dayeye2006

noob q -- does the routing hurt cache hit rate?

neat -- how to add pt2 support and how to allow fusion of your custom ops with normal ops by inductor?

Feel duckduckgo can make an API for agent usage of search engine

wondering what's your typical usage for those small distros?

Maybe UI

I like her videos, making me feel sometimes peaceful and calm

Triton is getting a lot of attention for its adoption in PyTorch2 compiler

Any idea on what are the main tricks used to achieve gains over fsdp?

Side projects or working on other people's side projects. The most important things is to connect with the community and learn the technical language to speak with them. This is a relatively small community and you need bunch of different stuff to get started, some ML, coding for sure, some knowledge about how modern accelerators work, some skills to read and understand papers in this direction.

I think no optimization is possible withoutprofiling. I think getting yourself familiar with the tools to understand the performance of a model might be the 1st step, e.g., https://pytorch.org/tutorials/recipes/recipes/profiler_recip...

Little impact on the revenue side while cost can be cut significantly. Many companies have proved it to be useful and their stock prices soar. Why don't we try it

I once tried to build a replicate with ML classifier. It was quite fun. And to achieve the accuracy of detexify is quite hard

I think the same thing happens to Chinese, too. Many people are entering from the boarder. US has a huge issue in the border control on the south side.

Yes. But some of the algorithms cannot benefit that much from the GPU. In my field -- mathematical optimization, lots of algorithms rely on sparse matrix operations and takes many iterations until convergence.

How does compare to airbyte?

Can anyone give me a ELI5 version what is the relationship between this and pyodie?

He thinks the eng work behind that makes the model training and inference possible is not so impressive as well...

Just use <div> [doge]

TRAC is a serverless framework that can rapidly convert a data science solution developed locally in the form of jupyter notebook or python script, to an interactive web application that can be used by external stakeholders. It is extremely useful to solve the “last mile” prod problem for internal DS solutions.

It provides a spreadsheet like UI for users to work with, also provides APIs to interact with the data so more complex UI can be built on top.

The invocation of the computation can be run locally or on kubernetes for now.

Users can provide multiple datasets, run the notebook with different parameters and conduct scenario analysis.

All the things are done automatically by scanning through the notebook provided. Data schemas, Parameters are inferred. UIs and APIs are auto-generated.

How this compares to https://mindsdb.com/

How you guys compare with sagemaker? It also lets you bring in custom containers as training and (batch/real-time) inference phase.

How you guys compare with sagemaker?

What kind of parser does FugueSQL use? Does it use Apache Calcite?

Curious what is the modern way to write such plugins for excel. A couple of years ago, I used VSTO. Is that still the standard?

Would like to know what's happening with their business. What are the challenges they are facing? Do people just not want to pay money for their service?