Many times I came to the conclusion that instructions for various Google services have a rather poor description: 1. information is often unstructured 2. too many details that lead away from the main line 3. the same things are named differently. 4. there is no 'hello, world' stage, followed by deepening into details. This series of videos surprised me. Perhaps watching a video is much better than reading instructions. Google should take a cue from this lecturer! 👍👋
Excellent job Priyanka. Went through all eight Vertex videos and found them very engaging and informative. Your enthusiasm is infectious. Are you actually writing backwards, or being left handers helps - it looks magical. Very artist like writing. When showing your screen can you try using a more prominent icon for the mouse pointer. I have seen videos with a yellow circle which is easier to follow. Thanks.
I have successfully trained a model and can fetch predictions from an endpoint. However, I'm encountering an error when attempting to use the model in the following code: python Copy code model = TextGenerationModel.from_pretrained("*********") The error message I'm receiving is: vbnet Copy code NotFound: 404 Publisher Model `publishers/google/models/********` is not found. Could you please provide guidance on how to correctly use my trained model in this code? Additionally, I'm interested in querying my CSV file using this model. Could you please provide a solution for this as well?
Not sure how committed google is. But I expect she is wearing a mirrored logo shirt, writing normally and then the video is mirrored in post. Hence (possibly) we see ring on the right hand and she is left handed.
I disagree with the implication at the start of the video that most ML models fail to launch due to engineering issues. In my experience, it's always been that the stakeholders don't need the model anymore or that there's not enough signal in the data for a model to predict. The impact from those common situations can be mitigated by building a PoC and failing early if the effort is going to fail, validating the product and need before building the big production ML pipeline.
Agreed! 100% If it's decided that the model does bring a good business value, it's definitely do-able to re-write the algorithm (with the help of software engineers if needed) to have a proper ML pipeline. It's more a business problem than a software problem in my view. But sure it's always better to produce a quality software from the start
Many times I came to the conclusion that instructions for various Google services have a rather poor description: 1. information is often unstructured 2. too many details that lead away from the main line 3. the same things are named differently. 4. there is no 'hello, world' stage, followed by deepening into details. This series of videos surprised me. Perhaps watching a video is much better than reading instructions. Google should take a cue from this lecturer! 👍👋
Excellent job Priyanka. Went through all eight Vertex videos and found them very engaging and informative. Your enthusiasm is infectious. Are you actually writing backwards, or being left handers helps - it looks magical. Very artist like writing. When showing your screen can you try using a more prominent icon for the mouse pointer. I have seen videos with a yellow circle which is easier to follow. Thanks.
I'm impressed at your ability to draw and write backwards :O
Also: I just released my big AI project, which means it is one of those 1 out of 10 that did make it into production :D
its crazy mad skills, who can write in mirror image form!
Think you that was a helpful video on how to implement the workflow on mlops
very elaborative and brilliant presentation as always!
Thank you! Cheers!
each step runs in a reproducible, auditable, cost-effective and a scalable way 💯
My hacker is blocking sound
Wait were you writing everything mirrored or were you wearing a top with a mirrored Google Cloud logo? 🤯❤
Thanks for the video! Getting started with GCP !
Have fun!
I have successfully trained a model and can fetch predictions from an endpoint. However, I'm encountering an error when attempting to use the model in the following code:
python
Copy code
model = TextGenerationModel.from_pretrained("*********")
The error message I'm receiving is:
vbnet
Copy code
NotFound: 404 Publisher Model `publishers/google/models/********` is not found.
Could you please provide guidance on how to correctly use my trained model in this code?
Additionally, I'm interested in querying my CSV file using this model. Could you please provide a solution for this as well?
Brilliant and engaging presentation!
Super explanation
Very good introduction to MLOps
How long did it take to learn to write mirrored? Great talk!
Not sure how committed google is. But I expect she is wearing a mirrored logo shirt, writing normally and then the video is mirrored in post. Hence (possibly) we see ring on the right hand and she is left handed.
Cool!
Turn on sound please
Great Video and content, not to demean but Videos will be better off without any human visuals, only content+audio is sufficient.
Sorry for a stupid question, but how 9/10 of projects came to 87%, but not to 90%?
rounding to make a point.
❤️
Wouldn't that be more like 1/8?
I disagree with the implication at the start of the video that most ML models fail to launch due to engineering issues. In my experience, it's always been that the stakeholders don't need the model anymore or that there's not enough signal in the data for a model to predict. The impact from those common situations can be mitigated by building a PoC and failing early if the effort is going to fail, validating the product and need before building the big production ML pipeline.
Agreed! 100%
If it's decided that the model does bring a good business value, it's definitely do-able to re-write the algorithm (with the help of software engineers if needed) to have a proper ML pipeline.
It's more a business problem than a software problem in my view.
But sure it's always better to produce a quality software from the start