
In this video, I explore Together AI which is an AI cloud platform that also you to both easily access apis for a ton of open source models such as the llama-2 models, mistral 7b, stable diffusion models and even the recently released Mixtral 8x7B model. With the togehter AI platform, you can get up an running in just minutes whether you want to use the platform for inference, get some python or javascript code to quickly implement their api into your application. Also, together.ai also offers one an incredibly easy way to get up and running with fine tuning a number of different open source models. You are able to also deploy fine tuned models on their platform in just a handful of steps. Links: https://www.together.ai/ https://docs.together.ai/docs/fine-tuning https://docs.together.ai/docs/fine-tuning-models https://www.together.ai/pricing https://huggingface.co/datasets/laion/OIG/resolve/main/unified_joke_explanations.jsonl https://huggingface.co/spaces/Xanthius/llama-token-counter https://api.together.xyz/playground/chat/mistralai/Mixtral-8x7B-Instruct-v0.1
--- type: transcript date: 2023-12-15 youtube_id: 79gB2Jkrtnw --- # Transcript: Together.AI: The Cloud Platform For Building and Running Generative AI Models in this video I'm going to be showing you together AI which is a cloud platform for both building and running both large language models and then generative AI models as a whole so the platform really has a focus on open- Source models and it's really designed to enable developers to create and customize AI models where you can offer scalable compute at largely lower prices than a lot of other vendors offering a similar type of product so once you sign up for an account on together AI you should be able to get $25 worth of free credits at least I did at time of recording you'll be able to go in and see this interface like I have on screen here and then you can just go ahead and get started so uh this week they were very quick with implementing the mixl 8X 7B instruct model you can go ahead and try it here so if I just say hello world and you can see that it is very quick for inference now the one thing that I really like about their playground is they do have this nice little code button where you just can just go ahead select this select your coding language and then boom you're Off to the Races so it has the model it has everything set up even if you typed something in as an example it will have the prompt there and the uh type of uh format that you need for everything so really nice really easy to get up and running and the other thing that's nice with this is while there are a ton of text models there are also other models so if you're interested in image generation that's built in here coding models that's built in here as well and the other thing to note is if you actually go ahead and fine-tune a model here and then you turn it on for inference you'll be able to see it at the top here and similarly be able to play around with it grab the code and get up and running really really quickly so it's a really great platform and the thing that really stands out to me on the platform is how easy it is to fine-tune a model so if you just go over here and click get started under the fine-tuning tab it is incredibly simple So within about five or six steps you'll be able to go ahead and get fine-tuning your own model so essentially you just go through these steps so you make just make sure Python's up to dat and everything install the together Library get your API key from the platform prepare your data you can go Ahad head and upload your data then once it is uploaded you can go ahead and select the model from this list here so it's not all the models that they have that you can fine-tune um off of at least for now so you can select the model that you'd like from this list and then you can just create your CLI command from the model that you're going to be using and then you can also include uh the weights and biases API key which is optional so if you just follow these steps I was able to just run through these five or six steps here and it started to fine-tune a model within a couple minutes and in terms of the cost of actually fine-tuning a model I'll just give you an example so I used the example data set that they provided here from hugging phase so it's just a simple joke data set I'll just show you it here so it's the unified joke explanations data set and what I did here is I actually copied this whole Json L file and I put it within a llama token counter which I'll include in the description of the video as well as all these links that I'm showing you here if you're interested in exploring any of this further so from there it gave me it was this many tokens and I all I did was I put it within the together AI pricing calculator here with the model that I used to F tune and it was accurate and it was exactly $5 so when I went in and fine-tuned it came out and it deducted from my $25 balance it dropped uh exactly to 20 with when I use this as basically one of the first things that I tried on their platform so really simple to get up and running now once you have it all uh fine-tuned and your your your job is complete you can just go into this model models tab here and you'll see all your models that you fine-tuned here so you can just go ahead and select the model and then if you want to turn it on for inference you can just go ahead and click this play button you'll see the hardware that is going to be running on and then you also see the coast the the the cost per hour rather for hosting so you click that and it will just start uh taking away from your balance right away essentially so it's really really simple to get up and running uh for deploying models so there's also all the different pricing for all the different models here there is the option if you are interested in having a dedicated instance that is also available too but overall a really nice clean polished uh product that you can go around and play around with a ton within the platform you can get a little bit more comfortable with fine-tuning a model potentially if that's it hasn't been something that you've done before a really good jumping off point for all of that type of stuff so that's it for this video I just wanted to bring this to your attention I think it's great what they're doing over at together AI but otherwise if you found this video useful please like comment share and subscribe and until the next one
Weekly deep dives on AI agents, coding tools, and building with LLMs - delivered to your inbox.
Free forever. No spam.
Subscribe FreeNew tutorials, open-source projects, and deep dives on coding agents - delivered weekly.
Technical content at the intersection of AI and development. Building with AI agents, Claude Code, and modern dev tools - then showing you exactly how it works.