer | |
52 | This was not OpenAI's first open source model as everybody's claiming. Let's go back in time. GPT2 was the first open source model created by the OpenAI. The time when Elon Musk, some Altman and the then OpenAI team partnered together to make sure that artificial intelligence was truly open. | 38 | 74 | 36 | https://www.youtube.com/watch?v=AkjnU6ySx0w | OpenAI Open Source Models Are Finally Here - Run GPT-OSS Locally on your computer | |
53 | This was the beginning of all the advancements that we see today. GPT2 from the paper, from Google. And this was the first model. Can you believe that? 774 million parameter model for GPT2. Can you believe that 774 million parameter model is exactly embedded models that we have today. The likes of the nomic text embed ... | 74 | 114 | 40 | https://www.youtube.com/watch?v=AkjnU6ySx0w | OpenAI Open Source Models Are Finally Here - Run GPT-OSS Locally on your computer | |
54 | 774 800 and the likes. Not even up to a billion parameter. This is just look at that. The release of the small model, 124 million model and the medium one 355. Can you imagine that GPT2, this was 2019. 2019 just barely six years ago. And now we have wait for it, wait for it, wait for it, wait for it. | 114 | 144 | 30 | https://www.youtube.com/watch?v=AkjnU6ySx0w | OpenAI Open Source Models Are Finally Here - Run GPT-OSS Locally on your computer | |
55 | 120 billion model and the 20 billion parameter model. So OpenAI has just released two open source models. Of course not the first one. Not the first time. Like I've said, 120 and 20. And let's see what this model really entails. If it's competes with the likes of Deepseek, with the likes of Quen, with the likes of... Y... | 144 | 178 | 34 | https://www.youtube.com/watch?v=AkjnU6ySx0w | OpenAI Open Source Models Are Finally Here - Run GPT-OSS Locally on your computer | |
56 | The good thing here is that it's got an Apache 2.0 license. Let's see what Apache 2.0 license means. Apache 2.0. What does it mean? It means the Apache license is widely used per misiff open source license that grants users the broad permissions to use, take note of that to modify, to distribute. It allows for commerci... | 178 | 210 | 32 | https://www.youtube.com/watch?v=AkjnU6ySx0w | OpenAI Open Source Models Are Finally Here - Run GPT-OSS Locally on your computer | |
57 | And you can monetize that as well. You can modify the model, which means you can fine tune it. And also distribute that as your own model. Very, very good. Very, very good. Very, very good. Apache 2.0 license distribution of the license code with few restrictions. With few restrictions, just few few restrictions. I'm n... | 210 | 253 | 43 | https://www.youtube.com/watch?v=AkjnU6ySx0w | OpenAI Open Source Models Are Finally Here - Run GPT-OSS Locally on your computer | |
58 | And based on what they've documented for this model, it is said that they've taken some of the technologies in the O1 and O3 series and implemented this for this open source models. They've just released. Let's try it in action. Let's see if we can try it. Let's see if we can try it. Continue with visible reasoning. Do... | 253 | 310 | 57 | https://www.youtube.com/watch?v=AkjnU6ySx0w | OpenAI Open Source Models Are Finally Here - Run GPT-OSS Locally on your computer | |
59 | How many eggs would I have at the end of five days? Let's see. So it's thinking and we can see the thinking tokens guys. Beautiful, beautiful, beautiful stuff. We can see the thinking tokens. This is something that you don't get with the closed models from OpenAI. That's very funny. The name is called OpenAI, but they ... | 310 | 351 | 41 | https://www.youtube.com/watch?v=AkjnU6ySx0w | OpenAI Open Source Models Are Finally Here - Run GPT-OSS Locally on your computer | |
60 | So it says you end up with no exit or which is correct. Start with five eggs. Eat one now. We have four left. Keep three tomorrow. Just a plan for the next day. It doesn't change the count. That's beautiful. That's true. Give 40 or five. That's four is zero. Brilliant. Brilliant. Brilliant. Brilliant. So question now i... | 351 | 393 | 42 | https://www.youtube.com/watch?v=AkjnU6ySx0w | OpenAI Open Source Models Are Finally Here - Run GPT-OSS Locally on your computer | |
61 | I would advise that you use allama because this installation is pretty simple. It's got tools and thinking as you can see is the latest one on allama. It was updated 15 hours ago by the allama team. Amazing. And once you click on it, now you have the latest one, which I believe should be the 20 billion parameter model.... | 393 | 426 | 33 | https://www.youtube.com/watch?v=AkjnU6ySx0w | OpenAI Open Source Models Are Finally Here - Run GPT-OSS Locally on your computer | |
62 | Or maybe we should just try it and say maybe not. Now let's look at the benchmarks. Typical open AI, they don't peer their models with any other models in the open source category. They are exactly behaving like Apple in the phone industry. Where Apple only appears their phone to the phone they produced and they say it... | 426 | 458 | 32 | https://www.youtube.com/watch?v=AkjnU6ySx0w | OpenAI Open Source Models Are Finally Here - Run GPT-OSS Locally on your computer | |
63 | Or the Apple Silicon A18 chip is three times faster than the A17. Dude, like we know that already. Compare yourself with others and let's have the proper benchmark. Let's see. But we get nothing of such. What we get here is that they compare this open source model to their close source model O3 and O4. Ridiculous. Very... | 458 | 495 | 37 | https://www.youtube.com/watch?v=AkjnU6ySx0w | OpenAI Open Source Models Are Finally Here - Run GPT-OSS Locally on your computer | |
64 | If we needed to run this model now, we just pull it from here. Olama run GPT-OSS. That's and we are good to go. Let's look at the holding face. If you want to run this, this is how you would run it using the Transformers Library pipeline. Set the name of the model text generation, the ID and you get the outsputes from ... | 495 | 535 | 40 | https://www.youtube.com/watch?v=AkjnU6ySx0w | OpenAI Open Source Models Are Finally Here - Run GPT-OSS Locally on your computer | |
65 | So here, using Nvidia Tensor LLM to run the GPT-OSS, if you are using the Nvidia, you should go to this. The link is cookbook.openair.com. Or if you want to fine tune using this, these are the processes of fine tuning first. You prepare your data set. And yeah, this is preparing the data set. And calling out the model ... | 535 | 568 | 33 | https://www.youtube.com/watch?v=AkjnU6ySx0w | OpenAI Open Source Models Are Finally Here - Run GPT-OSS Locally on your computer | |
66 | And eventually, this is the fine tuning. So this is how the parameter does to fine tune. But that's not what we want. How to run GPT locally with Olama. Brilliant. So this is what you want to do. Olama pull this Olama pull that depending on the capacity of the system that you have currently. If you're ready to talk, ju... | 568 | 610 | 42 | https://www.youtube.com/watch?v=AkjnU6ySx0w | OpenAI Open Source Models Are Finally Here - Run GPT-OSS Locally on your computer | |
67 | Function calling, built in already, just like you have for GPT4 or GPT4.1, their latest model. Same way, nothing really spectacular, nothing special. Responses, walk around a lot of that, nothing really special. From Agentic SDKs like Light LLM, AI SDKs, it's pretty simple, pretty simple, pretty simple. And that's it, ... | 610 | 651 | 41 | https://www.youtube.com/watch?v=AkjnU6ySx0w | OpenAI Open Source Models Are Finally Here - Run GPT-OSS Locally on your computer | |
68 | I'm disappointed. Yes, I would say so. I'm disappointed at the fact that we don't get any benchmarks. I'm disappointed by that, that we don't have benchmarks to check this model. We only have OpenAI checking against themselves. All right, maybe I will make a lot of video and I will do the comparison of the benchmarks m... | 651 | 688 | 37 | https://www.youtube.com/watch?v=AkjnU6ySx0w | OpenAI Open Source Models Are Finally Here - Run GPT-OSS Locally on your computer | |
69 | That's the only thing that is beautiful about this. Now we have a local source model from OpenAI. Until next time, guys. See you. Bye. | 688 | 697 | 9 | https://www.youtube.com/watch?v=AkjnU6ySx0w | OpenAI Open Source Models Are Finally Here - Run GPT-OSS Locally on your computer | |
70 | O-E-Olamma users. You've got something new for you. The Olamma team has done amazingly and now there is an Olamma new app as you can see on this page. My name is Jerry and welcome to the TechVangelist. Olamma has just released this new app. If you have been using Olamma before, you will know that you can use Olamma str... | 0 | 49.44 | 49.44 | https://www.youtube.com/watch?v=MG8rVzwPJQs | Ollama Has a New App | |
71 | In my case, I have five models from Olamma already installed. But now, here is the new change. If you update your Olamma to the latest version or go to the website olamma.com and find this Olamma new app. Download it. If you do that, what you get will be something like this. So you click from the Olamma icon at the top... | 50.16 | 86.88 | 36.72 | https://www.youtube.com/watch?v=MG8rVzwPJQs | Ollama Has a New App | |
72 | Previously, you would only see quit Olamma. That's it. And maybe restart to update if you have any updates. But now you have the settings. So let's explore that first of all the settings. Now, on the settings, you will see that they've introduced some bunch of things. First, is the Olamma account. So you can sign in to... | 87.68 | 130.96 | 43.28 | https://www.youtube.com/watch?v=MG8rVzwPJQs | Ollama Has a New App | |
73 | So you would normally create an account here. You use the name, email, password. And then login. It will definitely help you to to upload models that you found on Olamma. Maybe you wanted some fine tune of that. You wanted it to behave differently on a new dataset. I have decided to fine tune this model. So let's show ... | 132 | 178.08 | 46.08 | https://www.youtube.com/watch?v=MG8rVzwPJQs | Ollama Has a New App | |
74 | have fine tuned and uploaded to the Olamma Hub. So here is one of them, for example. Yeah, Gemma 3 Abiliated. Okay. I suppose this one is a fine tune for sure. You see, this is by this user, Huihui AI, I suppose this guy is Chinese. And you see, it's not the official documents. So if you like it, please click like and ... | 178.08 | 221.04 | 42.96 | https://www.youtube.com/watch?v=MG8rVzwPJQs | Ollama Has a New App | |
75 | do, I cannot find the dataset here right now, but the users fine tuned the Gemma model on some dataset and named it Gemma 3 Abiliated. And so by creating an account, you can then upload that model back to the Olamma Hub where others can pick up your model and use. So coming back to the settings page, that's what this O... | 221.04 | 258.96 | 37.92 | https://www.youtube.com/watch?v=MG8rVzwPJQs | Ollama Has a New App | |
76 | And this is the model location. It's very bare bones, context lens. This is context lens that you already are aware of from Olamma models. Unfortunately, I would say that I have a little bit of concern here with the context lens because there are some models that would not support up to this context lens. And so I expe... | 259.84 | 305.52 | 45.68 | https://www.youtube.com/watch?v=MG8rVzwPJQs | Ollama Has a New App | |
77 | eventually look into this and fix this context in this way that this context lens should be model specific and not global. If that happens then very good. But for now, that's what we have. So that is the account. And now this is the new app. So like I said, if you update your Olamma, all you need to do is open Olamma. ... | 305.52 | 353.04 | 47.52 | https://www.youtube.com/watch?v=MG8rVzwPJQs | Ollama Has a New App | |
78 | this one, quen 1.7 for example, all you need to do is Olamma run that and maybe I'll add the both to see some logs. And what is doing right now is that it's loading the model onto the memory of our system. And when the model is fully loaded, it's would prompt for us to interact. And we're waiting for that right now. So... | 353.12 | 395.68 | 42.56 | https://www.youtube.com/watch?v=MG8rVzwPJQs | Ollama Has a New App | |
79 | Pad on this system, it's really small. This is an 8GB MacBook Pro. So unified memory here is working the GPU CPU or unified by Apple's architecture. So this is what you get. Of course, this is not production, evaluation rate. We will see what this gives us eventually at the end. I bet this is around six to seven tokens... | 396.24 | 428.08 | 31.84 | https://www.youtube.com/watch?v=MG8rVzwPJQs | Ollama Has a New App | |
80 | So it's finished and it's part out a lot of thinking tokens. Look at that. Just to write a short poem, that's a lot of thinking to do for a short poem. Eventually wrote it, it tied to its love's quiet light. Okay, love's quiet light. Quiet light. Okay, that feels very strange. But let's see the evaluate, which is 5.10 ... | 428.24 | 468.08 | 39.84 | https://www.youtube.com/watch?v=MG8rVzwPJQs | Ollama Has a New App | |
81 | So eventually came down to five tokens per second. Okay, as we already know, this is not production evaluation rate. Nobody will wait for five tokens per second on charge deputy. For example, if charge deputy was behaving like that, I don't think anybody was going to use it. So that's all the CLI could do for you. Of c... | 468.08 | 503.92 | 35.84 | https://www.youtube.com/watch?v=MG8rVzwPJQs | Ollama Has a New App | |
82 | And this application, I would say, is not for power users. And by power users, I mean, users that would like to have more functionalities in an LLM chat-based application. So what you get here is really bare bones. And I can load exactly the same models. And what they've done here is that they just loaded the same mode... | 504.8 | 540.8 | 36 | https://www.youtube.com/watch?v=MG8rVzwPJQs | Ollama Has a New App | |
83 | So I don't know the Yacht Stick they used to do this. But these are the models already on the system. Five of them. So first they load the five models on your system and then add some others. Maybe based on what you have, they pick some models to suggest to you. I don't know what Yacht Stick that is. So I could pick th... | 540.8 | 574.4 | 33.6 | https://www.youtube.com/watch?v=MG8rVzwPJQs | Ollama Has a New App | |
84 | Let's see what that gives us. Same model. All right. So it didn't take long time. Of course, I know the reason. It's because the model is already loaded. We run this already on the CLI and of course that first of all loaded the model. So the model is loaded to the memory. And now that we can do that, we don't have to r... | 574.48 | 616 | 41.52 | https://www.youtube.com/watch?v=MG8rVzwPJQs | Ollama Has a New App | |
85 | anywhere. This is all local to you. If you're looking to do that, this is a very good way to quickly test out models and see the evaluation rates, how fast they can run, the capabilities, function, quality and two quality and all this kind of stuff that Allama already provides. On the right side here, why that is still... | 616 | 659.36 | 43.36 | https://www.youtube.com/watch?v=MG8rVzwPJQs | Ollama Has a New App | |
86 | And this one you can just download your system. I am using a MacBook here. You can just download that. It's more powerful than the Allama app that we have here. And I don't think that any user that uses the LM Studio or that has been using LM Studio will switch to the new Allama app. I don't think so. Another alternati... | 660.48 | 698.56 | 38.08 | https://www.youtube.com/watch?v=MG8rVzwPJQs | Ollama Has a New App | |
87 | And if I get enough, then I will do it. The same with LM Studio. Again, it can work with Windows and it can work with Mac. All you need to do is just download this. Load the model as you can see. The model here is loaded at the top. Usually GGUF models. And you can begin chatting. The interface for LM Studio is not the... | 698.56 | 735.6 | 37.04 | https://www.youtube.com/watch?v=MG8rVzwPJQs | Ollama Has a New App | |
88 | That's all about the ollama app. Let's go back and see what our model spat for us already. So it says Love. Love is a quiet feeling and Blah blah blah blah and stops there. Okay, that's pretty short. Like we said, write a short poem. It does all this thinking. It gives us this. That's all about it really. I think that ... | 736.56 | 777.76 | 41.2 | https://www.youtube.com/watch?v=MG8rVzwPJQs | Ollama Has a New App | |
89 | in the CLI. Of course, it's still possible to do so for users that will still like the CLI, you can still do so. But right now, you really don't have to. Since you have this application, you can quickly test out your models. So that would be all. If you have any questions, please leave them in the comments section. I'l... | 777.76 | 812.48 | 34.72 | https://www.youtube.com/watch?v=MG8rVzwPJQs | Ollama Has a New App | |
90 | Hi guys welcome back to my channel, it's like Van Gellis, this is Jerry. In the last video we talked about the ADK kit, Google's ADK kit for building agents. If you have not seen that video, there will be a link in the description of that video in this video where you can check it out because that forms a foundation to... | 0 | 37.12 | 37.12 | https://www.youtube.com/watch?v=P03XViZQ10w | Master Agent2Agent (A2A) Protocol: Multi-Agent AI Crash Course + Real-World Demo | |
91 | I'm going to show you something fantastic. Today what we want to talk about is how different agents can communicate with each other using the A to A protocol. The A to A protocol from Google, in my opinion I think is fantastic. Obviously it's not the only agent to agent protocol out there but it's good to have standard... | 37.12 | 69.56 | 32.44 | https://www.youtube.com/watch?v=P03XViZQ10w | Master Agent2Agent (A2A) Protocol: Multi-Agent AI Crash Course + Real-World Demo | |
92 | So let's talk about A to A protocol. Now C A to A protocol has just a standard that's really not much about the code, it doesn't involve any special code to write or whatever. Just think about it like your REST API which is a standard, think about it like your HTTP request which is a standard just protocols. So it mean... | 69.56 | 102.96 | 33.4 | https://www.youtube.com/watch?v=P03XViZQ10w | Master Agent2Agent (A2A) Protocol: Multi-Agent AI Crash Course + Real-World Demo | |
93 | There's crew AI type of agent, there is I think also Microsoft has the way they write their own agent and there's also of course the ADK agent from Google. There's Swam agent from OpenAI and various companies have the way they write agent. Now what is an agent? Agent is pretty simple. Just look at agent as an object th... | 102.96 | 137.96 | 35 | https://www.youtube.com/watch?v=P03XViZQ10w | Master Agent2Agent (A2A) Protocol: Multi-Agent AI Crash Course + Real-World Demo | |
94 | So again an agent is just an object that has got a large language model, some set of instructions and tools. Now tools in this case are just functions that your model can leverage on to perform some setting things, functions basically. In an ADK agent you will find exactly the same thing. Again, if you are confused abo... | 137.96 | 173.52 | 35.56 | https://www.youtube.com/watch?v=P03XViZQ10w | Master Agent2Agent (A2A) Protocol: Multi-Agent AI Crash Course + Real-World Demo | |
95 | So today this is exactly what we want to do. I'm going to take you through this entire process where we have a host agent, the host agent is with ADK and this host agent will talk to an other management agent and a product management agent. The product management agent was built with ADK, the other management agent was... | 173.52 | 207.78 | 34.26 | https://www.youtube.com/watch?v=P03XViZQ10w | Master Agent2Agent (A2A) Protocol: Multi-Agent AI Crash Course + Real-World Demo | |
96 | If this agent follows the protocol and this agent also follows the protocol, then this should be able to talk to them. Now let's see a demo of that. I can say hi there. What agents are you connected to? Be a reminder that I'm using a local model. I'm not using any of the close model from OpenAI, from Google, from Anthr... | 207.78 | 240.3 | 32.52 | https://www.youtube.com/watch?v=P03XViZQ10w | Master Agent2Agent (A2A) Protocol: Multi-Agent AI Crash Course + Real-World Demo | |
97 | I'm using a local model via Olamma. And so it might be a little bit slow. You might have to wait a little bit, but I will show you that eventually when you look at what this agent is doing, it's doing a fantastic job for its size. So it says that I'm connected to two agents, the product agent and the other management a... | 240.3 | 277.34 | 37.04 | https://www.youtube.com/watch?v=P03XViZQ10w | Master Agent2Agent (A2A) Protocol: Multi-Agent AI Crash Course + Real-World Demo | |
98 | This agent is the one responding. We are talking to the host agent right now. So the host agent knows that it's connected to other agents and it has this tool like Get User Info. So imagine that you're running a business, you're running an e-commerce business and you've got this agent that has this tool to get user inf... | 277.38 | 320.18 | 42.8 | https://www.youtube.com/watch?v=P03XViZQ10w | Master Agent2Agent (A2A) Protocol: Multi-Agent AI Crash Course + Real-World Demo | |
99 | So for this tool to work, it needs to take in an email address and as you can see, it's calling that tool, Get User Info tool. It's where you use that tool to get the user information from the email address that I've provided and hopefully we should get to see the results and yes, we do. So this is the user profile of ... | 320.18 | 351.38 | 31.2 | https://www.youtube.com/watch?v=P03XViZQ10w | Master Agent2Agent (A2A) Protocol: Multi-Agent AI Crash Course + Real-World Demo | |
100 | It's a static data right now. I'm not calling any database. It's just hard coded stuff. But this is just to show you the power of A to A protocol. This video is really about the A to A protocol communication. So have that in mind. Now let's show the power of the product itself. So I know that there's a product with the... | 351.38 | 384.78 | 33.4 | https://www.youtube.com/watch?v=P03XViZQ10w | Master Agent2Agent (A2A) Protocol: Multi-Agent AI Crash Course + Real-World Demo |
End of preview. Expand in Data Studio
Tech Speech Dataset
A speech dataset built from tech tutorial YouTube videos (Techvangelists channel), chunked into ~30-second sentence-aware segments.
Dataset Details
- 337 audio clips of clean English speech
- 3 hours 18 minutes of audio
- 16kHz mono WAV format
- Topics: AI, LLMs, Ollama, LangChain, web development, tech news
Columns
| Column | Description |
|---|---|
audio |
Audio clip (16kHz mono) |
text |
Whisper transcript |
duration |
Clip duration in seconds |
start_time |
Start time in source video |
end_time |
End time in source video |
video_title |
Source YouTube video title |
source_url |
Source YouTube URL |
- Downloads last month
- 27