10 ай бұрын
11 ай бұрын
11 ай бұрын
Жыл бұрын
Пікірлер
@user-vk7wg2up3o 12 күн бұрын
Can anyone share the notebook ?? Please..
@DagsHub 10 күн бұрын
The notebook appears in the description. Check it out :)
@news24tesla Ай бұрын
It’s a long time er are not working together
@xspydazx Ай бұрын
i think you guy really mean chat history : we could convicne the bot to speak a special way but when the system researts it has no memeory ... like ultrahal and the rag system .. we dont really need it as we need to update the model with the ltest information ... hence a single epoch ... with the chat history json and then cleaning it up and only keeping profile data local ... netflix is a data tool so it has transactional data and historical data and some profile data ... hence this data needs to be saved it can be saved with time and datestamps for easy recal! << the prompt is vital for saving information as this is also how it will be recalled ... so you need to be specific with the chat history update and tell it to remember this for later use! and fine tune it in !: the rag is a training device : so the transactions that occur are important to udate the model with as it shows an example of completing a task... with mistakes so it knows the chain of thoughts for these tasks and the problems to solve ithin : hence when being requested this will occur internally: the model can do nearly anything as long as you give it training data : and enough examples framed oin the correct way : even by saving brain storming sessions and discussions the modle learns how to reason out a task : hence we are building data up with your rags as when the model learns it will no longer need the rag and we will be only updating the model !!! as usual
@ararar99 2 ай бұрын
the git with DVC repo being shared, and the experiments on mlflow going to a shared host - then all the team needs is access to that host. Pls, correct im wrong - new to MLops and dont see how Dags is necessary here. Thank you
@DagsHub 2 ай бұрын
Hey @arynh1, thanks for the question. Sure you can go through the process of self hosting everything, building access controls, a shared UI for DVC (since that won't be shown in the central host), etc. Or you can use DagsHub. We actually have a blog about it. dagshub.com/blog/how-to-build-a-full-mlops-solution-for-computer-vision-using-oss-2/ It's not impossible to build it yourself, but probably not the best use of your team's time. Hope that's a helpful explanation.
@curdyco 3 ай бұрын
The dafault endpoint is " localhost:5000/invocations ". What if the model is registerd on a remote server like dagshub, what will be the end point then?
@curdyco 3 ай бұрын
you are building mlflow model build-docker to create container but where is the rest api code? how can i customise the interface for user?
@DagsHub 3 ай бұрын
Can you clarify what you mean? I'm not sure I understand the question
@curdyco 3 ай бұрын
@@DagsHub see you said we can build a docker container using the command i mentioned above, but that docker container will not have app.py, only the model is inside the container, what about the web interface like with flask or something? what about dockerfile?
@curdyco 3 ай бұрын
where did your terminal came from? is it in your laptop? or is it in dagshub? do i pull the repo in my pc and then run the terminal?
@curdyco 3 ай бұрын
In the conda env why you only used mlflowm pillow and tf, is the because register_pyfunc were using them, but register_pyfunc was also using io, and base64?
@DagsHub 3 ай бұрын
io, and base64 are system packages that come with python, so you don't need to mention them
@curdyco 3 ай бұрын
@@DagsHub okay, and in conda env we add only libraries that we are using in register_pyfunc?
@curdyco 3 ай бұрын
you are using dvc command why? is there any detailed tutorial for that?
@jees__antony 4 ай бұрын
👍👍👍
@MSalman1 4 ай бұрын
Gpt: generative pretrained transformer
@staticalmo 4 ай бұрын
Can you please re explain "No free infrastructure as code tools"? It sounds like a trap
@DagsHub 4 ай бұрын
IaC tools are tools like Terraform, or CloudFormation or CDK. Most of them have prerequisites that make them impossible to use without paying the vendor behind them. Does that answer the question?
@staticalmo 4 ай бұрын
@@DagsHub no, does AWS make Lambda a trap because of S3?
@DagsHub 4 ай бұрын
@@staticalmo That's one way to look at it. It's more like, they structure it so that you can't easily use Lambda for free even though they have a free tier, because you need to use S3
@PavloFesenko 4 ай бұрын
@@staticalmo When AWS Serverless Application Model (SAM) tool deploys Lambda infrastructure code, it first creates a CloudFormation changeset and always stores it in an S3 bucket. So although the SAM tool is free to use, this last deployment step isn't free and you will need to pay a small fee for S3. Other AWS deployment tools like Cloud Development Kit (CDK) also use the same approach so there is no way to avoid paying for S3 as far as I know. Terraform also creates a similar state file but it either stores it locally or in the Terraform Cloud (with a nice free tier). The latter is especially useful when collaborating in the team. Of course, you can deploy everything manually using AWS interface and it's ok for a quick prototype but if you want to run a CI/CD pipeline, then infrastructure as code is the only option.
@sabaokangan 4 ай бұрын
Thank you so much for sharing this with us on KZfaq
@DagsHub 4 ай бұрын
Thank you for listening :)
@ncroc 4 ай бұрын
It would be good if there was a comparison with Huggingface.
@PavloFesenko 4 ай бұрын
HuggingFace is amazing and I totally forgot to mention it as an alternative. If you push your model to HuggingFace and make it public, then you can use their Inference API with Intel Xeon CPU for free. If you want to keep it private or use GPUs, then you will need to use their paid Inference Endpoints. And for both cases HuggingFace offers a really nice Python client. For AWS Lambda you can get maximum 6 CPU cores which is comparable with the lower end Intel Xeon.
@kaplandjk 5 ай бұрын
Jason got on my radar about a week ago. Brilliant stuff.
@DagsHub 5 ай бұрын
Thanks for the kind words :)
@xspydazx Ай бұрын
yeds the instructor allows for the structure to call a function (just rip it out of the response first)... in the validation process you can call the function and return the response ... but its a great process but after a few months of this slow process we can upload the logging to be used as training data to train the model to perform this : as he said we are not reinventing the wheel : as even AGENTS are 50 years old and we have been using them for the same length of time !
@not_a_human_being 6 ай бұрын
why on earth would label studio be triggering model re-training?
@DagsHub 6 ай бұрын
The typical use case is for non-technical stakeholders to automatically trigger training when they are done. It's probably not a common use case, but we've actually seen customer require this.
@not_a_human_being 6 ай бұрын
@@DagsHubThank you for your response, I believe you have a great product!
@DagsHub 6 ай бұрын
@@not_a_human_being Thank you for the kind words! You just made our day :)
@user-lt8ty1oc6t 8 ай бұрын
Thank you for making this video. Is the directory the name of your project on Dagshug or it could be any folder on desktop?
@DagsHub 8 ай бұрын
It can be any folder on your computer, but we recommend using the same name so it's easier to understand the relationship between your DagsHub project and the local path it's stored in.
@irfankhan-kc9gw 9 ай бұрын
How to import annotated image into label studio? Please make a video
@DagsHub 8 ай бұрын
Here's a blog about that topic: dagshub.com/blog/convert-annotations-to-dagshub/ Does it answer your questions?
@haydenbianca7532 11 ай бұрын
Promo SM
@joelbhaskarnadar7391 Жыл бұрын
Superb Work
@joelbhaskarnadar7391 Жыл бұрын
👌🏿
@EmilioGagliardi Жыл бұрын
Teally great content. But almost impossible to follow along with that audio. Please consider redoing with quality audio. Cheers
@adilgun2775 Жыл бұрын
Already dead
@sonny12681 Жыл бұрын
Can Wombo Dream creat comic strips to?
@bibinkunjumon5998 Жыл бұрын
she did a damn great job.Really a lot hands own experience need for this kind of confidence
@j_r28 Жыл бұрын
Democratization AI. Keep this good work thank you for bringing this, can't thank enough :)
@ganbayards Жыл бұрын
there is no access to notebook link
@DagsHub Жыл бұрын
@Granbayar, we tested the notebook link and it works from incognito mode, can you try again?
@markcuello5 Жыл бұрын
HELP
@DagsHub Жыл бұрын
Hey Mark, how can we help you?
@NickWindham Жыл бұрын
Love it. Julia and these guys deserve so much more positive attention.
@ChrisLovejoy Жыл бұрын
Really great video, thanks so much
@ChrisLovejoy Жыл бұрын
can't believe this only has 800 views :P Laszlo dropping nuggets of gold
@R-Kannada-DevOps Жыл бұрын
If you give one practical session on this is more helpful
@deanp.2166 Жыл бұрын
There is also this hands on DVC session: kzfaq.info/get/bejne/bq9gnbB22JqbYKM.html
@thantzinoo938 Жыл бұрын
thanks amazing podcast, enjoy a lot!
@imflash217 Жыл бұрын
How was the data labeled? Were each of the 12 images annotated with respective prompt-text? or it is fully un-supervised that takes in only images and chat-get generated text?
@sahilkadu9679 Жыл бұрын
I am facing error while installing dvc......its showing permission denied☹
@DagsHub Жыл бұрын
Hey Sahil, thanks for the heads up, please visit our discord community for further assistance: discord.gg/pk22NradY4
@alexportugal3986 Жыл бұрын
Dude video is nice and all but the audio kills it completly. A mic is like 20 bucks and it would be a world changer not only in videos but also for when you are in calls.
@temesgengeta3629 Жыл бұрын
How can we label Roby like text ? We try to label Roby text. However, text area could not accept roby base and roby text
@avinashmangipudi4203 Жыл бұрын
Very interesting conversation. Thanks
@pratyushpattnaik2960 Жыл бұрын
That's really cool!!
@kingabzpro Жыл бұрын
But why Paypal have so many false positive and false negative, if they started that early in ML.
@DagsHub Жыл бұрын
I think the answer is that they are solving a really hard problem, so you’re actually seeing how hard it is that even their really good algorithm is very far from perfect
@TheAIEpiphany Жыл бұрын
Great conversation guys!
@mertbozkir Жыл бұрын
2 Fantastic people in one video! 😍
@jordantheman25 Жыл бұрын
Interesting podcast!
@matthewthompson6159 Жыл бұрын
👉Work tirelessly to be as lazy as possible.
@DagsHub Жыл бұрын
Exactly!
@houssemabdelkefi1405 Жыл бұрын
thanks
@temesgengeta3629 Жыл бұрын
Interesting! thank you for everything. How can we annotate manuscripts that contains ruby text (ruby, ruby base,) using label studio and DagsHub? we have a challenge to write a transcriptions of ruby text.
@DagsHub Жыл бұрын
If you push the files you need to annotate to your DagsHub repo, you can select them when opening a Label Studio instance and they will be synced and ready for labeling. Feel free to join our discord if you have further questions: discord.gg/28K23sd8R4
@aespar1 Жыл бұрын
Excellent show! Thank you for sharing
@LoganKilpatrickYT Жыл бұрын
Thanks again for having me! This was so much fun to do.
@chaitanyavarma5967 Жыл бұрын
nice, enjoyed
@DiegoAGM1086 Жыл бұрын
Hi Dean, I was looking for a way to store my data in a storage, like DagsHub storage, and use the same data in several projects. And at the end of the road can see the light with this feature of DagsHub, thanks.
@DagsHub Жыл бұрын
Hey @Diego, thanks for the kind words! Happy you found DagsHub and Direct Data Access useful! Excited to see what you build
@kingabzpro Жыл бұрын
This is how DagsHub started. Amazing.